This is a new subject dealing with apparent gains seen in diode based capacitor charging circuits utilizing the reverse bias parametric capacitance change of the diode or equivalent. This all came about from the study of Dr. Stiffler's work and his single wire experiments and disclosures.
This thread was originally private but is now for public viewing.
Regards,
Pm
Edit: 9.25.2018
Dear PM
Thank you for the invite. I am very interested in your work and hope to be able to make some positive contributions.
Regards
All,
This is a continuation of my previous tests with the latest result of the parametric charger using a rather large CSR of 100 ohms which is reducing the circuit Q but does give very consistent and accurate input power measurements without affecting the circuit performance greatly. The current probe just does not yield consistent measurements at these low currents! The schematic below shows the circuit used which has 4 paralleled 1N4148s per leg.
The scope pix below shows a measurement cycle over 100ms with a voltage across C1 at 50 ms of 38.45v mean. I would like to point out that the circuit has two phases of operation which is important as will be pointed out later on. This is seen in the large current draw from the start to the ~30ms point where the current then begins to taper off in a linear fashion.
The first data table below was taken with a 40ms horizontal scan and the second was taken over a 100ms horizontal scan with the same start conditions. These were taken at two different times and I would like to point out the close agreement of the input power measurements when comparing between these tables at the 10-40ms marks.
There appears to be gain in this device and the question is, what is the cause if this is true?
Regards,
Pm
The parametric charging (PC) circuit shown previously with an applied constant frequency will exhibit two phases as mentioned. These are due to the resonant frequency differences between C1 being shorted which results in a lower self resonant frequency and a fully charged C1 which will result in a higher self resonant frequency. This difference is created by the decreasing capacitance of the diodes as the reverse bias voltage increases across them due to the charging of C1 and this diode capacitance change is in series with C2 creating the varying resonant frequency over time.
Normally, the frequency from the generator is greater than the starting resonant frequency with C1 shorted or discharged. As C1 charges, the circuit's self resonance frequency increases to a point of being equal to the generator frequency (Phase 1) and then continues increasing to be greater than the generator frequency (Phase 2) and eventually stabilizes at some level depending on circuit Q.
In measuring and analyzing the previous tests during Phase 2, it seemed apparent that if the proper load value was placed across C1, a gain would be realized in continuous operation. The following scope pix demonstrate the results.
The first pix is the continuous power input to the circuit which is seen to be 11.54mw.
The second pix shows the voltage across C1 with a 46.5k load resistor attached to be 35.85v mean or dc. This equates to an output power level of 35.85^2/46.5e3 = 27.64mw.
This results in an apparent COP = 27.64/11.54 = 2.395. This is still flea power due to the small capacities and voltages involved but it does show promise IMO.
I might add that the probe measuring the voltage across the CSR uses the probe tip and ground spring directly connected to the resistor leads as is good RF measurement practice.
Regards,
Pm
ION had suggested adding a primary to L1 and driving the primary with the mosfet driver which allows a higher operating voltage level to the circuit. L1 is currently an air core wound on a 50mm dia pvc form with the primary wound over the secondary and is not an ideal transformer at these frequencies.
Anyway, this test is using the same 46.5k load resistor as previously and at the same frequency of 1.42MHz as can be seen in the schematic below. The resonance curve has changed but I thought the results would be interesting to post as the power level is creeping upwards.
The first scope pix is the Pin measurement which is 56.52mw. Note the sample rate of 2.5GS/s.
The second scope pix is the voltage across C1 with the 46.5k ohm load attached and is 75.81vdc. The output power is 75.81^2/46.5e3 = 124mw.
Therefore, the COP = 124/56.52 = 2.19.
Pm
Quote from: partzman on 2018.08.22, 22:17:25
ION had suggested adding a primary to L1 and driving the primary with the mosfet driver which allows a higher operating voltage level to the circuit. L1 is currently an air core wound on a 50mm dia pvc form with the primary wound over the secondary and is not an ideal transformer at these frequencies.
Anyway, this test is using the same 46.5k load resistor as previously and at the same frequency of 1.42MHz as can be seen in the schematic below. The resonance curve has changed but I thought the results would be interesting to post as the power level is creeping upwards.
The first scope pix is the Pin measurement which is 56.52mw. Note the sample rate of 2.5GS/s.
The second scope pix is the voltage across C1 with the 46.5k ohm load attached and is 75.81vdc. The output power is 75.81^2/46.5e3 = 124mw.
Therefore, the COP = 124/56.52 = 2.19.
Pm
Pm
If the 100r CSR is placed between Ls and the diode array,dose this still give you the same P/in value ?.
Im just thinking maybe most of the P/in is via single wire/avamenco plug type deal here-maybe ? via the positive path.
Added-also,have you included the dissipated power of the 100r CVR in any of your P/out calculations at any point in time ?
Brad
Ok,so now you have my attention Pm O0
Below is a pic of my setup,and the schematic to go with it.
I decided to give the 1n4007s a go as my diode array,and larger capacitor values to try and get the frequency down.
Well seems i have achieved that,as the frequency is down to just 27.9KHz O0
The wave forms on the scope are very clean,but making the calculations from the scope is going to be trick,as some of the forward voltage is being recycled through the transformer,and so average cannot be used to make the calculation.
In the schematic,you will see i have added another CVR on the front end,as i had my suspicions that the CVR in position 2 (CVR 2) would not be giving the correct information to use as the current CVR.
The wave form across CVR1 is very different to that of CVR 2.
For my P/in calculations i have used CVR1.
In saying all that,i think we must be missing something here,as my COP is very high :o
P/in(tonight) was calculated by using my DMMs,as at this low frequency they are very accurate.
V/in was taken by placing the DMM across the primary of L1.
I/In was taken by placing the DMM across CVR1,and measuring the voltage. Then current was calculated using ohms law.
The scope shows(oddly enough)that the voltage and current are dead in phase,and so the power factor is 1 O0
Both DMMs are a different brand,and both show the same voltage readings across CVR 1 and the primary of L1
Scope was used to confirm measurements.
So P/in was calculated to be--
I/in=165mV across 10 ohms=16.5mA
V/in=165mV across L1 primary.Scope average did not agree with DMM in this case--something to look into there.
Scope gave an average voltage of 355mV
Yes,oddly enough the voltage across the primary was the same as the voltage across CVR1 :o
So P/in =2.722mW
P/out is 26.2v over the 15k resistor.
The voltage across R1 C1 was taken with my DMM and scope. Both agree with each other.
P/out=45.762mW
Apparently the COP is 1681%-->!!Apparently!!
I find it hard to believe that both DMMs would be out the very same amount.
My top end DMM has been used many times,and is accurate with pulsed DC far above the frequency being used here.
!BUT! something is not right some where?
What we need is to brainstorm the most accurate way to calculate P/in.
P/out is no problem,as it is smoothed DC across a known resistance.
I will be doing the long hours on this tomorrow night for sure.
Brad
Added
I am driving the circuit with my SG.
Just out of curiosity,i removed L1,C2,and CVR 2,and without changing anything on the SG,i placed the SG across the diode array.
The maximum voltage achieved across C1/R1 was 12.4 volts,for a dissipated power of just 10.25mW
Interesting stuff O0
Brad
Quote from: TinMan on 2018.08.23, 02:40:23
Pm
If the 100r CSR is placed between Ls and the diode array,dose this still give you the same P/in value ?.
Im just thinking maybe most of the P/in is via single wire/avamenco plug type deal here-maybe ? via the positive path.
Added-also,have you included the dissipated power of the 100r CVR in any of your P/out calculations at any point in time ?
Brad
Brad,
Excellent question and one that I had considered but talked myself out of because normally, the current entering a network will equal the current exiting the network. This is true however, at these frequencies, there are obvious capacitive conduction paths that I have not accounted for because the current entering the diode/capacitor network is greater than the current exiting the same network. Therefore, the power measurements taken on the loaded device continuously running with the CSR at the output of the network are too low. In fact, when the CSR is placed at the network input and the current measured differentially, the COP = .79 or so.
Now the question is, has this affected all my measurements on all the previous tests? I'm afraid that perhaps this is the case but I will do the tests to find out. I have seen gains in the lower frequency range of 500-600kHz but these too will be suspect to error.
I've attached a pix of the 3.94uf cap used along with the attached diodes as one can see, I've at least made a nice 1-2MHz antenna array! :-[ There is a lot of surface area to this composite cap that is coupled to any ground surface in the area and I believe this is the basic problem involving the measurement error.
pm
This is a lower frequency test at 687kHz with a few value changes plus I'm back to the current probe placed at the input of L1. In order to raise the current measurement well above the noise floor, I have passed the sense lead thru the probe 4 times to raise the measurement level 4x and then compensated in the Math channel with a division by 4. This gives consistency of the current measurement with the probe and all the input current is accounted for.
The schematic below shows another change and that is the parametric device is now an RFP14N0L5 mosfet. This device has a BVdss breakdown voltage of ~50 volts but they typically will reach 65-70v as is the case with the devices used here. When the circuit reaches that level across C1, the voltage increase stops and the pin stays relatively constant.
The data table for a 100ms scan is also attached and can be seen, we still have apparent OU. The goal from here is to attempt reaching lower frequencies with gain.
Pm
Pm
I am down to a frequency of just 27.9KHz,and measured as you have,i get very high COPs
I believe that the measurement across your CSR 1 should be the RMS value,and not the mean value you have been using,as your transformer secondary will be feeding the circuit an AC .
The diode array is just a double diode FWBR,and i think in my case,it is working just as that-an FWBR.
I will spend some more time on it tonight,and try and work out why the COP seems so high.
Brad
Quote from: TinMan on 2018.08.23, 14:51:57
Ok,so now you have my attention Pm O0
Below is a pic of my setup,and the schematic to go with it.
I decided to give the 1n4007s a go as my diode array,and larger capacitor values to try and get the frequency down.
Well seems i have achieved that,as the frequency is down to just 27.9KHz O0
The wave forms on the scope are very clean,but making the calculations from the scope is going to be trick,as some of the forward voltage is being recycled through the transformer,and so average cannot be used to make the calculation.
In the schematic,you will see i have added another CVR on the front end,as i had my suspicions that the CVR in position 2 (CVR 2) would not be giving the correct information to use as the current CVR.
The wave form across CVR1 is very different to that of CVR 2.
For my P/in calculations i have used CVR1.
In saying all that,i think we must be missing something here,as my COP is very high :o
P/in(tonight) was calculated by using my DMMs,as at this low frequency they are very accurate.
V/in was taken by placing the DMM across the primary of L1.
I/In was taken by placing the DMM across CVR1,and measuring the voltage. Then current was calculated using ohms law.
The scope shows(oddly enough)that the voltage and current are dead in phase,and so the power factor is 1 O0
Both DMMs are a different brand,and both show the same voltage readings across CVR 1 and the primary of L1
Scope was used to confirm measurements.
So P/in was calculated to be--
I/in=165mV across 10 ohms=16.5mA
V/in=165mV across L1 primary.Scope average did not agree with DMM in this case--something to look into there.
Scope gave an average voltage of 355mV
Yes,oddly enough the voltage across the primary was the same as the voltage across CVR1 :o
So P/in =2.722mW
P/out is 26.2v over the 15k resistor.
The voltage across R1 C1 was taken with my DMM and scope. Both agree with each other.
P/out=45.762mW
Apparently the COP is 1681%-->!!Apparently!!
I find it hard to believe that both DMMs would be out the very same amount.
My top end DMM has been used many times,and is accurate with pulsed DC far above the frequency being used here.
!BUT! something is not right some where?
What we need is to brainstorm the most accurate way to calculate P/in.
P/out is no problem,as it is smoothed DC across a known resistance.
I will be doing the long hours on this tomorrow night for sure.
Brad
Added
I am driving the circuit with my SG.
Brad,
If you have time, could you take a scope shot of your input waveforms. Maybe we can work out from there how you are getting such a high COP. I can't seem to find a reverse bias capacitance curve for the 1N4000 series diodes and you are at such a low frequency that I'm not sure if you are seeing a parametric effect or not.
Does your scope have math capability?
Pm
Quote from: partzman on 2018.08.24, 00:50:33
Brad,
If you have time, could you take a scope shot of your input waveforms. Maybe we can work out from there how you are getting such a high COP. I can't seem to find a reverse bias capacitance curve for the 1N4000 series diodes and you are at such a low frequency that I'm not sure if you are seeing a parametric effect or not.
Does your scope have math capability?
Pm
I believe the 1n4000 series diode capacitance value is 15pf at 4v-1MHz.
My new 4ch scope has all the fruit,just need to learn how to use it.
I will be doing more testing tonight,and get you some wave forms.
The wave forms are preaty clean,and we can calculate the instantaneous P/in,then devide by duty cycle.
But there is a question with the voltage due to feedback,which i will point out tonight.
I'll get a video up,and show you what is going on.
Brad
OK Pm,let me see if i can get all this in order.
First pic is of a quick test i did.
Using two of the 1n4007s reversed biased,in series with an LED polarity correct,i connected the SG across the circuit.
The SG was left as it was when driving the diode array circuit.
The aim was to see if the diodes had enough junction capacitance at that lower frequency(30KHz)to run the LED.
Turns out the LED lit up quite well. Even at 10KHz,the LED will light.
Second is the schematic with scope probe positions,and component values--except inductance values of the transformer.
OK,the first scope shot shows the current(blue trace) across the 10 ohm CVR,and the voltage(yellow trace)across both the 10 ohm CVR,and the secondary of the transformer.
So this is where our P.in is taken from,as it is the input to the circuit.
The second scope shot is where i have removed everything other than the 20% on time.
What you see left in that scope shot is what is used to calculate P/in.
In the third scope shot,i have drawn a red box around the 20% on time of the SG.
The forth scope shot shows everything that is going on in the remaining 80% of each cycle,where i have removed the 20% P/in part of each cycle.
You will note that !some how! the value of current is higher in both directions during the off time,than it is during the on time--through the same inductor(the secondary).
You can see that the flyback current value is far greater than the current value during the on time,and then the bounce back current just before the next cycle starts,is also higher than that during the on time.
Now,i was always told that the maximum current you could get during the flyback period would never be higher than that of the maximum current flowing through the inductor during the on time period.
Seems they were wrong ?.
Anyway,my calculations are as follows.
I have calculated the voltage across the 10 ohm CVR during the 20% on period,to be an average of 35mV-or close to.
The voltage seems to be a steady 50v during the on time period.
So P/in is 50v X 3.5mA=175mW.
This is the power for the 20% on time,and so we divide that by 5 to get average power in.
So P/in is 35mW.
P/out is 28.6v over the 15kohm resistor
P/out = 54.53mW
COP=155.8%
Brad
I offer this Real Time Tuning Method to aid in finding optimum parameters.
I liken it to engine tuning, where you help the engine to run a bit with an external current source and adjust parameters for less dependency on the external current source. (External current source is like the "starter" in a ICE.)
A breakdown device is used to dump some energy from the C1 cap to shock excite a tank coil comprised of the L1 primary and the variable cap C2, which may be deleted if you want to run solely on the self resonance of L2.
The firing point of the breakdown device is another parameter that is part of the tuning, and it should occur as the system is approximately at the maximum parametric charging of C1.
I wish I had time to build this right now but am bogged down with other stuff. I think it would be fun to play with the tuning to see how great a dip in external supplied current could be achieved towards self running.
Regards
edit: post updated to show suggested starter circuit / external current source
Brad,
OK ,this helps. At first glance I question whether it is accurate to sample the secondary output and use the 20ms input pulse period to arrive at the input energy consumption but I'm probably wrong. I think it might be better to compute the entire secondary output voltage and current to arrive at the input to the diode/cap circuit ( this would require using your math channel) or, place the CSR back in the primary and measure the primary voltage and current for input power.
In any case, you may have something here placing the charging of a cap in a series resonant loop. I will try a single coil version of your circuit at low frequency to hopefully replicate what you've done as I don't have a transformer on hand that matches yours which has a leakage inductance of ~23mH.
Pm
All,
Here is a continuous run test based on the parameters and data for the circuit shown in post #9. After analyzing the data, I felt that there should be an energy gain running continuously at 687kHz with a load resistor across C1 that resulted in a C1 voltage of ~24v dc. Experimentally this value was found to be 15k ohms and the scope shots below show the results.
The first scope pix shows the input power to be 18.81mw.
The second scope pix shows the voltage across C1 with the 15k attached to be 24.61v for an output power of 24.61^2/15e3 = 40.3mw.
This results in an apparent COP = 40.3/18.81 = 2.14.
This is an easy method of proving or disproving the circuits ability to produce a gain. I have been trying different mosfets and diodes at these lower frequencies and this is the best combo thus far.
I think that the circuit can be reduced even further by using a full wave center tapped secondary with on two diode elements which should increase the parametric capacitance 2x as compared to the bridge circuit due to the fact that the parametric capacitances are connected in series for each polarity. This should possibly allow a lower operating frequency.
Regards,
Pm
Quote from: ion on 2018.08.24, 16:11:30
I offer this Real Time Tuning Method to aid in finding optimum parameters.
I liken it to engine tuning, where you help the engine to run a bit with an external current source and adjust parameters for less dependency on the external current source. (External current source is like the "starter" in a ICE.)
A breakdown device is used to dump some energy from the C1 cap to shock excite a tank coil comprised of the L1 primary and the variable cap C2, which may be deleted if you want to run solely on the self resonance of L2.
The firing point of the breakdown device is another parameter that is part of the tuning, and it should occur as the system is approximately at the maximum parametric charging of C1.
I wish I had time to build this right now but am bogged down with other stuff. I think it would be fun to play with the tuning to see how great a dip in external supplied current could be achieved towards self running.
Regards
ION,
For some reason I didn't see your post until now but I think it is a great idea. There seems to be too much to do and too little time at the moment but I plan to implement your circuit. On thing to note is the circuit can get unstable in the sense that if the load resistance is too high or the frequency too low, it will snap to produce maximum voltage across C1 and is very inefficient in this mode. The breakdown voltage of the 14N05L presently clamps the output to ~68v or it would head to........?
Anyway, more later.
Regards,
Pm
author=ion link=topic=3655.msg69355#msg69355 date=1535127090]
QuoteI offer this Real Time Tuning Method to aid in finding optimum parameters.
I liken it to engine tuning, where you help the engine to run a bit with an external current source and adjust parameters for less dependency on the external current source. (External current source is like the "starter" in a ICE.)
A breakdown device is used to dump some energy from the C1 cap to shock excite a tank coil comprised of the L1 primary and the variable cap C2, which may be deleted if you want to run solely on the self resonance of L2.
The firing point of the breakdown device is another parameter that is part of the tuning, and it should occur as the system is approximately at the maximum parametric charging of C1.
I wish I had time to build this right now but am bogged down with other stuff. I think it would be fun to play with the tuning to see how great a dip in external supplied current could be achieved towards self running.
Regards
.
Thanks ION
If you have time,could you join in and check my measurements in my next post O0.
Pm's circuit is very simple,and i recommend those here that can-join in.
Brad
First off,i do agree with Pm that my setup may not be working as intended(parametric charging),but the circuit is the same,and interesting results are being had.
This will be a long post,as i am at the point where great care must be taken in power measurements,and the events actually taking place at each point in time throughout the circuit.
My measurements are taken from continuous running of the circuit.
The P/out measurements are very simple,as it is a smooth DC voltage across a 15kohm resistor(value checked with 3 different DMMs)
OK,the first schematic is in reference to scope probe placement for P/in measurements.
Scope shot one shows current(blue)and voltage(yellow) traces for P/in measurements.
Scope shot 2 shows a reverse current flow through the primary of the transformer,even though there is a blocking diode on the positive input to the primary winding of the transformer. This will be explained a little later down.
Scope shot 3 shows a time base expanded view of the traces,and where i have averaged both the current and voltage values for the 25% on time(P/in time)
You will note that the current and voltage are in phase,and this can be explained easily a little further on down.
So my P/in is calculated as such
Average current for on time is 1.4v across the 10 ohm CVR (CVR1)
So our current during the on time is 140mA.
Our average voltage across the primary is calculated as being an average of 3.9v- the 1.4v across CVR1 for an average of 2.5v across the primary winding for the 25% on time.
So the P/in for 25% of each cycle is 2.5v X 140mA=350mW
From this we subtract the power dissipated by CVR1,which is 1.4v across 10 ohms=196mW.
This gives us the P/in to the circuit as being 350mW-196mW=154mW.
P/in is 154mW for the 25% on time for each cycle.
Average P/in is 154mW/4-->or 154mW X 25%.
Our average P/in is 38.5mW.
Ok,now we look at the reverse current flow through CVR1,even though the primary has a blocking diode on the positive rail.
The reverse current flow seen across CVR1 is current being transferred via capacitive coupling between the primary and secondary windings of the transformer.
I confirmed this by placing ch1 of the scope across CVR2,while leaving ch2 across CVR1--see test circuit 1B.
In scope shot 4,we can clearly see the reverse current flow shown across CVR1(the transformers primary) also existing across CVR2(the transformers secondary current).
This confirms current being transferred from primary to secondary,via capacitive coupling,and not current being returned to the source through the blocking diode on the primaries positive input rail.
Now the vertical rise of current during the 25% on time,and the current and voltage being in phase.
When we look at the scope shots,we will notice that we already have a forward voltage across the primary of the transformer of close to 2.2 volts just before the 25% on time part of the cycle starts. You will also notice at that time there is 0 volts across the CVR. This means the actual forward voltage across the primary just before P/in starts is actually 2.2 volts.
This is the reason that we see a vertical current trace as P/in starts,and also the reason the voltage and current are in phase.
P/out is easy,as it is a stable DC voltage across the 15kohm resistor (R1)
P/out is 36.8 volts across 15kohms=90.28mW
COP=234.49%
My next step is to increase the size of the DUT.
Im thinking along the lines of a MOT,and driving the input via mosfet or transistor.
Brad
OK,today was another full day of testing,and trying out some other transformers on new setups for higher power conversions.
What i found was this--
Using transformers where the primary and secondary are wound separately(such as MOTs),did not yield any OU results.
High frequency transformers that have the primary and secondary wound together(one over the top of the other,or bifilar type) all yielded OU results.
I put this down to the capacitive coupling between primary and secondary windings playing a vital roll in these circuits. The capacitive coupling between primary and secondary windings on transformers with separate windings will have a value of next to 0,and probably why they do not show OU results.
Going back to the original circuit(as it yields the best results so far),i carried out 3 different tests.
I have made a couple of modifications to the circuit.
I have increased C1s value to 2200uF,and added another diode in series on the input rail.
This was to halve the junction capacitance of the input diode(well,i think it would ?),so as to reduce any feedback to the source(my SG),and sending it to C1 instead--see circuit below for mods.
The needed information and calculations are attached to each scope shot.
Blue chanel across CVR1,and yellow chanel across CVR1 and primary coil.
Brad
Post #14 updated to include suggested starter circuit / external current source.
Note also that other types of oscillators can be tried such as single transistor or push pull types. I chose to start with negative resistance oscillators for practically no loading of C1 up to the firing point, after which bursts are generated due to capacitor discharge into tank circuit.
Quote from: TinMan on 2018.08.25, 12:57:50
OK,today was another full day of testing,and trying out some other transformers on new setups for higher power conversions.
What i found was this--
Using transformers where the primary and secondary are wound separately(such as MOTs),did not yield any OU results.
High frequency transformers that have the primary and secondary wound together(one over the top of the other,or bifilar type) all yielded OU results.
I put this down to the capacitive coupling between primary and secondary windings playing a vital roll in these circuits. The capacitive coupling between primary and secondary windings on transformers with separate windings will have a value of next to 0,and probably why they do not show OU results.
Going back to the original circuit(as it yields the best results so far),i carried out 3 different tests.
I have made a couple of modifications to the circuit.
I have increased C1s value to 2200uF,and added another diode in series on the input rail.
This was to halve the junction capacitance of the input diode(well,i think it would ?),so as to reduce any feedback to the source(my SG),and sending it to C1 instead--see circuit below for mods.
The needed information and calculations are attached to each scope shot.
Blue chanel across CVR1,and yellow chanel across CVR1 and primary coil.
Brad
Brad,
These are very interesting results, O0 but I think you forgot to attach the modified schematic. I agree with your calculations on your post previous to the modified version and I also agree with your capacitive coupling analysis between primary and secondary.
Late last night I tried a single inductor version of your circuit and it did not produce your waveforms or any OU. This would support the importance of capacitive coupling between windings rather than the importance of the leakage inductance.
What is the voltage step up or down ratio of the transformer that gives you the best performance? Also, what is the core material?
Carry on-
Pm
Edit
TM and PM,
i used the diagram on post #9 above to replicate using:
V1 = IXDN614PI MOSFET driver on 40V (driven by my FG)
L1 = 322uH (measured) on 5cm OD former
C1 = 5.015uF (measured) 2x 10uF/50V WIMA caps in series
C2 = 3274pF (measured)
Mx = 4x IRF3205 MOSFETs
R1 = 14.86KOhm (measured) load resistor across C1
I think the resonance frequency is around the shown (screenshots) 466Khz.
Probes are as indicated on the post #9 diagram (including the 4 turns for the current probe).
Blue marker is underneath the red marker on the left side
Not sure the current probe is in the righ way, see picture, but this gives positive values (inverted the green trace).
Anyway, is this what i have at resonance, not sure what PM means on his table by sweep (ms).
screenshot 1 is with input calcs (yellow x green = red) 6.2w / 4 = 1.55W (/ 4 because the 4 turns on the current probe)
screenshot 2 is with output calcs (blue - purple= red) 91V across 14.86KOhm = 560mW
Itsu
Hi Brad
Should not your circuit be as I have posted below with the addition of the red cables?
Just to clarify :)
Interesting stuff now coming out from all directions but in a similar area.
Regards
Mike 8)
author=partzman link=topic=3655.msg69375#msg69375 date=1535205791]
QuoteThese are very interesting results, O0 but I think you forgot to attach the modified schematic.
Oop's
Modified circuit now attached to previous post O0
QuoteWhat is the voltage step up or down ratio of the transformer that gives you the best performance? Also, what is the core material?
Ok,the core material is ferrite.
Primary to secondary ratio unknown.
I will get turn ratio,inductance and resistance values for you tomorrow.
I believe the small step up transformers found in mains CFL bulbs would do the job.
Also he small step up transformers in the cheap cameras that are used to charge up the flash cap may also work.
Maybe one of the same type from an old CPU power supply would work.
The best results so far are achieved by the original one i tried,and are using for my posted results--that being the one from the 6 volt CFL(small fluro tube) camping lantern.
So many to try,so little spare time. C.C
QuoteI agree with your calculations on your post previous to the modified version and I also agree with your capacitive coupling analysis between primary and secondary.
Late last night I tried a single inductor version of your circuit and it did not produce your waveforms or any OU. This would support the importance of capacitive coupling between windings rather than the importance of the leakage inductance.
It would seem that the capacitive coupling between primary and secondary is needed,as none of the transformers i tried void of capacitive coupling between primary and secondary windings worked--not even close.
In those cases,all power could be accounted for accurately.
This circuit of yours Pm could be the one,as measurements have been taken carefully,and the OU results are consistent O0
What we need is a 555 PWM circuit with adjustable frequency that could be run from the energy being sent to C1.
The output from that is then looped back into the input.
I cant do much more with the measurements,or be more accurate than i have been,so a looped DUT is the next step.
Brad
Quote from: Itsu on 2018.08.25, 14:35:21
TM and PM,
i used the diagram on post #9 above to replicate using:
V1 = IXDN614PI MOSFET driver on 40V (driven by my FG)
L1 = 322uH (measured) on 5cm OD former
C1 = 5.015uF (measured) 2x 10uF/50V WIMA caps in series
C2 = 3274pF (measured)
Mx = 4x IRF3205 MOSFETs
R1 = 14.86KOhm (measured) load resistor across C1
I think the resonance frequency is around the shown (screenshot) 441Khz.
Probes are as indicated on the post #9 diagram (including the 4 turns for the current probe).
Blue marker is underneath the red marker on the left side
Not sure why i have a negative input power calculation (probe looks the righ way, see picture).
Could be the scope / probe needs some warmup time.
Anyway, this is what i have at resonance, not sure what PM means on his table by sweep (ms).
Itsu
Hi Itsu
Perhaps make the turns on current probe the opposite direction.
Am i correct that the P/in is near 800mW ?,once inverted,or is it divided by the 4 turn ratio to give P/in of 200mW?.
Brad
Quote from: Centraflow on 2018.08.25, 14:46:28
Hi Brad
Should not your circuit be as I have posted below with the addition of the red cables?
Just to clarify :)
Interesting stuff now coming out from all directions but in a similar area.
Regards
Mike 8)
Hi mike.
It is,but i forgot to add them in the schematics.
Thanks for pointing that out,and sorry if this has screwed anyone around--my bad.
Below is a pic of my actual circuit(as posted in reply 6),which shows all diodes soldered together.
Amended schematic below as well.
OK Brad
thought it was a mistake, no problem this end, used to use diodes in this form to make variable capacitors for oscillators. As you change the voltage between them, so changes the capacitance between them ;)
Regards
Mike 8)
Quote from: TinMan on 2018.08.25, 14:55:19
Hi Itsu
Perhaps make the turns on current probe the opposite direction.
Am i correct that the P/in is near 800mW ?,once inverted,or is it divided by the 4 turn ratio to give P/in of 200mW?.
Brad
I fauled up with the circuit, so i redid it on that same post.
Itsu
Quote from: Itsu on 2018.08.25, 14:35:21
TM and PM,
i used the diagram on post #9 above to replicate using:
V1 = IXDN614PI MOSFET driver on 40V (driven by my FG)
L1 = 322uH (measured) on 5cm OD former
C1 = 5.015uF (measured) 2x 10uF/50V WIMA caps in series
C2 = 3274pF (measured)
Mx = 4x IRF3205 MOSFETs
R1 = 14.86KOhm (measured) load resistor across C1
I think the resonance frequency is around the shown (screenshots) 466Khz.
Probes are as indicated on the post #9 diagram (including the 4 turns for the current probe).
Blue marker is underneath the red marker on the left side
Not sure the current probe is in the righ way, see picture, but this gives positive values (inverted the green trace).
It is too hard to tell for me so I'll describe what should be the correct direction. The current probe arrow shows the direction of conventional current flow or the positive flow. Picture the positive current leaving your generator and entering the start of your 4 turn loop and exiting the finish of the same loop that would connect to the coil. The arrow should point in the direction of your finish wind or toward the coil. I originally had the sense loop wound as you have but when I measured it's inductance it was like .15uH. So, I wound a new one like in the pix attached with the loops separated and the inductance is not measurable. I also profiled the sense loop with a 100 ohm 1% non inductive resistor from 10kHz to 5MHz to determine the correct divisor to use in the math calculation. I realize you don't have that capability in your scope but that's OK since you can do the same thing as you have been by dividing the power input by 4 but just use the corrected number instead. My setup measures a little low so my divisor is 3.883 at 500kHz for example.
Quote
Anyway, is this what i have at resonance, not sure what PM means on his table by sweep (ms).
When you run continuously with a proper load across C1, you want to run at a frequency that is higher than the resonance frequency you are now reaching. I have determined the correct operating frequency and load from my 100ms scans and there is a maximum power level that will produce OU. I will explain this later in another post. I would recommend that you try increasing your frequency slowly while watching input power on your math channel and you should see the power level begin to reduce. Eventually you should see the input drop below the output if the parametric capacitance curve is proper for the circuit values and frequency thus providing a gain. The curve for this is dependent on the mosfets and will vary from my results as you used a different device.
The sweep (ms) indicates the horizontal time that the measurements were taken during a total sweep of 100ms. IOW, the complete sweep was taken and stored or frozen, and then vertical cursors were placed at 10ms intervals from the start and all measurements were taken at those points in time and recorded into the table. So a sweep (ms) of 30 means that the measurements of Pin and C1 volts were taken at 30ms from the start of the sweep and the rest is calculated.
Quote
screenshot 1 is with input calcs (yellow x green = red) 6.2w / 4 = 1.55W (/ 4 because the 4 turns on the current probe)
screenshot 2 is with output calcs (blue - purple= red) 91V across 14.86KOhm = 56mW
Itsu
Thanks PM,
i will check on the current probe, guess i had it the wrong way (top of the loop), so therefor i had to
invert the trace. Also will spread the 4 turns evenly.
By the way, the output was not 56mW, but 557mW (P=U²/R = 91²/14860 = 8281/14860 = 0.557W (557mw).
Itsu
PM,
i still do not get what you mean by sweep, sweeping what?
Not the frequency (what i normally understand by sweeping).
You sweep time? Is that not what a scope normally does?
Anyway, i increased my time base to 20ms/div. and used Single Seq. on the trigger menu, then started my FG.
This way i get a 20ms/div (about 9 div = 180ms) sweep from start of signal.
Is that what you mean?
I see a gradual increase in the blue and purple signals as well in the red math trace (blue - purple).
Also the current increases untill around 131ms where it levels of, see screenshot between blue cursors.
This is for a fixed frequency at 480Khz, going up further in frequency, the signals (blue and purple) drops strongly.
Same when going down in frequency, at a certain point (say 420Khz), the blue and purple sigs drop, so we seem
to have a narrow bandwidth to work with (resonance?).
Itsu
Quote from: Itsu on 2018.08.25, 19:23:12
PM,
i still do not get what you mean by sweep, sweeping what?
Not the frequency (what i normally understand by sweeping).
You sweep time? Is that not what a scope normally does?
Anyway, i increased my time base to 20ms/div. and used Single Seq. on the trigger menu, then started my FG.
This way i get a 20ms/div (about 9 div = 180ms) sweep from start of signal.
Is that what you mean?
I see a gradual increase in the blue and purple signals as well in the red math trace (blue - purple).
Also the current increases untill around 131ms where it levels of, see screenshot between blue cursors.
This is for a fixed frequency at 480Khz, going up further in frequency, the signals (blue and purple) drops strongly.
Same when going down in frequency, at a certain point (say 420Khz), the blue and purple sigs drop, so we seem
to have a narrow bandwidth to work with (resonance?).
Itsu
Itsu,
You basically have it except I used 10ms/div for the horizontal sweep to be able to reach a 100MS/s sample rate with my 350MHz scope. IIRC, you have a higher bandwidth scope so you should be able to reach a higher sample rate.
Try using the 100ms window and adjust the frequency until the voltage across C1 peaks at or near the end of the sweep. That frequency is what you will be using for the continuous loaded C1 testing.
Once you have captured the screen at that point, then measure the Pin and C1 volts at each 10ms division and document so you can calculate the COP for each 10 ms division. Hopefully you will find that at the shorter sweep times you will have COP>1 and possible even clear to the end of the 100ms sweep although this is difficult to achieve. I might add that my scope seems to take forever to make the individual calculations so it requires a lot of patience at this point.
An approximation of the amount of output power you can achieve can be done in this way. Let's say you have a maximum COP at 40ms with an output energy calc of 1mJ. Then potentially a Pout ~ 1e-3/40e-3 ~ 25mw can be achieved. At this point determine the best resistor value experimentally by starting with say a 10k. You can keep increasing the value until you see the voltage on C1 go to maximum which you don't want. Due to the non-linearity of the circuit, it gets really touchy near this point as you will see I hope.
Pm
Thanks PM,
i will continue tomorrow, i think the optimum will be around 470Khz then (voltage across C1 peaks at the end of the sweep (100ms))
Meanwhile i ran across this screenshot at 450Khz: :)
Itsu
Sorry to interrupt but I'm posting in this thread since I can't post, or even SEE, the Dr. Jones thread when I'm logged in. The screenshots below illustrate the issue and problem.
When I first encountered this I got mad, I admit, because I thought I was being excluded from Yet Another Secret Thread for some reason. I actually stared at the "Delete Account" button for many minutes last night. But now I see that others are experiencing the same problem.
1. When I'm not logged in I see the Jones thread listed, and I can view the thread normally.
2. If I'm viewing the thread and try to log in, I get this error message. Login succeeds but I am forbidden from seeing the thread at all.
3. When I'm logged in the thread doesn't even appear at all on the top contents page.
How many other threads are like this, I wonder. How many other members are also experiencing this problem? Would someone not even _know_ that some threads are totally invisible to that someone?
Quote from: Itsu on 2018.08.25, 20:31:41
Thanks PM,
i will continue tomorrow, i think the optimum will be around 470Khz then (voltage across C1 peaks at the end of the sweep (100ms))
Meanwhile i ran across this screenshot at 450Khz: :)
Itsu
OK, this looks good and tells me you could raise the frequency say 5-10kHz at a time because the IRF3205s have hit their limit at ~60ms. Increasing the frequency will move this limit to the right and ideally the peak voltage on C1 would be at the 100ms mark but it's OK to go a little beyond. I think in doing this you will have OU somewhere perhaps at the lower sweep times because the IRF3205 is reasonably close to the specs for the 14N05L.
Pm
Quote from: TinselKoala on 2018.08.25, 22:00:56
Sorry to interrupt but I'm posting in this thread since I can't post, or even SEE, the Dr. Jones thread when I'm logged in. The screenshots below illustrate the issue and problem.
When I first encountered this I got mad, I admit, because I thought I was being excluded from Yet Another Secret Thread for some reason. I actually stared at the "Delete Account" button for many minutes last night. But now I see that others are experiencing the same problem.
1. When I'm not logged in I see the Jones thread listed, and I can view the thread normally.
2. If I'm viewing the thread and try to log in, I get this error message. Login succeeds but I am forbidden from seeing the thread at all.
3. When I'm logged in the thread doesn't even appear at all on the top contents page.
How many other threads are like this, I wonder. How many other members are also experiencing this problem? Would someone not even _know_ that some threads are totally invisible to that someone?
TK,
I don't know but it would appear to me that there is a problem with the forum software in regards to certain members yourself included. I don't think it is anything malicious but purely coincidental. I'm sure Peter is looking into it.
Regards,
Pm
TK, you sort of have it in reverse so it seems, how odd, don't login and you see all, login and don't see all, hmmmmm
Regards
Mike 8)
PS. now I know why you did not answer my questions :(
I was busy today evening noting down the values of a 100ms sweep and catalog them for every 10ms which is a painstaking task.
But it does not look good.
I never came across a point where output is even close to the input.
Something like this (437KHz):
sweep time ms input mW voltage across 14.86KOhm output mW
10 127 8.25 4.6
20 151 13.3 12
30 374 18.5 23
40 632 23.9 38
50 1168 31.5 67
60 2202 42.9 124
70 3712 57 219
80 4986 72.5 354
90 4988 83.5 469
100 3778 87.8 519
Will try some more different frequencies in the next few days.
Itsu
Quote from: Itsu on 2018.08.26, 20:40:48
I was busy today evening noting down the values of a 100ms sweep and catalog them for every 10ms which is a painstaking task.
But it does not look good.
I never came across a point where output is even close to the input.
Something like this (437KHz):
sweep time ms input mW voltage across 14.86KOhm output mW
10 127 8.25 4.6
20 151 13.3 12
30 374 18.5 23
40 632 23.9 38
50 1168 31.5 67
60 2202 42.9 124
70 3712 57 219
80 4986 72.5 354
90 4988 83.5 469
100 3778 87.8 519
Will try some more different frequencies in the next few days.
Itsu
Itsu,
I know how painstaking it is to take these measurements so my hat is off to you for doing so. O0 O0
Now, I must admit that I wasn't very clear in my explanation of the overall measurements as it appears you have made the 100ms sweep and taken the measurements correctly, but this should be done without any load resistor connected across C1. I apologize for my lack of clarity on this point. :-[ The load resistor is connected only during a continuous test at the frequency where OU is seen somewhere during a sweep.
If you would please, re-run the same test without the 14.86k ohm resistor and log your results and let's see what your results are.
Regards,
Pm
All,
This is a change in the circuit design from full wave bridge to full wave center tap reducing the parametric elements to two. The results are interesting and require much more analysis but I'll post the info now anyway.
In the schematic below, T1 is trifilar wound with 15x34 litz wire separated out into 3 windings of 5x34 each on a T225-26B powdered iron core. This results in a winding-to-winding capacitance of ~3.5nf. The CSR used is a Caddock MP820 series 10 ohm 1% non-inductive film resistor. This gives really accurate current measurements but at the cost of wasted energy. This is added into the data table for the sweep measurements to arrive at a total net COP.
It should be noted that the input voltage measurement is taken ahead of the coupling cap and T1 so these losses are included in the measurements.
The operating frequency of 151kHz is the lowest to date and is partly due to the increase in parametric capacitance with the full wave center tap topology. We are now approaching an area where really efficient oscillator/drivers can be built!!!
Anyway, from the data table we see the apparent COP without the loss in the CSR considered, and an apparent COPnet with the CSR lost energy added to the calculation.
The scope pix shows the 8ms sweep which is shorter than previous sweeps due to the charge curve difference from slight different circuit operation. More on this later.
Regards,
Pm
Edit: Corrected missing L1 in schematic.
Ok,a little info on the small transformer that is giving the best results in my setup.
Primaries inductance value is very low at .1mH
Secondaries inductance value is 40.2mH
Resistance value across primary is 1.6 ohms
Resistance across secondary is 12.9 ohms
Below is a scope shot showing a supplied AC voltage(blue channel),and the resultant ac voltage across the secondary(yellow channel)
The turn ratio seems to be around 21:1
I have not yet come across another transformer that will give the results this one dose,although i am yet to try one with the same turn ratio-dont have one.
So,will be winding one tonight,but on a larger core,with larger wire,in the hope of upping the power ratio's.
Brad
Quote from: TinMan on 2018.08.26, 23:26:19
Ok,a little info on the small transformer that is giving the best results in my setup.
Primaries inductance value is very low at .1mH
Secondaries inductance value is 40.2mH
Resistance value across primary is 1.6 ohms
Resistance across secondary is 12.9 ohms
Below is a scope shot showing a supplied AC voltage(blue channel),and the resultant ac voltage across the secondary(yellow channel)
The turn ratio seems to be around 21:1
I have not yet come across another transformer that will give the results this one dose,although i am yet to try one with the same turn ratio-dont have one.
So,will be winding one tonight,but on a larger core,with larger wire,in the hope of upping the power ratio's.
Brad
Brad,
Thanks for taking these measurements. O0
Pm
All,
I guess I'm the one responsible for screwing up the members that were having login problems. I was adding members to this list through the moderation tab and somehow the forum software changed individual membership parameters. So, I sincerely apologize to any and all that had a problem :-[
Regards,
Pm
Quote from: partzman on 2018.08.27, 00:45:51
Brad,
Thanks for taking these measurements. O0
Pm
I have found a large ferrite E core in one of my old computer power supplies,so I'll wind that tonight,and see what happens.
Brad
Quote from: partzman on 2018.08.27, 00:51:03
All,
I guess I'm the one responsible for screwing up the members that were having login problems. I was adding members to this list through the moderation tab and somehow the forum software changed individual membership parameters. So, I sincerely apologize to any and all that had a problem :-[
Regards,
Pm
I was hoping for more people to check my measurements,maybe this is the reason only you (Pm) have checked and confirmed my measurements?.
No blame of fault on anyones part.
Was just a software problem apparently.
Brad
There is definitely no blame towards partzman for the software problem, it's a shame the function does not work correctly ;)
All,
Here is an updated data table in reference to the test device in post #41 which shows an additional COPb calculation which I believe to be correct.
The original COP measurement does not take into consideration the loss in the CSR. So, COPa adds the CSR loss to the output and then calculates the gain while COPb subtracts the CSR loss from the input before calculating the gain. I believe the latter is correct if one considers KPL and the overall input power minus the CSR loss leaves the actual energy supplied to the remaining circuitry.
Regards,
Pm
Quote from: partzman on 2018.08.27, 14:58:49
All,
Here is an updated data table in reference to the test device in post #41 which shows an additional COPb calculation which I believe to be correct.
The original COP measurement does not take into consideration the loss in the CSR. So, COPa adds the CSR loss to the output and then calculates the gain while COPb subtracts the CSR loss from the input before calculating the gain. I believe the latter is correct if one considers KPL and the overall input power minus the CSR loss leaves the actual energy supplied to the remaining circuitry.
Regards,
Pm
It is odd that we both seem to be showing COPs of about the same value Pm
Are we both making the same mistake somewhere,or have we finally found the secret to a true energy amplifier ?
Brad.
Quote from: TinMan on 2018.08.28, 00:19:17
It is odd that we both seem to be showing COPs of about the same value Pm
Are we both making the same mistake somewhere,or have we finally found the secret to a true energy amplifier ?
Brad.
Brad,
Yes it is odd! There is more to this simple circuitry than meets the eye and at this point in time, I don't think were experiencing measurement error!
Pm
Still on my initial 4 MOSFET setup.
I redid my measurements without the load resistor and found a certain frequency (475KHz) where the max voltage
across C1 was almost covering 100ms (88ms or so), but where i found some strange effects on the signals, see screenshot.
The white trace (ontop of the yellow input voltage trace) represents the input calculations yellow x green where
yellow is the input voltage and green the input current)
The red trace represents the calculated voltage across C1 (blue - purple).
What we can see is that between the red vertical cursors, the input power (white) seems negative (-1.906W), while the
voltage across C1 is positive (14.7V).
So this means that we have more out then in at this point, but the fact that i see a negative input power makes
me thinking we are looking to some kind of artifact.
If this is what you mean by COP >1 then i have serious doubts.
by the way, i have setup gating between the vertical cursors.
Itsu
Quote from: Itsu on 2018.08.28, 20:11:59
Still on my initial 4 MOSFET setup.
I redid my measurements without the load resistor and found a certain frequency (475KHz) where the max voltage
across C1 was almost covering 100ms (88ms or so), but where i found some strange effects on the signals, see screenshot.
The white trace (ontop of the yellow input voltage trace) represents the input calculations yellow x green where
yellow is the input voltage and green the input current)
The red trace represents the calculated voltage across C1 (blue - purple).
What we can see is that between the red vertical cursors, the input power (white) seems negative (-1.906W), while the
voltage across C1 is positive (14.7V).
So this means that we have more out then in at this point, but the fact that i see a negative input power makes
me thinking we are looking to some kind of artifact.
If this is what you mean by COP >1 then i have serious doubts.
by the way, i have setup gating between the vertical cursors.
Itsu
Itsu,
Since I have gone to the larger than normal CSR values, I have not seen any negative input power levels so there is something not right with the scan. Since you are now measuring between the vertical cursors, have you placed cursor A to the far left of the screen and placed cursor B at the various 10ms divisions. If yes and the power measurements are still negative, then your 100kS/s sample rate is affecting the accuracy IMO. Otherwise the circuit operation as you describe and is shown in the scope shot agrees with my results. IOW, if you now raise the frequency slightly, the time period from start to maximum C1 voltage will increase.
I will again setup the same test you are trying to replicate and reduce my sample rate to 100kS/s instead of 100MS/s to see if I get close to the results you are in the 100ms scan.
Thanks for doing the work! O0
Regards,
Pm
Quote from: Itsu on 2018.08.28, 20:11:59
Still on my initial 4 MOSFET setup.
I redid my measurements without the load resistor and found a certain frequency (475KHz) where the max voltage
across C1 was almost covering 100ms (88ms or so), but where i found some strange effects on the signals, see screenshot.
The white trace (ontop of the yellow input voltage trace) represents the input calculations yellow x green where
yellow is the input voltage and green the input current)
The red trace represents the calculated voltage across C1 (blue - purple).
What we can see is that between the red vertical cursors, the input power (white) seems negative (-1.906W), while the
voltage across C1 is positive (14.7V).
So this means that we have more out then in at this point, but the fact that i see a negative input power makes
me thinking we are looking to some kind of artifact.
If this is what you mean by COP >1 then i have serious doubts.
by the way, i have setup gating between the vertical cursors.
Itsu
Well Itsu,never seen a scope trace like that before.
Looks to me like the scope is not triggering correctly-maybe?
Brad
Quote from: partzman on 2018.08.28, 21:54:34
Itsu,
Since I have gone to the larger than normal CSR values, I have not seen any negative input power levels so there is something not right with the scan. Since you are now measuring between the vertical cursors, have you placed cursor A to the far left of the screen and placed cursor B at the various 10ms divisions. If yes and the power measurements are still negative, then your 100kS/s sample rate is affecting the accuracy IMO. Otherwise the circuit operation as you describe and is shown in the scope shot agrees with my results. IOW, if you now raise the frequency slightly, the time period from start to maximum C1 voltage will increase.
I will again setup the same test you are trying to replicate and reduce my sample rate to 100kS/s instead of 100MS/s to see if I get close to the results you are in the 100ms scan.
Thanks for doing the work! O0
Regards,
Pm
PM,
this screenshot/effect only happens at this specific frequency, any higher or lower its much smoother.
This circuit almost act like a PLL system as within a 100Khz range (430 - 530KHz) it shows the sine wave like
signals across C1 (blue and purple traces) around 91V, below that it starts to hickup, above it it the signals collaps.
So yes, i guess this specific frequency is indeed some kind of artifact due to the 100KS/s sample setting.
So still not able to see the COP >1.
Itsu
Quote from: TinMan on 2018.08.28, 23:13:54
Well Itsu,never seen a scope trace like that before.
Looks to me like the scope is not triggering correctly-maybe?
Brad
Tinman,
could be, but at any other frequency the triggering is ok, which seems odd to me.
Itsu
Quote from: Itsu on 2018.08.29, 09:28:25
PM,
this screenshot/effect only happens at this specific frequency, any higher or lower its much smoother.
This circuit almost act like a PLL system as within a 100Khz range (430 - 530KHz) it shows the sine wave like
signals across C1 (blue and purple traces) around 91V, below that it starts to hickup, above it it the signals collaps.
So yes, i guess this specific frequency is indeed some kind of artifact due to the 100KS/s sample setting.
So still not able to see the COP >1.
Itsu
Itsu,
In your post #39 and also in this test, were you using the 4x current probe loop? If yes, did you compensate manually by dividing your input power by 4?
Regards,
Pm
PM,
yes and no on those questions :-[, but even then, it still does not look good looking at post #39,
not sure how that influences the last (negative input power) test.
To much things to think of in these measurements :o
OK, i tested with the current set 5x lower then actual (so not 4 but 5) and still at 475KHz i get negative power calculated.
Itsu
Itsu,
Here are two scope pix of my circuit in post#9 taken at different sample rates for comparison. P9A used a sample rate of 100MS/s and p9B used 100kS/s. P9B loses all definition of the high frequency but accurately shows the Math for CH2-CH3 voltage across C1 which is a slow event. In essence, p9B is useless. These were taken in the hi-res mode with a single sample.
What I don't understand is the complete lack of the generator voltage on CH1(yel) and current on CH4(grn) as compared to your scope shot on your TDS!?! Perhaps the TDS series uses a different sampling algorithm to compensate for the slow sample rates during longer sweeps.
IMO, the 100ms sweep measurements taken with a 100kS/s sample rate can not be relied on for accuracy due to aliasing with the input frequency of ~500kHz.
Regards,
Pm
Itsu,
I fired up my TDS 3034 and took measurements over 100ms of the same circuit above and came up with Math power measurements all over the place including negative taking single shots. One thing to look at is to expand the captured view with the horizontal scale knob and see what happens to the traces. There are just not enough samples of the 500-600kHz waveforms taken at the 100kS/s rate to rely on their measurements over a 100ms sweep.
I will later post some results of a modified circuit that will lower the frequencies down to a level that shouldn't require the higher sampling rates which hopefully will allow you and others to replicate.
Regards,
Pm
Ok PM,
so there is no use to hunt for COP >1 with my present setup and used scope.
I wonder how other scopes manage as mine is old, but still fast.
Perhaps Tinman can show similar traces to see how his Rigol performs.
Anyway, i guess i will be switching over to your latest circuit which hopefully is no
problem for my scope.
Thanks, Itsu
All,
Here is the latest update to the circuit in post #41. This circuit is providing apparent gain but not with parametric means as the mosfet rectifiers have been replaced with schottky diodes for improved performance. Therefore, the apparent gain means (which is similar IMO to what Tinman has seen) should be researched assuming that the gains are real.
The schematic is shown below with the circuit operating at 102kHz plus an input bypass cap C2 is added to remove any dc from L1 and P1.
The first scope pix is taken over a 20ms horizontal sweep with the cursors measuring between start and 10ms. Note the 250MS/s sample rate.
The second scope pix is an expanded view of the 50ms cursor position to show detail of the waveforms.
The data table is shown next with the various COP calculations and notes.
If one takes an analysis of the input power and C1 charge levels between ~8ms to 20ms (not shown or detailed here), it becomes apparent that a loaded continuous power level of ~1.3w with gain can be achieved at 102kHz. The last scope pix shows the results of this test using a 3.71k ohm load resistor across C1.
The measurements show an output dc level of 69.91 volts across C1 and the load which equates to a Pout of 69.91^2/3.71e3 = 1.317 watts. The Pin is seen to be 777.1mw for an apparent COP = 1.317/.7771 = 1.695. However, this does not account for the loss in the CSR which is Pcsr = .9116^2/10 = 83.1mw resulting in an adjusted apparent COP = 1.317/(.7771-.0831) = 1.90 . Note how this compares to the COPb value in the data table at the 10-12ms sweep mark.
This topology lends itself to scaling much better than the parametric design due to the ease of building a tightly coupled transformer verses large parametric capacitors. I anticipate being able to lower the frequency of this design considerably depending on the actual theory of operation which is yet to be determined.
Regards,
Pm
Edit
PM,
looking much simpler, also in measuring, so will be trying to replicate in the next days.
Meanwhile i have removed the current probe and used a 100 Ohm 1% csr in the return line from C2 to ground like post #9 diagram.
Now the negative input power values are gone and i see more steady signals.
But still no COP >1 is noted across 2 different frequencies checked (455 and 464KHz), below table is for 464Khz
ms input mw / uJ C1 V / uJ c1 = 5.015Uf
10 118 1180 10.9 298
20 129 2580 15.6 610
30 139 4170 20.5 1054
40 185 7400 25.3 1605
50 303 15150 31.8 2536
60 522 31330 40.2 4093
70 745 52150 51.7 6703
80 1027 82160 66.1 10956
90 1361 122490 82.7 17150
100 1361 136100 93 21687
Will leave this setup and start building PM's latest.
Itsu
Hi Pm,
Thanks for including me in your list.
Would like to ask on this latest test: how the input DC power draw changes for the IXDD609 when you remove the 3.71 kOhm load resistor for a moment from C1? I understand that you measure AC input power but for applying a practical looping circuit it would be useful to learn about the total DC input power to the IXDD609 and how it changes when the output load is removed from C1 for some moments.
The looping circuit Ion suggested is interesting and simple. A little bit more involved alternative circuit would be the use of an isolated DC-DC converter circuit like for instance the LT830* family ((about 7-8 USD at Digikey, Mouser, etc). They draw some hundred microAmper for their internal circuits from a wide range input supply voltage and the output is adjustable with a single series feedback resistor. The isolation is done with a flyback transformer (5-6 USD). Here is the IC family:
No-Opto, Monolithic Isolated Flyback Converters
Part Number Internal Power Switch Input Voltage Range (V) Max Output Power (W)
LT8301 65V/1.2A 2.8 to 42 6
LT8302 65V/3.6A 2.8 to 42 18
LT8304 150V/2A 3 to 100 24
LT8300 150V/260mA 6 to 100 2.5
LT8303 150V/450mA 5.5 to 100 5
LT8315 630V/300mA 18 to 560 15
I attached one typical circuit for IC type LT8301, a step up converter though, to show the relative simplicity. The individual data sheets include several schematics for different Vin and Vout ranges, efficiency changes mainly between 80% to 90%.
I mention looping because it rules out the perpetual dilemma of the accuracy of scope sampling. Maybe it is time to attempt looping, that is all.
Of course a low power timer circuit is also needed to provide for a clock signal to drive the IXDD IC but a CMOS timer like LMC555 or TLC555 would draw max 1-2 mA from say at 10 V at the clock frequencies involved.
Gyula
Dear Gyula
With the circuit I offered, I was seeking a method of as direct a conversion as possible from the charge on the accumulator capacitor to an oscillatory sinusoidal waveform with a minimum of intermediate steps. I reasoned the burst oscillator tank circuit fed from a negative resistance device would be most direct.
One could also go from the charge on the accumulator cap to any form of conventional oscillator of high efficiency, class C or even a push pull or Mazili type.
I did not see the need to go to a DC-DC converter, as you still would need to go to the HF AC required to directly drive the system without needing the mosfet, and we are already starting with DC. I reasoned a DC-DC converter would add an unnecessary conversion loss, in addition to the DC to AC oscillator.
On the other hand if the devices you suggested could be arranged to drive the system with DC to AC conversion by eliminating the rectification step, there might be merit.
A complete diagram of your thoughts on the subject would be welcome , including the DC to DC then DC to AC conversion.
Where an impedance match is desired, the DC to DC converter could have considerable merit.
I believe partzman may have some tricks up his sleeve using his "pulsed saturating oscillator" that may be most efficient.
QuoteI mention looping because it rules out the perpetual dilemma of the accuracy of scope sampling. Maybe it is time to attempt looping, that is all.
I completely agree with your sentiments in the quote.
Thanks for your suggestions
Regards
Tonight i went about rearranging the circuit,so as i could confirm the voltage across the primary winding in my DUT.
As my scope and SG share a common ground(pain in the butt),i had to make the common rail of the circuit the positive,and the input to the transformer the negative--see modified circuit below.
This allows me to measure the voltage across the coil before the 10 ohm CVR.
Turns out my calculated voltage across the coil from previous tests,where i subtracted the voltage across the CVR from the total voltage,was correct,and the results are the same.
The first scope shot shows the pulse input,which is a 20% duty cycle.
The second scope shot shows the current and voltage trace,but where we now see a voltage across the coil by it self(yellow trace),and then the voltage across the coil and CVR series(blue trace).
So this time when we make our current calculations,we subtract the coil voltage from the total voltage to get out voltage across the CVR.
So V/in is an average of 1.8v for the 20% on time.
I/in is 2.9v minus 1.8v= 1.1v across 10 ohms=110mA for our 20% on time.
So P/in for the 20% on time is 198mW
The 10 ohm CVR during this 20% on time dissipates 121mW
So our average P/in is 198mW minus 121mW X 20% =15.4mW
P/out is 22v across 15k ohms
P/out = 32.26mW
COP= 32.26/15.4 X 100 = 209.48%
The COPs are consistent,no matter which way i take the measurements.
Brad
P.S
Added the missing input pulse scope shot--Thanks Itsu for the reminder O0
Quote from: TinMan on 2018.08.30, 13:39:08
Tonight i went about rearranging the circuit,so as i could confirm the voltage across the primary winding in my DUT.
As my scope and SG share a common ground(pain in the butt),i had to make the common rail of the circuit the positive,and the input to the transformer the negative--see modified circuit below.
This allows me to measure the voltage across the coil before the 10 ohm CVR.
Turns out my calculated voltage across the coil from previous tests,where i subtracted the voltage across the CVR from the total voltage,was correct,and the results are the same.
The first scope shot shows the pulse input,which is a 20% duty cycle.
The second scope shot shows the current and voltage trace,but where we now see a voltage across the coil by it self(yellow trace),and then the voltage across the coil and CVR series(blue trace).
So this time when we make our current calculations,we subtract the coil voltage from the total voltage to get out voltage across the CVR.
So V/in is an average of 1.8v for the 20% on time.
I/in is 2.9v minus 1.8v= 1.1v across 10 ohms=110mA for our 20% on time.
So P/in for the 20% on time is 198mW
The 10 ohm CVR during this 20% on time dissipates 121mW
So our average P/in is 198mW minus 121mW X 20% =15.4mW
P/out is 22v across 15k ohms
P/out = 32.26mW
COP= 32.26/15.4 X 100 = 209.48%
The COPs are consistent,no matter which way i take the measurements.
Brad
Brad,
Good work, I'm glad to see your results! O0
I do have two suggestions that may improve your performance and output. Change the diode pairs to a single schottky rectifier on each side and try a 50% duty cycle. We are no longer operating in a parametric mode and I will go into this in a later post.
Regards,
Pm
Guys,
a word of caution, seeing those values of COP = 209% should trigger you to look for any (measurement) errors,
as normally this would not be possible.
Not sure if this any measurement error, calculation error or even the circuit itself that does not lend itself for
doing measurements with our equipment.
Tinman, could you use the math function on your scope to calculate the power in?
By the way, i miss your "first scope shot" in your last post.
Itsu
All,
These test results are with the same basic circuit as my post #61 with the following changes: L1 is 2.47mH wound with litz wire on a gapped 1/4" ferrite core and the frequency is 64kHz so no schematic is shown.
The first scope pix is a 40ms sweep of the device and I would like to point out that Cursor B indicates a C1 voltage that is higher than indicated in the data table. The reason for this is because I use the mean measurement of the C1 voltage by placing both A and B cursors close to the measuring point rather than relying on the single cursor measurement. More accurate this way.
Some might wonder why I keep running these sweeps and the reason is that from the data, we can determine the power level that can be reached with the circuit. This is done by treating the C1 charge curve as an energy slope over time and then converting to power. For example, let's calculate the energy slope differential (Esd) from 16ms to 32ms by using the C1 voltage levels at these points. With this we calculate Esd = (74.1^2-47.31^2)*3.94e-6/2 = 6.408nJ. To convert the Esd to power available (Pa), we simply divide Esd by the time period so, Pa = 6.408e-9/(32e-3-16e-3) = 400mw. This is the approximate power level we can expect with these circuit values and operating parameters. We then use this value to determine the load resistance at any given C1 voltage. In this case we chose the C1 voltage to be ~62v (highest COP level) which resulted in using a 7.461k ohm which was on hand.
Next is the data table with all the calcs and circuit values.
Last is a scope shot of the device running in the continuous mode with the 7.461K load. Compare the power output to the above calculations as well as the COP to the data table. I believe these numbers are real!
Edit: OOps, I meant to calculate the COP in this mode so the output power is Pout = 55.16^2/7.461e3 = 408mw. The input power is 62.14mw so the COP = 408/62.14 = 6.57
At Gyula,
I will check and post the differential power measurements of the Ixys driver after this has soaked in.
So, this leaves us with the question of where the extra energy is coming from. After losing lots of sleep and pondering on this, I've concluded that it is supplied by the aether as Arie DeGues describes in his patent NL1032750. The tightly coupled bifilar secondary meets all the criteria that Arie describes. If you are not familiar with this patent I've attached a copy below.
At Peter,
I've done a current search for this patent thru WIPO and can't seem to find it. Where did you find the original copy that OUR translated?
Regards,
Pm
PM, Brad
Great work 8)
Yes I think if we can shake the aether and collect a charge over and above, we are in business.
I've got that buzz feeling
Regards
Mike 8)
As a follow up on my previous post, below is a scope pix for comparison of the same circuit running continuously with the Tek TCP0020 current probe measuring the current rather than the 10 ohm CSR in case there were any doubts on the input current measurements and calcs.
The 10 ohm CSR is removed so the output voltage is slightly higher and we see the power in is 74.36mw. The power output is 57.7^2/7.461e3 = 446.2mw so the COP = 446.2/74.36 = 6.00.
At Guyla,
I believe you wanted these measurements taken on the input from the power supply feeding the circuitry. I'm using a Rigol DP832 with the hi-res option and CH1 is set at 30vdc and CH2 is 10vdc to arrive at the 40vdc total. While running with the load attached for the above results, CH1 reads .66 watts and CH2 reads .23 watts. With the circuit disconnected from the output of the IXYS 609P1 driver, CH1 reads .09 watts and CH2 reads .05 watts. This results in a delta power level of .75 watts.
Regards,
Pm
Hi Pm,
Thanks!
Regarding your patent question, there seems to be a typo at the end of the patent number, it is not zero but a 9 i.e. NL1032759
At the European Patent Office here are Arie DeGues's patent list:
https://worldwide.espacenet.com/searchResults?ST=singleline&locale=en_EP&submitted=true&DB=&query=GEUS+ARIE+MELIS&Submit=Search (https://worldwide.espacenet.com/searchResults?ST=singleline&locale=en_EP&submitted=true&DB=&query=GEUS+ARIE+MELIS&Submit=Search)
and the 3rd in the list is the patent you asked, a direct link to it is here:
https://worldwide.espacenet.com/publicationDetails/biblio?II=2&ND=3&adjacent=true&locale=en_EP&FT=D&date=20080429&CC=NL&NR=1032759C1&KC=C1# (https://worldwide.espacenet.com/publicationDetails/biblio?II=2&ND=3&adjacent=true&locale=en_EP&FT=D&date=20080429&CC=NL&NR=1032759C1&KC=C1#)
Gyula
Quote from: gyula on 2018.08.30, 16:38:23
Hi Pm,
Thanks!
Regarding your patent question, there seems to be a typo at the end of the patent number, it is not zero but a 9 i.e. NL1032759
At the European Patent Office here are Arie DeGues's patent list:
https://worldwide.espacenet.com/searchResults?ST=singleline&locale=en_EP&submitted=true&DB=&query=GEUS+ARIE+MELIS&Submit=Search (https://worldwide.espacenet.com/searchResults?ST=singleline&locale=en_EP&submitted=true&DB=&query=GEUS+ARIE+MELIS&Submit=Search)
and the 3rd in the list is the patent you asked, a direct link to it is here:
https://worldwide.espacenet.com/publicationDetails/biblio?II=2&ND=3&adjacent=true&locale=en_EP&FT=D&date=20080429&CC=NL&NR=1032759C1&KC=C1# (https://worldwide.espacenet.com/publicationDetails/biblio?II=2&ND=3&adjacent=true&locale=en_EP&FT=D&date=20080429&CC=NL&NR=1032759C1&KC=C1#)
Gyula
That last patent is where I am atm, but I'm using 3 phase same frequency, not two, the first phase is converted into a floating neutral and any positive wave is crossed over on the output.
Regards
Mike 8)
Quote from: gyula on 2018.08.30, 16:38:23
Hi Pm,
Thanks!
Regarding your patent question, there seems to be a typo at the end of the patent number, it is not zero but a 9 i.e. NL1032759
At the European Patent Office here are Arie DeGues's patent list:
https://worldwide.espacenet.com/searchResults?ST=singleline&locale=en_EP&submitted=true&DB=&query=GEUS+ARIE+MELIS&Submit=Search (https://worldwide.espacenet.com/searchResults?ST=singleline&locale=en_EP&submitted=true&DB=&query=GEUS+ARIE+MELIS&Submit=Search)
and the 3rd in the list is the patent you asked, a direct link to it is here:
https://worldwide.espacenet.com/publicationDetails/biblio?II=2&ND=3&adjacent=true&locale=en_EP&FT=D&date=20080429&CC=NL&NR=1032759C1&KC=C1# (https://worldwide.espacenet.com/publicationDetails/biblio?II=2&ND=3&adjacent=true&locale=en_EP&FT=D&date=20080429&CC=NL&NR=1032759C1&KC=C1#)
Gyula
Gyula,
Ah, thanks for this info! O0
Pm
Quote from: Centraflow on 2018.08.30, 17:25:32
That last patent is where I am atm, but I'm using 3 phase same frequency, not two, the first phase is converted into a floating neutral and any positive wave is crossed over on the output.
Regards
Mike 8)
Mike,
The last patent in the list doesn't seem to be available. Do you have a copy?
Regards,
Pm
Good work indeed PM, TM
If I remember correctly Arie said that the power increases as the frequency increases.
I am wondering how the cop varies relative to drive voltage, does it scale linearly or squared.
Quote from: Centraflow on 2018.08.30, 17:25:32
That last patent is where I am atm, but I'm using 3 phase same frequency, not two, the first phase is converted into a floating neutral and any positive wave is crossed over on the output.
Regards
Mike 8)
Hi Mike,
It is the Brasil patent BR0205594, right? (Método para incrementar a energia cinética de elétrons utilizando uma combinação de campos magnéticos e eletromagnéticos permanentes)
Maybe it cannot be retrived freely via the web I guess? Did you have to pay for it?
Gyula
Quote from: Peterae on 2018.08.30, 20:16:34
Good work indeed PM, TM
If I remember correctly Arie said that the power increases as the frequency increases.
I am wondering how the cop varies relative to drive voltage, does it scale linearly or squared.
Peter,
Thanks!
In my testing the output power does seem to decrease with frequency but I'm not sure if it is due to impedance mismatch or what Arie describes.
The COP tends to decrease slightly with an increase in supply voltage so it follows a more linear than square function IMO at this point.
Regards,
Pm
Quote from: Itsu on 2018.08.30, 15:30:45
Guys,
a word of caution, seeing those values of COP = 209% should trigger you to look for any (measurement) errors,
as normally this would not be possible.
Not sure if this any measurement error, calculation error or even the circuit itself that does not lend itself for
doing measurements with our equipment.
Tinman, could you use the math function on your scope to calculate the power in?
By the way, i miss your "first scope shot" in your last post.
Itsu
Hi Itsu
First,i have added the missing input pulse scope shot to my last post--Thanks for the reminder O0
Regarding the measurements-
I have been very lenient toward the P/in,as in using higher than actual values to make the calculations.
If you look at the circuit i am using,you will see that i have also not accounted for the dissipated power of CVR two,and so this would also increase the COP calculations. In fact,the two CVRs them self dissipate more power than what the circuit consumes--!!apparently!!
My scope shots are quite clean and clear,and the frequency is very low.
We can use the scopes math function,due to the input not being an RMS value,nor can we use mean of average,due to the traces across the primary showing an AC type waveform,but where what you see across the primary is really power flowing through the secondary via capacitive coupling.
P/out is simple,a DC voltage across the 2200uF cap and 15k resistor combo.
P/dissipated by the two CVRs is also straight forward ohms law calculations.
But like you said--this shouldnt be happening,so we will continue on until an explanation is found.
Brad
Pm and all here.
I believe we are looking at capacitive acceleration of electrons here--not sure how else to put it?
Seems to me that the electric field is causing a flow of electrons (current) and not the magnetic field here,and i believe i can show this.
More to come tonight.
Brad
Quote from: Peterae on 2018.08.30, 20:16:34
Good work indeed PM, TM
If I remember correctly Arie said that the power increases as the frequency increases.
I am wondering how the cop varies relative to drive voltage, does it scale linearly or squared.
Thanks Peter.
ATM i am concentrating on the power measurements i am getting,along with the !how! is this possible,or !where! is this extra energy coming from.
Im leaning toward an electric field effect due to capacitive coupling.
Much to do yet to confirm our findings.
Where is Darren (Poynt) when you need him ?
Brad
Brad
well
Darren is always around if you need him
??
Shall I ring him ?
I've Pm'd Darren ;)
Quote from: TinMan on 2018.08.30, 22:42:43
Hi Itsu
First,i have added the missing input pulse scope shot to my last post--Thanks for the reminder O0
Regarding the measurements-
I have been very lenient toward the P/in,as in using higher than actual values to make the calculations.
If you look at the circuit i am using,you will see that i have also not accounted for the dissipated power of CVR two,and so this would also increase the COP calculations. In fact,the two CVRs them self dissipate more power than what the circuit consumes--!!apparently!!
My scope shots are quite clean and clear,and the frequency is very low.
We can use the scopes math function,due to the input not being an RMS value,nor can we use mean of average,due to the traces across the primary showing an AC type waveform,but where what you see across the primary is really power flowing through the secondary via capacitive coupling.
P/out is simple,a DC voltage across the 2200uF cap and 15k resistor combo.
P/dissipated by the two CVRs is also straight forward ohms law calculations.
But like you said--this shouldnt be happening,so we will continue on until an explanation is found.
Brad
Brad,
thanks for the screenshot, i was hoping it showed how you calculated the input power, but it does not make it clear to me.
It does not show any RMS or MEAN values (also i don't know where your probes are).
Not sure what you are saying above in bold, is it we can or we cannot?)
the screenshot shows an input square wave of 13.4V max DC (does your FG is capable of doing so?)
So no idea how you came to the mentioned:
QuoteSo V/in is an average of 1.8v for the 20% on time.
I/in is 2.9v minus 1.8v= 1.1v across 10 ohms=110mA for our 20% on time.
So P/in for the 20% on time is 198mW
But no problem, it will pan out eventually, i guess.
Here a screenshot from a 13.4V (max) 28.9KHz 20% duty cycle all DC square wave signal (like yours) using
my MOSFET driver as my Rigol FG does not produce more the 10V DC square wave (20pp) signals.
The RMS value and the mean (average) are shown.
Regards Itsu
Quote from: partzman on 2018.08.30, 18:26:41
Mike,
The last patent in the list doesn't seem to be available. Do you have a copy?
Regards,
Pm
Hi PM
I did have, but it seems it was on the computer that was taken when I was robbed. I had it in Dutch, not in English.
It was very much in the style of Jaños, The thing is with most of these patent applications, you have to read between the lines, not that there are things missing, just not in the order to make a replication. I'm talking about how to get the negative or ground ref: on the output, and why I'm going for 3 phase in a "type of" wye configuration.
Regards
Mike 8)
author=Itsu link=topic=3655.msg69485#msg69485 date=1535705587]
Quotethanks for the screenshot, i was hoping it showed how you calculated the input power, but it does not make it clear to me.
I thought the attached schematic would have made that clear--maybe not.
Here is the description from that post,the schematic and circuit are posted below.
The second scope shot shows the current and voltage trace,but where we now see a voltage across the coil by it self(yellow trace),and then the voltage across the coil and CVR series(blue trace).
QuoteNot sure what you are saying above in bold, is it we can or we cannot?)
Sorry-typo. Should say cant (cannot).
QuoteIt does not show any RMS or MEAN values
As i said,you dont use RMS for measuring a DC current or voltage.
RMS is for AC only.
You will see in your scope shot that the mean and RMS values are very different.
When it is pulsed DC,mean or average should be used,and when it is AC,RMS should be used.
Now,in this setup,we are using a pulsed DC,but we can use mean or average due to the reverse current flow caused by the capacitive coupling to the secondary.
The only way to do it is manually,unless my scope can measure between cursors ?,which i dont know as of yet.
But non the less,the manual calculations are accurate.
Quote(also i don't know where your probes are).
Sorry,i should get into that habit.
Schematic with scope probe placement below also.
Quotethe screenshot shows an input square wave of 13.4V max DC (does your FG is capable of doing so?)
Yes,open voltage. But when loaded,the max voltage depends on the impedance of the load,as my SG has a max current of 150mA at 1v.
QuoteSo no idea how you came to the mentioned:
The same way we always measure power.
The average current during the 20% on time X the average voltage during the 20% on time,minus the power dissipated by the 10 ohm CVR during the 20% on time,and then divide the total by 5,or times the total by 20%.
This gives us an average P/in for continuous running.
Brad
Quote from: Chet K on 2018.08.31, 06:04:10
well
Darren is always around if you need him
??
Shall I ring him ?
Hey Chet.
I see Peter has messaged him,so see how that go's.
I think we are on a winner here O0
Brad
Quote from: TinMan on 2018.08.30, 22:42:43
Hi Itsu
First,i have added the missing input pulse scope shot to my last post--Thanks for the reminder O0
Regarding the measurements-
I have been very lenient toward the P/in,as in using higher than actual values to make the calculations.
If you look at the circuit i am using,you will see that i have also not accounted for the dissipated power of CVR two,and so this would also increase the COP calculations. In fact,the two CVRs them self dissipate more power than what the circuit consumes--!!apparently!!
My scope shots are quite clean and clear,and the frequency is very low.
We can use the scopes math function,due to the input not being an RMS value,nor can we use mean of average,due to the traces across the primary showing an AC type waveform,but where what you see across the primary is really power flowing through the secondary via capacitive coupling.
P/out is simple,a DC voltage across the 2200uF cap and 15k resistor combo.
P/dissipated by the two CVRs is also straight forward ohms law calculations.
But like you said--this shouldnt be happening,so we will continue on until an explanation is found.
Brad
Brad,
In your highlighted statement above did you really mean "
can't use the scopes math function" because it fits the context that follows? If yes, actually we can use the scopes math function on these complex voltage and current wave forms to calculate the input power using the Math channel to display the 'mean' or average of say CH1(volts) x CH2(current). If not, then ignore the following. :-[
Our digital sampling scopes take instantaneous amplitude measurements of each waveform many times per second so they can be stored and displayed later on the screen as you know. This is perfect because it provides a means to perform calculations using these sampled values and in the case of power measurements, the voltage and current sampled values are multiplied times each other and the product is saved. Then integration is performed on these product samples which simply means they are all added together or summed and then averaged by dividing by the number of samples over the measurement time period. This is why it is important to do these measurements between the vertical cursors because they define the measurement time period. The scope knows how many samples have been taken over that period so all can be calculated from there.
The results from this use of the Math channel is as accurate as the scope's vertical and horizontal resolution. Most scopes are 8 bit on vertical which is 1 part in 256 with full screen deflection. At half screen deflection, it is 1 part in 128 or 0.78% and so on which is why it is important to have the waveform being measured at the highest possible amplitude on the screen. The horizontal sample rate determines the faithful reproduction of the wave being measured along with phase so it too is very important to have the highest sample rate possible for the measurement window.
IMO, at these low frequencies we are currently measuring, the scope math is reasonably accurate.
Regards,
Pm
Edit
Pm--check this out :o
Ok,i finally found another transformer that works much the same--in fact,even better.
Turns out i have 10 of them inside the battery operated CFL safety lights i scored from work when they changed to LEDs O0
So,below is the scope shot of the new circuit,with the new transformer-->yes,the original circuit is still together,and will remain so O0.
From this scope shot we get our values to make our P/in calculations.
The scope traces are very clean,and so accurate measurements are quite easy.
The blue trace (CH2)is across our 10 ohm CVR,and the yellow trace(CH1) is across the 10 ohm CVR and primary coil.
Our current is calculated using ohms law,where we have a clear 1.6 voltas across our 10 ohm CVR.
This gives us a current of 160mA
Now,our voltage across the coil is the value of CH1 minus the value of CH2
So our voltage across the coil is 1.9v(and you will see that i am using the max voltage here),minus 1.6v(voltage across the CVR),give us .3v across the coil.
At first i thought this could not be right,so i checked it using the reverse circuit(see circuit below,where the common rail is used as positive in). This lets me scope the coil voltage by it self,and it was indeed .3 volts(300mV) during the 25% on time.
So now we have 300mV X 160mA for our P/in during the 25% on time.
So during the 25% on time,our P/in is 48mW,where i have left the dissipated power by the CVR out.
Now we have our P/in to the transformer,which is 48mW for 25% of 1 full cycle.
Our average continuous P/in is 48/4,which is 12mW.
Our P/out is the voltage across R1(15k ohm),which for this test was 22 volts.
So our P/out is 32.2mW
Our COP is then 32.2/12x100 =268.3%
Brad
Quote from: partzman on 2018.08.31, 13:54:35
Brad,
In your highlighted statement above did you really mean "can't use the scopes math function" because it fits the context that follows? If yes, actually we can use the scopes math function on these complex voltage and current wave forms to calculate the input power using the Math channel to display the 'mean' or average of say CH1(volts) x CH2(current). If not, then ignore the following. :-[
Our digital sampling scopes take instantaneous amplitude measurements of each waveform many times per second so they can be stored and displayed later on the screen as you know. This is perfect because it provides a means to perform calculations using these sampled values and in the case of power measurements, the voltage and current sampled values are multiplied times each other and the product is saved. Then integration is performed on these product samples which simply means they are all added together or summed and then averaged by dividing by the number of samples over the measurement time period. This is why it is important to do these measurements between the vertical cursors because they define the measurement time period. The scope knows how many samples have been taken over that period so all can be calculated from there.
The results from this use of the Math channel is as accurate as the scope's vertical and horizontal resolution. Most scopes are 8 bit on vertical which is 1 part in 256 with full screen deflection. At half screen deflection, it is 1 part in 128 or 0.78% and so on which is why it is important to have the waveform being measured at the highest possible amplitude on the screen. The horizontal sample rate determines the faithful reproduction of the wave being measured along with phase so it too is very important to have the highest sample rate possible for the measurement window.
IMO, at these low frequencies we are currently measuring, the scope math is reasonably accurate.
Regards,
Pm
Edit
Hi Pm
Yes,i did mean !cant!,unless the math can be carried out between cursors.
We cant use RMS,because it is a DC pulse input,where the voltage never falls below the 0 volt line during that input pulse.
We cant use mean or average,because we have an AC component to the input voltage and current waveform,due to the capacitive coupling between primary and secondary.
We only want to calculate the on time,while the scope will see a complete cycle,and try to calculate using the whole of the cycle.
My wave forms are very clean,and easy to get accurate voltage values from for the on time.
Now,i made a couple of errors earlier on,where i subtracted the dissipated power of the CVR from the calculated P/in,when i had already subtracted the CVR voltage from the total voltage to get the voltage across the coil,and used that voltage to make my calculations.
In my last post,the calculations where made where the dissipated power of the CVR was disregarded ,and not subtracted from the P/in value.
Even then,the result was a COP of 268.3%.
I am not seeing where i am going wrong here, !but! ???
Maybe Poynt will see it straight away ?
Brad
Quote from: TinMan on 2018.08.31, 14:22:30
Pm--check this out :o
Ok,i finally found another transformer that works much the same--in fact,even better.
Turns out i have 10 of them inside the battery operated CFL safety lights i scored from work when they changed to LEDs O0
So,below is the scope shot of the new circuit,with the new transformer-->yes,the original circuit is still together,and will remain so O0.
From this scope shot we get our values to make our P/in calculations.
The scope traces are very clean,and so accurate measurements are quite easy.
The blue trace (CH2)is across our 10 ohm CVR,and the yellow trace(CH1) is across the 10 ohm CVR and primary coil.
Our current is calculated using ohms law,where we have a clear 1.6 voltas across our 10 ohm CVR.
This gives us a current of 160mA
Now,our voltage across the coil is the value of CH1 minus the value of CH2
So our voltage across the coil is 1.9v(and you will see that i am using the max voltage here),minus 1.6v(voltage across the CVR),give us .3v across the coil.
At first i thought this could not be right,so i checked it using the reverse circuit(see circuit below,where the common rail is used as positive in). This lets me scope the coil voltage by it self,and it was indeed .3 volts(300mV) during the 25% on time.
So now we have 300mV X 160mA for our P/in during the 25% on time.
So during the 25% on time,our P/in is 48mW,where i have left the dissipated power by the CVR out.
Now we have our P/in to the transformer,which is 48mW for 25% of 1 full cycle.
Our average continuous P/in is 48/4,which is 12mW.
Our P/out is the voltage across R1(15k ohm),which for this test was 22 volts.
So our P/out is 32.2mW
Our COP is then 32.2/12x100 =268.3%
Brad
Brad,
At this point I seen nothing wrong with your calculations. IMO, you have a different method of operation with your device but I could be wrong on that. In any case, continue on! 8)
Pm
Hey guys. Good to see you're still having fun with this stuff.
It is a bit of a shame however that apparently some important info I've offered in the past has seemingly been forgotten or just not utilized. Anyway, I would start with the following:
1) Replace the input side CVR scope measurement with a DC milli-ammeter measurement (DMM voltage across the CVR). This will give you an accurate average current going into the transformer: I(avg)=Vcvr(avg)/Rcvr
2) Using the scope, get an accurate peak and duty cycle measurement of the input voltage (after the diodes).
3) Calculate your P(in) like this: P(in)= (Vp x duty-cycle) x I(avg).
Is the P(in) measured this way still lower than your measured P(out)? If so, then there are other considerations we can look at. If not, then what might that indicate about some of the P(in) measurements being taken?
btw, When performed correctly, Input power measurements should come out Negative (for example when measured directly near a battery). The "Input Power" measurement being taken as per the schematics is not an input power measurement per se; it is a measure of the power dissipated in some of the components external to the source, and are actually part of the "load", as far as the source is concerned.
All,
Errrr, hold the phone on my circuit as described in post #61 and all that follows! The input current measurement taken on the low side of the primary is not an accurate representation of the current drawn from the signal generator as I have assumed. Due to the tight coupling with the trifilar windings, current is conducted thru the primary winding capacitance to the secondaries and is not seen in the primary return line to ground. Therefore the current waveform measured at the input of the primary winding is greater than the current measured at the return or low side of the primary. See the two continuous running scope shots below to compare the input power levels with the different measurement points.
The first is with the current probe measuring the current in the input side of the primary and the second is with the probe in the low side or ground side of the primary. As can be seen and calculated, the COP goes away using the high side current measurement in the pin calculation. With the 3710 ohm load resistor across C1, the high side COP = 73.81^2/3710/2.03 = 1.468/2.03 = .723 and the low side COP = 73.32^2/3710/.807 = 1.45/.807 = 1.8 . The high side is real and accurate. :-[
When I changed the circuit topology for a supposed improvement, I was careless in my assumptions and this is the result. This does not negate the performance of the original FWB parametric circuit but leaves the question, "can the FWB be improved topology wise?" .
Poynt,
Good to see you back!
Regards,
Pm
Quote from: poynt99 on 2018.08.31, 17:13:50
QuoteHey guys. Good to see you're still having fun with this stuff.
Great to see you still around Darren,and thanks for taking a look at this O0
QuoteIt is a bit of a shame however that apparently some important info I've offered in the past has seemingly been forgotten or just not utilized. Anyway, I would start with the following:
Never forgotten,and you will see i have also used the means you describe below--always do to confirm measurements. O0
Quote1) Replace the input side CVR scope measurement with a DC milli-ammeter measurement (DMM voltage across the CVR). This will give you an accurate average current going into the transformer: I(avg)=Vcvr(avg)/Rcvr
2) Using the scope, get an accurate peak and duty cycle measurement of the input voltage (after the diodes).
3) Calculate your P(in) like this: P(in)= (Vp x duty-cycle) x I(avg).
Indeed O0
See video's below.
QuoteIs the P(in) measured this way still lower than your measured P(out)? If so, then there are other considerations we can look at.
Yes,way below.
Maybe time for those further considerations ?.
Quotebtw, When performed correctly, Input power measurements should come out Negative (for example when measured directly near a battery). The "Input Power" measurement being taken as per the schematics is not an input power measurement per se; it is a measure of the power dissipated in some of the components external to the source, and are actually part of the "load", as far as the source is concerned.
And when the total power dissipated by the circuit exceeds that supplied by the source ?.
The fist schematic below is for the first video.
The second schematic below is for the second video.
If you do not have time to watch both video's,then watch video 2,as that is the higher powered device.
https://www.youtube.com/watch?v=AnGD4BXsFyQ
https://www.youtube.com/watch?v=M5oYdyEGOAo
Both video's are unlisted
Brad
And a third confirmation video.
https://www.youtube.com/watch?v=yOfgl9Uov3U
Good job Brad triple checking the Pin measurements! It's good to make your measurements in more than one way to ensure one is not fooling one's self.
Along that train of thought, I went back through the thread and I couldn't find any Pin measurement you've done using the scope and the tried and true method of AVG[v(t) x i(t)]. I think you are familiar with this method, and I would encourage you to do this on one of your DUTs in question. I have confidence you will see quite different results.
The problem with the technique you have been using for Pin, is that assumptions are being made about the voltage and current used in the calculation. If the load driven by the signal generator was purely resistive, then the technique is valid. However, you are driving the primary of a transformer, and all bets are off in terms of peaks and duty cycles. Picking a Vp from the input voltage and ignoring everything else present on that wave form is not valid here. Using the tried and true AVG[v(t) x i(t)] method takes care of all the issues present with this type of scenario.
The Pin measurement for Rosemary's circuit is different in that the power supply "Input" is an assumed steady DC voltage. With that scenario, we can measure the average current with the DMM and multiply the result by the DC input voltage. So I led you astray somewhat in my last post, apologies. With your circuit one can see that the "Input Voltage" is anything but a steady DC state, and there is a reactance involved.
Have you yet purchased a decent meter that can do RMS measurements to say 100kHz? One cool thing that can be done with a true RMS meter is to determine the exact power being transferred to the secondary of your transformer. If you use a very low inductance CVR you can then determine the RMS current in the primary circuit. From this, you can now calculate the power dissipated1 in the primary of your transformer (in your case a primary resistance of 1.6R). Subtract that from the AVG[v(t) x i(t)] measurement, and the result is the remaining power (VA) going to the secondary. Of course you have the fancy scope that can do RMS of a trace, so buying a separate fancy true RMS meter is not necessary.
You have the scope and I would suggest you use it for these measurements, even as a double-check for those that seem like no-brainers.
1 This is the power dissipated (W) in the transformer primary, not the total power (W + VA) in the primary. P(W)(prim)= I(rms)2 x R(prim).
All,
Before proceeding on with the parametric charger, I decided to profile the RFP14N05L mosfet I've been using and also some other devices I have on hand. The schematic used for the testing the is attached including the math used to generate the traces to allow the determination of the capacitance values with increasing voltage.
Scope shots with samples of the measurements for the FRP14N05L, IRF740, and a 24mm piezo are also included for comparison plus a data table for the devices tested thus far.
Regards,
Pm
Edit 9/3/18- Replaced data table with corrected info on 24mm piezo.
Quote from: poynt99 on 2018.09.01, 18:26:45
Good job Brad triple checking the Pin measurements! It's good to make your measurements in more than one way to ensure one is not fooling one's self.
Along that train of thought, I went back through the thread and I couldn't find any Pin measurement you've done using the scope and the tried and true method of AVG[v(t) x i(t)]. I think you are familiar with this method, and I would encourage you to do this on one of your DUTs in question. I have confidence you will see quite different results.
The problem with the technique you have been using for Pin, is that assumptions are being made about the voltage and current used in the calculation. If the load driven by the signal generator was purely resistive, then the technique is valid. However, you are driving the primary of a transformer, and all bets are off in terms of peaks and duty cycles. Picking a Vp from the input voltage and ignoring everything else present on that wave form is not valid here. Using the tried and true AVG[v(t) x i(t)] method takes care of all the issues present with this type of scenario.
The Pin measurement for Rosemary's circuit is different in that the power supply "Input" is an assumed steady DC voltage. With that scenario, we can measure the average current with the DMM and multiply the result by the DC input voltage. So I led you astray somewhat in my last post, apologies. With your circuit one can see that the "Input Voltage" is anything but a steady DC state, and there is a reactance involved.
Have you yet purchased a decent meter that can do RMS measurements to say 100kHz? One cool thing that can be done with a true RMS meter is to determine the exact power being transferred to the secondary of your transformer. If you use a very low inductance CVR you can then determine the RMS current in the primary circuit. From this, you can now calculate the power dissipated1 in the primary of your transformer (in your case a primary resistance of 1.6R). Subtract that from the AVG[v(t) x i(t)] measurement, and the result is the remaining power (VA) going to the secondary. Of course you have the fancy scope that can do RMS of a trace, so buying a separate fancy true RMS meter is not necessary.
You have the scope and I would suggest you use it for these measurements, even as a double-check for those that seem like no-brainers.
1 This is the power dissipated (W) in the transformer primary, not the total power (W + VA) in the primary. P(W)(prim)= I(rms)2 x R(prim).
Hi Darren
Yes,i am a little confused here regarding this AVG[v(t) x i(t)] method you are talking about.
I thought that is what i had done all through the thread here,where the voltage during the 20% on time was averaged,and the current during the on time was averaged,and then the two multiplied to give our average P/in.
Should not the average P/in be calculated by only using the values during the 20% on time,as that is the only portion of the complete cycle the DUT is receiving energy from the SG. The rest of the waveforms through the remaining 80% of the cycle is recycled through the primary via the circuit as a whole.
If we look at the scope shot below,we see the on time where power is delivered to the transformer from the SG(between the red lines).
Between the green line's shows the inductive kickback cycle,where the voltage is inverted,but the current continues to flow in the same direction,and the last small portion before on time is C2 discharging
Now,the average voltage in that scope shot is the average voltage across both the CVR and primary of the transformer,not just the transformer it self,and that is why it cannot be used to calculate average power dissipated by the transformers primary only--the joys of having the scope and SG sharing a common ground :(
So,when making my calculations during the 20% on time,i subtract the voltage across the CVR from the total voltage to get my voltage across the transformer. I confirmed this value by using the circuit below,where i could measure the voltage across the transformers primary only during the 20% on time portion of the cycle.
This showed me that my calculated voltage across the transformer in previous tests was correct.
Im not quite sure what you mean by this AVG[v(t) x i(t)] ,as i thought that was what i was doing throughout the whole thread here,but you say you dont see where i have done it ?.
If you mean average V over time X average current over time,then that is what i have done throughout the thread here,where i averaged the input voltage across the coil during the 20% on time to the full cycle time,and the same for the current.
Now,if you mean using the average voltage X average current from the scope's measurements,then lets use the average values shown in the supplied scope shot.
Average voltage is 150mV,and average current is 316mV across 10 ohms for an average current of 31.6mA
Now our average P/in would be 4.74mW
Our P/out is 20v across 15kohm=26.66mW
We now have a COP of 562%
I am happy to test the DUT anyway you see fit,as all my testing so far gives COPs that should not be.
Brad
Brad,
I'm no math guru, but what I am trying to convey with AVG[v(t) x i(t)] in words is:
Take the average (AVG) of the instantaneous voltage v(t) times instantaneous current i(t).
You do this by setting up one probe (CH1) measuring the input voltage, just as you have been, and one probe (CH2) measuring across the low-inductance (not wirewound power resistors) cvr for current. Set up a MATH trace for CH1xCH2, then set a measurement for the average value (or MEAN) of the MATH trace. You will have to scale the current based on the value of your CVR resistor. Some scopes allow you to scale, otherwise you do it in your head by dividing by 10 (for a 10R resistor).
That should now ring a bell with you...
This method has been discussed and recommended many times over the years, not only by myself, but by the true masters out there, TK, ION, etc. etc (apologies for those I have missed). It was used extensively during the Rosemary Ainslie saga as well. And I am fairly certain that we had you doing this as well on some test, or at least attempting to until we couldn't figure out how to get the AVG working on your MATH trace as I recall.
Anyway, give that a shot. It should be no problem with your new scope, and you very well may have already done it in recent times.
Quote from: poynt99 on 2018.09.02, 00:52:19
Brad,
I'm no math guru, but what I am trying to convey with AVG[v(t) x i(t)] in words is:
Take the average (AVG) of the instantaneous voltage v(t) times instantaneous current i(t).
You do this by setting up one probe (CH1) measuring the input voltage, just as you have been, and one probe (CH2) measuring across the low-inductance (not wirewound power resistors) cvr for current. Set up a MATH trace for CH1xCH2, then set a measurement for the average value (or MEAN) of the MATH trace. You will have to scale the current based on the value of your CVR resistor. Some scopes allow you to scale, otherwise you do it in your head by dividing by 10 (for a 10R resistor).
That should now ring a bell with you...
This method has been discussed and recommended many times over the years, not only by myself, but by the true masters out there, TK, ION, etc. etc (apologies for those I have missed). It was used extensively during the Rosemary Ainslie saga as well. And I am fairly certain that we had you doing this as well on some test, or at least attempting to until we couldn't figure out how to get the AVG working on your MATH trace as I recall.
Anyway, give that a shot. It should be no problem with your new scope, and you very well may have already done it in recent times.
OK,this is how i have been doing it all throughout the thread,but manually,as i cannot measure the voltage across the primary only,and current through the primary via the CVR separately in the one go,due to the common grounds between the scope and SG. The voltage across the primary has to include the voltage across the CVR as well,so we cannot use RMS or average value shown on the scope to give us the voltage across the primary only to make any power calculations,as that voltage value will be higher than the actual value across the primary coil.
I can bring up the math trace of CH1 x CH2 and get the average of that --that is no problem.
But it will give a wrong value,as CH1 will be showing the voltage across both the primary and CVR,while CH2 will give the voltage across the CVR only. This result will give us the power dissipated by both the CVR and primary--not the primary alone.
This is the method i have been using throughout the thread,where the total power in is calculated,and then the calculated dissipated power of the CVR is subtracted from the total P/in to give us the P/in for the primary of the transformer.
I would like to do it the way you stipulate,but simply cannot due to the common ground issue between scope and SG.
I have tried a 1:1 isolation transformer,and an inverter to run the SG,in the hopes that i would be able to scope across circuits without any common ground issue's--but that dose not work as people say it dose,as i can always get an AC voltage reading between the two grounds/commons--even when !apparently! isolated from each other.
So,the best we can do is get the total P/in,and then subtract the power dissipated by the CVR,and use the remainder as our P/in value--which i have been doing--unless you have another idea i have overlooked?.
Brad
It is a fairly straight forward setup and procedure Brad. Here's what you do:
1) Use the CVR1 setup as in the attached pic. Same common ground for both probes.
2) CH1 is v(t), and CH2 is i(t) (and is v(cvr1)/10).
3) Set up a MATH trace with CH1 x CH2.
4) Set up a measurement for the MEAN (or average) of the MATH trace. Divide the indicated measurement by 10 to correct for the current. This is the total corrected power going into the circuit after the diode.
5) Set up a measurement for the RMS value of CH2, and divide that by 10. This is the corrected RMS input current.
6) Make note of the total corrected average power in step 4).
7) Calculate the power dissipated in CVR1 by using the corrected RMS current in step 5), as: P(cvr1)=I(rms)2 x 10
8] Subtract the CVR1 power calculated in step 7) from the total power noted in step 6).
9) The subtraction in step 8] results in you obtaining the net power going into the primary of the transformer, aka P(in).
Quote from: poynt99 on 2018.09.02, 04:39:09
It is a fairly straight forward setup and procedure Brad. Here's what you do:
1) Use the CVR1 setup as in the attached pic. Same common ground for both probes.
2) CH1 is v(t), and CH2 is i(t) (and is v(cvr1)/10).
3) Set up a MATH trace with CH1 x CH2.
4) Set up a measurement for the MEAN (or average) of the MATH trace. Divide the indicated measurement by 10 to correct for the current. This is the total corrected power going into the circuit after the diode.
5) Set up a measurement for the RMS value of CH2, and divide that by 10. This is the corrected RMS input current.
6) Make note of the total corrected average power in step 4).
7) Calculate the power dissipated in CVR1 by using the corrected RMS current in step 5), as: P(cvr1)=I(rms)2 x 10
8] Subtract the CVR1 power calculated in step 7) from the total power noted in step 6).
9) The subtraction in step 8] results in you obtaining the net power going into the primary of the transformer, aka P(in).
Ok,i have done what you suggested(see scope shot below),but i disagree with the highlighted above,where this is not showing total corrected P/in. This is showing total power dissipated by the CVR and Primary of the transformer,which is 69.4mW
By doing it this way,the scopes math is including power being sent back to the source-i would suspect via the diodes junction capacitance for a brief time,and then once the voltage inverts across the coil,current then flows to the SG via the ground rail,through the SGs circuit,and then back into the DUT.
How do we know this?--
Well,lets have a look at the scope shot below.
See the time period i have circled in green.
If this was power being sent to the circuit by the SG,how come the voltage is inverted across the primary?
This area in green is the primary coils magnetic field collapsing(the flyback period).
So,we know from years of testing,that this is the time when the supply is cut off,and we get our inductive kickback,where the voltage inverts. During the 80% off time of the SG,the transformer is now the source--not the SG,but the scope math still sees this as power flowing through the primary of the transformer,and so is adding that to it's calculations.
What we want to know is the power being delivered to the DUT by the SG only,during it's 20% delivery cycle.
Please see next post,where i show clearly when power is being delivered by the SG to the DUT.
Brad
Ok,in this test i added a lightbox,which has an LED and solar panel inside.
It is housed in a sealed box to eliminate any outside light influence.
The circuit is as below.
I hooked CH3 across the solar panel,so as we can see exactly when and for how long current is flowing into the DUT.
If we look at the scope shot below,and take note of CH3,we can clearly see when current is flowing into the circuit,as we see a sharp rise in voltage across the solar panel,due to the LED lighting up.
As we can see,this rise on CH3 coincides exactly with the 20% on time shown via the voltage and current traces.
This is the only time power is delivered to the DUT by the FG.
You will also see a blue line showing peak current value,and a yellow line showing peak voltage value during the 20% on time. I think you would have to agree that i have been generous in my averaging,where actual values would be slightly less.
To get our P/in from the SG only,first we calculate the instantaneous power.
You will also note that the current and voltage are exactly in phase,which makes the calculations easy.
The instantaneous current is 1.1v across 10 ohms=110mA
Instantaneous power dissipated by the CVR is 121mW
The instantaneous voltage across the CVR and primary is 2.6 volts.
The total instantaneous power is there for 286mW
From that we subtract the 121mW dissipated by the CVR-->286mW-121mW=165mW
This 165mW is the instantaneous power dissipated by the primary of the transformer.
We now multiply that by our 20% duty cycle to get an average dissipated power of the primary.
165mW X 20%=33mW.
The voltage across R1/C1 is 22.1v,for a dissipated power of 32.8mW
Our COP is therefor 99.3%
Poynt
Are you happy with the way those measurements were calculated?.
Brad
Brad,
i replicated your circuit as much as possible, i only use a different transformer (Prim. 200uH / 0.5 Ohm,
Sec. 400mH / 106 Ohm) out of a flyswatter (ferrite core), see picture.
My signals are not that clean as yours (28.8KHz), especially the input diode causes massive ringing on the voltage.
As i can omit CVR1 by using my current probe, i can directly measure/calculate the input power using my scope.
First measurements show 9mW input for a 2.7mW across the 15K resistor for a cop = 0.3
Not sure if this transformer is messing things up here (to low ohmage primary for my 50 Ohm FG)
let me know if you want me to setup/measure/calculate things the way you do it to compare.
Itsu
Quote from: Itsu on 2018.09.02, 09:32:01
Brad,
i replicated your circuit as much as possible, i only use a different transformer (Prim. 200uH / 0.5 Ohm,
Sec. 400mH / 106 Ohm) out of a flyswatter (ferrite core), see picture.
My signals are not that clean as yours (28.8KHz), especially the input diode causes massive ringing on the voltage.
As i can omit CVR1 by using my current probe, i can directly measure/calculate the input power using my scope.
First measurements show 9mW input for a 2.7mW across the 15K resistor for a cop = 0.3
Not sure if this transformer is messing things up here (to low ohmage primary for my 50 Ohm FG)
let me know if you want me to setup/measure/calculate things the way you do it to compare.
Itsu
Hi Itsu
That is a big step up you have with that transformer,which may be the problem ?
Also,dose your FG have a function where you can set the impedance to 0 ,such as mine has?
Would be good to see what type of waveforms you have across the scope as well.
As seen in all my video's,i get very clean waveforms,and the results are very easy to see.
You should be able to get 20v across the 15kohm resistor with no problem at all--maybe some tuning is needed?,due to different components.
After testing many different transformers,i finally found another that yields the COPs i have been showing-im not sure why very few give me these results,and where others of the same type(E core ferrite)show below COP+ results.
Brad
Hey Brad.
Is it possible that some of your Ferite cored transformers might have cracks in them?
Just thinking out loud based upon some of Verpies hypothesis.
Cheers Graham.
Quote from: Itsu on 2018.09.02, 09:32:01
...
i only use a different transformer (Prim. 200uH / 0.5 Ohm,
Sec. 400mH / 106 Ohm) out of a flyswatter (ferrite core), see picture.
...
Not sure if this transformer is messing things up here (to low ohmage primary for my 50 Ohm FG)
...
Dear Itsu,
As you know the primary coil inductance of 200 uH represents roughly 36 Ohm inductive reactance load to your FG at 28.8 kHz.
EDIT: Brad's measured transformer data is in his post #42, page 2 of this thread.
If you see feasable to separate the ferrite cores of the transformer from each other, then perhaps you could remove the many turns of the original HV winding ? I know the two E cores are permanently glued together and the ferrite may break easily, perhaps heating up the glued area at the two sides may be successful. If you see the separation is not possible without the risk of core break, then let's forget this.
Gyula
This is what MH had to say about calculating power.
From the !Lawrence Tseung sent a prototype to test! thread
When it comes to a "power waveform," that is an abstraction derived from two other real-world variables. The power waveform already factors in the "squared" factor associated with the voltage and current waveforms. Therefore it makes no sense at all to "square power" and then take the mean value of that and then calculate the square root. The only thing that you have to do is average out the power over an entire periodic cycle.
When you want to calculate energy over a cycle, you integrate the power waveform over that cycle to get your energy calculation. You do not integrate over the square of the power waveform and then take the square root. Again, that makes no sense.
You have to multiply the instantaneous voltage by the instantaneous current and then take the average of the power waveform.Therefore the correct way of measuring the output power from the LTJT is for Poynt to use his DSO and measure the instantaneous voltage and current values and then do the associated calculations. The same principle applies to measuring the input power because the LTJT itself is a non-linear load.
And from Poynt
I believe the issue at hand is even more basic. My understanding of the disagreement is simply regarding which computation should be used to obtain the REAL power in the circuit*, either:
Method 1: The Power value obtained from the MEAN of the product of instantaneous current and instantaneous voltage.
or
Method 2: The Power value obtained from the RMS of the product of instantaneous current and instantaneous voltage.
Again from Poynt
[I should mention, and I trust we are all in agreement on this, that all measurements being argued here are based on one that is obtained by first acquiring the instantaneous voltage and current wave forms, then multiplying them together to produce the instantaneous power trace. For circuits of this nature, there is no other sure method (using an oscilloscope) to obtain the value of average power in a circuit.]
At this time,i would like to confirm that the above method is the one i have been using--instantaneous voltage X instantaneous current. That result is then multiplied by the duty cycle to give us an average power input over each full cycle.
To Quote Poynt-->tthere is no other sure method (using an oscilloscope) to obtain the value of average power in a circuit
Brad
Quote from: Grumage on 2018.09.02, 10:51:47
Hey Brad.
Is it possible that some of your Ferite cored transformers might have cracks in them?
Just thinking out loud based upon some of Verpies hypothesis.
Cheers Graham.
Well,im not sure Graham
I have noticed that some of these cores are glued together,and some are not,they are just held together by the tape wound around them.
I have many of the second transformer i found that is giving the OU results,so i will check out another one,and see if it gives the same result as the identical one i have used. If not,then i will start looking for them cracks or loose glue joints.
I too remember Verpies mentioning something about the two halves of the core should be free from each other,and vibrate when being pulsed.
Any idea as to which thread that was in ?
Brad
Quote from: gyula on 2018.09.02, 10:56:48
Dear Itsu,
As you know the primary coil inductance of 200 uH represents roughly 36 Ohm inductive reactance load to your FG at 28.8 kHz.
If you see feasable to separate the ferrite cores of the transformer from each other, then perhaps you could remove the many turns of the original HV winding ? I know the two E cores are permanently glued together and the ferrite may break easily, perhaps heating up the glued area at the two sides may be successful. If you see the separation is not possible without the risk of core break, then let's forget this.
Gyula
Hi Gyula
I have tried many times to separate the two glued halves of E core ferrite transformers using heat to soften the glue,only to have them crack/explode due to localized expansion in the heated area,while no expansion in the non heated area. Ferrite has no forgiveness ,and simply will not flex.
On occasion i get lucky,and find a core now and then which is not glued together,but only held together by the tape wrapped around them.
Brad
Hi Brad.
I'm pretty sure I launched the topic on my bench. CARA, or something like that.
http://www.overunityresearch.com/index.php?topic=2751.msg44488#msg44488
Do you have a means of mechanically stimulating the transformers ultrasonically? Might be interesting to see what output you get on the scope. ;)
Cheers Grum.
Quote from: Grumage on 2018.09.02, 12:03:06
Hi Brad.
I'm pretty sure I launched the topic on my bench. CARA, or something like that.
http://www.overunityresearch.com/index.php?topic=2751.msg44488#msg44488
Do you have a means of mechanically stimulating the transformers ultrasonically? Might be interesting to see what output you get on the scope. ;)
Cheers Grum.
Actually i do indeed.
I have a few of them 45 watt ultrasonic water misters from old home fogger cooling fans O0
So i just sit the transformer on top of the pizo ?
Brad
Ah, great. O0
Yes. I did several videos on the subject perhaps 3 years ago?
Aided by Verpies we found that mechanical stimulation of Ferite cores can produce a voltage in a coil wound around them. Admittedly voltage ain't power and I didn't subject the coils to a load resistance, perhaps you could take things a step further?
I've just realised, this perhaps isn't the best place for discussion as we're getting a little off topic, my apologies PM.
Cheers Grum.
Poynt
Please check the below.
This is a test carried out on another transformer that gave COP+1 results.
This is how P/in to the primary of the transformer only was calculated.
The measurements are taken from the maximum values shown in the scope shot below.
These values are slightly high than actual values,and so will yield a higher than actual P/in value.
I do this to eliminate any incorrect values that would be lower than actual P/in values.
As we can see from the scope shot,the voltage and current are in phase.
This is due to the very low inductance value of the primary coil in relation to the frequency.
So in this case,no power factor correction is needed.
First i get the instantaneous current value,which in this case is the V/max across the 10 ohm CVR
V/max is 1.6 volts over 10 ohms for an instantaneous current of 160mA
Now we need the instantaneous voltage across the primary coil of the transformer.
This is gained by the total instantaneous voltage across the CVR/primary coil series,minus the voltage across the CVR.
This gives us an instantaneous voltage across the coil of 1.9v - 1.6v=300mV
We now have a means of calculating the instantaneous power consumed by the primary coil,which is 300mV x 160mA=48mW.
The 48mW is for 25% of a complete cycle,and so is multiplied by the duty cycle to get an average P/in over a complete cycle.
So P/in to the primary of the transformer is--48 x 25%=12mW
P/out from the secondary is very easy to calculate,as it is 22 volts over our 15kohm resistor/C1 combo.
P/in=12mW
P/out=32.26mW
COP=268.8%
Is there anyone here that can carry out the manual calculations ,using the scope shot below,to confirm my results?.
Brad
Quote from: Grumage on 2018.09.02, 12:31:23
Ah, great. O0
Yes. I did several videos on the subject perhaps 3 years ago?
Aided by Verpies we found that mechanical stimulation of Ferite cores can produce a voltage in a coil wound around them. Admittedly voltage ain't power and I didn't subject the coils to a load resistance, perhaps you could take things a step further?
I've just realised, this perhaps isn't the best place for discussion as we're getting a little off topic, my apologies PM.
Cheers Grum.
Well if you can find the old thread related to this Graham,then we can re-boot it.
Now that you mentioned this,i seem to remember doing some quick tests,where i just hit a MOT core with a hammer,and measured a voltage across the coils using my scope.
This has to mean that energy can be produced by vibrating the core of an inductor/transformer.
I might do a quick trip out to the workshop for a quick test O0
Back soon.
Brad
Quote from: TinMan on 2018.09.02, 11:53:05
Hi Gyula
I have tried many times to separate the two glued halves of E core ferrite transformers using heat to soften the glue,only to have them crack/explode due to localized expansion in the heated area,while no expansion in the non heated area. Ferrite has no forgiveness ,and simply will not flex.
On occasion i get lucky,and find a core now and then which is not glued together,but only held together by the tape wrapped around them.
Brad
Yes, I have had more failure than success in separating glued E cores.
A question on your present setup: do you plan to build a circuit around the setup to loop the output from C1 back to the input ? Such circuit should include the 28-29 kHz frequency source (with the 20-25% duty cycle) and should have roughly a 70-80% efficiency (to be on the safe side) if the setup has a COP between 1.7 to 2 as you measured. The circuit should run from the DC output available across the output capacitor.
Gyula
Quote from: gyula on 2018.09.02, 12:59:55
Yes, I have had more failure than success in separating glued E cores.
A question on your present setup: do you plan to build a circuit around the setup to loop the output from C1 back to the input ? Such circuit should include the 28-29 kHz frequency source (with the 20-25% duty cycle) and should have roughly a 70-80% efficiency (to be on the safe side) if the setup has a COP between 1.7 to 2 as you measured. The circuit should run from the DC output available across the output capacitor.
Gyula
Sure
If you or some one can design a loop circuit,operating under the calculated excess energy output,where the output of that circuit is a 25% duty cycle,with a voltage output of around 4 to 5 volt's,and current of around 160mA,then i will build it,and give it a go. O0
Now,here is the problem--
The current through the 15kohm resistor at 22 volts is only 1.46mA :o
Brad
Quote from: Grumage on 2018.09.02, 12:31:23
Ah, great. O0
Yes. I did several videos on the subject perhaps 3 years ago?
Aided by Verpies we found that mechanical stimulation of Ferite cores can produce a voltage in a coil wound around them. Admittedly voltage ain't power and I didn't subject the coils to a load resistance, perhaps you could take things a step further?
I've just realised, this perhaps isn't the best place for discussion as we're getting a little off topic, my apologies PM.
Cheers Grum.
Ok.ill throw this in here quickly.
The quick and dirty test.
A MOT with a 100 ohm resistor across the primary,and me hitting the core with a small ring spanner,yielded the below results.
Looks like i hit at about 6 Hz :D :D
Now,lets imagine this being two core halves being pulled together via the magnetic attraction force during the input pulse at a frequency of say 28KHz ;)
Brad
Quote from: TinMan on 2018.09.02, 12:53:13
...
Is there anyone here that can carry out the manual calculations ,using the scope shot below,to confirm my results?.
...
Brad,
Your calculations numerically are correct from the data available from your measurements. Due to the debate on measurement protocols, it seems the only convincing method would be the use of a correct looping circuit, that is why I asked.
Thanks for the answer,
Gyula
Quote from: TinMan on 2018.09.02, 13:11:24
...
Now,here is the problem--
The current through the 15kohm resistor at 22 volts is only 1.46mA :o
Yes, it sounds a problem. If you reduce the value of the load resistance so that the C1 voltage go down to say as low as 6-7V only, then
1) what reistor value is needed for that?
2) how the input current changes then?
Quote from: gyula on 2018.09.02, 13:32:00
Yes, it sounds a problem. If you reduce the value of the load resistance so that the C1 voltage go down to say as low as 6-7V only, then
1) what reistor value is needed for that?
2) how the input current changes then?
OK,as C1 is charging,the current and voltage in value decreases.
If i decrease the value of the load resistor to 100 ohms,the P/in value go's up.
Can we use IONs circuit below,or a modification of it?
Brad
Well, Ion's circuit needs a negative resistance (breakdown) device that fires in the 10-20V range or so in your case. A diac needs at least 30-35V to fire, more suitable for Pm setup.
I am thinking on a DC-DC converter circuit as previously I indicated in this thread but the energy available in C1 may prove low.
EDIT it is in my post #63, for your setup a similar but a step down converter could be attempted.
Quote from: TinMan on 2018.09.02, 07:06:06
Ok,i have done what you suggested(see scope shot below),but i disagree with the highlighted above,where this is not showing total corrected P/in. This is showing total power dissipated by the CVR and Primary of the transformer,which is 69.4mW
Yes, exactly. I know this, and this is part of the procedure. There is a step that perhaps you skipped which subtracts out the CVR power.
Quote
By doing it this way,the scopes math is including power being sent back to the source-i would suspect via the diodes junction capacitance for a brief time,and then once the voltage inverts across the coil,current then flows to the SG via the ground rail,through the SGs circuit,and then back into the DUT.
How do we know this?--
Well,lets have a look at the scope shot below.
See the time period i have circled in green.
If this was power being sent to the circuit by the SG,how come the voltage is inverted across the primary?
This area in green is the primary coils magnetic field collapsing(the flyback period).
So,we know from years of testing,that this is the time when the supply is cut off,and we get our inductive kickback,where the voltage inverts. During the 80% off time of the SG,the transformer is now the source--not the SG,but the scope math still sees this as power flowing through the primary of the transformer,and so is adding that to it's calculations.
What we want to know is the power being delivered to the DUT by the SG only,during it's 20% delivery cycle.
Please see next post,where i show clearly when power is being delivered by the SG to the DUT.
Brad
If you wish for my help I'd like to remain focused on the procedure so that there can be an understanding of the what and why.
So you obtained a total power of 69.4mW (or is it 68.4mW?).
Can you carry out the rest of the steps in the procedure and
let us know how the result compares with P(out)? I see you obtained the RMS of the input current (59.4mA), so you have the data you need.
Quote from: gyula on 2018.09.02, 13:47:20
Well, Ion's circuit needs a negative resistance (breakdown) device that fires in the 10-20V range or so in your case. A diac needs at least 30-35V to fire, more suitable for Pm setup.
I am thinking on a DC-DC converter circuit as previously I indicated in this thread but the energy available in C1 may prove low.
EDIT it is in my post #63, for your setup a similar but a step down converter could be attempted.
Diacs and Trigger diodes are available in a wide range of firing voltages. You can even build your own from a pnp-npn regenerative pair and tailor the firing voltage with b-e resistors.
At this point I am running sims on such regenerative firing circuits and will post the results later. I would hold off on this method at present.
When power measurements of the presently evolved OU circuits of PM and Tinman are proven with alternative measurement techniques we can then with confidence proceed to closing the loop.
If you wish to skip the alternative measuring methods, then to loop, I would recommend one of the many proven class C single transistor oscillators as a simple DC to AC converter.
Regards
P.S Thanks to Poynt for stepping in to clarify the measurement protocol. We enjoy your expert posts, so please continue.
Quote from: TinMan on 2018.09.02, 10:24:22
Hi Itsu
That is a big step up you have with that transformer,which may be the problem ?
Also,dose your FG have a function where you can set the impedance to 0 ,such as mine has?
Would be good to see what type of waveforms you have across the scope as well.
As seen in all my video's,i get very clean waveforms,and the results are very easy to see.
You should be able to get 20v across the 15kohm resistor with no problem at all--maybe some tuning is needed?,due to different components.
After testing many different transformers,i finally found another that yields the COPs i have been showing-im not sure why very few give me these results,and where others of the same type(E core ferrite)show below COP+ results.
Brad
Brad,
yes, could be the step up ratio is to much.
My Rigol FG can dial in any impedance (1 Ohm being the lowest), but i don't think it will physically change
its output impedance, it will only adjust the amplitude readout (it will always stay at 50 Ohm).
Will take some screenshots tonight.
When using 20Vpp (AC) square wave signal (with input diode), i get about 6.5V across the 15K resistor.
But this input diode messes up the signal (ringing), so i also used 10Vpp (DC) square wave without the diode
which gives much cleaner signals.
So i will be looking / testing some different (ferrite cored) transformers to see if i can obtain a
lower ratio and somewhat higher primary impedance.
Gyula,
thanks for the reminder of the 36 Ohm reactance at 28.8KHz, it should be enough then for a clean signal.
Like Brad mentioned, those flittle ferrite x-formers are unforgiveable when messing with them.
Itsu
1st screenshot is voltage across the prim. in yellow, using 20Vpp AC square wave 20% duty cycle (see white trace)
and using an UF4007 diode in the red supply lead.
Notice the ringing and spikes on the yellow signal.
2th screenshot is voltage across the prim. in yellow, using 10V DC square wave 20% duty cycle (see white trace)
and using NO diode in the red supply lead.
Notice the ringing is gone and less spikes.
3th screenshot is an increased vertical signal shot of the yellow voltage signal from screenshot 2 situation
Itsu
I found another ferrit cored transformer:
prim. 6mH / 2 Ohm
sec. 2.5H / 30 Ohm
Using the 10V DC 20% duty cycle 28.8KHz signal without input diode.
I have a DC mA DMM in the red plus lead showing the (37.2mA rms? mean?) current going in, see picture.
(a 2th Fluke DMM across a temporary installed 1 Ohm csr shows 37mV).
The screenshot shows;
yellow: the voltage across the primary (still large spikes) 911mV rms
green: the current from my current probe in the red FG supply lead 105mA rms
red: the math function yellow x green presenting the average input power 20.8mW
So we see 20.8mW input.
A 3th fluke DMM shows the voltage across the 14860 Ohm resistor to be 11.2V
This represents a power out of 8.7mW indicating a cop = 0.42.
Strange is the difference between the input current DMM (37.2mA) backed up by the temp 37mV across a 1 Ohm resistor
and the current probe input current measurement of 105mA rms.
Are the DDM's unable to accuratly display the 28.8KHz current/voltage? It looks like it.
Using another methode of obtaining the input current (2 probes in differential setup across the temp. 1 Ohm csr)
shows a similar current signal as with the current probe, but due to the low level of signals the specific value
is hard to obtain, but i tend to believe the current probe is showing the corrent input current.
Regards Itsu
Dear Ion,
Okay, I understand your take on the looping conditions.
Very briefly: For me, diacs mainly mean the gate trigger devices for triacs used for mains power control so I automatically wrote to Brad the 30-35V trigger voltage they have for such purposes. Okay on building trigger diodes with quasi programmable firing voltages: I simply do not call them diacs. I agree that trigger diodes (like Sidac) and other bilateral switches are available in wide firing voltage ranges. These are my take on these, no offense intended. Brad asked me on your suggested regenerative circuit and I wanted to keep it short: it is your baby, you run with it.
Thanks for recommending the Class C oscillator, I agree and I would add here the use of Class E oscillators to get high efficiency DC to AC conversion. Efficiency wise I think such oscillators in practice are in par with the DC-DC converters I already referred to in my post #63. But Brad's setup involves lower voltages than Partzman's, so careful selections are needed. I agree that a direct DC to AC conversion can be simpler.
Gyula
Quote from: TinMan on 2018.09.02, 13:11:24
...
Now,here is the problem--
The current through the 15kohm resistor at 22 volts is only 1.46mA :o
Well, you have 22V DC stored in a 2200 uF capacitor. So there is 0.5324 Joule in it.
What is needed is to reproduce the FG's output pulses that has a 20Vpp amplitude across 50 Ohm internal resistance (when unloaded) and the pulse ON time is about 6.7-7 usec (i.e. duty cycle is 25%), frequency is 28.8 kHz.
Is the 20Vpp correct when the FG output is unloaded? I mean is that what reduces to 4-5 Vpp when the transformer input is hooked up to the FG? I take the 20Vpp from your videos, it is displayed on the FG.
I mention these because we can take out in theory 0.5324 Watt in one second from C1, this would involve using (V
2/0.5324) = 909 Ohm load resisitor and the load current would be roughly 24.2 mA. Now for your transformer input the pulse width needs to be 'only' roughly 7 usec, so this is what is needed to consider when thinking on whether C1 stores enough energy for looping. So taking-out the pulses from C1 should be in a balance with the looped back pulses at least for a COP=1. I ask: Is this way of thinking correct?
Gyula
Quote from: Itsu on 2018.09.02, 19:25:15
I found another ferrit cored transformer:
prim. 6mH / 2 Ohm
sec. 2.5H / 30 Ohm
Using the 10V DC 20% duty cycle 28.8KHz signal without input diode.
I have a DC mA DMM in the red plus lead showing the (37.2mA rms? mean?) current going in, see picture.
(a 2th Fluke DMM across a temporary installed 1 Ohm csr shows 37mV).
The screenshot shows;
yellow: the voltage across the primary (still large spikes) 911mV rms
green: the current from my current probe in the red FG supply lead 105mA rms
red: the math function yellow x green presenting the average input power 20.8mW
So we see 20.8mW input.
A 3th fluke DMM shows the voltage across the 14860 Ohm resistor to be 11.2V
This represents a power out of 8.7mW indicating a cop = 0.42.
Strange is the difference between the input current DMM (37.2mA) backed up by the temp 37mV across a 1 Ohm resistor
and the current probe input current measurement of 105mA rms.
Are the DDM's unable to accuratly display the 28.8KHz current/voltage? It looks like it.
Using another methode of obtaining the input current (2 probes in differential setup across the temp. 1 Ohm csr)
shows a similar current signal as with the current probe, but due to the low level of signals the specific value
is hard to obtain, but i tend to believe the current probe is showing the corrent input current.
Regards Itsu
Itsu
A few things to try.
1-try using a 10 ohm cvr,so as the signal amplitude is larger.
2-set you scope to average a high number of samples--this should clean up the waveforms.
3-adjust input frequency to point where highest voltage is obtained across the 15kohm resistor--that is what i do.
4-with the CVR on the ground rail,you should be able to use your scope to obtain the voltage across it during the 20% on time.
This gives you your current in. You then average that current over a whole cycle. If your current probe gives you a different value,then the current probe is wrong,as we have seen many times in the past.
DMMs do a very good job at averaging out pulsed DC current,as Poynt and myself have shown many times-even at higher frequencies.
Brad
Quote from: poynt99 on 2018.09.02, 14:09:31
Yes, exactly. I know this, and this is part of the procedure. There is a step that perhaps you skipped which subtracts out the CVR power.
If you wish for my help I'd like to remain focused on the procedure so that there can be an understanding of the what and why.
So you obtained a total power of 69.4mW (or is it 68.4mW?).
Can you carry out the rest of the steps in the procedure and let us know how the result compares with P(out)? I see you obtained the RMS of the input current (59.4mA), so you have the data you need.
Why is an RMS value used to calculate a pulsed DC waveform?
RMS is used for AC only,or am i wrong there?
Brad
Quote from: gyula on 2018.09.02, 21:31:16
Dear Ion,
Okay, I understand your take on the looping conditions.
Very briefly: For me, diacs mainly mean the gate trigger devices for triacs used for mains power control so I automatically wrote to Brad the 30-35V trigger voltage they have for such purposes. Okay on building trigger diodes with quasi programmable firing voltages: I simply do not call them diacs. I agree that trigger diodes (like Sidac) and other bilateral switches are available in wide firing voltage ranges. These are my take on these, no offense intended. Brad asked me on your suggested regenerative circuit and I wanted to keep it short: it is your baby, you run with it.
Thanks for recommending the Class C oscillator, I agree and I would add here the use of Class E oscillators to get high efficiency DC to AC conversion. Efficiency wise I think such oscillators in practice are in par with the DC-DC converters I already referred to in my post #63. But Brad's setup involves lower voltages than Partzman's, so careful selections are needed. I agree that a direct DC to AC conversion can be simpler.
Gyula
Dear Gyula
Sorry, I should not have used the term DIAC as a bidirectional trigger device is not needed when DC is the source. I should have used a more generic unidirectional device name.
I just happened to have a few hundred 32 V devices on hand along with other types so loosely threw it on my drawing.
I'm not 100% happy with my testing of the trigger diode approach. Although you do get a nice damped oscillation at the tank frequency with each trigger pulse, the repetition rate is not what I had hoped for.
Class E would probably be better, as you stated.
Thanks for your input
Kind Regards
Quote from: Itsu on 2018.09.02, 19:25:15
I found another ferrit cored transformer:
prim. 6mH / 2 Ohm
sec. 2.5H / 30 Ohm
Using the 10V DC 20% duty cycle 28.8KHz signal without input diode.
I have a DC mA DMM in the red plus lead showing the (37.2mA rms? mean?) current going in, see picture.
(a 2th Fluke DMM across a temporary installed 1 Ohm csr shows 37mV).
The screenshot shows;
yellow: the voltage across the primary (still large spikes) 911mV rms
green: the current from my current probe in the red FG supply lead 105mA rms
red: the math function yellow x green presenting the average input power 20.8mW
So we see 20.8mW input.
A 3th fluke DMM shows the voltage across the 14860 Ohm resistor to be 11.2V
This represents a power out of 8.7mW indicating a cop = 0.42.
Strange is the difference between the input current DMM (37.2mA) backed up by the temp 37mV across a 1 Ohm resistor
and the current probe input current measurement of 105mA rms.
Are the DDM's unable to accuratly display the 28.8KHz current/voltage? It looks like it.
Using another methode of obtaining the input current (2 probes in differential setup across the temp. 1 Ohm csr)
shows a similar current signal as with the current probe, but due to the low level of signals the specific value
is hard to obtain, but i tend to believe the current probe is showing the corrent input current.
Regards Itsu
Itsu
Looking at your picture,i see you are connected straight to the transformer,and have bipassed the diode ?
Your voltage trace also shows this,and it looks like the FG is clamping the flyback.
The diode is a must,and your FGs offset should be at 0 volts during the 80%off time-as per the scopeshot i provided showing the pulsed waveform from my FG.
Brad
Quote from: TinMan on 2018.09.03, 00:40:05
Why is an RMS value used to calculate a pulsed DC waveform?
RMS is used for AC only,or am i wrong there?
Brad
If you are measuring a voltage across a pure resistance, no matter what type of wave form, the RMS voltage across it will always allow one to calculate the real power dissipated in that resistor, or calculate the RMS current through it.
The scope traces indicate that there are reactances involved in this circuit and therefore one can not presume a DC type measurement will produce an accurate result. How much average power would be calculated with a 2Vpp sine wave across a 10R resistor using the average of the voltage? How much would be calculated using the RMS of the voltage?
Disabuse yourself of any preconceived notions regarding pulsed DC or AC, average or RMS, and give the measurement a shot. If you don't wish to pursue this line of testing, then that's ok, just let me know. I came here only because Peter asked me to take a look and see if I could figure things out. I realize however that I was not invited by the experimenters, and therefore will gracefully bow out if asked. I don't wish to waste anyone's time, nor my own.
If results are posted from the procedure I will continue, otherwise I will gracefully leave it to your capable hands.
Cheers.
Quote from: poynt99 on 2018.09.03, 03:52:00
If you are measuring a voltage across a pure resistance, no matter what type of wave form, the RMS voltage across it will always allow one to calculate the real power dissipated in that resistor, or calculate the RMS current through it.
The scope traces indicate that there are reactances involved in this circuit and therefore one can not presume a DC type measurement will produce an accurate result. How much average power would be calculated with a 2Vpp sine wave across a 10R resistor using the average of the voltage? How much would be calculated using the RMS of the voltage?
Disabuse yourself of any preconceived notions regarding pulsed DC or AC, average or RMS, and give the measurement a shot. If you don't wish to pursue this line of testing, then that's ok, just let me know. I came here only because Peter asked me to take a look and see if I could figure things out. I realize however that I was not invited by the experimenters, and therefore will gracefully bow out if asked. I don't wish to waste anyone's time, nor my own.
If results are posted from the procedure I will continue, otherwise I will gracefully leave it to your capable hands.
Cheers.
Poynt,please hang around,and your time here is much valued by us all.
I will get on with your method of measuring tonight when i get home,and post my results ASAP.
One question--
Should your measuring method still produce a COP of 200%+,then what?.
Please forgive me if i seem a little stubborn here,but i remember all the testing you (and myself)did using DMMs to measure pulsed DC current and voltages,and found them to be very accurate.
But after carrying out that test with great care,and tripple checking the results,it now seems that DMMs are no good ??? .
Is this just because they too returned an OU result?
Also,if the current and voltage waveforms show an inphase relationship during the 20% on time,can this not be seen as a pure resistance?--disregarding the other 80% the cycle.
Brad
Poynt, yes, please stick around somewhat longer to sort this thingy out.
Brad,
QuoteItsu
Looking at your picture,i see you are connected straight to the transformer,and have bipassed the diode ?
Your voltage trace also shows this,and it looks like the FG is clamping the flyback.
The diode is a must,and your FGs offset should be at 0 volts during the 80%off time-as per the scopeshot i provided showing the pulsed waveform from my FG.
I explained about this diode in my posts #125/126 above (it introduces massive ringing), but if that diode is
a must for this circuit to work then i will reinstall it.
Quote1-try using a 10 ohm cvr,so as the signal amplitude is larger.
2-set you scope to average a high number of samples--this should clean up the waveforms.
3-adjust input frequency to point where highest voltage is obtained across the 15kohm resistor--that is what i do.
4-with the CVR on the ground rail,you should be able to use your scope to obtain the voltage across it during the 20% on time.
This gives you your current in. You then average that current over a whole cycle. If your current probe gives you a different value,then the current probe is wrong,as we have seen many times in the past.
1- I want to eliminate the cvr completely as it complicates the calculations.
3- So are you hunting for resonance then?
Thanks Itsu
Quote from: Itsu on 2018.09.03, 09:03:52
Poynt, yes, please stick around somewhat longer to sort this thingy out.
Brad,
Thanks Itsu
Hi Itsu
QuoteI explained about this diode in my posts #125/126 above (it introduces massive ringing), but if that diode is
a must for this circuit to work then i will reinstall it.
I do not understand as to why you are getting this ringing ???,as all my waveforms are very clean,no matter what transformer i use. Maybe it is that diode--have you tried a simple 1n4007?.
Below i have added two scope shot's,which are taken from two different circuits,using different transformers.
As you can see(as in my 3 video's),my waveforms are all quite clean,and have no ringing.
Quote3- So are you hunting for resonance then?
In my years of tinkering,i have never found that resonance has some form of hidden magic,or a means to gain energy. So no,i am not looking for resonance. I just tune the circuit to gain the highest output for the lowest input.
QuoteI want to eliminate the cvr completely as it complicates the calculations.
I would keep the CVR,and get rid of the troublesome current probe.They seem to be quite unreliable when trying to measure very small currents.
But first,i would work on why your traces are so noisy
Brad
Quote from: poynt99 on 2018.09.03, 03:52:00
If you are measuring a voltage across a pure resistance, no matter what type of wave form, the RMS voltage across it will always allow one to calculate the real power dissipated in that resistor, or calculate the RMS current through it.
The scope traces indicate that there are reactances involved in this circuit and therefore one can not presume a DC type measurement will produce an accurate result. How much average power would be calculated with a 2Vpp sine wave across a 10R resistor using the average of the voltage? How much would be calculated using the RMS of the voltage?
Disabuse yourself of any preconceived notions regarding pulsed DC or AC, average or RMS, and give the measurement a shot. If you don't wish to pursue this line of testing, then that's ok, just let me know. I came here only because Peter asked me to take a look and see if I could figure things out. I realize however that I was not invited by the experimenters, and therefore will gracefully bow out if asked. I don't wish to waste anyone's time, nor my own.
If results are posted from the procedure I will continue, otherwise I will gracefully leave it to your capable hands.
Cheers.
OK,the results following your measurement procedure Poynt
Data from the scope shot below.
Our average from the math trace is 675mW/10=67.5mW--our total corrected power in.
RMS value across the 10 ohm CVR is 765mV
Power dissipated by the CVR is 58.52mW
Current through the primary coil is 76.5mA
P/in for the primary winding is there for 67.5-58.52= 8.98mW
Is this correct Poynt?
Brad
Quote from: TinMan on 2018.09.03, 05:09:43
One question--
Should your measuring method still produce a COP of 200%+,then what?.
Then we look deeper to verify the results.
Quote
Please forgive me if i seem a little stubborn here,but i remember all the testing you (and myself)did using DMMs to measure pulsed DC current and voltages,and found them to be very accurate.
But after carrying out that test with great care,and tripple checking the results,it now seems that DMMs are no good ??? .
Is this just because they too returned an OU result?
For P(in) measurements, DMM's are excellent for finding the average on a signal/source that is/should be flat DC. Ainslie's circuit used a battery source for the input. The load was switched, therefore the current pulsed. So when measuring power back at the battery (the source), the voltage was a no-brainer, as it is simply the loaded battery voltage. The current on the other hand is pulsed, and this is where the DMM measurement shines above the rest, because it does an excellent job averaging the current from the fixed DC source. One simply multiplies the fixed loaded DC voltage times the measured average current, and you have the input power being delivered by the battery. In your case the source voltage is not fixed, as it is a pulsed DC output. So we have to use the tried and true method with the scope doing AVG[v(t) x i(t)]. BTW, this is an instantaneous measurement done by the scope, as it is taking many samples and performing the math on each sample across the full cycle. Doing this manually as you have been is not the same thing.
Quote
Also,if the current and voltage waveforms show an inphase relationship during the 20% on time,can this not be seen as a pure resistance?--disregarding the other 80% the cycle.
That's kinda the problem; there is other stuff happening in that wave form, and it all has to be accounted for. The best method we have available to achieve this is the one I've been asking you to use.
Quote from: TinMan on 2018.09.03, 12:01:39
OK,the results following your measurement procedure Poynt
Data from the scope shot below.
Our average from the math trace is 675mW/10=67.5mW--our total corrected power in.
RMS value across the 10 ohm CVR is 765mV
Power dissipated by the CVR is 58.52mW
Current through the primary coil is 76.5mA
P/in for the primary winding is there for 67.5-58.52= 8.98mW
Is this correct Poynt?
Brad
Your math looks correct.
This seems low, and I am assuming that the P(out) is a fair bit higher in this case.
In my experience working with the scope and this method, it is important to use a high number of cycles on the display when utilizing the "measurements" available in the scope, especially for MEAN, RMS measurements. Also, be careful with the selection of MEAN options, as there is often a "Cycle MEAN", and just "MEAN". I prefer to use just "MEAN", and have many cycles showing on the screen, at least 25 cycles. This should produce the most accurate measurement. So I would suggest you experiment with this as well, and note how or if the numbers_in_boxes changes as you vary the number of cycles on the screen.
Another important adjustment, is the samples per second, and number of points in the screen. Your screen shot indicates 500Ms/s, and 120k points. This will provide a fast screen update, but may not provide the most accurate measurement. See if you can increase the samples and points and observe the effect (if any) on the measurements.
The other potential source of error is the CVR. Even though the frequency is relatively low (~28kHz), the quick rise and fall times of the source signal will bring into effect the circuit reactances, particularly across the CVR. Because the CVR is a relatively low value of 10 Ohms, only a tiny bit of inductance in the CVR is necessary to significantly skew the resulting voltage measured across it. I strongly suggest that those pretty gold power resistors
not be used as CVR's. This has been discussed many many times over the years, and several CVR options suggested, esp by TK. Because you are dealing with such low power, why not use a simple 1/2W or 1W carbon resistor? The value/tolerance is not too critical, as you can measure that accurately, and make the necessary allowance in your calculations. At least then the inductance will be at a minimum. Also,
ensure that the scope probes are directly across the CVR as close to the resistor body as possible.
All,
Yesterday, I was trying to use some of the other mosfets using the newly measured capacitance change data and the results were terrible. So, I thought I would replicate my previous circuit and data in post #9 for comparison and all attempts throughout the day to replicate the performance failed with all circuit values and measurement procedures identical. The COPs were ~.8-.9 at best. The weather here was overcast and rainy off and on all day during these test attempts!
This morning, all the weather has moved on and we have a bright sunny day at the moment and when running the replication test at this moment in time, the COPs with gain have returned. Any suggestions as to why this is happening? I might add that my lab is basically climate controlled so I believe there is no test equipment variables to deal with. Is it possibly diurnal although I tested both day and night? Or is the frequency close to the first odd integer frequency of 710,611 Hz as calculated with Dr. Stifflers lattice energy formula he disclosed in this video-
https://www.youtube.com/watch?v=oPikYP40NS8&t=87s
You probably won't be able to read the formula but thanks to iQuest and Mikrovolt on OU, the following is an accurate representation.
Lv=(Zo*(c/1e6)*2pi)*n
Lv - Lattice Vibration Frequency (Frequency At Which Energy Lattice Will Release Energy)
Zo - Free Space Impedance (120*PI), may be rounded to 377
c - Speed Of Light (300M meters/sec)
n - Every Odd Integer
Lv odd integer frequencies:
1 - 710,611 3 - 2,131,834 5 - 3,553,057 7 - 4,974,280 9 - 6,395,503 11 - 7,816,726
13 - 9,237,949 15 - 10,659,172 17 - 12,080,395 19 - 13,501,618 21 - 14,922,841 23 - 16,344,064
Any thots?
At Itsu: What is your weather like at the time or when you attempted to replicate this circuit?
Regards,
Pm
This is a follow up of sorts to my last post in regards to Dr. Stiffler's energy lattice formula in which the first odd integer calc results in 710,611Hz. This is a test of the circuit used in my post #9 except the frequency is now 710,611Hz, C2 is changed to 2108pf to change the transfer function, and the scan is now 200ms with a sample rate of 50MS/s.
The first scope pix shows the measurements at 100ms and the second pix is the same area expanded. The scope shot was taken with a single hi-res sample which has 16 bit vertical resolution.
The data table is also shown.
Regards,
Pm
Checking my current probe seems that its in good working order, however while doing some comparing measurments,
it seems that the DMM's in DC mA mode and in DC mV mode show the MEAN (Average) current/voltage.
While the scope (my scope) uses for the math calculations x (and probably / ) the both RMS values of the signals.
When using the - or + math function it uses the mean values of the signals.
This seems to be in my TDK3000 series of scope.
So be aware, while you can display the signals in mean on your scope, the math function x and / will probably
take the rms value of those signals and display the result in what you set it, normally in mean.
This means that 3 methods of obtaining the current via a cvr (DC mA DMM in the supply line, DC mV DMM across the cvr and the differential probe setup CHx - CHY) all use the MEAN
values, while the Voltage x current math function uses the rms values.
Can this be right?
Itsu
Quote from: partzman on 2018.09.03, 17:17:48
All,
Yesterday, I was trying to use some of the other mosfets using the newly measured capacitance change data and the results were terrible. So, I thought I would replicate my previous circuit and data in post #9 for comparison and all attempts throughout the day to replicate the performance failed with all circuit values and measurement procedures identical. The COPs were ~.8-.9 at best. The weather here was overcast and rainy off and on all day during these test attempts!
This morning, all the weather has moved on and we have a bright sunny day at the moment and when running the replication test at this moment in time, the COPs with gain have returned. Any suggestions as to why this is happening? I might add that my lab is basically climate controlled so I believe there is no test equipment variables to deal with. Is it possibly diurnal although I tested both day and night? Or is the frequency close to the first odd integer frequency of 710,611 Hz as calculated with Dr. Stifflers lattice energy formula he disclosed in this video-
https://www.youtube.com/watch?v=oPikYP40NS8&t=87s
You probably won't be able to read the formula but thanks to iQuest and Mikrovolt on OU, the following is an accurate representation.
Lv=(Zo*(c/1e6)*2pi)*n
Lv - Lattice Vibration Frequency (Frequency At Which Energy Lattice Will Release Energy)
Zo - Free Space Impedance (120*PI), may be rounded to 377
c - Speed Of Light (300M meters/sec)
n - Every Odd Integer
Lv odd integer frequencies:
1 - 710,611 3 - 2,131,834 5 - 3,553,057 7 - 4,974,280 9 - 6,395,503 11 - 7,816,726
13 - 9,237,949 15 - 10,659,172 17 - 12,080,395 19 - 13,501,618 21 - 14,922,841 23 - 16,344,064
Any thots?
At Itsu: What is your weather like at the time or when you attempted to replicate this circuit?
Regards,
Pm
PM,
interesting, i have seen that formula, but i understand that the jury is still out on that formula untill some math guru's
have given the OK on it.
Our weather was unusually dry and warm this summer, only untill 3 weeks ago or so it changed back to our normal
Dutch summer weather meaning humid (70%), cooler, partly clouded and some rain now and then.
It was in those last 3 weeks that i was doing the replication on your circuit.
Itsu
Quote from: TinMan on 2018.09.03, 11:33:46
Hi Itsu
I do not understand as to why you are getting this ringing ???,as all my waveforms are very clean,no matter what transformer i use. Maybe it is that diode--have you tried a simple 1n4007?.
Below i have added two scope shot's,which are taken from two different circuits,using different transformers.
As you can see(as in my 3 video's),my waveforms are all quite clean,and have no ringing.
In my years of tinkering,i have never found that resonance has some form of hidden magic,or a means to gain energy. So no,i am not looking for resonance. I just tune the circuit to gain the highest output for the lowest input.
I would keep the CVR,and get rid of the troublesome current probe.They seem to be quite unreliable when trying to measure very small currents.
But first,i would work on why your traces are so noisy
Brad
I changed the following on my setup:
using a 10 Ohm CVR
using a 1N4007 diode in the supply line
lowering the FG frequency to a point of max. voltage across the load resistor (2.2Khz).
We have a much stable signals (also due to the lower frequency) but the 1N4007 introduces much less
ringing compared the the (better?) UF4007.
The input current seems to be 12.5mA on the DC mA DMM in the supply lead, 125mV on the DC mV DDM across
the 10 Ohm cvr, 130mV mean using 2 probes in differential mode and using math CH2 - CH3 and 31.2mA RMS using
the current probe (9.7mA mean).
The voltage across the primary is 3.64V RMS.
The power in calculated by the math function shows 50mW.
The voltage across the load resistor (14860 Ohm) is 24.64V for a power out of 40.7mW.
COP therefor is now 0.82
Itsu
Quote from: Itsu on 2018.09.03, 19:53:03
Checking my current probe seems that its in good working order, however while doing some comparing measurments,
it seems that the DMM's in DC mA mode and in DC mV mode show the MEAN (Average) current/voltage.
While the scope (my scope) uses for the math calculations x (and probably / ) the both RMS values of the signals.
When using the - or + math function it uses the mean values of the signals.
This seems to be in my TDK3000 series of scope.
So be aware, while you can display the signals in mean on your scope, the math function x and / will probably
take the rms value of those signals and display the result in what you set it, normally in mean.
This means that 3 methods of obtaining the current via a cvr (DC mA DMM in the supply line, DC mV DMM across the cvr and the differential probe setup CHx - CHY) all use the MEAN
values, while the Voltage x current math function uses the rms values.
Can this be right?
Itsu
Itsu,
Well, the best way to tell is to do an experiment and although I'm using my Tek MDO, your TDS should measure in the same manner.
The test setup is a signal generator set at 500kHz sine wave driving a 100 ohm 1% non-inductive resistor with the current being measured with a Tek TCP0020 current probe. I could have used a CSR and the results would have been more accurate on the second test.
The first scope pix shows the voltage and current waveforms with very little phase difference and the resulting mean Math channel calculation. So, we multiply the rms values and get Pin = 4.415*42.47e-3 = 187.5mw. neglecting the small phase angle. This compares favorably with the Math channel value of 187.3w.
In the next scope pix, we've added a ~100uh inductor to create some phase shift to move the power factor off unity so we'll be forced to use cosine correction. Therefore now the Pin = 6.118*16.64e-3*cosine(72.48) = 30.65mw. The Math shows 28.17mw for an error of 9% which I attribute to the fact that we are in the minimum current range of 10ma for the TCP0020 probe rather than the internal math calculations of the Tek scope.
Rather than repeat here, one can look at my post #88 for an explanation of how the "mean" or average measurements are taken on complex waveforms with the current Tek scopes.
Regards,
Pm
Interesting!
PM writes: "Yesterday, I was trying to use some of the other mosfets using the newly measured capacitance change data and the results were terrible. So, I thought I would replicate my previous circuit and data in post #9 for comparison and all attempts throughout the day to replicate the performance failed with all circuit values and measurement procedures identical. The COPs were ~.8-.9 at best. The weather here was overcast and rainy off and on all day during these test attempts!
This morning, all the weather has moved on and we have a bright sunny day at the moment and when running the replication test at this moment in time, the COPs with gain have returned. Any suggestions as to why this is happening? "
Air humidity is clearly suspect as the culprit. Possibly in your lab, but possibly humidity in the air outside...
WHY humidity might matter so much, is another question.
I would say it is very valuable to be able to turn the effect "off", and you have evidently found one way to do this, with increased air humidity.
Quote from: Itsu on 2018.09.03, 19:53:03
Checking my current probe seems that its in good working order, however while doing some comparing measurments,
it seems that the DMM's in DC mA mode and in DC mV mode show the MEAN (Average) current/voltage.
While the scope (my scope) uses for the math calculations x (and probably / ) the both RMS values of the signals.
When using the - or + math function it uses the mean values of the signals.
This seems to be in my TDK3000 series of scope.
So be aware, while you can display the signals in mean on your scope, the math function x and / will probably
take the rms value of those signals and display the result in what you set it, normally in mean.
This means that 3 methods of obtaining the current via a cvr (DC mA DMM in the supply line, DC mV DMM across the cvr and the differential probe setup CHx - CHY) all use the MEAN
values, while the Voltage x current math function uses the rms values.
Can this be right?
Itsu
Hi Itsu.
What indications are you seeing that leads you to conclude the scope is performing any conversion (MEAN or RMS) on the v(t) x i(t) components?
When you set up the MATH channel/trace to perform CH1 x CH2 (for example), there is no processing before or after the result. You can see the result in Brad's scope shots (the purple trace), and it is just the raw output or result of the v(t) x i(t) process. When you set up a measurement on a channel/trace,
then processing is performed and the result is indicated in the box labeled AVG or MEAN, or RMS etc.
In the case of the MATH trace set up for measuring power, the raw product is processed with the MEAN measurement function in order to take the instantaneous power trace and convert it to the average power.
Quote from: poynt99 on 2018.09.03, 16:03:30
Your math looks correct.
This seems low, and I am assuming that the P(out) is a fair bit higher in this case.
In my experience working with the scope and this method, it is important to use a high number of cycles on the display when utilizing the "measurements" available in the scope, especially for MEAN, RMS measurements. Also, be careful with the selection of MEAN options, as there is often a "Cycle MEAN", and just "MEAN". I prefer to use just "MEAN", and have many cycles showing on the screen, at least 25 cycles. This should produce the most accurate measurement. So I would suggest you experiment with this as well, and note how or if the numbers_in_boxes changes as you vary the number of cycles on the screen.
Another important adjustment, is the samples per second, and number of points in the screen. Your screen shot indicates 500Ms/s, and 120k points. This will provide a fast screen update, but may not provide the most accurate measurement. See if you can increase the samples and points and observe the effect (if any) on the measurements.
The other potential source of error is the CVR. Even though the frequency is relatively low (~28kHz), the quick rise and fall times of the source signal will bring into effect the circuit reactances, particularly across the CVR. Because the CVR is a relatively low value of 10 Ohms, only a tiny bit of inductance in the CVR is necessary to significantly skew the resulting voltage measured across it. I strongly suggest that those pretty gold power resistors not be used as CVR's. This has been discussed many many times over the years, and several CVR options suggested, esp by TK. Because you are dealing with such low power, why not use a simple 1/2W or 1W carbon resistor? The value/tolerance is not too critical, as you can measure that accurately, and make the necessary allowance in your calculations. At least then the inductance will be at a minimum. Also, ensure that the scope probes are directly across the CVR as close to the resistor body as possible.
Ok,i will up the sample rate on the screen,and give that a shot.
A few things--
This test was done on my second DUT,and the frequency is around 14.7KHz.
The resistors used on this DUT are the 1/2 watt carbon resistors..
On that note,the large gold ones were described as being carbon resistors by the seller,not the wire wound cement type,such as some of those white block ones i have,which is why i bought them-the gold ones.
I will redo the test tonight,post the results,and touch on a !hiccup! using this test method.
Brad
Quote from: TinMan on 2018.09.04, 00:28:16
Ok,i will up the sample rate on the screen,and give that a shot.
A few things--
This test was done on my second DUT,and the frequency is around 14.7KHz.
The resistors used on this DUT are the 1/2 watt carbon resistors..
On that note,the large gold ones were described as being carbon resistors by the seller,not the wire wound cement type,such as some of those white block ones i have,which is why i bought them-the gold ones.
I will redo the test tonight,post the results,and touch on a !hiccup! using this test method.
Brad
Sounds good Brad.
I hope those goldies are indeed low L. You could certainly measure their inductance, and compare to the 1/2W carbons.
Quote from: poynt99 on 2018.09.03, 15:37:22
Then we look deeper to verify the results.
For P(in) measurements, DMM's are excellent for finding the average on a signal/source that is/should be flat DC. Ainslie's circuit used a battery source for the input. The load was switched, therefore the current pulsed. So when measuring power back at the battery (the source), the voltage was a no-brainer, as it is simply the loaded battery voltage. The current on the other hand is pulsed, and this is where the DMM measurement shines above the rest, because it does an excellent job averaging the current from the fixed DC source. One simply multiplies the fixed loaded DC voltage times the measured average current, and you have the input power being delivered by the battery. In your case the source voltage is not fixed, as it is a pulsed DC output. So we have to use the tried and true method with the scope doing AVG[v(t) x i(t)]. BTW, this is an instantaneous measurement done by the scope, as it is taking many samples and performing the math on each sample across the full cycle. Doing this manually as you have been is not the same thing.
That's kinda the problem; there is other stuff happening in that wave form, and it all has to be accounted for. The best method we have available to achieve this is the one I've been asking you to use.
Well i must say that it is odd the manually calculating the average current using the scope values,using the scope to calculate the average current value,and the DMMs average current value all result in the same amount,while using the RMS value results in near twice the value--such as Itsu has also found using his current probe RMS value.
But anyway,if i use a battery,and switch the circuit on and off by way of the SG switching a mosfet at the same frequency and duty cycle,we can then trust our DMM to give us an accurate average current value-correct?
This being the case,and our scopes calculated average current value is the same as that of the DMM,and my manually calculated current value is also the same,then that would mean that my methods of calculating the P/in are correct-yes?
But how did you average the voltage in the Ainslie circuit,as it would be the supply voltage x duty cycle to get the average voltage over a full cycle. That is how the math works on a scope-is it not?,where V and I are averaged,and then multiplied together ?
Brad
Quote from: Itsu on 2018.09.02, 19:25:15
I found another ferrit cored transformer:
prim. 6mH / 2 Ohm
sec. 2.5H / 30 Ohm
Using the 10V DC 20% duty cycle 28.8KHz signal without input diode.
I have a DC mA DMM in the red plus lead showing the (37.2mA rms? mean?) current going in, see picture.
(a 2th Fluke DMM across a temporary installed 1 Ohm csr shows 37mV).
Both of these will be AVG current.
Quote
The screenshot shows;
yellow: the voltage across the primary (still large spikes) 911mV rms
green: the current from my current probe in the red FG supply lead 105mA rms
red: the math function yellow x green presenting the average input power 20.8mW
So we see 20.8mW input.
A 3th fluke DMM shows the voltage across the 14860 Ohm resistor to be 11.2V
This represents a power out of 8.7mW indicating a cop = 0.42.
Did you subtract the CVR power from the 20.8mW? To obtain the CVR power, you need to know the RMS voltage across it.
Quote
Strange is the difference between the input current DMM (37.2mA) backed up by the temp 37mV across a 1 Ohm resistor
and the current probe input current measurement of 105mA rms.
It isn't strange, because the DMM measures average current.
Quote
Are the DDM's unable to accuratly display the 28.8KHz current/voltage? It looks like it.
The DMM displays the average very accurately. In general the only time we use the DMM for input power measurement is when the source is a fixed DC supply, and in that case it would be to measure the average current. I(avg) x Vdc = P(in). Since we are not using a fixed DC supply in this device, we don't use the DMM.
Quote
Using another methode of obtaining the input current (2 probes in differential setup across the temp. 1 Ohm csr)
shows a similar current signal as with the current probe, but due to the low level of signals the specific value
is hard to obtain, but i tend to believe the current probe is showing the corrent input current.
Regards Itsu
As I believe Brad suggested, you could increase your CVR to 10R rather than 1R. This will give you 10x the signal level to measure across the CVR. Set up an RMS measurement on the CVR voltage and you can then obtain the CVR power to subtract from the total (if your circuit configuration requires it).
Quote from: TinMan on 2018.09.04, 01:02:23
Well i must say that it is odd the manually calculating the average current using the scope values,using the scope to calculate the average current value,and the DMMs average current value all result in the same amount,while using the RMS value results in near twice the value--such as Itsu has also found using his current probe RMS value.
Indeed. Using the average current would result in a significant error, wouldn't you agree?
Quote
But anyway,if i use a battery,and switch the circuit on and off by way of the SG switching a mosfet at the same frequency and duty cycle,we can then trust our DMM to give us an accurate average current value-correct?
Yes, using the average DC input current is valid in this case. There is a consideration with this setup however: The device is only using milliwatts of power so there is a chance the SG driving the MOSFET gate could skew the P(in) measurement. In the RA case we were dealing with 30W or so of power, therefore not really a concern. You could insert a hefty series gate resistance to limit the effect, but that could slow down your switching as well.
Quote
This being the case,and our scopes calculated average current value is the same as that of the DMM,and my manually calculated current value is also the same,then that would mean that my methods of calculating the P/in are correct-yes?
Not quite. I suspect you are assuming you will be correcting the 12V battery (for example) by the duty cycle, when in fact that would not be correct. The duty cycle correction already takes place during the current measurement via the DMM's averaging. Since we are already using the DMM to obtain the average current, we don't need nor want to apply the duty cycle correction twice, which is what you would be doing if you corrected the voltage also.
Assuming minimal power loss in the wiring and the MOSFET, and assuming that the source power will then equal the power into the transfomer primary, the goal is to determine the power being supplied by the source, agreed? If a CVR is in series with the battery, the loss there can easily be subtracted out. Quick review: The power from a DC source is determined by the average voltage across it times the average current through it.
So, you will be measuring voltage across the battery, not the primary of the transformer. As such, you could use 3 DMMs and turn the scope off now that both your source and your load are DC.
Quote
But how did you average the voltage in the Ainslie circuit,as it would be the supply voltage x duty cycle to get the average voltage over a full cycle. That is how the math works on a scope-is it not?,where V and I are averaged,and then multiplied together ?
As per above, you don't correct for duty cycle when measuring the battery supply voltage, the DMM current measurement already automatically does the correction on the current (which is the parameter that is pulsing) for you. Performing the duty cycle correction twice (once on the battery voltage and once on the battery current) would be really un-cool. 8)
Quote from: poynt99 on 2018.09.04, 01:56:50
Indeed. Using the average current would result in a significant error, wouldn't you agree?
Yes, using the average DC input current is valid in this case. There is a consideration with this setup however: The device is only using milliwatts of power so there is a chance the SG driving the MOSFET gate could skew the P(in) measurement. In the RA case we were dealing with 30W or so of power, therefore not really a concern. You could insert a hefty series gate resistance to limit the effect, but that could slow down your switching as well.
Not quite. I suspect you are assuming you will be correcting the 12V battery (for example) by the duty cycle, when in fact that would not be correct. The duty cycle correction already takes place during the current measurement via the DMM's averaging. Since we are already using the DMM to obtain the average current, we don't need nor want to apply the duty cycle correction twice, which is what you would be doing if you corrected the voltage also.
Assuming minimal power loss in the wiring and the MOSFET, and assuming that the source power will then equal the power into the transfomer primary, the goal is to determine the power being supplied by the source, agreed? If a CVR is in series with the battery, the loss there can easily be subtracted out. Quick review: The power from a DC source is determined by the average voltage across it times the average current through it.
So, you will be measuring voltage across the battery, not the primary of the transformer. As such, you could use 3 DMMs and turn the scope off now that both your source and your load are DC.
As per above, you don't correct for duty cycle when measuring the battery supply voltage, the DMM current measurement already automatically does the correction on the current (which is the parameter that is pulsing) for you. Performing the duty cycle correction twice would be really un-cool. 8)
Ok,i will get straight onto it when i get home from work.
Thanks for you time Poynt,great to have you here.
Hope you stick around till the end of this.
Brad
Post removed--silly question C.C
Quote from: poynt99 on 2018.09.04, 00:33:24
Sounds good Brad.
I hope those goldies are indeed low L. You could certainly measure their inductance, and compare to the 1/2W carbons.
I will indeed do that,just incase the ebay seller was telling a small fib :-\
In the mean time,i will stick with the 1/2 watt carbon resistors until i have a look at the big ones.
Ok,well im glad you suggested increasing the point's,as it actually reduces the calculated measurement values,and not increase it. So my COP just went up O0--see the difference in the two scope shots below.
Well,on with the math.
So we have the average of the math trace-804mW /10 ohms=80.4mW for total corrected P/in
RMS across 10 ohm CVR is 838mV
Power dissipated by CVR is 70.22mW
And a current value of 83.8mA
P/in for primary of transformer is there for a whopping 10.18 mW O0
Are you happy with these calculations Poynt ?
Brad
I don't know if this helps but when making true duty cycle measurements, the input diode may skew results by not allowing the SG to actively "pull down" the input.
In other words when the input "floats down" essentially disconnected from the SG by the diode, the duty cycle may not be accurate. It may act as a peak detector in a circuit with no inductance. With inductance, the primary reactance may alter input power measurements.
The reactive component cannot be absorbed by the generator as the diode blocks such action.
Of course i could be wrong about all this.
Regards
Quote from: ion on 2018.09.04, 12:21:45
I don't know if this helps but when making true duty cycle measurements, the input diode may skew results by not allowing the SG to actively "pull down" the input.
In other words when the input "floats down" essentially disconnected from the SG by the diode, the duty cycle may not be accurate. It may act as a peak detector in a circuit with no inductance. With inductance, the primary reactance may alter input power measurements.
The reactive component cannot be absorbed by the generator as the diode blocks such action.
Of course i could be wrong about all this.
Regards
Hi ION
I dont think that is quite correct.
Once the on cycle has finished(duty cycle),we then have the flyback period.
If we look at the scope shot below,we can see that there is still current flowing through the CVR during this flyback period.
If we have say a 25% duty cycle,then for 25% of the cycle the FG is the source.
During the 75% off time,the inductor/transformer then becomes the source.
We can see the voltage inverts,and current continues to flow in the same direction.
This current is now flowing through the common(0 volts or ground as we call it),through the FG,and back into the circuit.
This is why i have a problem using the scopes math to work out P/in,as it will be accounting for power that is not only being delivered by the FG,but also the power being recycled by the transformer it self.
The scope shots clearly show an inductive kickback cycle during the off time,where the voltage is inverted,and current continues to flow in the same direction-->through the CVR on the ground rail.
Brad
Quote from: TinMan on 2018.09.04, 13:37:28
Hi ION
I dont think that is quite correct.
Once the on cycle has finished(duty cycle),we then have the flyback period.
If we look at the scope shot below,we can see that there is still current flowing through the CVR during this flyback period.
If we have say a 25% duty cycle,then for 25% of the cycle the FG is the source.
During the 75% off time,the inductor/transformer then becomes the source.
We can see the voltage inverts,and current continues to flow in the same direction.
This current is now flowing through the common(0 volts or ground as we call it),through the FG,and back into the circuit.
This is why i have a problem using the scopes math to work out P/in,as it will be accounting for power that is not only being delivered by the FG,but also the power being recycled by the transformer it self.
The scope shots clearly show an inductive kickback cycle during the off time,where the voltage is inverted,and current continues to flow in the same direction-->through the CVR on the ground rail.
Brad
Yes, my thinking may not be worked out on this, although I do notice a float down on the yellow trace, although I'm not sure what the colored traces represent. Maybe they are the same as some earlier description, I will have to go back and check.
Can you try it without the diode just to be sure? But this could be dangerous to the SG if it is not internally clamped so maybe not a good idea..
Quote from: TinMan on 2018.09.04, 11:16:36
Well,on with the math.
So we have the average of the math trace-804mW /10 ohms=80.4mW for total corrected P/in
RMS across 10 ohm CVR is 838mV
Power dissipated by CVR is 70.22mW
And a current value of 83.8mA
P/in for primary of transformer is there for a whopping 10.18 mW O0
Are you happy with these calculations Poynt ?
They appear correct.
Quote from: TinMan on 2018.09.04, 13:37:28
This is why i have a problem using the scopes math to work out P/in,as it will be accounting for power that is not only being delivered by the FG,but also the power being recycled by the transformer it self.
The scope shots clearly show an inductive kickback cycle during the off time,where the voltage is inverted,and current continues to flow in the same direction-->through the CVR on the ground rail.
Brad, this is in fact one of the beautiful things about this measurement method, it accounts for any energy going back to the source.
That means that during the kickback phase the power product reverses in polarity (notice the purple trace dips below zero?), and when this is averaged with the forward power, it actually decreases the over all forward power result.
So what you seem to think is working against you, is actually working for you in this case.
Quote from: poynt99 on 2018.09.04, 14:08:49
They appear correct.
Poynt
We need to be 99% + accurate here.
Is there anywhere mistakes could have been made here?
Is there another way to calculate P/in,so as to back up the shown results using your method?.
Looking at the circuit below,is there any way power can get to R1 without passing through CVR1 ?
Brad
Brad
Maybe you have said before, but is your ground floating or real ground as shown in your diagram, if real ground then what is it connected to? apart from the scope ground.
Regards
Mike 8)
Quote from: TinMan on 2018.09.04, 15:16:19
Poynt
We need to be 99% + accurate here.
Is there anywhere mistakes could have been made here?
Is there another way to calculate P/in,so as to back up the shown results using your method?.
Looking at the circuit below,is there any way power can get to R1 without passing through CVR1 ?
Brad
I don't see any problems in the protocol, but your last question is a good one. Perhaps through capacitive coupling of the windings...
Remove your transformer and short the input leads, and output leads and treat it as a capacitor. See what your capacitor meter measures across the now two leaded component.
I created a simulation of the circuit, and just 100pF from pri-sec has a fairly significant impact on the CVR wave form.
Out of curiosity, have you tried driving this with a proper sine wave at about the same frequency?
Quote from: PhysicsProf on 2018.09.03, 23:04:28
Interesting!
PM writes: "Yesterday, I was trying to use some of the other mosfets using the newly measured capacitance change data and the results were terrible. So, I thought I would replicate my previous circuit and data in post #9 for comparison and all attempts throughout the day to replicate the performance failed with all circuit values and measurement procedures identical. The COPs were ~.8-.9 at best. The weather here was overcast and rainy off and on all day during these test attempts!
This morning, all the weather has moved on and we have a bright sunny day at the moment and when running the replication test at this moment in time, the COPs with gain have returned. Any suggestions as to why this is happening? "
Air humidity is clearly suspect as the culprit. Possibly in your lab, but possibly humidity in the air outside...
WHY humidity might matter so much, is another question.
I would say it is very valuable to be able to turn the effect "off", and you have evidently found one way to do this, with increased air humidity.
Today is another very wet and rainy day and the COPs have again dropped! For comparison, the two data tables below are of the same circuit tested in the same manner but the first was taken on yesterday 9-3-18 (sunny) and the second taken today 8-4-18 (rainy). Obviously the permittivity of the outside air is affected by the humidity but locally at my bench, the relative humidity couldn't be changing by much due to being enclosed and air conditioned.
So, is the apparent energy gain due do an aetheric source or...?
Regards,
Pm
PM, Is it a temperature related thing?
I would suggest running a calibration on your scope before taking your measurements, and then compare.
It is suspicious to me that in one case the P(in) are all over the map, both pos and neg, while the next time they are all pos. Suggests to me a change in the offset in the scope somewhere.
PM I noticed in an earlier post you were using a plug board to build on, these are well known to induce pf capacitance onto the traces, which would very likely be humidity responsive.
Although if air conditioned then maybe not, maybe affecting your house ground rod if it has one.
In regards to testing another way, we could try the following as per the pic. This places the input CVR before the primary, which is ideally where it should be located. Without a current probe, it is a bit of a challenge, especially if we want to do further MATH on it.
If you can somehow multiply the MATH trace CH1-CH2 across the CVR x the CH3 probe, then this would be another way to measure the power input. I don't know if you can set up a MATH trace that already involves another MATH trace however.
Quote from: partzman on 2018.09.03, 22:31:51
Itsu,
Well, the best way to tell is to do an experiment and although I'm using my Tek MDO, your TDS should measure in the same manner.
The test setup is a signal generator set at 500kHz sine wave driving a 100 ohm 1% non-inductive resistor with the current being measured with a Tek TCP0020 current probe. I could have used a CSR and the results would have been more accurate on the second test.
The first scope pix shows the voltage and current waveforms with very little phase difference and the resulting mean Math channel calculation. So, we multiply the rms values and get Pin = 4.415*42.47e-3 = 187.5mw. neglecting the small phase angle. This compares favorably with the Math channel value of 187.3w.
In the next scope pix, we've added a ~100uh inductor to create some phase shift to move the power factor off unity so we'll be forced to use cosine correction. Therefore now the Pin = 6.118*16.64e-3*cosine(72.48) = 30.65mw. The Math shows 28.17mw for an error of 9% which I attribute to the fact that we are in the minimum current range of 10ma for the TCP0020 probe rather than the internal math calculations of the Tek scope.
Rather than repeat here, one can look at my post #88 for an explanation of how the "mean" or average measurements are taken on complex waveforms with the current Tek scopes.
Regards,
Pm
PM,
thanks for the setup, i used this to, so this confirms that the x (multiply) function of the math is using
the rms values of the both channels used.
Now see what the + (add) and - (subtract) functions do with 2 signals, what values do they use?
Itsu
Quote from: poynt99 on 2018.09.04, 00:16:04
Hi Itsu.
What indications are you seeing that leads you to conclude the scope is performing any conversion (MEAN or RMS) on the v(t) x i(t) components?
When you set up the MATH channel/trace to perform CH1 x CH2 (for example), there is no processing before or after the result. You can see the result in Brad's scope shots (the purple trace), and it is just the raw output or result of the v(t) x i(t) process. When you set up a measurement on a channel/trace, then processing is performed and the result is indicated in the box labeled AVG or MEAN, or RMS etc.
In the case of the MATH trace set up for measuring power, the raw product is processed with the MEAN measurement function in order to take the instantaneous power trace and convert it to the average power.
Poynt99,
see post #146 from PM above, thats how i did it too, you can see that the math trace with x is using the rms values.
Its different when using the + or - functions, they seem to use the mean values, see screenshot below
where i subtract CH1 (yellow) from CH3 (purple) 3.28V - 1.65V = 1.63V like the red math trace shows.
If i would of have used the rms values of CH1 and CH3 to subtract them it would not match the red math trace value.
Itsu
Itsu,
Did you miss my two posts to you? I think there is some confusion here regarding what the scope is doing when we utilize a AVG(CH1 * CH2) type measurement.
The scope is NOT using RMS nor any other process on the raw data being sampled on CH1 and CH2 in a "multiply" or any other MATH function trace.
What PM is saying is that the sampled AVG(CH1 * CH2) operation we set up for measuring average power is equivalent to multiplying the RMS values of the voltage and current, if there is no significant phase shift between the two. If there is significant phase shift between the voltage and current, then we can not simply multiply the two RMS values, unless we also incorporate the phase angle in the calculation. But since we are often dealing with non-sinusoidal wave forms, we can not do that, so we let the scope do all the work in determining the average power by setting up the CH1 * CH2 MATH trace. The result is RAW data, i.e. AVG[(sample1 vi product) + (sample2 vi product) + (sample3 vi product) +...]. The scope is sampling at a very high rate, as in 500Ms/s, and the average is a running average of these samples, because we apply a MEAN measurement to the samples.
So the samples are the RAW products of v(t) x i(t), nothing more.
I hope that helps.
Quote from: Itsu on 2018.09.04, 19:47:11
PM,
thanks for the setup, i used this to, so this confirms that the x (multiply) function of the math is using
the rms values of the both channels used.
Now see what the + (add) and - (subtract) functions do with 2 signals, what values do they use?
Itsu
Itsu,
Anytime you use the Math channel for a calculation let's say the "mean" or average, it uses the same method for all four math functions. The instantaneous values at each sample along the horizontal time window of the waveforms you've selected are added or subtracted or multiplied or divided on at that moment in time and the result is stored in the scopes memory. Then, depending on the cursor measurement mode you have selected say "between the vertical cursors", the scope knows that you wish to compute with the samples taken only between those cursors. It then "integrates" or "sums" all those samples together to get one sum and then divides by the number of samples it took between your cursors. This is of course for the "mean" measurement. Long answer for your question but I hope it helps.
Pm
Quote from: poynt99 on 2018.09.04, 01:11:51
QuoteQuoteQuote from: Itsu on 2018-09-02, 20:25:15
I found another ferrit cored transformer:
prim. 6mH / 2 Ohm
sec. 2.5H / 30 Ohm
Using the 10V DC 20% duty cycle 28.8KHz signal without input diode.
I have a DC mA DMM in the red plus lead showing the (37.2mA rms? mean?) current going in, see picture.
(a 2th Fluke DMM across a temporary installed 1 Ohm csr shows 37mV).
Both of these will be AVG current.
OK, good to know.
QuoteQuoteThe screenshot shows;
yellow: the voltage across the primary (still large spikes) 911mV rms
green: the current from my current probe in the red FG supply lead 105mA rms
red: the math function yellow x green presenting the average input power 20.8mW
So we see 20.8mW input.
A 3th fluke DMM shows the voltage across the 14860 Ohm resistor to be 11.2V
This represents a power out of 8.7mW indicating a cop = 0.42.
Did you subtract the CVR power from the 20.8mW? To obtain the CVR power, you need to know the RMS voltage across it.
I did not use a cvr in that setup, so nothing to subtract. (i used momentarely a 1 Ohm cvr with a DC voltage DMM
across it to backup the current reading on the DC mA DMM).
QuoteQuoteStrange is the difference between the input current DMM (37.2mA) backed up by the temp 37mV across a 1 Ohm resistor
and the current probe input current measurement of 105mA rms.
It isn't strange, because the DMM measures average current.
Right, it was strange to me as i thought that the DMM's would also show rms values, their not, thanks.
QuoteQuoteAre the DDM's unable to accuratly display the 28.8KHz current/voltage? It looks like it.
The DMM displays the average very accurately. In general the only time we use the DMM for input power measurement is when the source is a fixed DC supply, and in that case it would be to measure the average current. I(avg) x Vdc = P(in). Since we are not using a fixed DC supply in this device, we don't use the DMM.
QuoteQuoteUsing another methode of obtaining the input current (2 probes in differential setup across the temp. 1 Ohm csr)
shows a similar current signal as with the current probe, but due to the low level of signals the specific value
is hard to obtain, but i tend to believe the current probe is showing the corrent input current.
Regards Itsu
As I believe Brad suggested, you could increase your CVR to 10R rather than 1R. This will give you 10x the signal level to measure across the CVR. Set up an RMS measurement on the CVR voltage and you can then obtain the CVR power to subtract from the total (if your circuit configuration requires it).
I did that in my latest setup showing a COP =0.81
Itsu
Quote from: poynt99 on 2018.09.04, 20:06:55
Itsu,
Did you miss my two posts to you? I think there is some confusion here regarding what the scope is doing when we utilize a AVG(CH1 * CH2) type measurement.
The scope is NOT using RMS nor any other process on the raw data being sampled on CH1 and CH2 in a "multiply" or any other MATH function trace.
What PM is saying is that the sampled AVG(CH1 * CH2) operation we set up for measuring average power is equivalent to multiplying the RMS values of the voltage and current, if there is no significant phase shift between the two. If there is significant phase shift between the voltage and current, then we can not simply multiply the two RMS values, unless we also incorporate the phase angle in the calculation. But since we are often dealing with non-sinusoidal wave forms, we can not do that, so we let the scope do all the work in determining the average power by setting up the CH1 * CH2 MATH trace. The result is RAW data, i.e. AVG[(sample1 vi product) + (sample2 vi product) + (sample3 vi product) +...]. The scope is sampling at a very high rate, as in 500Ms/s, and the average is a running average of these samples, because we apply a MEAN measurement to the samples.
So the samples are the RAW products of v(t) x i(t), nothing more.
I hope that helps.
Yes I agree. It seems there could be a fruitful discussion somewhere just on the use of the Math channel as all the newer scopes seem to include this feature. It's really cool when they allow you to create your own math expression! :)
Pm
Quote from: poynt99 on 2018.09.04, 18:46:01
PM, Is it a temperature related thing?
I would suggest running a calibration on your scope before taking your measurements, and then compare.
It is suspicious to me that in one case the P(in) are all over the map, both pos and neg, while the next time they are all pos. Suggests to me a change in the offset in the scope somewhere.
I normally recal the scope after a warmup period and the TCP0020 current has a degauss/auto zero that I initiate before doing a scan or sweep. This doesn't mean there isn't some kind of offset or other equipment problem but those variations seem to be due to the outside weather conditions. The first time I experienced this, I fought measurements all day that were consistent but different from the previous day. The difference was the weather. When it clears off, I'll run the same tests again for comparison.
Pm
Quote from: Itsu on 2018.09.04, 20:13:09
Right, it was strange to me as i thought that the DMM's would also show rms values, their not, thanks.
A DMM can indicate accurate RMS values under 2 conditions only:
1) for standard DMM's, when the wave form is pure, or close to pure sine wave, and it is below say 10kHz.
2) when the DMM is a "True RMS" meter (can be expensive).
Quote from: Peterae on 2018.09.04, 19:22:54
PM I noticed in an earlier post you were using a plug board to build on, these are well known to induce pf capacitance onto the traces, which would very likely be humidity responsive.
Although if air conditioned then maybe not, maybe affecting your house ground rod if it has one.
Yes, I have used the plug boards for certain circuits but not in this case as everything is soldered and wired. I have given the inside humidity some thot and I presently don't have a meter in my lab to measure it but I guess I need to do that. At the moment it still remains a mystery!
Pm
Poynt99, PM,
thanks for the replies, i was trying to get a handle on what signals i see and what the math function is doing with
them to get the result.
Seems its more complicated then i expected.
Guess it needs some further study from me, so lets leave it that so we don't further clutter up this thread.
Itsu
Quote from: poynt99 on 2018.09.04, 20:25:07
A DMM can indicate accurate RMS values under 2 conditions only:
1) for standard DMM's, when the wave form is pure, or close to pure sine wave, and it is below say 10kHz.
2) when the DMM is a "True RMS" meter (can be expensive).
OK, i have a Fluke 8060A/AA which says TRUE RMS MULTIMETER, so i will use that in the supply lead to see how it compares to the cheaper DMM's i use now.
Itsu
Ah the 8060A, an oldie but a goodie. Nice, and good for you Itsu!
- RMS measurements to 100kHz
- Crest Factor 3:1 (peak to average ratio capability, determines ability to accurately handle spikey signals.)
OK, this is a continuous running loaded test of the circuit that produced the 9418A data table in post #165. The schematic can be seen in post #9 except the frequency is 710,611Hz and C2 is 2108pf.
The first scope pix shows the power input to be 6.664mw. Make note of the mean current offset of +150.4ua which would typically influence the average power input to a higher level than would be seen with a perfect 0.00ua offset. The yellow trace is the input pulse to L1, the green trace is the input current, the red trace is the Math, and the blue and pink traces are the differential voltages across C1 and the load. The current probe has a 4 turn loop for amplification but the correction factor is 3.883 in the math calculation which is correct according to the profiling of the probe with said loop at this frequency.
The second pix shows the mean output of 18.87 volts across C1 with a 32.1k carbon film resistor as a load. From this we calculate the output power = 18.87^2/32.1e-3 = 11.1mw. This is flea power but real IMO.
Therefore, the apparent COP = 11.1/6.664 = 1.67 which is a little higher than I anticipated according to the 9418A data table.
Again, using the delta energy taken from the data table say between 40ms-140ms, the energy differential of C1 is (33.51^2-17.06^2)*3.94e-6/2 = 1639uJ. This equates to a potential power level of 1639e-6/100e-3 = 16.39mw. If we desire this power to be produced at the voltage level at the peak gain point in the table at 80ms, the voltage would be ~24v. Therefore, the load resistor should be Rl = E2/P = 24^2/16.39e-3 = 35.14k. I used a 32.1k which lowered the output but raised the COP.
I will use this test as a benchmark for the humid weather and compare it to an identical test run with a dryer set of conditions in the upcoming days.
Pm
Quote from: Centraflow on 2018.09.04, 15:26:04
Brad
Maybe you have said before, but is your ground floating or real ground as shown in your diagram, if real ground then what is it connected to? apart from the scope ground.
Regards
Mike 8)
Hi Mike
The ground represents the neutral of our power grid,which has the same potential as earth ground,but isolated from it.
Brad
Quote from: poynt99 on 2018.09.04, 18:33:56
I don't see any problems in the protocol, but your last question is a good one. Perhaps through capacitive coupling of the windings...
Remove your transformer and short the input leads, and output leads and treat it as a capacitor. See what your capacitor meter measures across the now two leaded component.
I created a simulation of the circuit, and just 100pF from pri-sec has a fairly significant impact on the CVR wave form.
Out of curiosity, have you tried driving this with a proper sine wave at about the same frequency?
I have not tried AC input yet,but will do as soon as i work out what is wrong here.
Tried to measure capacitance the way you said,and cant seem to get a stable reading on the meter.
The battery symbol is not flashing on the DMM,but i will get some new batteries tomorrow,and see if that helps.
At the moment,we are looking for a 192.3% mistake in the power measurements.
Putting aside what i would like it to be,i must stick to what we !believe! to be real,and find out where this measurement error is.
So far i have tried 11 different transformers-all with the E type ferrite core,but only two out of the eleven give these COP>+ results,even though the waveforms all look much the same.
There is nothing special about this circuit. It is just a pulsed transformer,where the secondary go's through a FWBR,and charges a cap. The only thing added is C2--the small 1.4nF cap that is in series with ground and one AC leg of the FWBR.
Tonight i did a self cal on the scope,and nothing changed in way of power measurements--so the scope seems fine.
One question Poynt
Is there a difference between mean and average on the scope,as this new fangdangled 4 channel scope dose not have mean-only average,while my old 2 channel Atten has both.
Too many times in the past,when i though there was really something there,it bit me on the backside,as it always turned out to be something missed that was staring me right in the eyes all along. So this time im looking for a mistake-->there must be a mistake somewhere--right ?.
I have about 8 or 9 of the second transformer i found that showed the !!apparent!! OU results,so ill put another circuit together tomorrow night,and see if we get matching result's,or different results from the same type of transformer.
Anyway,need sleep.
Will report back tomorrow night.
Brad
All,
This is the variation of the original PC circuit which takes advantage of several things, one of which is there is only two parametric elements, another is the fact that C1 is now ground referenced making current measurements simpler, and yet another is that the fixed series capacitance is integrated into the bifilar windings distributed capacitance, and last but not least is the absence of any magnetization current that would be present with an additional primary winding.
Although the circuit appears to be simple, the operation is not. Referring to the schematic below, if the 14N05Ls were replaced with a non-parametric diode, C1 would charge to 1/2 the power supply minus one diode drop with a 50% duty cycle. So, if a voltage across C1 is greater than 1/2 the supply voltage, it is a result of the parametric pumping action.
Also what may not be obvious is the action from driving the center tap of the bifilar windings. Since the are bifilar, they are tightly coupled magnetically and electrically. As a result, when the end of P1 is conducting thru the substrate diode to the 14N05L in a somewhat normal fashion, the other end of P2 is driving the parametric capacitance of the other 14N05L with an equal magnitude but opposite polarity voltage.
The first scope pix shows a test made with this circuit using a 4x loop and a current probe that compares the current taken at the input with CHR1(wht) to the current taken at the ground side of C1 with CH4(grn). From this it is determined there is no significant difference so a 10 ohm 1% non-inductive film resistor will be used for the CSR.
The last scope pix is the result of the test taken at 764kHz. With a load resistor of 32.1k ohm, the output power is 17.2mw resulting in an apparent COP = 17.2/6.673 = 2.58 .
Pm
Quote from: TinMan on 2018.09.05, 13:32:47
I have not tried AC input yet,but will do as soon as i work out what is wrong here.
Tried to measure capacitance the way you said,and cant seem to get a stable reading on the meter.
The battery symbol is not flashing on the DMM,but i will get some new batteries tomorrow,and see if that helps.
Yeah, I suspected the measurement might be wonky.
Quote
One question Poynt
Is there a difference between mean and average on the scope,as this new fangdangled 4 channel scope dose not have mean-only average,while my old 2 channel Atten has both.
MEAN and AVERAGE are exactly the same thing. Different scope manufacturers use one or the other. No problem, the RIGOL uses AVERAGE.
Quote
Too many times in the past,when i though there was really something there,it bit me on the backside,as it always turned out to be something missed that was staring me right in the eyes all along. So this time im looking for a mistake-->there must be a mistake somewhere--right ?.
I have about 8 or 9 of the second transformer i found that showed the !!apparent!! OU results,so ill put another circuit together tomorrow night,and see if we get matching result's,or different results from the same type of transformer.
Anyway,need sleep.
Will report back tomorrow night.
Brad
I like your attitude Brad, optimistic, yet skeptical and cautious when things don't appear quite as expected. O0
Quote from: TinMan on 2018.09.05, 13:32:47
[snip]
I have about 8 or 9 of the second transformer i found that showed the !!apparent!! OU results,so ill put another circuit together tomorrow night,and see if we get matching result's,or different results from the same type of transformer.
[snip]
Brad
Brad,
What would be interesting IMO would be to take one of those transformers that don't seem to work that are like those that do, and take as many measurements as possible for comparison and then carefully disassemble it so we can see the construction of the windings, etc. No need to do this at the moment but whenever time permits somewhere down the road!
Pm
Quote from: partzman on 2018.09.05, 17:19:05
All,
This is the variation of the original PC circuit which takes advantage of several things, one of which is there is only two parametric elements, another is the fact that C1 is now ground referenced making current measurements simpler, and yet another is that the fixed series capacitance is integrated into the bifilar windings distributed capacitance, and last but not least is the absence of any magnetization current that would be present with an additional primary winding.
Although the circuit appears to be simple, the operation is not. Referring to the schematic below, if the 14N05Ls were replaced with a non-parametric diode, C1 would charge to 1/2 the power supply minus one diode drop with a 50% duty cycle. So, if a voltage across C1 is greater than 1/2 the supply voltage, it is a result of the parametric pumping action.
Also what may not be obvious is the action from driving the center tap of the bifilar windings. Since the are bifilar, they are tightly coupled magnetically and electrically. As a result, when the end of P1 is conducting thru the substrate diode to the 14N05L in a somewhat normal fashion, the other end of P2 is driving the parametric capacitance of the other 14N05L with an equal magnitude but opposite polarity voltage.
The first scope pix shows a test made with this circuit using a 4x loop and a current probe that compares the current taken at the input with CHR1(wht) to the current taken at the ground side of C1 with CH4(grn). From this it is determined there is no significant difference so a 10 ohm 1% non-inductive film resistor will be used for the CSR.
The last scope pix is the result of the test taken at 764kHz. With a load resistor of 32.1k ohm, the output power is 17.2mw resulting in an apparent COP = 17.2/6.673 = 2.58 .
Pm
PM,
looking promising, using my trifilar toroid (1 winding unused) 2nF each between the used windings,
5.5uF C1 cap, 32.8K load resistor. 330uH L1 coil and 2x IRF3205 MOSFETs.
Screenshot shows input voltage from IXDN614PI at 40V driven at 764KHz in yellow,
current probe in SINGLE lead between MOSFET driver and L1 coil in green,
math trace in red (yellow x green) = 12.4mW (fluctuating a lot), and voltage across
the 32.8K load resistor and 10 Ohm csr in blue (19.9V).
So input calculated to be 12.4mW, output is 19.9V across 32.8K = 11.95mW for a COP= 0.96
Still wet and rainy overhere.
Itsu
Quote from: Itsu on 2018.09.05, 20:41:04
PM,
looking promising, using my trifilar toroid (1 winding unused) 2nF each between the used windings,
5.5uF C1 cap, 32.8K load resistor. 330uH L1 coil and 2x IRF3205 MOSFETs.
Screenshot shows input voltage from IXDN614PI at 40V driven at 764KHz in yellow,
current probe in SINGLE lead between MOSFET driver and L1 coil in green,
math trace in red (yellow x green) = 12.4mW (fluctuating a lot), and voltage across
the 32.8K load resistor and 10 Ohm csr in blue (19.9V).
So input calculated to be 12.4mW, output is 19.9V across 32.8K = 11.95mW for a COP= 0.96
Still wet and rainy overhere.
Itsu
Itsu,
I have discovered a problem with the circuit in that it is not operating as it should. You will notice that if you lower the frequency to a point, the circuit goes unstable. It has hysteresis so you will have to raise the frequency considerably to exit that unstable mode. This is not the only problem however so I'm working on solutions and will post as soon as I find a solution.
Regards,
Pm
Quote from: PartzMan
...if you lower the frequency to a point, the circuit goes unstable.
Are you able to identify the frequency at which the instability
begins? What do you suspect may be causing this?
Quote from: partzman on 2018.09.05, 18:53:17
Brad,
What would be interesting IMO would be to take one of those transformers that don't seem to work that are like those that do, and take as many measurements as possible for comparison and then carefully disassemble it so we can see the construction of the windings, etc. No need to do this at the moment but whenever time permits somewhere down the road!
Pm
Hi Pm
I have already tried to take apart some of those transformers,but there all glued together,and dam near impossiable to get apart without breaking the core.
Actually,i should have a look and see if the two i have that show this OU effect are glued together. Graham actually mentioned something along these lines before.
Brad
Quote from: muDped on 2018.09.06, 00:59:37
Are you able to identify the frequency at which the instability
begins? What do you suspect may be causing this?
The frequency is variable depending on the supply voltage and parametric device(s) used. To better describe the instability, one can carefully reduce the frequency and as a result, the voltage across the loaded C1 will increase accordingly but at some critical point, the output across C1 with load will suddenly rise to it's maximum which is dependent on many factors. This is due to the negative delta "C" or decrease in capacitance of the parametric device due to the increasing voltage across it. The device is always resonating as some frequency but the device becomes a highly non-linear resonant oscillator at the critical point and rises in frequency to"lock" into it's highest frequency of oscillation depending on circuit parameters and voltage. It is not very efficient in this mode much less OU.
I don't know if many have noticed but in my profiling of the mosfets I included a profile of a 24mm piezo element which has a positive delta "C". Solar cells also exhibit a positive delta "C" and I plan to test both of these in the near future. Wouldn't it be wonderful if a solar panel could produce energy in the dark!
Again depending on circuit values and critical adjustment, I have seen a device have bursts in and out of this critical mode and if captured, the steady state portion below the critical point is highly OU but uncontrollable at this point.
Hope this helps.
Pm
Edit: As I re-read post I really didn't communicate this very well. The basic working principle is that the device has a self resonant frequency (SRF) frequency based on the series inductance and the series parametric capacitance. As the voltage rises across C1, the parametric capacitance rises and creates a higher self resonant frequency. If this SRF rises above the generator frequency, we have instability.
Quote from: Itsu on 2018.09.05, 20:41:04
PM,
looking promising, using my trifilar toroid (1 winding unused) 2nF each between the used windings,
5.5uF C1 cap, 32.8K load resistor. 330uH L1 coil and 2x IRF3205 MOSFETs.
Screenshot shows input voltage from IXDN614PI at 40V driven at 764KHz in yellow,
current probe in SINGLE lead between MOSFET driver and L1 coil in green,
math trace in red (yellow x green) = 12.4mW (fluctuating a lot), and voltage across
the 32.8K load resistor and 10 Ohm csr in blue (19.9V).
So input calculated to be 12.4mW, output is 19.9V across 32.8K = 11.95mW for a COP= 0.96
Still wet and rainy overhere.
Itsu
Itsu,
Keep the circuit intact and try it again when you have dryer conditions. I'll be curious to see your results at that time.
Pm
PM,
ok, will do, will be somewhere around next summer ;D
I see what you mean by unstable, lowering from 764 to about 450Khz the current starts to increase in bursts
and around 250KHz its stable high current again, but at terrible COP.
Will play around some more.....
Itsu
Itsu,
Here is an optional circuit if you wish to try it. This will not be as touchy around the unstable point so should be able to see positive results even in your humidity.
The schematic is below. Remove T1 and leave your two IRF3205s in parallel and drive directly with your 330uH L1. Split your power supply so you also have +20v dc and connect the end of C1 and the load as shown if you wish to use your current probe. Make sure your pulse generator is set at 50% duty cycle for this as what we are doing is eliminating any dc bias on the load and/or any coupling cap.
If you use your current probe, use your Math channel to multiply the output of your mosfet driver times the current waveform and have the Math measure in "mean". For the output measurement, I prefer to do a differential measurement across the load resistor but you could use just the "mean" measurement of CH2 for example, then subtract 20v dc from that reading and do the calculation manually with the difference.
If you wish to use a CSR, then connect as shown. In your case, you will need to do a differential measurement across the CSR and save in a Rx or reference channel. You can then do a separate Math multiplication with this measurement and your driver output to get the power input.
In either case, have the "mean" measurement of current displayed so you have the offset current measured with the probe or CSR as this value must be multiplied times the 20v dc supply for a power corrention to the input. If the value is positive, you are actually supplying power to the +20v and if negative, you are drawing power from the +20v. The idea here is to get that value as low as possible.
To tune the device, start at say 600kHz and lower the frequency by 10kHz steps until you see the output voltage across C1 rise reasonably quick to it's maximum. It should be a rather 'soft' instability that you will see. Whatever frequency this happens at, you want to run slightly higher to do the measurements. There won't be as much hysteresis but you will need to raise the frequency some to get out of that unstable mode. Repeat this until you have as high a voltage on C1 as you can without the instability and then capture the screen and do your tests.
I've attached scope pix of my own tests with this circuit using IRF740s which appear to be close to the IRF3205s but you should run at a slightly lower frequency than the 560kHz I used. Note that the offset current is consuming ~2mw from the 20v dc supply.
Pm
Edit: I haven't tried this but the 20v dc supply could be replaced two electrolytics of equal value between the 40v supply and ground. They should self adjust the offset current to zero. The 'lytics should have a value of least 10x the value of C1.
PM.
i rebuilded my circuit as to your last post.
Starting at 600KHz i get the screenshot below.
Yellow / green are input voltage / current (probe)
Red is math yellow x green for input power which is negative.
blue and purple are across the 32.8K resistor, a Fluke DDM reads 0.325V.
So my initial problem is that the input power calculations show negative values.
Itsu
Quote from: Itsu on 2018.09.06, 15:54:12
PM.
i rebuilded my circuit as to your last post.
Starting at 600KHz i get the screenshot below.
Yellow / green are input voltage / current (probe)
Red is math yellow x green for input power which is negative.
blue and purple are across the 32.8K resistor, a Fluke DDM reads 0.325V.
So my initial problem is that the input power calculations show negative values.
Itsu
Itsu,
And the problem is... :)
Seriously, this is possible with this circuit but as you can see, the voltage and current waveforms are close to 90 degrees from one another so it now takes very careful means of taking the measurements to be sure what the input power is. You should have a higher voltage across C1 so I would reduce the frequency and increase the this voltage which will increase the input power and hopefully show an overall gain for you.
Pm
Itsu,
Here is another test set of the same circuit. I discovered that I had not taken the "divide by 3.88" off my math equation in the first test which would incorrectly reduce the input power by that amount so this is a re-run with the 4x loop in place. I am using the split power supply in this setup.
The input is 3.508mw with an offset of -36.63ua which results in .73mw drawn from the 20v supply for a total pin = 3.515mw. The output voltage across C1 is 25.77v with a 32.1k resistor which results in a pout = 20.7mw for an apparent COP = 20.7/3.515 = 5.89.
Pm
Edit: The clipping seen on Ch4 is from very high speed transients that do not materially affect the overall current readings.
Quote from: partzman on 2018.09.06, 17:35:09
Itsu,
And the problem is... :)
Seriously, this is possible with this circuit but as you can see, the voltage and current waveforms are close to 90 degrees from one another so it now takes very careful means of taking the measurements to be sure what the input power is. You should have a higher voltage across C1 so I would reduce the frequency and increase the this voltage which will increase the input power and hopefully show an overall gain for you.
Pm
Well, negative input should not happen i guess.
Anyway, from 600KHz i went down in 10KHz steps and at 430KHz there is a sudden change, the voltage across
the load resistor jumps up to 14V, and the input power drops to -76mW, see screenshot 1.
So -76mW in, 6mW out.
When dropping frequency even more, a 2th jump happens at 370Khz where the voltage across the load resistor
jumps to 31.6V, and the input power is 331mW, see screenshot 2.
So 331mW in, 30mW out.
Finally continuing on down, max voltage across the load resistor is at 240Khz at 35.6V
But the input is 1.15W, see screenshot 3.
So 1.15W in, 39mW out.
By the way, my IXDN614 gets rather hot after some time.
Using 2PS's, one at 40V, one at 20V, grounds connected together.
Itsu
When going UP in 10KHz steps from the 1st jump (370Khz), i can go up to 530Khz with increasing voltage from 14 to 19V
across the load resistor (next step drops the voltage to 2V or so) with an input of 14,2mW, see screenshot 1
So 14.2mW in, 11mW out (this means a COP = 0.77)
But there seems to be a ripple (14Hz) on the blue trace which if i reduce the time setting looks like screenshot 2.
Itsu
Quote from: Itsu on 2018.09.06, 19:07:15
Well, negative input should not happen i guess.
Anyway, from 600KHz i went down in 10KHz steps and at 430KHz there is a sudden change, the voltage across
the load resistor jumps up to 14V, and the input power drops to -76mW, see screenshot 1.
So -76mW in, 6mW out.
When dropping frequency even more, a 2th jump happens at 370Khz where the voltage across the load resistor
jumps to 31.6V, and the input power is 331mW, see screenshot 2.
So 331mW in, 30mW out.
Finally continuing on down, max voltage across the load resistor is at 240Khz at 35.6V
But the input is 1.15W, see screenshot 3.
So 1.15W in, 39mW out.
By the way, my IXDN614 gets rather hot after some time.
Using 2PS's, one at 40V, one at 20V, grounds connected together.
Itsu
Itsu,
OK, in your first scope pix or when you reach 430kHz, you have gone into a first unstable state. You can tell this by looking at the current waveform symmetry. Look carefully at the slope and magnitude of each current cycle and also notice the difference in the Math trace. Also change your "CH4 rms" measurement to "CH4 mean" as it is more meaningful at this point to know the average offset cuurent. You will want to raise the frequency until you come out of that state and lower back down to say 420kHz or just prior to being unstable and take measurements.
In your second scope pix, you have entered a second unstable mode which is OK but do the same thing as above- Raise the frequency until you are out of that mode and then lower to say to 360kHz or whatever frequency gets you as close as possible without "tripping" the circuit and then do your measurements. I think this region will give you positive results from what I can tell but I'm not sure.
Everything else at 350kHz and below is essentially out of the optimum operational area for the device so that data can be discounted.
Don't be too surprised if you see negative inputs as we are dealing with highly non-linear functions in the circuits and therefore it may be possible IMO.
Pm
Itsu,
I should have added that the Ixys drivers do not seem to like seeing negative current when the output pulse is high or near the supply voltage. Their output device(s) do not handle reverse conduction so they will get hotter that normal under these conditions and if left on too long, they will short!
Pm
PM,
ok, i can't say i am with you right now, but this is what i did:
Green trace set to mean
From 600Khz lowered frequency in 10KHz steps until unstable state (430KHz).
Then in 10KHz steps up in frequency to get out of unstable state (540KHz).
Then in steps back to just before unstable state (440Khz), taking screenshot 1.
DDM across load resistor reads 2.4V = 0.175mW.
I cannot get out of 2th unstable mode as it goes right through the 1st unstable mode (to arrive at 540Khz again)
Itsu
Quote from: Itsu on 2018.09.06, 20:27:37
PM,
ok, i can't say i am with you right now, but this is what i did:
Green trace set to mean
From 600Khz lowered frequency in 10KHz steps until unstable state (430KHz).
Then in 10KHz steps up in frequency to get out of unstable state (540KHz).
Then in steps back to just before unstable state (440Khz), taking screenshot 1.
DDM across load resistor reads 2.4V = 0.175mW.
I cannot get out of 2th unstable mode as it goes right through the 1st unstable mode (to arrive at 540Khz again)
Itsu
Itsu,
I think your negative input measurement is due to your -3.05ma mean current. If you are using an amplifier with your probe, try to de-gauss or zero the amplifier just before taking any measurements. You may not be able to get any closer to zero than this but zero would be the ideal value so the closer the better. A positive offset would more than likely take the input into the positive range but you don't have enough output voltage across C1 to really matter at this point.
Maybe it would help the instability if you increased L1 to say something close to 500uh-1mh. This will lower the the frequency before hitting instability and may allow a higher C1 voltage range with your mosfets. I don't have anything else to suggest at this point but if I think of something I'll let you know.
Pm
Quote from: PartzMan
...the Ixys drivers do not seem to like seeing negative current
when the output pulse is high or near the supply voltage.
Their output device(s) do not handle reverse conduction so
they will get hotter that normal under these conditions and if
left on too long, they will short!
Excellent observation.
MosFet Driver Chips are able to deliver an impressive high
current impulse into their normal load of Gate Capacitance
but cannot handle that high current for prolonged periods of
'on-time' without overheating or destruction. They are rather
fragile.
Perhaps the driver for the circuit needs beefing up?
Quote from: partzman on 2018.09.06, 22:07:06
Itsu,
I think your negative input measurement is due to your -3.05ma mean current. If you are using an amplifier with your probe, try to de-gauss or zero the amplifier just before taking any measurements. You may not be able to get any closer to zero than this but zero would be the ideal value so the closer the better. A positive offset would more than likely take the input into the positive range but you don't have enough output voltage across C1 to really matter at this point.
Maybe it would help the instability if you increased L1 to say something close to 500uh-1mh. This will lower the the frequency before hitting instability and may allow a higher C1 voltage range with your mosfets. I don't have anything else to suggest at this point but if I think of something I'll let you know.
Pm
PM,
i use the AM 503B as amplifier for the probe, let it warm up at least one hour and normally i degauss before
every measurement.
I will try to increase L1 and/or use another current measurement method (differential or cvr).
Thanks, itsu
Quote from: muDped on 2018.09.07, 02:29:15
Excellent observation.
MosFet Driver Chips are able to deliver an impressive high
current impulse into their normal load of Gate Capacitance
but cannot handle that high current for prolonged periods of
'on-time' without overheating or destruction. They are rather
fragile.
Perhaps the driver for the circuit needs beefing up?
Yes I think so too. I was hoping to achieve a lower range of frequencies showing apparent OU which would open up more possibilities to do this, but no success thus far.
Pm
Changed L1 from 330uH to 1mH.
Unstable mode happens now at 230Khz, so stopped at 240Khz.
Input current / power seems stable positive now, around 7mW
Power across the load resistor 2.61V (DMM) for a Pout of 0.2mW
Weather still humid allthough dry.
Itsu
Quote from: Itsu on 2018.09.07, 14:29:51
Changed L1 from 330uH to 1mH.
Unstable mode happens now at 230Khz, so stopped at 240Khz.
Input current / power seems stable positive now, around 7mW
Power across the load resistor 2.61V (DMM) for a Pout of 0.2mW
Weather still humid allthough dry.
Itsu
Itsu,
The current offset is much better now but the voltage across C1 is really low. I think the problem is (if the device works) that since we do not the have a sweep of your circuit or a profile of your mosfets, your load resistance is not ideal and possibly L1. You may also be affected by the weather but this circuit seemed to overcome that issue on my end as we still have rain and overcast with 95-100% humidity.
I need to spend more time trying to do a network analysis on one of these circuits so it would be a little easier to calculate component values for a given mosfet family. I'm also working on other topologies that hopefully may yield a more replicable circuit.
Pm
No problem, i have some RFD14N05L's on order, so will change to these MOSFETs when arrived.
Meanwhile i can change the load resistor to see if some other value will increase COP.
Itsu
Itsu,
Just an update. It is a sunny day in the morning here and while being very wet outside with dew all over everything, the OU has returned in the 14N05L test setup. It seems to be related to the sunshine or perhaps cloud cover. I thought I had run tests after dark but not sure now so will test for that later today.
Pm
PM,
thanks for the update, i cannot imagine that the OU noted could have anything to do with the type of weather.
But time (and further tests) will tell.
I lost my last IXDD driver, so need to search around on other PCB's for another one or wait for the slow
boat from China.
Regards Itsu
Could it be an effect caused by the Earth's electric field. This varies considerably and is influenced by the weather. With charged clouds overhead the E field can drop to zero and even reverse polarity. On a good clear day it is something like 300V/m pointing downwards (the Earth is charged negatively).
Smudge
Quote from: Smudge on 2018.09.09, 18:13:08
Could it be an effect caused by the Earth's electric field. This varies considerably and is influenced by the weather. With charged clouds overhead the E field can drop to zero and even reverse polarity. On a good clear day it is something like 300V/m pointing downwards (the Earth is charged negatively).
Smudge
Smudge,
Now that's an idea! How could I maybe confirm this? Would a faraday shield or similar shielding work?
Pm
Suggest covering a large cardboard box with Aluminum foil, top and sides. Then place this over your device, and see what happens.
If it still gives ou, then you might put Al foil on the box top, sitting on the table (thinking of a Xerox-paper box or equivalent), put device on that - then the rest of the box over that. Covering as much of your experiment as you can with the Faraday cage...
This would be an interesting test...
Hmmm, the "Carnegie curve" comes in mind then for the daily fluctuations:
https://link.springer.com/article/10.1007/s10712-012-9210-2
Itsu
Found some IXDN614PI's, so started testing again, now using severall load resistors in the 10K to 156K range.
Also added another 10uF in series with the 2 i already had lowering the C1 capacitance from 5 to 3.3uF again
in combination with severall resistors.
Staying just above "unstable mode" which is around 240KHz but the output is very low.
L1 still 1mH.
No sign of even a reasonable COP, let alone a COP > 1.
I notice some oscillations in the circuit when the FG drive is removed spanning a vast part (0 - 500Mhz) of
the frequency spectrum.
Itsu
If you watch the sun series here https://www.youtube.com/user/Suspicious0bservers You will see that each day is a different story as far as the energy around the earth which comes from the Sun.
I watch this guy every day, it will open your eyes.
Regards
Mike 8)
Quote from: PhysicsProf on 2018.09.10, 03:42:17
Suggest covering a large cardboard box with Aluminum foil, top and sides. Then place this over your device, and see what happens.
If it still gives ou, then you might put Al foil on the box top, sitting on the table (thinking of a Xerox-paper box or equivalent), put device on that - then the rest of the box over that. Covering as much of your experiment as you can with the Faraday cage...
This would be an interesting test...
Thanks, I'll give this a try. I now think it is varying throughout the day but need more testing to confirm this. I see there are other post relative to daily variations so will take a look at those links.
Pm
Quote from: Itsu on 2018.09.10, 08:41:42
Hmmm, the "Carnegie curve" comes in mind then for the daily fluctuations:
https://link.springer.com/article/10.1007/s10712-012-9210-2
Itsu
Interesting link Itsu, thanks. I'm more convinced than ever that the effect is diurnal.
Pm
Quote from: Itsu on 2018.09.10, 09:13:26
Found some IXDN614PI's, so started testing again, now using severall load resistors in the 10K to 156K range.
Also added another 10uF in series with the 2 i already had lowering the C1 capacitance from 5 to 3.3uF again
in combination with severall resistors.
Staying just above "unstable mode" which is around 240KHz but the output is very low.
L1 still 1mH.
No sign of even a reasonable COP, let alone a COP > 1.
Well, maybe go back try the full bridge configuration again as that is what I'm presently using ATM. I have seen times thru the day where I may have COPs ranging from .6 to 3 or greater!?! I'm also using a 1 ohm CSR instead of the current probe and all probes are de-skewed identically and set to full bandwidth. Use the same tuning technique as you have been.
Quote
I notice some oscillations in the circuit when the FG drive is removed spanning a vast part (0 - 500Mhz) of
the frequency spectrum.
Now that is interesting! I'll have to take some measurements for this.
Pm
Quote
Itsu
Quote from: Centraflow on 2018.09.10, 10:51:08
If you watch the sun series here https://www.youtube.com/user/Suspicious0bservers You will see that each day is a different story as far as the energy around the earth which comes from the Sun.
I watch this guy every day, it will open your eyes.
Regards
Mike 8)
Mike,
Thanks for the link, lots of info there. I'm beginning to think that if the effect is due to potential variations in the atmosphere for whatever reason, it is interesting to study but not too useful for energy production on a practical scale at least with this circuitry.
Regards,
Pm
Itsu and all,
Update- On this nice bright sunny morning with cloudless skies, the COPs are ~0.7! ???
Pm
Itsu and all,
For comparison, these are the data results from testing yesterday and today at approximately the same time with the identical circuit and layout.
Pm
Partzman do you have any energy saving lamps any where, I once had loads of problems measuring waveforms and the problem came and went, it turned out to be noisy energy saving spotlights.
Quote from: Peterae on 2018.09.10, 17:20:59
Partzman do you have any energy saving lamps any where, I once had loads of problems measuring waveforms and the problem came and went, it turned out to be noisy energy saving spotlights.
Peter,
Well, that's a good thought! Let's see, I have two LED can lights above the bench and two 2-lamp T8 solid state fluorescent fixtures for general room light. I ran two consecutive tests, one with the lights off and the other with the lights on and the tests were essentially identical. These last tests BTW had a lower COP than the previous test taken earlier today with data in my last post. The 100ms COP ~ 0.801.
Pm
PM
I have asked for the full text here
https://www.researchgate.net/publication/326962959_Temporal_and_spatial_variations_of_the_equatorial_electrojet_during_storm_times_from_CHAMP_observations
I am a full member of Researchgate. Will post it when I have it.
Regards
Mike 8)
Quote from: Centraflow on 2018.09.10, 18:17:19
PM
I have asked for the full text here
https://www.researchgate.net/publication/326962959_Temporal_and_spatial_variations_of_the_equatorial_electrojet_during_storm_times_from_CHAMP_observations
I am a full member of Researchgate. Will post it when I have it.
Regards
Mike 8)
Mike,
That's great! I'm a registered member but not full.
Thanks,
Pm
All,
Here is a spectrum analysis from 0 to 20Mhz with the circuit running with a 15k load. With no signal or power to the circuit, the floor is at -72dBm with no signals present. The signal generator frequency is 687kHz.
Pm
Quote from: partzman on 2018.09.10, 16:41:46
Itsu and all,
For comparison, these are the data results from testing yesterday and today at approximately the same time with the identical circuit and layout.
Pm
PM,
thanks, those figures speak for them self, immense differences.
I understand you are again back to the circuit from post #9 including a 15K load resistor.
As my scope is not able to use your sweep measurement scheme i cannot do much here to replicate.
Itsu
Quote from: Itsu on 2018.09.10, 18:42:13
PM,
thanks, those figures speak for them self, immense differences.
I understand you are again back to the circuit from post #9 including a 15K load resistor.
As my scope is not able to use your sweep measurement scheme i cannot do much here to replicate.
Itsu
Itsu,
Yes, I think the differences are too great to be measurement error but it's possible I guess. You're right, I'm back to the circuit in post #9 and when you get the 14N05Ls, you could try that circuit running continuously with a 15k load and 40v supply without having to do a sweep. If the conditions are right, you should see OU, hopefully.
I have some other ideas on the circuitry but will have to wait until I can see some gain again.
Pm
Got my RFD14N05L's, so will be testing them in the circuit tonight.
Itsu
Itsu and all,
In an attempt to overcome the lack of gain during these past few days, I tried experimenting with FDD2572 mosfets which have a BVdss of 150v with a little higher capacitance than the 14N05Ls. The results are interesting in several ways. First, they seemed to show an apparent gain when the other devices do not at this time, and they have a large power output with gain when operating in what I call Mode 2.
The schematic shows the circuit and the first scope pix is the single hi-res 100ms sweep. Referring to the sweep shot, we see the delta energy from 90ms to 100ms is (131.9^2-104.4^2)*3.94e-6 = 12.8mJ. This equates to a power level of 12.8e-3/10e-3 = 1.28 watts. The average voltage between these points is (131.9+104.4)/2 ~118v. So, our load resistor to run continuous at these levels would be 118^2/1.28 = 10.878k ohms. The closest I had on hand was an 11.17k ohm. We can also see from this scope pix the change in shape of the Math trace near the end of the sweep. It is during this time and just prior to the 100ms end that the network's resonant frequency has increased beyond the generator frequency of 636kHz and has entered Mode 2 which previously had been unstable and very inefficient with the 14N05Ls.
The next two scope shots are of the power input seen as 1.294 watts and the output voltage of 123.8v across C1 loaded with the 11.17K resistor which results in a pout of 123.8^2/11.17e3 = 1.372w. The apparent COP is therefore 1.372/1.294 = 1.06.
Last is the data table for the sweep.
I think this shows that the previous circuits are not optimized due to lack of understanding the effects of the dynamic parametric changes. IMO, any device used for the parametric element should be profiled and from that data the other circuit values could be established for maximum gain and output. Any volunteers to help determine the math needed for this?
Pm
Quote from: Itsu on 2018.09.11, 16:39:31
Got my RFD14N05L's, so will be testing them in the circuit tonight.
Itsu
Itsu,
That's great! I have had a couple of thoughts as to why the performance has fallen off even in clear sky's. It is a possibility that the substrate diode is being degraded by the repeated conduction cycles on the parts I've been testing with. If this is the case, then it would be best to use a suitable schottky diode to bypass the mosfet's body diode. I plan to do some testing on this later today with a fresh set of 14N05Ls and will post the results when available.
Pm
PM,
I added BAT42's across source drain of the MOSFETs
I had to use a 100 Ohm csr else no voltage/current was registrated above the noise.
687Khz was to low as it went into unstable mode (distorted current traces).
740Khz is just before entering unstable mode.
In this situation, current (both with the probe as with the 100 Ohm csr) show negative mean.
The calculated input power is around 30mW and the voltage across the 14.8K load resistor is 19.9V (26mW).
I will do some further test tomorrow during the day to see if things change.
Itsu
Quote from: TinMan on 2018.09.02, 13:11:24
...
If you or some one can design a loop circuit,operating under the calculated excess energy output,where the output of that circuit is a 25% duty cycle,with a voltage output of around 4 to 5 volt's,and current of around 160mA,then i will build it,and give it a go. O0
...
Hi Brad,
Would like to show you a possible looping circuit I think would be able to maintain a continuous operation of your setup shown in your reply #65 (provided the measured COP=2 was correct, no offense intended with this). In that circuit you had 110 mA input current instead of 160 mA, probably due to a certain transformator and also to the 20% duty cycle versus your more recent data with another transformator and a 25% duty cycle.
You can see that the looping circuit has two decent parts: an isolated step down DC to DC converter and a pulse generator based on a CMOS timer IC, the TLC555C in plastic DIP8 (PDIP8) package, made by Texas Instruments. The absolute maximum output current for this IC is 150 mA, so the 110 mA current needed from its output pin 3 to the input of your transformer would not kill the IC. The LMC555C is also a CMOS timer made earlier by National Semiconductor but its absolute maximum output current is only 100 mA.
First I would suggest to build only the pulse generator with a TLC555C, fed from say around 5 VDC, either from a variable power supply or from two super capacitors in series (each charged up to around 2.5 V). If such pulse gen is able to feed your setup with a very similar pulse wave shape you kindly showed in your post #65, both amplitude and current wise, then energy wise there would be an even clearer insight into your setup and also you then eliminated the use of the FG. Then the purchase of the DC converter chip could be decided and I would return onto that part later on. The timer has a variable frequency and duty cycle control, independently from each other. The CMOS version has a 300-350 uA maximum current consumption (typically 170-200 uA) to operate its internal circuit from 5 V at such a low 28-29 kHz frequency.
Mouser and Digikey have the TLC555C and they have also the LT8301. Data sheet for the latter is here:
http://www.analog.com/media/en/technical-documentation/data-sheets/8301f.pdf (http://www.analog.com/media/en/technical-documentation/data-sheets/8301f.pdf)
Gyula
PS edited for additional text
Quote from: Itsu on 2018.09.11, 20:08:02
PM,
I added BAT42's across source drain of the MOSFETs
I had to use a 100 Ohm csr else no voltage/current was registrated above the noise.
687Khz was to low as it went into unstable mode (distorted current traces).
740Khz is just before entering unstable mode.
In this situation, current (both with the probe as with the 100 Ohm csr) show negative mean.
The calculated input power is around 30mW and the voltage across the 14.8K load resistor is 19.9V (26mW).
I will do some further test tomorrow during the day to see if things change.
Itsu
Itsu,
OK, things are getting simpler I think. Here is a test for you using your 14N05L's in Mode 2 which should allow you to see OU.
The schematic is the same as shown in post #232 with the following changes: L1 is now 700uH but not critical, C2 is a .22uf good high frequency cap, the load resistor is 3.71k ohms, the frequency with my setup is 246kHz, and the supply voltage is now 15v dc. The reason for the lower supply voltage is to keep the 14N05L out of avalanche to prevent instability while allowing the maximum delta C swing. C2 now fully bypasses the operating frequency which simplifies the overall analysis of the circuit's operation. I hope to have some math in the near future which should allow one to pick the best mosfets for the job.
Mode 2 now means that you will tune the input frequency until you have a half-sine current wave in-phase with the positive half of the driver pulse. This will yield the maximum gain. You can leave the schottky's in place as they will probably enhance the gain due to their lower forward drop (we'll assume) and still not affect the mosfet's reverse delta C.
The scope pix attached show the Pin at 86.4mw and the loaded C1 volts at 20.18 resulting in a Pout = 20.18^2/3.71e3 = 109.7mw for a COP = 109.7/86.4 = 1.27 . As you should find, the circuit values are not that critical and you could now test your other mosfets but be sure to operate them below their BVdss rating.
I think anyone who is interested should be able to replicate this test as sweep testing is no longer required but it would be helpful to have profiles of the mosfets used. The sensitivity to the ambient should not be a factor any longer either which I hope to be able to explain later.
Pm
Itsu and all,
It seems that the circuitry used in Mode2 in the last posts is also sensitive to the ambient so, I'm going to stop my research in this area and pursue other avenues unless some other ideas spring to life! :(
Regards,
Pm
PM,
bummer, but understandable. Thanks for the ride though, it was interesting nonetheless.
Itsu
Quote from: partzman on 2018.09.12, 15:49:23
Itsu and all,
It seems that the circuitry used in Mode2 in the last posts is also sensitive to the ambient so, I'm going to stop my research in this area and pursue other avenues unless some other ideas spring to life! :(
Regards,
Pm
When you say, "sensitive to the ambient," do you mean that you are picking up (50 or) 60-Hz noise - or what? could you explain a bit please?
Quote from: PhysicsProf on 2018.09.12, 19:03:27
When you say, "sensitive to the ambient," do you mean that you are picking up (50 or) 60-Hz noise - or what? could you explain a bit please?
Yes I should have explained it. The circuits I've tested have gain variations throughout the course of a day from <unity to >unity so they seem to be diurnal. However, there are times when the gains are consistent. I have tried faraday shielding to no avail so perhaps it is due to measurement errors or some other anomaly. In any case, it is very frustrating when this occurs as it hinders any logical progress so for me it's not worth pursuing any longer.
Pm
Hi Pm,
Whatever anomaly or say a measurement issue "tricks" you, have you considered looping when a circuit shows ou by the measurements? I mean you prepare for looping in advance with a suitable circuit you think proper for the job. Such looping test could help explore and eventually rule out a measurement issue, no? That would also be a step forward.
Anyway, many thanks for sharing your results.
Gyula
Quote from: gyula on 2018.09.12, 20:02:56
Hi Pm,
Whatever anomaly or say a measurement issue "tricks" you, have you considered looping when a circuit shows ou by the measurements? I mean you prepare for looping in advance with a suitable circuit you think proper for the job. Such looping test could help explore and eventually rule out a measurement issue, no? That would also be a step forward.
Anyway, many thanks for sharing your results.
Gyula
Yes, I agree. However, it would be nice if someone else could show some signs of replication in this case plus I think I have had other circuits in the past that may have been more likely candidates for looping.
Pm
Hi partzman.
A few years ago I was heavily into research of TH Moray.
I made a " detector " device using Sterling Silver tubes and a Lead tube cast from some Victorian water pipe. I had the Silver tubes placed top and bottom poking into the Lead tube. All parts were isolated from each other and the assembly was evacuated with a vacuum pump.
I connected the top Silver electrode to a sixty foot long aerial and the bottom Silver electrode to ground. I placed my oscilloscope probe onto the middle Lead electrode.
The scope was monitored over several hours showing a voltage that peaked around 10:30 AM each day. I'm also reminded of a conversation with a.king21 about the Kapanadse " Aquarium " device. The demonstration was cancelled because of an impending electric storm. No reason was offered by Kapanadse other than the device wouldn't work.
Finally, could these circuits deemed " Parametric " ??
http://www.overunityresearch.com/index.php?topic=2358.msg36649#msg36649
The top drawing was built, poorly, by me and the storage capacitor exploding both damaged my youngest son and the front of my frequency counter too.
Cheers Graham.
Quote from: Grumage on 2018.09.13, 13:01:08
Hi partzman.
A few years ago I was heavily into research of TH Moray.
I made a " detector " device using Sterling Silver tubes and a Lead tube cast from some Victorian water pipe. I had the Silver tubes placed top and bottom poking into the Lead tube. All parts were isolated from each other and the assembly was evacuated with a vacuum pump.
I connected the top Silver electrode to a sixty foot long aerial and the bottom Silver electrode to ground. I placed my oscilloscope probe onto the middle Lead electrode.
The scope was monitored over several hours showing a voltage that peaked around 10:30 AM each day. I'm also reminded of a conversation with a.king21 about the Kapanadse " Aquarium " device. The demonstration was cancelled because of an impending electric storm. No reason was offered by Kapanadse other than the device wouldn't work.
Finally, could these circuits deemed " Parametric " ??
http://www.overunityresearch.com/index.php?topic=2358.msg36649#msg36649
The top drawing was built, poorly, by me and the storage capacitor exploding both damaged my youngest son and the front of my frequency counter too.
Cheers Graham.
Graham,
Thanks for sharing this info on your previous Moray investigation and work. In the beginning, I thought my circuit was parametric due to the fact that I replaced the mosfets with diodes having fixed capacitors across them to construct the bridge. These showed no signs of any gain at any time as far as I could tell.
So, I was surprised when all the variations showed up and I asked myself, 'how could any outside field affect the parametric action' and I honestly don't have an answer for that.
Anyway, your experiments look interesting and I will spend some time on that thread to see what you were up to.
Regards,
Pm
Hi PM.
I've just noticed an error in my previous message.
I missed out " be " in the sentence " Finally, could these circuits ( be ) deemed Parametric?
Morays work fascinated me after reading that in 1910 he had managed to light a 1/4 candlepower lamp from air and ground. Was this real? I suppose we'll never know for sure.
Cheers Graham.
Quote from: partzman on 2018.09.12, 22:00:58
Yes, I agree. However, it would be nice if someone else could show some signs of replication in this case plus I think I have had other circuits in the past that may have been more likely candidates for looping.
Pm
Dear Pm,
I understand that you have become a bit disappointed and stop research in this area as you wrote because the circuitry used in Mode 2 was also sensitive to the ambient. I also understand that you had expected more replications, besides Itsu and Tinman.
I think Itsu used the same circuitry like you but he did not measure OU, while Tinman's circuit uses a input transformer he carefully selected out of several transformers to get a measured OU, recently a COP of 2 or so.
Regarding your other circuits in the past that may have been more likely candidates for looping, I do not know whether anyone attempted those circuits to loop back? No offense but I recall eventual measurement errors you yourself found and reported in some of the cases.
There was your setup last year with the bifilar pancake transformer that TinselKoala replicated ( https://www.youtube.com/watch?v=MNzbc-N-e9c (https://www.youtube.com/watch?v=MNzbc-N-e9c) ) and his measurements showed a COP of 2.9 which is remarkable of course. Now what I do not know is whether this setup was attempted to loop back with a suitable circuit by you (or by TinselKoala)? If not, then from your part, why not?
Regarding my lack of replication attempt of your parametric circuit, I apologize for not doing it. Simply, I am not fully convinced by the measurements and again I mean no offense with this, sorry. This is why I mentioned the looping for you, also for Tinman with a circuit suggestion, that would be the ultimate test.
However, my take on this is that nobody has any 'obligation' to show a looped circuit to prove he has an OU circuitry, this is completely voluntary, so I do not 'blame' you or anyone else for not doing so.
Maybe others think this differently, saying that this is a private group, you created it, so why not show looping?
When I suggest looping in case someone reports measured OU, then first of all I mean proving it for himself, not to send himself up the garden path with any wrong measurements that are not easy to detect. If anomaly causes OU and not measurement errors, then looping could maintain operation of course.
I've thought I write these thoughts of mine to you. Please keep up your excellent work in this or any other area.
Gyula
Quote from: gyula on 2018.09.16, 21:19:10
Dear Pm,
I understand that you have become a bit disappointed and stop research in this area as you wrote because the circuitry used in Mode 2 was also sensitive to the ambient. I also understand that you had expected more replications, besides Itsu and Tinman.
I think Itsu used the same circuitry like you but he did not measure OU, while Tinman's circuit uses a input transformer he carefully selected out of several transformers to get a measured OU, recently a COP of 2 or so.
Regarding your other circuits in the past that may have been more likely candidates for looping, I do not know whether anyone attempted those circuits to loop back? No offense but I recall eventual measurement errors you yourself found and reported in some of the cases.
There was your setup last year with the bifilar pancake transformer that TinselKoala replicated ( https://www.youtube.com/watch?v=MNzbc-N-e9c (https://www.youtube.com/watch?v=MNzbc-N-e9c) ) and his measurements showed a COP of 2.9 which is remarkable of course. Now what I do not know is whether this setup was attempted to loop back with a suitable circuit by you (or by TinselKoala)? If not, then from your part, why not?
You are referring to what I called MEI or Magneto Electric Induction which utilized the distributed capacitance between tightly coupled windings to convert reactive input power to real output power. They worked very efficiently and measured OU with some showing infinite OU at power levels around 35 watts but alas, they were actually COP<1. The reason I never attempted looping these circuits was due to a simple test which indicated to me that the measurements were in error and too critical to measure even with the best of oscilloscopes and probes under the most ideal conditions.
The test was this: With an MEI device that measured a Pin in the range of -2 to -6 watts, the circuit was driven with a set number of pulses at the correct operating frequency until the circuit stabilized and then the pulse driver was grounded. Given the fact that the input of the MEI was a capacitive reactance, it was driven by a pulse driver through a given inductance resulting in a series resonant load to the square wave driver. So, if the input was truly negative power, by grounding the driven end of the input inductor, the device should have not only self run but the resonant voltage should have increased until limited by the overall losses or Q. This is not what happened but rather the circuit oscillation decayed similar to normal resonance circuits with the drive removed. IMO this proved the measurements were not real.
Why were the measurements that far off? The answer is easy and applies to nearly all reactive to real generators. Refer to the scope shot in my post #29 in Brad's "AT=0" thread for an example of the above. Here we see a measure Pin = -5.067 with a power out of ~2.3 watts and notice the phase angles. Now also notice the peak reactive power that reaches ~240 watts! This is where the error creeps in in that the ratio of peak (or rms) reactive power to the input power is far greater than present oscilloscopes are capable of discerning. I discussed this with Tek engineers and they didn't/don't have any answers but in fact I did learn that my rated 8 bit vertical resolution MDO is actually capable of up to 16 bits resolution under certain conditions but even this didn't help. Using a 350MHZ rated scope measuring a 5MHz circuit to find out that just moving a probe over a little affects those phase angles and resultant power ratios.
This is why I didn't attempt any looping of these devices with additional circuitry. None of then were/are really worthy of the attempt IMO.
BTW, I did design very efficient class E amplifiers at the appropriate frequencies to drive the MEI's and attempted to measure OU at the class E power source and was not successful.
Quote
Regarding my lack of replication attempt of your parametric circuit, I apologize for not doing it. Simply, I am not fully convinced by the measurements and again I mean no offense with this, sorry. This is why I mentioned the looping for you, also for Tinman with a circuit suggestion, that would be the ultimate test.
I am personally convinced that some of the parametric charging circuits are truly OU but appear to be random as stated. This randomness may be due to measurement error for the very same reason as stated above for the MEI's because if one looks closely at the scope traces, this is a reactive to real generator. The problem with attempted looping in my mind is the randomness coupled with the relatively low power and the operating frequencies.
I have continued to work with variations of these circuits in an attempt to resolve the issues but have not posted at this point in time due to lack of anything important to report.
Quote
However, my take on this is that nobody has any 'obligation' to show a looped circuit to prove he has an OU circuitry, this is completely voluntary, so I do not 'blame' you or anyone else for not doing so.
Maybe others think this differently, saying that this is a private group, you created it, so why not show looping?
When I suggest looping in case someone reports measured OU, then first of all I mean proving it for himself, not to send himself up the garden path with any wrong measurements that are not easy to detect. If anomaly causes OU and not measurement errors, then looping could maintain operation of course.
I've thought I write these thoughts of mine to you. Please keep up your excellent work in this or any other area.
Gyula
Thanks for your kind words!
Pm
Quote from: partzman on 2018.09.12, 00:08:29
Itsu,
OK, things are getting simpler I think. Here is a test for you using your 14N05L's in Mode 2 which should allow you to see OU.
The schematic is the same as shown in post #232 with the following changes: L1 is now 700uH but not critical, C2 is a .22uf good high frequency cap, the load resistor is 3.71k ohms, the frequency with my setup is 246kHz, and the supply voltage is now 15v dc. The reason for the lower supply voltage is to keep the 14N05L out of avalanche to prevent instability while allowing the maximum delta C swing. C2 now fully bypasses the operating frequency which simplifies the overall analysis of the circuit's operation. I hope to have some math in the near future which should allow one to pick the best mosfets for the job.
Mode 2 now means that you will tune the input frequency until you have a half-sine current wave in-phase with the positive half of the driver pulse. This will yield the maximum gain. You can leave the schottky's in place as they will probably enhance the gain due to their lower forward drop (we'll assume) and still not affect the mosfet's reverse delta C.
The scope pix attached show the Pin at 86.4mw and the loaded C1 volts at 20.18 resulting in a Pout = 20.18^2/3.71e3 = 109.7mw for a COP = 109.7/86.4 = 1.27 . As you should find, the circuit values are not that critical and you could now test your other mosfets but be sure to operate them below their BVdss rating.
I think anyone who is interested should be able to replicate this test as sweep testing is no longer required but it would be helpful to have profiles of the mosfets used. The sensitivity to the ambient should not be a factor any longer either which I hope to be able to explain later.
Pm
PM, All,
To wrap things up, i was the last days still playing with the setup with the suggested changes above.
My L1 was 1mH, C2 220nF ceramic and load resistor 3.8K.
15V was used for the supply voltage.
Frequency was tuned so the current half-sine wave was in phase with the drive pulse.
That means at about 170KHz.
This setup turns out to be very close to unity at an effecienty of around 96%.
I still can try different MOSFETs as i was only using the RFD14N05L's with Bat42's.
I never experienced dramatic changes due to time of day or weather changes.
Thanks, Itsu
Quote from: Itsu on 2018.09.17, 08:59:41
PM, All,
To wrap things up, i was the last days still playing with the setup with the suggested changes above.
My L1 was 1mH, C2 220nF ceramic and load resistor 3.8K.
15V was used for the supply voltage.
Frequency was tuned so the current half-sine wave was in phase with the drive pulse.
That means at about 170KHz.
This setup turns out to be very close to unity at an effecienty of around 96%.
I still can try different MOSFETs as i was only using the RFD14N05L's with Bat42's.
I never experienced dramatic changes due to time of day or weather changes.
Thanks, Itsu
Itsu,
I'm curious about one thing in your measurement setup, do you measure the voltage across C1 and the load with a meter or are you taking differential measurements with the scope probes? The meter leads may be affecting the measurements because there is a common mode resonant voltage swing at C1 at the driving frequency.
Pm
PM,
i did it mixed, always with the DMM, and from time to time checked with the 2 probes method.
They agreed almost all of the time.
Itsu
Itsu and all,
Here is another optional test that we will re-visit which uses two paralleled mosfets and in this case it is the 14N05L's. See the attached schematic below. If you were to substitute the mosfets with say Schottky diodes, the maximum voltage across C1 with a load would be slightly less than 1/2 the supply voltage at a 50% duty cycle. Here however, we see a voltage across C1 with the load that is greater than 1/2 the supply due to the parametric pumping of the mosfet's non-linear capacitance. This is the simplest configuration that demonstrates the principle I can think of at the moment. This should be easy to replicate by anyone who desires and the results will vary with the devices and the component values used.
The tuning is easy. Apply a frequency that is high enough to generate a voltage that is less than 1/2 the supply across C1 with the load. Then, carefully lower the frequency until the point is reached where the C1 voltage increases to be >1/2 the supply and then suddenly jumps to a higher stable value. Now raise the frequency until the voltage across C1 again drops to <1/2 the supply and lower the frequency again until you have the maximum output across C1 without the sudden jump. IOW, we will now be operating in Mode1 not Mode2. This may require several tries to find the "sweet" spot.
The load resistance can be found by experimentation if one does not the ability to do a 100ms or equivalent sweep as it is not that critical.
In the scope shot we see a C1 voltage of 22.28v with the 7.451 ohm load for a Pout = 22.28^2/7451 = 66.62mw. Pin is seen to be 46.67mw for a COP = 66.62/46.67 = 1.43. I used my current probe for this test but the results should be the same if a CSR is used. Note the mean current value as it alone does not account for the load's output voltage. The balance of output energy is supplied by the AC parametric pumping.
What I like about this circuit is the possibility of looping by using a low power highly efficient boost converter to build the supply voltage from the voltage across C1 with the load although it would help if the power levels were higher IMO.
Pm
All,
Here is a different driving scheme for the parametric charging circuitry. IMO, this is just a reactive to real converter with a parametric "C" for resonance. Other that that , I'm not going to comment except to say that the load resistance is 7.451k ohms.
Pm
All,
Please ignore the previous post as there is measurement error which cancels the apparent gain.
Pm
Quote from: Itsu on 2018.09.17, 15:13:14
PM,
i did it mixed, always with the DMM, and from time to time checked with the 2 probes method.
They agreed almost all of the time.
Itsu
PM,
i went back to that older circuit of post #248 and double checked the voltage across the load resistor (3902 Ohm).
It was 13.91V with my DMM (Fluke 179).
The blue and purple traces in the screenshot shows the means, so a very close difference (14.046V).
The input calculated in red shows 50.82mW, and the outputs:
(13.91² / 3902) is 49.5mW.
(14.046² / 3902) is 50.5mW
Both less (a fraction) then the input.
I repeated this severall times with similar outcomes.
Itsu
Quote from: Itsu on 2018.09.18, 20:15:14
PM,
i went back to that older circuit of post #248 and double checked the voltage across the load resistor (3902 Ohm).
It was 13.91V with my DMM (Fluke 179).
The blue and purple traces in the screenshot shows the means, so a very close difference (14.046V).
The input calculated in red shows 50.82mW, and the outputs:
(13.91² / 3902) is 49.5mW.
(14.046² / 3902) is 50.5mW
Both less (a fraction) then the input.
I repeated this severall times with similar outcomes.
Itsu
Itsu,
Thanks for continuing to try this setup! O0 It must boil down to measurement error on my part so I don't think I will waste anymore time on it. I did notice on your scope shot the slight amount of power consumption shown in the Math channel during the "off" time of your driver when the input current is negative. If this was not there, you would probably have a slight gain but if it is due to offset in your CH1, it would be added on the positive half of the cycle so all would cancel.
Pm
Quote from: gyula on 2018.09.11, 22:53:54
Hi Brad,
Would like to show you a possible looping circuit I think would be able to maintain a continuous operation of your setup shown in your reply #65 (provided the measured COP=2 was correct, no offense intended with this). In that circuit you had 110 mA input current instead of 160 mA, probably due to a certain transformator and also to the 20% duty cycle versus your more recent data with another transformator and a 25% duty cycle.
You can see that the looping circuit has two decent parts: an isolated step down DC to DC converter and a pulse generator based on a CMOS timer IC, the TLC555C in plastic DIP8 (PDIP8) package, made by Texas Instruments. The absolute maximum output current for this IC is 150 mA, so the 110 mA current needed from its output pin 3 to the input of your transformer would not kill the IC. The LMC555C is also a CMOS timer made earlier by National Semiconductor but its absolute maximum output current is only 100 mA.
First I would suggest to build only the pulse generator with a TLC555C, fed from say around 5 VDC, either from a variable power supply or from two super capacitors in series (each charged up to around 2.5 V). If such pulse gen is able to feed your setup with a very similar pulse wave shape you kindly showed in your post #65, both amplitude and current wise, then energy wise there would be an even clearer insight into your setup and also you then eliminated the use of the FG. Then the purchase of the DC converter chip could be decided and I would return onto that part later on. The timer has a variable frequency and duty cycle control, independently from each other. The CMOS version has a 300-350 uA maximum current consumption (typically 170-200 uA) to operate its internal circuit from 5 V at such a low 28-29 kHz frequency.
Mouser and Digikey have the TLC555C and they have also the LT8301. Data sheet for the latter is here:
http://www.analog.com/media/en/technical-documentation/data-sheets/8301f.pdf (http://www.analog.com/media/en/technical-documentation/data-sheets/8301f.pdf)
Gyula
PS edited for additional text
Thanks Gyula
Once im back on my feet,i will see if i can put it all together.
Brad
Quote from: partzman on 2018.09.17, 15:30:47
Itsu and all,
Here is another optional test that we will re-visit which uses two paralleled mosfets and in this case it is the 14N05L's. See the attached schematic below. If you were to substitute the mosfets with say Schottky diodes, the maximum voltage across C1 with a load would be slightly less than 1/2 the supply voltage at a 50% duty cycle. Here however, we see a voltage across C1 with the load that is greater than 1/2 the supply due to the parametric pumping of the mosfet's non-linear capacitance. This is the simplest configuration that demonstrates the principle I can think of at the moment. This should be easy to replicate by anyone who desires and the results will vary with the devices and the component values used.
The tuning is easy. Apply a frequency that is high enough to generate a voltage that is less than 1/2 the supply across C1 with the load. Then, carefully lower the frequency until the point is reached where the C1 voltage increases to be >1/2 the supply and then suddenly jumps to a higher stable value. Now raise the frequency until the voltage across C1 again drops to <1/2 the supply and lower the frequency again until you have the maximum output across C1 without the sudden jump. IOW, we will now be operating in Mode1 not Mode2. This may require several tries to find the "sweet" spot.
The load resistance can be found by experimentation if one does not the ability to do a 100ms or equivalent sweep as it is not that critical.
In the scope shot we see a C1 voltage of 22.28v with the 7.451 ohm load for a Pout = 22.28^2/7451 = 66.62mw. Pin is seen to be 46.67mw for a COP = 66.62/46.67 = 1.43. I used my current probe for this test but the results should be the same if a CSR is used. Note the mean current value as it alone does not account for the load's output voltage. The balance of output energy is supplied by the AC parametric pumping.
What I like about this circuit is the possibility of looping by using a low power highly efficient boost converter to build the supply voltage from the voltage across C1 with the load although it would help if the power levels were higher IMO.
Pm
Hmmm,
PM, i tried the above circuit, using 40V on the ixdd614pi,
L1 is the 322uH one,
2x 14N05L with schottky diodes (bat42)
C1 = 3.3uF
Rload is 8.14K
R1 = 10 Ohm optional csr.
Screenshot shows input calculated in red (from yellow x green) = 17.17mW
output across 8140 Ohm resistor is 23.55V meaning output power (23.55² / 8140) = 68mW.
Purple is voltage across 10 Ohm csr, so 1.25mA so confirms the green current probe value of 1.128mA.
Went from 650Khz slowly down to 638Khz after which the voltage across the load resistor switch to 33V or so.
So 638Khz gave the 23.55V across the Rload.
Tried severall times, and got similar results allthough the current (green) is not stable over long periods.
17.17mW input, 68mW output, COP = 3.96.
I did not calculate the power loss across the 10 Ohm csr.
Itsu
Quote from: Itsu on 2018.09.21, 20:39:35
Hmmm,
PM, i tried the above circuit, using 40V on the ixdd614pi,
L1 is the 322uH one,
2x 14N05L with schottky diodes (bat42)
C1 = 3.3uF
Rload is 8.14K
R1 = 10 Ohm optional csr.
Screenshot shows input calculated in red (from yellow x green) = 17.17mW
output across 8140 Ohm resistor is 23.55V meaning output power (23.55² / 8140) = 68mW.
Purple is voltage across 10 Ohm csr, so 1.25mA so confirms the green current probe value of 1.128mA.
Went from 650Khz slowly down to 638Khz after which the voltage across the load resistor switch to 33V or so.
So 638Khz gave the 23.55V across the Rload.
Tried severall times, and got similar results allthough the current (green) is not stable over long periods.
17.17mW input, 68mW output, COP = 3.96.
I did not calculate the power loss across the 10 Ohm csr.
Itsu
Itsu,
Excellent results! O0 The input current waveform looks really good as the slopes are really linear rather than having a slight curving near the peaks which as a visual usually indicates low gain or non at all.
Thank you for being persistent with this and now you get to play with the variable gain over time situation! ???
Regards,
Pm
Ok, no problem O0
Input from PS is 40V @ 44mA = 1.76W.
Also we have some input from the FG, like 4V @ 50% duty cycle over 50 Ohm = 80mA.
The IXDD614PI gets pretty hot, with heatsink (copper tape strip) about 40°C (104°F), so we dissipate some (most?) there.
Today again similar values as yesterday (weather similar, cold, rainy, windy).
Itsu
I added a boost converter (max 30V out) with its input across the loadR from PM's circuit and set its output to 24V.
This boost converter powers with its 24V a series connection of 8 leds.
This converters input return line and its output return line are connected
After powering on and adjusting the frequency to about 247Khz now (just before the jump to higher voltage)
i got again a stable situation with about calculated 134mW input and about 384mW (24V x 16mA) output into the leds.
The 16mA was found by using a 1 Ohm csr in the return line of the leds
Screenshot shows again the red trace input power (yellow x green) = 134mW
Blue is the voltage across the loadR and the input to the boost converter
Purple is the voltage across the 1 Ohm csr in the leds return line.
Video showing this setup here: https://www.youtube.com/watch?v=3iUWYbY4ujg&feature=youtu.be
Need a boost converter capable of 40V to try a feedback loop.
Itsu
I wonder if replacing the IX fet driver with the following circuit would work, still need a FG to drive it, but higher voltage fets could be used.
http://www.discovercircuits.com/DJ-Circuits/pushpul.htm
Thanks Peter,
i can give it a try, the 30V would be perfect for my present boost converter, not sure if i can get
the same COP with 30V though.
Itsu
Tried Peters circuit using IRF4905 and IRF3205 MOSFETs, but cannot get any higher as 30Khz else the output signal
is just peaks, no more square wave.
Guess i need the specified MOSFETs ZVP2106 / ZVN2106.
Itsu
All,
I would like to move the contents of this thread to public view on the forum so if you have contributed in a post that you would not like to have shared publicly, then delete that post or let me know and I will delete it for you.
I would hope that from this action, others will attempt replication like Itsu and Brad which perhaps could result in some practical application.
This thread will remain private over the next several days and then I will ask Peter to move it into public view. Thanks for everyone's input up to this point. O0
Regards,
Pm
Quote from: partzman on 2018.09.24, 13:05:49
All,
I would like to move the contents of this thread to public view on the forum so if you have contributed in a post that you would not like to have shared publicly, then delete that post or let me know and I will delete it for you.
I would hope that from this action, others will attempt replication like Itsu and Brad which perhaps could result in some practical application.
This thread will remain private over the next several days and then I will ask Peter to move it into public view. Thanks for everyone's input up to this point. O0
Regards,
Pm
OK - I support this move, given the progress you have made and the encouragement of replications that such a move would provide. Good idea, and my posts here can be included in the public arena. No problem with me.
Hi Pm,
It is ok for me to go public with this topic, including my posts too.
Gyula
Quote from: partzman on 2018.09.24, 13:05:49
I would like to move the contents of this thread to public view on the forum ...
This project has now been moved to an open part of the forum.
Peter,
Thanks you very much for all your work on this! O0 O0 O0
Regards,
Pm
Quote from: partzman on 2018.09.18, 18:13:08
All,
Here is a different driving scheme for the parametric charging circuitry. IMO, this is just a reactive to real converter with a parametric "C" for resonance. Other that that , I'm not going to comment except to say that the load resistance is 7.451k ohms.
Pm
First - thank you for making the thread open to the public and thus inviting replications.
In that same altruistic spirit, is the circuit below your "best" for this approach?
--> Itsu, is this the same circuit you used for your video above?
Or, what circuit would you all encourage would-be replicators to begin with?
I really appreciate it - and congratulate your progress (and that of Itsu and others) on this exciting project!
PS - is this device being discussed also on another forum?
Quote from: PhysicsProf on 2018.09.27, 00:56:02
First - thank you for making the thread open to the public and thus inviting replications.
In that same altruistic spirit, is the circuit below your "best" for this approach?
--> Itsu, is this the same circuit you used for your video above?
Or, what circuit would you all encourage would-be replicators to begin with?
I really appreciate it - and congratulate your progress (and that of Itsu and others) on this exciting project!
PS - is this device being discussed also on another forum?
Thanks for your comments! Actually the circuit I would recommend for replication is the one shown below or in my post #251. All the tuning instructions are called out there and are important to follow correctly. It is still unknown (to me at least) as to what makes one mosfet a better candidate over another due to all the variables involved so experimentation is still the rule however, one can parallel as many mosfets as desired which of course increases the magnitude of the parametric "C" thus lowering the frequency and increasing the available energy.
I have also tried 'P' channel mosfets following the thinking of some who value 'hole' or positron flow verses electron or negatron flow but saw no real advantage with the few devices I tested.
Another area of experimentation could be the proper use of powdered iron cores for L1 to take advantage of their inherent increase in permeability with an increase in H field which should broaden the gain operating area.
In regards to the other forums, I haven't seen any serious efforts to measure pin verses pout especially in charging a capacitor but maybe I'm not that thorough.
Regards,
Pm
Quote from: PhysicsProf on 2018.09.27, 00:56:02
First - thank you for making the thread open to the public and thus inviting replications.
In that same altruistic spirit, is the circuit below your "best" for this approach?
--> Itsu, is this the same circuit you used for your video above?
Or, what circuit would you all encourage would-be replicators to begin with?
I really appreciate it - and congratulate your progress (and that of Itsu and others) on this exciting project!
PS - is this device being discussed also on another forum?
PhysicsProf,
Quote--> Itsu, is this the same circuit you used for your video above?
No, that circuit you posted is not the one i used, its like Partzman also mentioned, the one from his post #251
and was linked by him just now.
I am waiting now for some parts (MOSFETs ZVP2106 / ZVN2106 (as in the suggested circuit by Peterae in post #261 above).
I hope i can ommit the power hungry IXDD614 MOSFET driver i am using now, so a feedback loop is more realistic.
Itsu
I hope that works out ok Itsu ;)
Thanks for opening the thread up to the public Partzman, a lot of hard work has been done and will it be interesting to see if other people get interested in the device, I think it must be stated that although there are COP > 1 results that it is still a good possibility there is some sort of measurement error causing these results which is why we have opened the thread up, with the hope others will find the circuit interesting enough to investigate and prove/disprove COP > 1 results, whether by looping or accurate power measurements.
Thank you PM and Itsu!
And well said, Peter!
Watching science unfold!
All,
This is a revised schematic of post #251 to help clarify some of the requirements.
Regards,
Pm
In the Revised schematic if you were to include the
MosFet Body Diodes in the diagram what would you
have?
PM
thanks for this revised schematic, it now clearly shows that the "input" is being measured "behind" the MOSFET driver stage.
So the input / output measurement is only done for that part behind the driver stage.
This makes it hard to create a feedback loop as the "input" is pulsed DC (square wave 50% duty cycle) and the output is DC.
If we include the FG and MOSFET driver stage then the input would be 100 times higher, see my post #259.
Perhaps we can reduce that significantly by using Peterae's mentioned circuit in his post #261.
I am waiting for the correct parts to try that.
Itsu
Quote from: muDped on 2018.09.28, 19:04:37
In the Revised schematic if you were to include the
MosFet Body Diodes in the diagram what would you
have?
What you would have and actually do have, is a half wave rectification of a unipolar pulse charging a capacitor. The difference is that the parametric capacitance of the mosfets provides the "C" that resonates with L1. It is considered that this parametric capacitance is the gain mechanism for the device.
Previously shown full wave circuits are actually more efficient but this circuit is the easiest to replicate IMO.
Pm
Quote from: Itsu on 2018.09.28, 19:11:52
PM
thanks for this revised schematic, it now clearly shows that the "input" is being measured "behind" the MOSFET driver stage.
So the input / output measurement is only done for that part behind the driver stage.
This makes it hard to create a feedback loop as the "input" is pulsed DC (square wave 50% duty cycle) and the output is DC.
If we include the FG and MOSFET driver stage then the input would be 100 times higher, see my post #259.
Perhaps we can reduce that significantly by using Peterae's mentioned circuit in his post #261.
I am waiting for the correct parts to try that.
Itsu
Itsu,
Yes, this is the challenge that is, to produce an efficient input switching circuit that is powered by the charge on C1. Several things that might help in this is to add additional paralleled mosfets to increase the available parametric charge while at the same time lower the frequency.
Increasing the pulse amplitude can also be experimented with but sometimes can actually prove to be detrimental depending on the delta C of the particular mosfet(s) used.
Pm
Fantastic work!
Either a gremlin to find or an effect to discover
Something that may be relevant from my experience using metal detectors where you spend many hours bush walking and listening to out of range signals which mean either a metal target or a ground variance which is usually a difference in water content. There is however a certain interference that affect detectors of the pulse induction kind i use where they become almost unusable for many hours at a time ,and on these days there is often a random fully saturated signal which stops you in your tracks and always, in hindsight is a ghost signal.
On a good day these ghost signals still occour but far less often.
A metal detector coil is 300uh +/- 10% they send pulses out and immediately switch to recieve after every pulse train looking for delayed eddy currents that metal objects leave behind .They are from 8" to 30" in diamater air core and use litz wire .
These coils also incorporate a faraday shield which i think i see as alluminium foil on the ftpu.
Without the faraday shield they are erratic and not useable at all near the ground .
That faraday shield is a full turn of conductive foil enclosing the whole coil like a shield which is joined to ground on one end but does not complete a full turn(2mm gap )which would short the coil.
I have never been able to predict that random signal but if there is agreement among detector users that this interference is called "shperics" and could be worth study .
If it is the semiconductors or the coil it may be worth changing the coils topology ,which i will do after im at first base .
If course we all would like it to be the coil and its environment!
Back on topic ,
I am keen to replicate .
Does any body have an economic source for the 14n05L?
Or are fast diodes the only requirement?
the other bits i have .
The 609p1 mosfer driver also seems to be elusive but there are other described methods.
Replicating needs to be as exact as possible to be of the most worth.
I live in Tinmans neck of the woods.
3D
Here a few pictures of my present setup using a breadboard.
I lost some MOSFET drivers as can be seen by the scorch marks, they do not like a too high frequency at 40V with
this L1 as a load. I have some copper tape as heatsink on now.
The covered part in the middle of picture 1 are some additional parts (see Peterae his post #261) not being used yet.
Guess the pictures speak for them self.
I do have extra Bat 42 schottky diodes across the source/drain of the MOSFETs
The RFD14N05L MOSFETs came from Ebay (Germany, which is close from here).
Itsu
Quote from: 3D Magnetics on 2018.09.29, 02:09:46
Fantastic work!
Either a gremlin to find or an effect to discover
Something that may be relevant from my experience using metal detectors where you spend many hours bush walking and listening to out of range signals which mean either a metal target or a ground variance which is usually a difference in water content. There is however a certain interference that affect detectors of the pulse induction kind i use where they become almost unusable for many hours at a time ,and on these days there is often a random fully saturated signal which stops you in your tracks and always, in hindsight is a ghost signal.
On a good day these ghost signals still occour but far less often.
A metal detector coil is 300uh +/- 10% they send pulses out and immediately switch to recieve after every pulse train looking for delayed eddy currents that metal objects leave behind .They are from 8" to 30" in diamater air core and use litz wire .
These coils also incorporate a faraday shield which i think i see as alluminium foil on the ftpu.
Without the faraday shield they are erratic and not useable at all near the ground .
That faraday shield is a full turn of conductive foil enclosing the whole coil like a shield which is joined to ground on one end but does not complete a full turn(2mm gap )which would short the coil.
I have never been able to predict that random signal but if there is agreement among detector users that this interference is called "shperics" and could be worth study .
If it is the semiconductors or the coil it may be worth changing the coils topology ,which i will do after im at first base .
If course we all would like it to be the coil and its environment!
Back on topic ,
I am keen to replicate .
Does any body have an economic source for the 14n05L?
Or are fast diodes the only requirement?
the other bits i have .
The 609p1 mosfer driver also seems to be elusive but there are other described methods.
Replicating needs to be as exact as possible to be of the most worth.
I live in Tinmans neck of the woods.
3D
3D,
Try Mouser for the mosfets and Ixys drivers. They have the 14N05L's for $0.82 in singles or $0.68 in 10 piece quantities. The Ixys drivers are available in 604, 609, and 614 which are 4, 6, and 9 amp devices respectively. They also come optionally with enable or inverted output and any of these will work for the driver although they are not very efficient for driving continuous loads.
Also, ordinary slow recovery diodes will work but at higher frequencies due to their higher reverse capacitance. Fast recovery diodes will not work well if at all.
Pm
Quote from: Itsu on 2018.09.29, 07:50:00
Here a few pictures of my present setup using a breadboard.
I lost some MOSFET drivers as can be seen by the scorch marks, they do not like a too high frequency at 40V with
this L1 as a load. I have some copper tape as heatsink on now.
The covered part in the middle of picture 1 are some additional parts (see Peterae his post #261) not being used yet.
Guess the pictures speak for them self.
I do have extra Bat 42 schottky diodes across the source/drain of the MOSFETs
The RFD14N05L MOSFETs came from Ebay (Germany, which is close from here).
Itsu
Itsu,
Thanks for posting your pix. I also find it interesting that in using the Bat 421 diodes across the mosfets and you achieved a higher COP than I did without using any bypass diodes. These apparently are increasing the forward conduction efficiency as compared to the 14N05L's substrate diodes.
Pm
Quote from: Itsu on 2018.09.28, 19:11:52
PM
thanks for this revised schematic, it now clearly shows that the "input" is being measured "behind" the MOSFET driver stage.
So the input / output measurement is only done for that part behind the driver stage.
This makes it hard to create a feedback loop as the "input" is pulsed DC (square wave 50% duty cycle) and the output is DC.
If we include the FG and MOSFET driver stage then the input would be 100 times higher, see my post #259.
Perhaps we can reduce that significantly by using Peterae's mentioned circuit in his post #261.
I am waiting for the correct parts to try that.
Itsu
I'm a little confused by your post Itsu, and I'm not 100% clear on PM's revised schematic.
What I see on the schematic, is an input voltage measurement (CH1, YEL), and in the previous schematic there was a current probe on the input. It appears the revised schematic is now using a CVR (CH4 voltage) at the output end of the circuit to provide the input current measurement, is this correct PM? (CH2-CH4 also provides for the voltage across Rload for Pout). Pin is computed via CH1 x CH4?
btw PM, does the sim show COP>1? Also, is the Pin result the same in the sim regardless if i(in) is taken at L1 or R1?
poynt99,
PM has stated earlier that both the current probe measurement taken just behind the CH1 (yellow) point should
give the same value as the CVR (1 ohm / 10 ohm) measurement.
This was confirmed by me and viewable in this post here:
http://www.overunityresearch.com/index.php?topic=3655.msg69766#msg69766
Purple is the CVR (10 Ohm in my case) trace, while green is the current probe trace.
In keeping the circuit replicable for many, i think PM has opted for the CVR only.
Itsu
Quote from: poynt99 on 2018.09.30, 01:48:04
I'm a little confused by your post Itsu, and I'm not 100% clear on PM's revised schematic.
What I see on the schematic, is an input voltage measurement (CH1, YEL), and in the previous schematic there was a current probe on the input. It appears the revised schematic is now using a CVR (CH4 voltage) at the output end of the circuit to provide the input current measurement, is this correct PM? (CH2-CH4 also provides for the voltage across Rload for Pout). Pin is computed via CH1 x CH4?
btw PM, does the sim show COP>1? Also, is the Pin result the same in the sim regardless if i(in) is taken at L1 or R1?
Poynt,
An LtSpice simulation yields similar results as a cloudy, rainy day in my location that is, less than unity efficiency but tweaking the sim frequency can yield efficiencies in the 98-99% range however.
Pm
I would like to add several points that I have discovered:
1) During "optimal" days with clear skies and sunshine, placing a tinfoil covered cardboard box shield over the circuitry actually improved the COP!?
2) I have found that even on cloudy, rainy days, very careful tuning downwards in frequency to achieve the lowest frequency possible prior to the sudden increase in the voltage across C1 can yield apparent COP's>1. This may not always be the case but at present our skies are overcast and the humidity is 100% with no rain falling and I can see COPs in the 1.15-1.25 range.
3) An increase in the inductance of L1 seems to soften or broaden the critical low frequency "break over" point.
Pm
All,
A test I've wanted to run for some time but have put off until now turns out to be quite interesting! It simply involves replacing the parametric capacitance device with a fixed capacitor paralleled by a schottky diode as shown below in the schematic. This scheme is now just "resonant" charging and seems to be able to provide smaller amounts of gain with tweaking.
I would like to add that this test was run immediately following a scope input re-calibration on the Tek MDO.
The advantage of this resonant charging verses the parametric version is it's ability to produce a greater voltage across C1 than the supply resulting in higher output power levels. The disadvantage at this point appears to be a much lower available gain.
The scope pix shows the input power pin to be 373.6mw mean and pout = 54.99^2/7.451e3 = 405.8mw. The apparent COP = 405.8/373.6 = 1.086 . This is low enough to be measurement error but I think it is worth pursuing to attempt higher gains and at lower frequencies.
Regards,
Pm
All,
Well, perhaps just plain resonant charging has some merit! The following is a charge sweep (for lack of a better term) of C1 over 100ms using the circuitry in the attached schematic seen below. L1 and C2 have been increased in value which results in an operating frequency of 62kHz.
The scope shot below shows the measurements at the 30.05ms mark in the charge sweep and from this we can determine the energy levels. The output voltage across C1 reaches 50.41 volts so Uout = 50.41^2 * 3.94e-6/2 = 5.006mJ . The input power required to raise C1 to this level is seen to be 104.8mw which equates to an input energy of Uin = 104.8e-3 * 30.05e-3 = 3.147mJ.
Therefore, the apparent COP = 5.006/3.147 = 1.59 .
If this is reasonably accurate and can be replicated, there should theoretically be no lower limit in frequency nor any limit in power output.
Regards,
Pm
Edit: I will add that the scope was in the hi-res mode (16 bit) with the sweep rate at 100Ms/S.
I like the simplified design.
Trying to catch up; would you kindly explain: "The scope pix shows the input power pin to be 373.6mw mean ".
Also, what is the current determined from the 1-ohm R1 and how is it used?
Hi Pm,
The question is how this latest setup behaves (input and output energy wise) when a real load is applied across C1?
Because at present there is no load and the stored energy in C1 is 'idle'. I would compare it to a flywheel which has no mechanical load on it, just a small amount of friction and / or air drag.
Gyula
Quote from: PhysicsProf on 2018.09.30, 20:10:26
I like the simplified design.
Trying to catch up; would you kindly explain: "The scope pix shows the input power pin to be 373.6mw mean ".
Also, what is the current determined from the 1-ohm R1 and how is it used?
The answer to both your questions is, the current measured by R1 or the CSR on CH4(grn) is multiplied by the voltage on CH1(yel) at every instantaneous point in time and summed to produce the average or mean power integral of 373.6mw which is the power input to the circuit.
Pm
Quote from: gyula on 2018.09.30, 21:08:26
Hi Pm,
The question is how this latest setup behaves (input and output energy wise) when a real load is applied across C1?
Because at present there is no load and the stored energy in C1 is 'idle'. I would compare it to a flywheel which has no mechanical load on it, just a small amount of friction and / or air drag.
Gyula
Gyula,
It is true there is no load and your flywheel analogy is close but not quite accurate. In this case, the energy to spin the flywheel up to a given speed is less than the energy contained in the spinning flywheel at that given speed.
IOW, we spend less energy to charge C1 than the energy contained in C1. We could use this circuit in continuous charge/discharge cycles to do work with gain.
I will do an analysis of the charge sweep in order to determine the ideal load resistance that will produce the maximum COP running continuously and will post the results later.
Pm
All,
Here is an analysis of the circuit shown in the schematic in post #288.
The first scope pix shows the charge sweep which is averaged over 8 samples during a 100ms horizontal sweep at a 100Ms/S sweep rate. The vertical cursors are placed at .905ms and 20.07 ms respectively resulting in a dt of 19.17ms. The respective voltage levels across C1 at these positions are 36.85v and 49.45v therefore, the resultant dU or differential energy in C1 between the cursors is (49.45^2-36.85^2) * 3.94e-6/2 = 2.142mJ. The equivalent power for this time period is then 2.142e-3/19.17e-3 = 111.7mw .
The mean or average voltage across C1 between the cursors is seen to be 45.22v so the ideal load resistance across C1 for continuous running would be 45.22^2/111.7e-3 = 18.3k ohm.
We also see the input power consumed between the cursors at 99.28mw so the gain is 111.7/99.28 = 1.125 . We would expect this to be the gain in continuous running at 62kHz with an 18.3k load.
The next scope pix shows the circuit running continuously with an 18.51k load across C1 which was the closest value on hand. As can be seen , the output voltage across C1 and the load is 45.69v resulting in a pout = 45.69^2/18.51e3 = 112.8mw . The Math channel calculation shows the pin to be 94.05mw resulting in an apparent COP = 112.8/94.05 = 1.20 which is higher than predicted from the charge sweep. I'm not sure why the difference at this point in time.
Note the close correlation between the charge sweep analysis and the continuous running data. IMO, this shows the importance of running charge sweeps to determine the performance and optimized load requirements for given circuit parameters. From what TK has stated, Rigol scopes have the capability of 50Ms/S scan rates for 100ms sweeps and should be greater for faster sweep rates. The important thing is to capture the charge slope so it can be analyzed.
Regards,
Pm
Looks promising but my trust in dso's higher functions has been low ever since the Rosemary incident ,many thanks to Poynt on that one.
Im not suggesring that you are in error ,your insights have got me off my arse !
All parts comming ..2 weeks away ...
I wonder if changing the coil topology ( like 22turns at 18" diameter,300uh ) has any effect on the results?
Does having the semis in the centre as opposed to outside make tuning easier?
3D
Quote from: 3D Magnetics on 2018.10.01, 23:02:31
Looks promising but my trust in dso's higher functions has been low ever since the Rosemary incident ,many thanks to Poynt on that one.
Im not suggesring that you are in error ,your insights have got me off my arse !
All parts comming ..2 weeks away ...
I wonder if changing the coil topology ( like 22turns at 18" diameter,300uh ) has any effect on the results?
Does having the semis in the centre as opposed to outside make tuning easier?
3D
L,
It wasn't the DSO or it's higher functions that became the nemesis for the RA team, it was their lack of understanding of the over all picture pertaining to the apparatus they were testing/measuring. Their scope performed flawlessly, but the assumptions made regarding their measurement points and methods is what ultimately led them down the garden path.
Quote from: 3D Magnetics on 2018.10.01, 23:02:31
...
I wonder if changing the coil topology ( like 22turns at 18" diameter,300uh ) has any effect on the results?
...
3D
Interesting question. 18" diameter is huge...
I wonder if you (PM) could show us a photo of your coil or inductor, with a ruler or something to show the size?
Quote from: poynt99 on 2018.10.02, 02:02:28
L,
It wasn't the DSO or it's higher functions that became the nemesis for the RA team, it was their lack of understanding of the over all picture pertaining to the apparatus they were testing/measuring. Their scope performed flawlessly, but the assumptions made regarding their measurement points and methods is what ultimately led them down the garden path.
Let's not forget the massive amount of outright lying that was emitted by RA.
If she had simply acknowledged the truth from the beginning --- that the original 555 circuit did not produce the duty cycle she thought it did but rather the exact inversion --- nobody would have been interested at all. But her lies percolated and even infected the principals at EF (not hard to do).
Sorry about the "off topic" but whenever I see Ainslie's folly blamed simply on measurement misunderstandings and errors... I have to point out the deliberate lying and obfuscation that contributed greatly to the "Ainslie" phenomenon.
I haven't heard from her in many months. I wonder if she's still alive. Lots of people aren't.
For those that have some time to tinker, and based on PM's research, I submit this first draft of a stand alone looped device with self oscillating feature.
I have not got the time to test it on the bench but am sure it can be made to self oscillate if the coupled inductor gain is sufficient.
The IXD601 device data sheet states a max logic input current of only 10 uA,so the input drive should not load the output coil very much.
Whether it will work without the starter circuit as a self powered unit is another matter.
Quote from: 3D Magnetics on 2018.10.01, 23:02:31
Looks promising but my trust in dso's higher functions has been low ever since the Rosemary incident ,many thanks to Poynt on that one.
Im not suggesring that you are in error ,your insights have got me off my arse !
All parts comming ..2 weeks away ...
I wonder if changing the coil topology ( like 22turns at 18" diameter,300uh ) has any effect on the results?
Does having the semis in the centre as opposed to outside make tuning easier?
3D
3D,
Hmmm, this would be an interesting experiment using a large diameter coil for L1 and placing a higher frequency parametric circuit in the center! I will try this and focus on one of Dr. Stiffler's lattice formula frequencies.
Regards,
Pm
Quote from: PhysicsProf on 2018.10.02, 02:04:47
Interesting question. 18" diameter is huge...
I wonder if you (PM) could show us a photo of your coil or inductor, with a ruler or something to show the size?
Sure, here is a pix of the coils used thus far. The air cores are used in the parametric circuit and the smaller ferrite cored inductor is used in the lower frequency resonant charging circuits.
Pm
Quote from: ion on 2018.10.02, 11:36:01
For those that have some time to tinker, and based on PM's research, I submit this first draft of a stand alone looped device with self oscillating feature.
I have not got the time to test it on the bench but am sure it can be made to self oscillate if the coupled inductor gain is sufficient.
The IXD601 device data sheet states a max logic input current of only 10 uA,so the input drive should not load the output coil very much.
Whether it will work without the starter circuit as a self powered unit is another matter.
Ion,
Thanks for the looping circuit. The newer resonant or more accurately off-resonant charging circuit would be an excellent candidate for attempted self running if the switching circuitry can be made efficient enough.
Pm
All,
This is a test of the circuit shown in post #288 running continuously with the frequency adjusted to 60.8kHz which then produces a voltage across loaded C1 that is slightly greater than the 40v dc supply voltage.
In the scope pix, we see the output voltage of 40.64v across C1 with a 7.451k ohm load resulting in a pout = 40.64^2/7.451e3 = 221.7mw. The pin is seen to be 195.5mw for an apparent COP = 221.7/195.5 = 1.13. This may be enough gain to achieve self looping but I'm not sure if the Ixd60x series is efficient enough though.
Also to clarify, the circuit is not operating at resonance but rather off-resonance to a higher than resonance frequency or inductive side. This is evidenced by the input current lag in regards to the input pulse seen on CH1(yel). It will also operate on the lower than resonance frequency or capacitive side but at a lower efficiency. At resonance, it is very inefficient.
Pm
Quote from: partzman on 2018.10.02, 14:17:42
Sure, here is a pix of the coils used thus far. The air cores are used in the parametric circuit and the smaller ferrite cored inductor is used in the lower frequency resonant charging circuits.
Pm
Thanks for the pix of the coils you have used - very nice!
All,
This is the same test run in post #302 but now operating at 23.6kHz or the low side off-resonance. Contrary to what I stated previously, this is actually more efficient!
In the scope pix we see the voltage across C1 with a 7.451k load is 40.27v resulting in a pout = 40.27^2/7.451e3 = 217.6mw. Pin is 187.1mw for a COP = 217.6/187.1 = 1.16.
Based on the current waveform, this may yet be another operating mode!
Pm
Edit: IMO at these low operating frequencies, the Rigol scopes should easily be able to accurately measure a 100ms charge sweep.
This is remarkable - running at just 23.6 kHz.
Just to double-check, in these runs - are you using the smaller ferrite-cored inductor?
Thanks again!
Quote from: PhysicsProf on 2018.10.02, 23:39:25
This is remarkable - running at just 23.6 kHz.
Just to double-check, in these runs - are you using the smaller ferrite-cored inductor?
Thanks again!
Yes, these low frequency tests use a 1/4" square ferrite E core with a 0.020" gap as they must operate as linearly as possible. IMO, there is no limit to the lower frequency that will work except possibly ELFs.
Attached is an example of lower frequency with a charge sweep and continuous run of a low off-resonance (LOR) circuit at 16.5kHz with L1 = 2.46mH, C1 = 3.94uf, and C2 = .01uf.
The first pix is the charge sweep measured over 100ms showing the C1 voltages with the A and B cursors. The energy differential is (58.28^2-43.03^2) * 3.94e-6/2 = 3.044mJ which equates to a pout = 3.044e-3/99e-3 = 30.7mw. The input power is seen to be 13.1mw resulting in an apparent COP = 30.7/13.1 = 2.34.
With the average voltage across C1 being 52.91v, we can now calculate the required load resistor RL = 52.91^2/30.7e-3 = 91.2k ohm.
The next scope pix shows the continuous run with a 90.7k (closest value) load resistor at 16.5kHz. The output voltage across C1 with the load is 53.30v so pout = 53.3^2/90.7e3 = 31.3mw. The pin is 12.21mw so the apparent COP = 31.3/12.21 = 2.56. Again note the correlation between the sweep and continuous data.
The downside to this mode is the low power which would be improved by higher operating voltages but perhaps there is a better way.....?
Regards,
Pm
I'll be doing at least 1 replication, but, as these things go, will likely build half a dozen once I see one do something O0
There are a bunch of questions though.
Is the 40V input based on raising the Pin and Pout to levels that can be more easily discerned as > COP 1 ?
Could some type of high secondary turns Joule Thief achieve similar end results, with a starting voltage much lower and bumping it to the 40V ?
A similar question for the looping thoughts.
Which components dictate the resonant frequency of the circuit ?
If it's the central WIMA cap bank, what happens if another brand or value of cap is used ?
Can these IXYS MOSFET drivers be used ? https://www.ebay.com/i/162811331846
Otherwise, the type that Itsu has unfortunately witnessed burning are about $7 each at the cheapest.
Is a MOSFET driver needed ?..apologies for my lack of knowledge on this section, if it's merely Gate drive then surely there are passive routes rather than spendy chips.
There was talk of close coupling within the coil windings, but i'm seeing large single winding crystal radio type coils in pics (for want of better wording). Could the coils be explained ?
Many thanks for any answers, have read all 13 pages but those keep niggling.
Quote from: Slider2732 on 2018.10.03, 14:39:19
I'll be doing at least 1 replication, but, as these things go, will likely build half a dozen once I see one do something O0
There are a bunch of questions though.
Is the 40V input based on raising the Pin and Pout to levels that can be more easily discerned as > COP 1 ?
No, the >1 COPs will be seen at lower voltage levels if the circuit is working properly. I run with 40v dc supply because it yields the highest power output which makes the scope measurements more accurate IMO. The Ixys drivers are actually rated with a max of 35v dc so they a little stressed at this level especially at the higher frequencies.
Quote
Could some type of high secondary turns Joule Thief achieve similar end results, with a starting voltage much lower and bumping it to the 40V ?
I don't see why not as long as it produces a square wave with reasonably fast rise and fall times. Sine waves will not work.
Quote
A similar question for the looping thoughts.
Which components dictate the resonant frequency of the circuit ?
If it's the central WIMA cap bank, what happens if another brand or value of cap is used ?
In the resonant charging circuit it is L1 and C2 that determine the resonant frequency. L1 should be a linear inductor and C2 can be any good quality film cap. The storage cap C1 can be any good quality capacitor and I just prefer the film caps over 'lytics.
Quote
Can these IXYS MOSFET drivers be used ? https://www.ebay.com/i/162811331846
Otherwise, the type that Itsu has unfortunately witnessed burning are about $7 each at the cheapest.
They have output fault protection which might get tripped when the circuits attempt to supply energy back to the power supply but otherwise they should work just fine.
Quote
Is a MOSFET driver needed ?..apologies for my lack of knowledge on this section, if it's merely Gate drive then surely there are passive routes rather than spendy chips.
There was talk of close coupling within the coil windings, but i'm seeing large single winding crystal radio type coils in pics (for want of better wording). Could the coils be explained ?
Actually any type of mosfet gate drive will work and I hope to do more investigating into this to allow looping if possible. At present there is only one inductor used but there could be many possible combinations of windings that would step up the drive voltage, provide full wave operation, and have L1 integrated in the leakage inductance. I don't think this is needed at this point to attempt looping with the LOF circuit, just efficient switching of the input pulses.
Pm
Quote
Many thanks for any answers, have read all 13 pages but those keep niggling.
All,
It seems that power output is not a problem with the LOR (low frequency off-resonance) mode it is just a matter of give and take. The lower the power out the higher the COP and vice-versa.
The first scope pix show the same circuit as previously shown except with a load of 3.71k ohm and operating at 16.3kHz with the supply at 40vdc. CH3(pnk) shows the dc voltage feeding the circuitry for comparison to the CH3(blu) showing the voltage across C1 with load. CH2 should be reading in mean rather than rms but with the dc voltage it is the same so we see 40.44v across the 3.71k load resistor for a pout = 40.44^2/3.71e3 = 441mw. The pin seen in the Math channel is 418mw for an apparent COP = 441/418 = 1.055. If we increase the load resistance, pout will decrease and the COP will increase so it is a matter of one's choice.
The highest apparent COP that can be obtained with the LOR is at 1/2 the resonant frequency or fo/2. The resonant frequency is determined by L1 and C2 so in this case fo = 1/((2.46e-3 * .01e-6)^.5 *2*3.14) = 32.08kHz. So, fo/2 = 16.04kHz.
The second scope pix shows the results of the identical circuit running at 16.0kHz with a 90.7k load at 40vdc. The output voltage is 40.0vdc so pout = 40^2/90.7e3 = 17.6mw. The input is seen to be -3.115mw however, due to the very small amount of current being measured during the pulse "on" time, the power measured by the Math channel is rather ambiguous so it will vary between what is seen and ~+2.0mw. It still equates to a rather high apparent COP.
The power can also be increased for a given load by increasing the frequency towards resonance but the output voltage across C1 will also increase. These tests were run with the loaded output across C1 to be slightly higher than the power supply voltage.
Also noteworthy is that the output voltage self regulates quite well in the LOR with changes in load but of course there are limits.
Regards,
Pm
Many thanks for the prompt answers Pm :)
I think i'm best ordering at least one of the spendier IXYS chips, but also to order the much cheaper versions as well.
As a side note, i've used IXYS 6V 22mA solar panels extensively over the past few years and know them to be a consistent high quality manufacturer.
PM: "The second scope pix shows the results of the identical circuit running at 16.0kHz with a 90.7k load at 40vdc. The output voltage is 40.0vdc so pout = 40^2/90.7e3 = 17.6mw. The input is seen to be -3.115mw however, due to the very small amount of current being measured during the pulse "on" time, the power measured by the Math channel is rather ambiguous so it will vary between what is seen and ~+2.0mw. It still equates to a rather high apparent COP."
Yes! O0
All,
This is a change in topology whereby the circuit connections are inverted that is, the common side of R1 and C1 is now connected to the positive supply rail instead of ground. This affects the circuit operation greatly by presenting the full resonant current to the positive supply. In the former common ground connection of R1 and C1, the resonant current is seen only by ground during the driving pulse "low" period.
The new schematic is shown below and it should be noted that the P-channel mosfet M2 operates in both reverse and normal conduction modes to allow energy to be fed back to the power supply. It should also be noted that the only energy not accounted for here is that required to drive the mosfet gates.
The scope pix shows the waveforms and a Tek current probe is used for the current measurements although a CSR could have been used as well. For clarity, CH3(pnk0 is the supply voltage and CH2(blu) is the voltage on the negative side of C1. The sum of the magnitudes of these two voltages is the total presented to the load. So, pout = (29.67+16.7)^2/7.451e3 = 288.6mw. The power consumed from the positive supply is seen to be 203.9mw for an apparent COP = 288.6/203.9 = 1.42.
Regards,
Pm
EDIT: There is a measurement error in the above so I will update as soon as the info is corrected.
EDIT2: The current probe position in the bench test was actually on the right side of mosfet M2 which gave erroneous measurements that included the recirculating current in M2 so the measurements stated are incorrect and the actual COP<1.
Intriguing progress, PM - awaiting also you update.
I do have a question about the new circuit - is it still the case that " The lower the power out the higher the COP and vice-versa"?
Quote from: PhysicsProf on 2018.10.04, 20:12:41
Intriguing progress, PM - awaiting also you update.
I do have a question about the new circuit - is it still the case that " The lower the power out the higher the COP and vice-versa"?
I'm still sorting out the details but the newest circuit has slightly different operating rules compared to the previous.
Pm
All,
In my previous post #312, I have added a second edit which explains the source of measurement error for that topology that when corrected gives a COP<1.
Instead of pushing ahead at this time, I will wait to see if any more replications similar to Itsu's come forward.
Regards,
Pm
PM,
In earlier tests you had found varying results based on something ,possibly humidity but local conditions are playing a part in some way.
You may need to have a reference of some kind, possibly the first apparent high cop circuit which remains un touched as a base line .
I apolagise for stating the obvious but a reference to a known point will help .After all, this all shoud be impossible right?
SM always pointed to rectifier tubes as an important element ,switching to mosfets after results were "what he wanted"
The best clues I have gleaned, relate to interaction between the local magnetic field and rectifier which may be what you have .
Whatever the case turns out to be , I thank you on behalf of all who are influenced for your dilligence, aptitude and pragmatisim even skepticisim.
3D
Which topology showed the greatest gain irrespective of operating frequency.
I would think the best approach is to try and use IONs looping circuit to run the best topology, at least the gain can be measured by the amount or lack of hopefully supply power, even if a pot was used with a marked dial to turn up or down the input power a graph could be plotted where the pot needs setting to maintain a set frequency or load, this approach could also be used to study different drive designs, each improvement would be measured against the previous design.
Quote from: 3D Magnetics on 2018.10.06, 01:57:24
PM,
In earlier tests you had found varying results based on something ,possibly humidity but local conditions are playing a part in some way.
You may need to have a reference of some kind, possibly the first apparent high cop circuit which remains un touched as a base line .
I apolagise for stating the obvious but a reference to a known point will help .After all, this all shoud be impossible right?
SM always pointed to rectifier tubes as an important element ,switching to mosfets after results were "what he wanted"
The best clues I have gleaned, relate to interaction between the local magnetic field and rectifier which may be what you have .
Whatever the case turns out to be , I thank you on behalf of all who are influenced for your dilligence, aptitude and pragmatisim even skepticisim.
3D
3D,
I totally understand what you are saying and agree 100%. I do have the elements of the original configurations and run them occasionally for comparison but have not documented the results other than what has been posted. My inherent problem is that I tend to look too far "forward" sometimes ending up down a path of distraction!
Thank you for your most kind comments!
Pm
Quote from: Peterae on 2018.10.06, 12:17:20
Which topology showed the greatest gain irrespective of operating frequency.
I would think the best approach is to try and use IONs looping circuit to run the best topology, at least the gain can be measured by the amount or lack of hopefully supply power, even if a pot was used with a marked dial to turn up or down the input power a graph could be plotted where the pot needs setting to maintain a set frequency or load, this approach could also be used to study different drive designs, each improvement would be measured against the previous design.
Peter,
The parametric charging circuits using mosfets for the non-linear element seem to show the largest apparent gains especially when driven with an increasing frequency sweep. The question then is if the amount of output power verses gain is sufficient to attempt looping. There are two types of running for these circuits and that is charge/discharge of a capacitance and continuous.
The topology easiest to loop is the off-resonant type with it's low apparent gains due to the convenience that the output voltage can easily be => the supply voltage. IOW, the looping circuit is simply achieved by replacing C1 with the power supply. I have tried this and not experienced self-running to date but I'm not using the most efficient switching circuits at this point in time. More experimentation needs to be done in this area.
Your suggestion of a variable load is an excellent idea for those unable to generate a charge sweep over time. I'm sure there are math relationships with various circuit values that would help determine the ideal operating conditions and loads but I'm unable to supply those unfortunately.
IF all else fails, I can see some value in having a simple, extremely efficient boost converter with reasonable regulation at low power levels. C.C
Pm
Hi partzman
Thanks for the reply, if only I had some free time, I am chomping at the bit to have a go at this device.
I am surprised there's not more people having a go.
Regards
Peter
Thanks, I'm gathering parts.
PM -- is the L or size/ferrite of the smaller ferrite-cored coil critical?
Quote from: PhysicsProf on 2018.10.08, 02:04:49
Thanks, I'm gathering parts.
PM -- is the L or size/ferrite of the smaller ferrite-cored coil critical?
I would not recommend a ferrite cored toroid as it may reach saturation which would reduce efficiency. Nearly any gapped ferrite core will work and powdered iron toroids as well. The inductance of L1 could be in the area of what I used and is not critical. Good luck in your replication!
Pm
Same here, though via the slow boat.
My question is also ferrite related. Have got a couple of CRT yoke ferrites, the big jobbies. I gather that a piece of paper between the halves and a number of windings will suffice ?
But the thinking is about litz and whether it's of use. Especially if the strands are separated and connected as per Romero's device from years ago. Such a method produces a 'super capacitance' coil and might be useful, rather than a single layer winding.
Quote from: Slider2732 on 2018.10.08, 15:06:08
Same here, though via the slow boat.
My question is also ferrite related. Have got a couple of CRT yoke ferrites, the big jobbies. I gather that a piece of paper between the halves and a number of windings will suffice ?
But the thinking is about litz and whether it's of use. Especially if the strands are separated and connected as per Romero's device from years ago. Such a method produces a 'super capacitance' coil and might be useful, rather than a single layer winding.
The CRT yoke ferrites should work with the paper providing the gap.
I have tried litz split into two windings for L1 which does result in a relatively large self capacitance but it didn't seem to perform well at all in the circuit.
Pm
Ah ok, was just a wondering about that circuit section.
Thanks for having tried it O0
Thank you for your replies, Partzman - very helpful as we approach replication.
One more question comes to mind - Itsu used an air core iirc; I'm thinking also of your photo in reply #300... Would you recommend an air core or gapped-ferrite core for a 1st attempt at replication?
Quote from: PhysicsProf on 2018.10.09, 03:05:43
Thank you for your replies, Partzman - very helpful as we approach replication.
One more question comes to mind - Itsu used an air core iirc; I'm thinking also of your photo in reply #300... Would you recommend an air core or gapped-ferrite core for a 1st attempt at replication?
If one is attempting to replicate the parametric charging circuits like Itsu has done, then I would recommend using
only some form of air cored coil for L1. I have tried various ferrite cored inductors and even with identical values to the air cored coils, they all degrade the performance. I really don't know why at this point but perhaps this has something to do with the gain changes with ambient changes or the gain mechanism itself!
Our weather currently is rainy and has been for four or five days and the best I can muster with the parametric circuits is about a 5% gain with very careful tuning. Most of the time however, the gains are <1.
It is fine to use ferrite cored coils if replicating the off resonant circuits but air coils may be used as well.
Pm
All,
I think I have a breakthrough regarding the inconsistency of the COP measurements. The value of C1 in all my previous parametric charging tests was 3.94uf but by reducing this value to .047uf, even in the very wet, rainy conditions we currently have, I can measure apparent COPs in the 2-3 range. The smaller value for C1 makes for much easier tuning in that the hysteresis is greatly reduced or non existent.
The values used in the circuit that produced the scope shot attached below are L1=374uH, C1=.047uf,Rl=7.451k,CSR=1ohm,f=619.1kHz,Vsply=30vdc, and the parametric element is 2-paralled RFP14N05Ls with a parallel MBR1100 schottky diode for forward conduction.
The output voltage is seen to be 16.64v so pout = 16.64^2/7451 = 37.2mw. The measured pin on the Math channel is 12.96mw so the apparent COP = 37.2/12.96 = 2.87.
I have asked Itsu to try this change so we'll hopefully see what his results are.
Regards,
Pm
All,
This is an update on the reduction of the value of C1 and it's effect on the parametric charging operation.
The first scope pix is the same circuit as the previous post except C1 has now been changed from .047uf to 4700pf. Here we see the output voltage across RL to be 17.42v mean so pout = 17.42^2/7451 = 40.7mw. The pin = 14.1mw for an apparent COP = 40.7/14.1 = 2.89.
With this change in C1 to 4700pf, more operating modes have become apparent. The next scope pix shows the result of lowering the frequency past the first stable point where the output voltage is slightly greater than 1/2 the supply voltage, thru the unstable portion until a second stable condition is met where the output voltage across the load is now slightly greater than the supply voltage. Here we see the output voltage is 30.74v mean so pout = 30.74^2/7451 = 126.8mw. Pin is seen to be 104.4 for an apparent COP = 126.8/104.4 = 1.21.
There is one other change in this circuit for this shot and that is L1 = 314uH wound on type 2 powdered iron core. Also, CH3(pnk) shows the resonant voltage on the input side of the RFP14N05Ls.
Weather conditions here are still very wet and rainy.
Regards,
Pm
Quote from: partzman on 2018.10.09, 16:17:16
All,
I think I have a breakthrough regarding the inconsistency of the COP measurements. The value of C1 in all my previous parametric charging tests was 3.94uf but by reducing this value to .047uf, even in the very wet, rainy conditions we currently have, I can measure apparent COPs in the 2-3 range. The smaller value for C1 makes for much easier tuning in that the hysteresis is greatly reduced or non existent.
The values used in the circuit that produced the scope shot attached below are L1=374uH, C1=.047uf,Rl=7.451k,CSR=1ohm,f=619.1kHz,Vsply=30vdc, and the parametric element is 2-paralled RFP14N05Ls with a parallel MBR1100 schottky diode for forward conduction.
The output voltage is seen to be 16.64v so pout = 16.64^2/7451 = 37.2mw. The measured pin on the Math channel is 12.96mw so the apparent COP = 37.2/12.96 = 2.87.
I have asked Itsu to try this change so we'll hopefully see what his results are.
Regards,
Pm
OK,
changed the setup to C1 = 47nF (was 3.3uF), both MOSFETs are paralleled by an extra one and schottky's still in place (bat 42).
L1 is still 322uH air coil, Rload is still 8140 Ohm, 30V on the driver chip, frequency 542Khz.
Screenshot shows an input power of 27.72mW, and Voltage across Rload is 16.33V (backed up be Fluke DMM) for an output power of 32.7mW.
This gives a cop of 1.18, so not very high, but given the fact that the original setup still gave cop <1, its better.
I see another update from PM for some more tests.
Itsu
Quote from: Itsu on 2018.10.09, 20:11:33
OK,
changed the setup to C1 = 47nF (was 3.3uF), both MOSFETs are paralleled by an extra one and schottky's still in place (bat 42).
L1 is still 322uH air coil, Rload is still 8140 Ohm, 30V on the driver chip, frequency 542Khz.
Screenshot shows an input power of 27.72mW, and Voltage across Rload is 16.33V (backed up be Fluke DMM) for an output power of 32.7mW.
This gives a cop of 1.18, so not very high, but given the fact that the original setup still gave cop <1, its better.
I see another update from PM for some more tests.
Itsu
Itsu,
Thanks for running this test so quickly! Do I understand that you have 3 RFP14N05Ls in parallel? Also, I don't see any cursors in your scope shot so are you measuring from screen edge to screen edge?
Regards,
Pm
All,
The change in lowering the value of C1 did affect the efficiency of the resonant circuits that are non-parametric. To be clear, both the parametric charge circuits and the high/lo off-resonant circuits utilize resonance. The parametric are obviously non-linear and the off-resonant are linear.
Below is a revised schematic shown in post #288 that has C1 reduced in value plus other changes. This is the circuit used in the test of the scope pix that is also attached.
As can be seen, the output voltage across the 18.51k load is 31.58v mean which is slightly higher than the 30vdc supply voltage. Therefore, pout = 31.58^2/18.51e3 = 53.9mw. The Math channel shows pin = 32.03mw for an apparent COP = 53.9/32.03 = 1.66 which is much better performance than previously achieved.
IMO, this is the preferred circuit for replication and it offers the direct connection to the supply for looping and the frequency is low. However, the newer version of the parametric charger also will provide a mode that reaches an output level greater than the supply with apparent gain, but the tuning is tricky.
Regards,
Pm
All,
This is an LOR or low off-resonance test of the same circuit as the previous post except the frequency is 26.6kHz or slightly above Fo/2.
From the scope shot we see the output voltage across the 18.51k load is 31.42v mean which is again slightly above the 30vdc power supply voltage. Therefore, pout = 31.42^2/18.51e3 = 53.3mw. Pin is 28.83mw for an apparent COP = 53.3/28.83 = 1.85. Again, much better performance than was previously achieved and also a preferred circuit to replicate.
Regards,
Pm
PM wrote: "Therefore, pout = 31.42^2/18.51e3 = 53.3mw. Pin is 28.83mw for an apparent COP = 53.3/28.83 = 1.85. Again, much better performance than was previously achieved and also a preferred circuit to replicate."
Again, wow! O0 Great progress!
Quick question, above L1= 2.47 mH, is that one of the air-core coils you showed in previous photo?
Quote from: partzman on 2018.10.09, 20:37:39
Itsu,
Thanks for running this test so quickly! Do I understand that you have 3 RFP14N05Ls in parallel? Also, I don't see any cursors in your scope shot so are you measuring from screen edge to screen edge?
Regards,
Pm
PM,
no, that was with 2 MOSFETs parallel, so in total 4 MOSFETs.
No cursors setup, so yes, i am measuring from screen edge to screen edge.
I could try the between cursors measurement to see if it gives some more repeatable results.
Itsu
Very nice. My parts are ordered. Would also like to know more about the coil. Be interesting to see what it does down here. Thanks for sharing
Quote from: PhysicsProf on 2018.10.10, 06:49:20
PM wrote: "Therefore, pout = 31.42^2/18.51e3 = 53.3mw. Pin is 28.83mw for an apparent COP = 53.3/28.83 = 1.85. Again, much better performance than was previously achieved and also a preferred circuit to replicate."
Again, wow! O0 Great progress!
Quick question, above L1= 2.47 mH, is that one of the air-core coils you showed in previous photo?
No, L1 in these low frequency circuits is the small 1/4" ferrite E-cored inductor shown in the photo along with the air coils. Again, any ferrite core used in these circuits must be gapped to provide a constant inductance. If un-gapped or a toroid, the inductance will decrease with increasing H field and can reduce the gain.
Pm
Quote from: Itsu on 2018.10.10, 08:43:25
PM,
no, that was with 2 MOSFETs parallel, so in total 4 MOSFETs.
No cursors setup, so yes, i am measuring from screen edge to screen edge.
I could try the between cursors measurement to see if it gives some more repeatable results.
Itsu
Itsu,
OK, thanks for clarifying that. You might try just 2-14N05Ls with the reduced C1 to see if your COP goes any higher. I think there is a relationship between the "mean" change in the capacitance of the mosfets and the optimum value of C1 and I'm working on that today.
I also think using your cursors will give more consistent results. You should be seeing a higher COP than your last results.
One thing I haven't mentioned along the way is that the 14N05Ls I'm using are actually NOS from IR and were manufactured in the late 80's! We don't know if there have been any die changes since then or not.
Pm
Quote from: JimBoot on 2018.10.10, 09:36:56
Very nice. My parts are ordered. Would also like to know more about the coil. Be interesting to see what it does down here. Thanks for sharing
For the parametric circuits with the mosfets, L1 is not that critical really and could be in the range of 300uH to 700uH for starters. I have used air core for most of the work but have recently tried powdered iron in type -2 which worked just fine. Powdered iron based coils should have a slight increase in inductance with an increasing H field which would tend to "soften" and stabilize the tuning point. Haven't tried any of the other types yet.
Pm
Quote from: partzman on 2018.10.10, 14:15:10
Itsu,
OK, thanks for clarifying that. You might try just 2-14N05Ls with the reduced C1 to see if your COP goes any higher. I think there is a relationship between the "mean" change in the capacitance of the mosfets and the optimum value of C1 and I'm working on that today.
I also think using your cursors will give more consistent results. You should be seeing a higher COP than your last results.
One thing I haven't mentioned along the way is that the 14N05Ls I'm using are actually NOS from IR and were manufactured in the late 80's! We don't know if there have been any die changes since then or not.
Pm
Thanks PM,
have removed the extra MOSFETs again, but still on the 47nF C1.
Measuring input now "between cursors" (yellow over 6 cycles).
I used the below new driver setup as presented by Peterae earlier instead of the IXDD614 driver chip.
Seems to work OK up to 30V with 5V (TTL) in.
But as can be seen in the screenshot, the input 45mW is high compared to the output 16.2V² / 8140 = 32mW
Power Supply used shows 30V @ 3mA = 90mW total input.
Not sure where the other 45mW or so is going, the driver MOSFETs are at room temperature.
Signals look cleaner though, probably due to the slower rise/fall times of the square input wave.
Will experiment with old driver setup and lower C1 now.
Itsu
Itsu and All,
I may have discovered a measurement problem that could be the source of gain in all the previous circuits. I am not absolutely sure yet but I am still researching the fact that any offset of the current channel greatly affects the Math channel calculations. It doesn't matter if a current probe is used or a CSR, ANY offset appears to determine if one has a COP>1. I will post as soon as I have confirmation of this or not.
Regards,
Pm
PM,
thanks for that info, would be nice to know what is going on, good or bad.
Meanwhile i went back to the IXDD614 drive setup but now with the even more reduced C1 (4700pF)
Using 2 MOSFETs only (no paralleling) and still the Bat 42 schottky's.
Rload still 8140 Ohm and 322uH aircoil for L1 and 30V (@ 30mA) from the PS.
The screenshot shows the result, input power calculations 38.89mW, and output 17.92² / 8140 = 39.4mW, so practical a COP = 1.
Itsu
Quote from: partzman on 2018.10.10, 14:21:39
For the parametric circuits with the mosfets, L1 is not that critical really and could be in the range of 300uH to 700uH for starters. I have used air core for most of the work but have recently tried powdered iron in type -2 which worked just fine. Powdered iron based coils should have a slight increase in inductance with an increasing H field which would tend to "soften" and stabilize the tuning point. Haven't tried any of the other types yet.
Pm
Thanks PM. I may truy printing something with my Iron PLA then. Ta.
Quote from: Itsu on 2018.10.10, 19:42:27
PM,
thanks for that info, would be nice to know what is going on, good or bad.
Meanwhile i went back to the IXDD614 drive setup but now with the even more reduced C1 (4700pF)
Using 2 MOSFETs only (no paralleling) and still the Bat 42 schottky's.
Rload still 8140 Ohm and 322uH aircoil for L1 and 30V (@ 30mA) from the PS.
The screenshot shows the result, input power calculations 38.89mW, and output 17.92² / 8140 = 39.4mW, so practical a COP = 1.
Itsu
Itsu,
Well, it is as you say "good or bad" and in this case I have concluded it is "bad and good"! ???
Bad in the fact that all of our previous COP results have been directly related to the offset in the scope channel measuring the voltage across the CSR or the offset of our current probes plus the scope channel.
Here is my analysis of the situation as I currently see it. The measured mean current by the CSR or current probe should equal the dc voltage across C1 divided by the load resistance. Let's take you last post #342 for example. We see that your output voltage is 17.92v mean and your load resistance is 8140 ohms so the mean current measured should be 17.92/8140 = 2.20ma but in fact is 1.864ma mean. This means that when corrected, the COP would be <1.
Now let's look at my test in post #328 that had a really good COP. Here we see the dc output is 16.64v with a 7.451k load so the mean current should be 16.64/7451 = 2.233ma but it is shown to be >138.1ua. This is the reason the input power is so low which gives us the high COP. When I check all my positive OU test results, I find that the measured mean current is less than the required amount. I have carefully (as possible) re-run some of my previous tests using a separate power supply and current resistor to trim the static offset as close to zero as possible and have concluded that the parametric circuits as they stand
do not exhibit OU.
If we were to take a look at your first OU replication of the parametric circuit in post #257, we will find similar results. I also think that at least all my variations that were coincident with weather variations were no more than current channel offset variations. I might add that doing scope re-cals didn't really give zero offset voltages plus the offset varies too much to be accurate at these low currents.
Now for the
good! In you test setup, remove C1 completely and tune for a resonant output waveform across the load resistor. Measure the rms of this waveform, calculate the output, and compare to the input power. What is your result?
Pm
PM,
thanks for this info, and good observation / thinking about the "real" current.
I remember some statements that scopes are not suitable to be used as a voltmeter, so this confirms it once again.
I normally used a Fluke DMM to confirm the Rload voltage and it was practically always the same as presented
by the scope, but that was in 10 - 30 volts range, not millivolts.
Perhaps the weather, humidity etc. could have effected the scopes ability to more or less accurate measure the
voltages which then produces these different outcomes we have seen.
Concerning your GOOD section, i removed C1 and tuned for resonance (1.76Mhz) across Rload, see screenshot.
Rload rms voltage is 28.55V, this squared and divided by 8140 gives 100mW.
Input as calculated by the scope as normal gives 128.9mW.
We still see that same current difference (calculated 3.5mA versus probe 2.9mA) as you explained above.
Itsu
Quote from: Itsu on 2018.10.11, 09:15:21
PM,
thanks for this info, and good observation / thinking about the "real" current.
I remember some statements that scopes are not suitable to be used as a voltmeter, so this confirms it once again.
I normally used a Fluke DMM to confirm the Rload voltage and it was practically always the same as presented
by the scope, but that was in 10 - 30 volts range, not millivolts.
Perhaps the weather, humidity etc. could have effected the scopes ability to more or less accurate measure the
voltages which then produces these different outcomes we have seen.
Concerning your GOOD section, i removed C1 and tuned for resonance (1.76Mhz) across Rload, see screenshot.
Rload rms voltage is 28.55V, this squared and divided by 8140 gives 100mW.
Input as calculated by the scope as normal gives 128.9mW.
We still see that same current difference (calculated 3.5mA versus probe 2.9mA) as you explained above.
Itsu
Itsu,
Thanks for doing the "Good" test! O0 Hopefully something can be salvaged from all this work for your sake and others who have ordered parts for replication as we are not 'dead in the water' yet.
This new circuit is actually a
Parametric
Series
Resonance with diode limiting and a series connected load. Later I hope to show a PSR with a parallel load and no diode limiting.
Anyway, in order to be able to determine if your current measurement is accurate (has any offset), we need to have the load voltage also measured in 'mean' volts along with the rms. This way we can determine the correct mean current by dividing the mean output voltage by the load resistance. Correction can then be made with a small current injection of the proper polarity in the CSR as shown in the attached schematic.
I would recommend using a 10 ohm CSR and then correct your current channel readings. The MDO scope allows this to be done internally with the input gain amplifiers so the offset is reduced by the gain reduction.
The attached scope shot the results of my test with the values shown in the schematic. The mean CSR current should be 15.14/7451 = 2.032ma and is measured to be 2.027ma for an error of -0.25%. We will correct pin by this amount so pin(cor) = 43.13mw * 1.0025 = 43.24mw. Pout = 22.05^2/7451 = 65.3 for an apparent COP = 65.3/43.24 = 1.51. This is good but not as easy to work with for looping.
I haven't tried removing the schottky diode across the mosfets as of this post, but the performance may actually be improved due to the fact that whatever diode is used limits the resonance so, a higher conduction level may improve the performance.
Regards,
Pm
Edit: Also notice that we are no longer dealing with a large Pin (reactive)/Pout (real) ratio. This means that we can use a lower bandwidth on the current channel without losing accuracy due to slight phase changes.
Edit2: Replaced schematic with corrected version-left out RL!
Itsu and all,
The PSR is load sensitive in regards to COP as can be seen in the attached scope pix. The required CSR mean current should be 20.6/14810 = 1.391ma but measures 1.398 so we are OK here and won't apply any correction. Pout = 42.34^2/14.81e3 = 121mw and with pin at 48.94mw, the apparent COP = 121/48.94 = 2.47.
I have tried pairs of various mosfets and they all produce COPs>1 dependent on load and tuning. Also, the efficiency is best with the schottky rectifier in place.
Pm
PM,
i have redone my above test and included the Rload mean voltage, see screenshot below.
But what i notice in your screenshot is that the yellow voltage and green current are not in phase, so
you are not on full resonance, right?
Anyway, my mean Rload voltage is 16.29V, so current through Rload is 16.29 / 8140 = 2mA
Input power math shows 126.4mW using the measured 2.405mA mean current.
2mA is 83.16% of 2.405mA, so input power corrected is then 126.4 x 0.8316 (83.16%) = 105.11mA
Cop is then 105.11 / 100.6 (28.62² / 8140) = 0.96.
Need some improvement i guess, or do i miss something here like "a small current injection of the proper polarity in the CSR"?
Itsu
ooops, lost another ixdd614, 1.7Mhz @ 30V with heatsink is still to much after some time.
Trying to get the "vertical offset" to work, but it does not seem to function in the mean value box nor the math
function picks it up (the signal trace does go up or down with the offset value).
Where is my scope manual?
Itsu
Quote from: Itsu on 2018.10.11, 20:12:56
ooops, lost another ixdd614, 1.7Mhz @ 30V with heatsink is still to much after some time.
Trying to get the "vertical offset" to work, but it does not seem to function in the mean value box nor the math
function picks it up (the signal trace does go up or down with the offset value).
Where is my scope manual?
Itsu
Itsu,
IIRC, adjusting the channel offset will not affect the Math calcs.
Pm
Itsu and all,
Here is a variation of the PSR that places the mosfets back-to-back, removes the schottky rectifier, and still retains the series load as seen in the schematic attached below. Ch3(pnk) shows the resonance waveform after L1 for reference.
In the scope shot we see the offset current of 22.52ua is on the high side so we will ignore. Pout = 42.72^2/14.81e3 = 123.2mw. With pin = 15.9mw, the apparent COP = 123.2/15.9 = 7.75.
Regards,
Pm
Quote from: partzman on 2018.10.11, 21:44:51
Itsu and all,
Here is a variation of the PSR that places the mosfets back-to-back, removes the schottky rectifier, and still retains the series load as seen in the schematic attached below. Ch3(pnk) shows the resonance waveform after L1 for reference.
In the scope shot we see the offset current of 22.52ua is on the high side so we will ignore. Pout = 42.72^2/14.81e3 = 123.2mw. With pin = 15.9mw, the apparent COP = 123.2/15.9 = 7.75.
Regards,
Pm
Damn!
Quote from: partzman on 2018.10.11, 21:44:51
Itsu and all,
Here is a variation of the PSR that places the mosfets back-to-back, removes the schottky rectifier, and still retains the series load as seen in the schematic attached below. Ch3(pnk) shows the resonance waveform after L1 for reference.
In the scope shot we see the offset current of 22.52ua is on the high side so we will ignore. Pout = 42.72^2/14.81e3 = 123.2mw. With pin = 15.9mw, the apparent COP = 123.2/15.9 = 7.75.
Regards,
Pm
That is a remarkable COP - how confident are you?
Would you say that this is your best design so far (despite the high freq)?
Also, what is the purpose of the variable pwr supply connected to R2?
Thank you for sharing as you proceed - and more power to you!
Quote from: partzman on 2018.10.11, 21:44:51
Itsu and all,
Here is a variation of the PSR that places the mosfets back-to-back, removes the schottky rectifier, and still retains the series load as seen in the schematic attached below. Ch3(pnk) shows the resonance waveform after L1 for reference.
In the scope shot we see the offset current of 22.52ua is on the high side so we will ignore. Pout = 42.72^2/14.81e3 = 123.2mw. With pin = 15.9mw, the apparent COP = 123.2/15.9 = 7.75.
Regards,
Pm
Hmmm, "when it sounds too good to be true, it probably is" is the saying i believe :).
Calculating the real current by 0.1773V / 14810 Ohm = 11.97uA (scope measures 22.52uA, so almost double).
This means the the real input power is about 50% less then the calculated 15.9mW meaning about 8mW.
Holding this input against the calculated output of 42.72^2/14.81e3 = 123.2mw means a COP of 15.4, Right? :D
If so, then its too good to be true in my opinion.
Do i understand it OK, that you are not using the correction PS at the csr?
Concerning this correction PS, could you clarify what it should do / does and what it consists of?
Will try to replicate tonight.
Itsu
Quote from: PhysicsProf on 2018.10.12, 01:41:53
That is a remarkable COP - how confident are you?
Would you say that this is your best design so far (despite the high freq)?
Also, what is the purpose of the variable pwr supply connected to R2?
Thank you for sharing as you proceed - and more power to you!
I hope today to run some tests that will raise the confidence level (my own included) of these recent tests. They are what they are and I certainly welcome any critical input as it is easy to overlook something.
This design is possibly the best as far as utilizing the parametric capacitance change in a mosfet but I'm thinking that it may go deeper than that. What we have is a very low Q series resonant circuit due to a relatively large value load resistance operating with a large amount of harmonics. Within the capacitance of this resonant circuit, we have diode decoupling at the zero current crossover point that in itself adds non-linearity and IMO has the effect of de-stabilizing the resonance effect. With tests today, I hope to find out if substituting the mosfets with regular caps produces similar results due to the reasoning in my previous statement.
The purpose of the power supply connected to R2 which is connected to the high side of the CSR, is to allow cancellation of the offset voltage in the scope's input amplifiers. These circuits in general at this time are operating at such low current levels that this offset becomes a major factor affecting the accuracy of the math measurements. This could also be accomplished with a pot connected between a + and - supply with the wiper connected to R2 which basically allows us to supply a small amount of current to the CSR creating the required voltage offset change.
Pm
Quote from: Itsu on 2018.10.12, 11:07:32
Hmmm, "when it sounds too good to be true, it probably is" is the saying i believe :).
Yes, and that saying may very well apply here. Further testing will hopefully tell us what we need to know.
Quote
Calculating the real current by 0.1773V / 14810 Ohm = 11.97uA (scope measures 22.52uA, so almost double).
This means the the real input power is about 50% less then the calculated 15.9mW meaning about 8mW.
Holding this input against the calculated output of 42.72^2/14.81e3 = 123.2mw means a COP of 15.4, Right? :D
If so, then its too good to be true in my opinion.
In this case and with this circuit, the offset voltage/current is not as critical as in the previous high ratio reactive/real circuits. The change in the micro amps of offset will not materially affect the Math calculations as they previously did because of the relatively small phase angle between the driving pulse and the circuit current. So, for this reason I just neglected the small offset however, I could be wrong in this regard.
IMO, mathematically correcting the offset as you suggested would not affect the COP.
Quote
Do i understand it OK, that you are not using the correction PS at the csr?
In the last posted scope pix no, I did not have the power supply turned on for any correction.
Quote
Concerning this correction PS, could you clarify what it should do / does and what it consists of?
My previous explanation was not the best so we'll see if I can get it right this time. I use a single variable supply that I adjust to provide the proper voltage level to create a current in R2 that is fed to the CSR to correct the "mean" current/voltage in the scope's current channel and therefore correct the scope's offset voltage. The magnitude of the "mean" current is dictated by the circuitry and could be zero or any other level. I change the polarity by simply flipping the leads to the supply.
A far better method would be to use a pot like I described to PhysicsProf.
Actually, the load could be used as the CSR in this case as this is the singular path for the current. However, this will require a simple formula to be written in the scope's math channel calculator which I know the TDS series does not have this capability. I am not sure about the Rigol's.
Quote
Will try to replicate tonight.
Great and good luck. O0
Pm
Quote
Itsu
All,
Here is a re-run of the previous "E" version test with the high COP showing two methods of calculating the pin. It should be noted that I have two pairs of RFP14N05L's paralleled for this and the previous test.
The first scope pix is using the same measurement across the CSR as in the previous test using CH1 * CH4 which shows the pin = 14.47mw. Pout = 40.68^2/14810 = 111.74mw for an apparent COP = 111.74/14.47 = 7.58.
The second scope pix uses the math formula pout = (CH2/Var1) * CH1 with Var1 = 14.81k. IOW, this uses the entire load resistance as the sense resistor and as can be seen, the pin = 14.76mw for an apparent COP = 111.74/14.76 = 7.57.
I would point out the horizontal resolution is at 1M points and the measurements are taken over 32 averaged samples. Also, varying the mean current in the first version by 10's of microvolts has seemingly little to no effect on the pin.
Regards,
Pm
OK, using your setup as in Post #351 with: 1mH air coil as L1, 14930 Ohm Rload (R3) and 10 Ohm csr at 40V.
No extra power supply at the csr.
I now see a big phase difference when using the current probe in its old position (before L1) or its new
position (between R3 and csr).
It causes the power input to change from 135mW to 6mW or so.
Probes (colors) are similar as in your drawing.
Strange is that the blue probe mean voltage is negative (mean voltage across R3).
So i cannot correct the current to its real value (i = -0.583 / 14930 = -39uA)
Input power calculated is 6.214mW
Power output is 39.12² / 14930 = 102.5mW
Using the current probe in the new position.
Not sure whats wrong with this.
By the way, i am onto my last IXDN614PI as the new batch IXDD614PI's i got from China have a problem as i already
blew up 3 of them.
looks like the are internally other/wrong connected as the old IXDN614PI i now got in works fine.
It seems these new IXDD614PI's i got cannot withstand any Vcc above 27V or so.
As the spec's say they could be used up to 40V Vcc (which we did without problems with the old ones), i guess
these are inferior parts from a 3th party manufacturer.
I have notified the seller about this.
Itsu
All,
The Mystery is solved and unfortunately there is no OU in the "E" series circuitry. I found it odd that by adding more mosfets in parallel, the capacitance increased but the resonant frequency did not. With the high load resistance, the current took the path of least resistance which happens to be the scope probe capacitance and the parasitic capacitance between components and ground. It is the energy of the parasitic current that is not being accounted for on the low side current measurements. The proof of this is when a current probe is used to measure the current from the pulse source and the Math channel uses this in the pin calcs, the COP <1.
So, I will continue to search for any gains using the mosfets but at this time, things look somewhat discouraging.
Regards,
Pm
Quote from: Itsu on 2018.10.12, 16:15:20
OK, using your setup as in Post #351 with: 1mH air coil as L1, 14930 Ohm Rload (R3) and 10 Ohm csr at 40V.
No extra power supply at the csr.
I now see a big phase difference when using the current probe in its old position (before L1) or its new
position (between R3 and csr).
It causes the power input to change from 135mW to 6mW or so.
Probes (colors) are similar as in your drawing.
Strange is that the blue probe mean voltage is negative (mean voltage across R3).
So i cannot correct the current to its real value (i = -0.583 / 14930 = -39uA)
Input power calculated is 6.214mW
Power output is 39.12² / 14930 = 102.5mW
Using the current probe in the new position.
Not sure whats wrong with this.
By the way, i am onto my last IXDN614PI as the new batch IXDD614PI's i got from China have a problem as i already
blew up 3 of them.
looks like the are internally other/wrong connected as the old IXDN614PI i now got in works fine.
Itsu
Itsu,
Thanks for your efforts in this replication! I was creating my post during the time you posted so we crossed each other but basically your measurement with the current probe at the very front is the correct one for the reason stated in my last post.
At this frequency, the parasitic capacitance paths can obviously greatly affect the measurements. I should have checked for this before posting that version so to you and all the others here, I apologize. :-[
Regards,
Pm
Partzman,
thanks for the info and no apologies needed, its all in this game.
I learned again a lot, so anytime you have something special, please put it out there for us replicators
to follow up.
Already looking forward to the next one O0
Itsu
Quote from: partzman on 2018.10.12, 16:16:42
All,
The Mystery is solved and unfortunately there is no OU in the "E" series circuitry. I found it odd that by adding more mosfets in parallel, the capacitance increased but the resonant frequency did not. With the high load resistance, the current took the path of least resistance which happens to be the scope probe capacitance and the parasitic capacitance between components and ground. It is the energy of the parasitic current that is not being accounted for on the low side current measurements. The proof of this is when a current probe is used to measure the current from the pulse source and the Math channel uses this in the pin calcs, the COP <1.
So, I will continue to search for any gains using the mosfets but at this time, things look somewhat discouraging.
Regards,
Pm
Hi Partzman,
It is very good you have found the explanation for the high COP for the 'E' series circuit. Would like to add the followings from an answer I was preparing the draft version but some other commitments prevented posting it earlier than your post. Perhaps other readers may find them useful.
"Because your latest setup (Reply #351) operates at 1.3 MHz frequency, the real impedance of the load resistor, R3 is questionable, unless it has a very, very low self inductance and self capacitance i.e. a non-inductive type (albeit in the increasing MHz range even the some nanoHenries start affecting the DC resistance of resistors).
Further, a 4 pF capacitor your blue probe puts across R3 has 30.6 kOhm reactance.
Further, a simple parallel RC of 14.81 kOhm and the 4 pF results in an impedance of Re(Z)=12 kOhm + Im(Z)= -5.8 kOhm at 1.3 MHz and the phase angle Ang(Z)= -25.8 degree." I used this calculator here:
http://www.cirvirlab.com/simulation/parallel_r_c_circuit_impedance_calculator.php (http://www.cirvirlab.com/simulation/parallel_r_c_circuit_impedance_calculator.php)
Also, I wanted to suggest to make a parallel LC circuit for 1.3 MHz and put it across R3, this would have tuned out any reactance in R3 itself and the ill effect of the 4 pF scope probe capacitance. By tuning the LC to 1.3 MHz or to whatever input frequency the setup needs, the voltage across this LC (and hence across R3) would be in phase with the current estimated by R1 10 Ohm csr, so output power could have been also calculated. And what should also be possible, your scope math function could be used for multiplying CH2 by CH4 to get output power after your input power already become known.
Lastly I assume that now the IXDD609 had a higher input current consumption from the 40V DC source than say in the some ten kHz range you used earlier. Was this so I wonder?
Anyway, thanks for sharing and no need for apology, just continue and have fun.
Gyula
PM - thanks for your diligence and honesty in quickly reporting the mistake. We all make them.
And for your continuing research.
"The Mystery is solved and unfortunately there is no OU in the "E" series circuitry. "
Does this apply to the earlier circuits as well? (Like those studied at <100 kHz )
Quote from: gyula on 2018.10.12, 18:33:16
Hi Partzman,
It is very good you have found the explanation for the high COP for the 'E' series circuit. Would like to add the followings from an answer I was preparing the draft version but some other commitments prevented posting it earlier than your post. Perhaps other readers may find them useful.
"Because your latest setup (Reply #351) operates at 1.3 MHz frequency, the real impedance of the load resistor, R3 is questionable, unless it has a very, very low self inductance and self capacitance i.e. a non-inductive type (albeit in the increasing MHz range even the some nanoHenries start affecting the DC resistance of resistors).
I was using an array of 15k MOX (metal oxide film) resistors for the loads in these tests and I'm sure they do have some inductance at the higher frequencies.
Quote
Further, a 4 pF capacitor your blue probe puts across R3 has 30.6 kOhm reactance.
Further, a simple parallel RC of 14.81 kOhm and the 4 pF results in an impedance of Re(Z)=12 kOhm + Im(Z)= -5.8 kOhm at 1.3 MHz and the phase angle Ang(Z)= -25.8 degree." I used this calculator here:
http://www.cirvirlab.com/simulation/parallel_r_c_circuit_impedance_calculator.php (http://www.cirvirlab.com/simulation/parallel_r_c_circuit_impedance_calculator.php)
Yes, even this low value capacitance probe did affect the measurements in the MHz range.
Quote
Also, I wanted to suggest to make a parallel LC circuit for 1.3 MHz and put it across R3, this would have tuned out any reactance in R3 itself and the ill effect of the 4 pF scope probe capacitance. By tuning the LC to 1.3 MHz or to whatever input frequency the setup needs, the voltage across this LC (and hence across R3) would be in phase with the current estimated by R1 10 Ohm csr, so output power could have been also calculated. And what should also be possible, your scope math function could be used for multiplying CH2 by CH4 to get output power after your input power already become known.
These are both good suggestions.
Quote
Lastly I assume that now the IXDD609 had a higher input current consumption from the 40V DC source than say in the some ten kHz range you used earlier. Was this so I wonder?
The Ixys 6xx series drivers when used to drive a continuous load in the MHz range have large internal losses and heat up really quick and will self destruct if unchecked. In the kHz range not so much, so the power drawn from the supply is considerably less. We (Itsu and I) were abusing them as well by operating them at 40vdc when they are rated at 35v max.
Quote
Anyway, thanks for sharing and no need for apology, just continue and have fun.
Gyula
Thanks,
Pm
Quote from: PhysicsProf on 2018.10.13, 02:50:07
PM - thanks for your diligence and honesty in quickly reporting the mistake. We all make them.
And for your continuing research.
"The Mystery is solved and unfortunately there is no OU in the "E" series circuitry. "
Does this apply to the earlier circuits as well? (Like those studied at <100 kHz )
I think the answer to your question is "yes" but just in case, I'm taking time to re-evaluate the various topologies to be absolutely sure.
Pm
Thanks.
Have just checked which drivers I have on order and they're the IXDD version.
Thoughts may well be correct of them being inferior, particularly from the 1:1 swapout.
Pm - It would be awesome to discover the best topologies use <20V in the low frequency ranges !
But am certainly in agreement with others. Your routing out of possible causes for effects is appreciated. Any new circuit design is taken a lot more seriously than from someone who would cover over mistakes O0
All,
I have done what I consider to be extensive testing of all the topologies presented in this thread and have concluded that unfortunately, none have the ability to produce COPs>1. I have learned that as good as our scopes are these days, they have their limits when measuring low level waveforms.
The Tek MDO series has the ability to scale the input levels if the probe is used for current measurements which in addition to being convenient means the channel offset is reduced by the scaling factor used. So, I used 20 ohm CSRs to measure the circuit current which meant the scope amplifier was adjusted to 50ma/volt reducing the offset by 1/20. This allowed consistent and relatively accurate low level current measurements to be taken which then resulted in measurements indicating COPs<1.
Thanks to all who participated in this effort and I truly wish the outcome had been different!
Regards,
Pm
Thanks Partzman
Been a very interesting project, even though the end result is COP <1 there's a wealth of information and data been produced which people can learn from for future use.
Quote from: partzman on 2018.10.15, 15:44:22
...
Thanks to all who participated in this effort and I truly wish the outcome had been different!
Regards,
Pm
Hi Pm,
I also thank all your efforts and willingness to share.
I hope Brad is ok and would also share his findings unless he also finds his scope has played some tricks with him.
Gyula
Quote from: gyula on 2018.10.16, 21:47:15
Hi Pm,
I also thank all your efforts and willingness to share.
I hope Brad is ok and would also share his findings unless he also finds his scope has played some tricks with him.
Gyula
If i calculate the instantaneous power measured by the scope,and divide by duty cycle,i get consistent COPs of 180-190 %
If i set the scope up as TK showed me,where ch1 is set to voltage,and ch2 is set to amps,the scope calculated watts in is within +/- 1% of the total calculated P/out.
So,i do not know if the manually calculated COPs are correct,and the scope is making a mistake when making it's calculation's,or the scope calculated P/in is correct,and manual calculations cannot be used in this system ?.
I have always thought that if the current and voltage traces are exactly in phase,and a very clean square wave,that instantaneous power calculated,divided by duty cycle, always gives you the correct answer. But in this case,the scope calculations seem to disagree.
As we know the laws today,it would seem that the scope's calculations are correct,and all is as it should be.
On a side note,i have been working on a little something else.
Who here has one of the !drinking birds! ?.
Brad
Hi Brad,
Well, I mainly meant the "inherent" problem Partzan's Tektronix scope manifested as he described in Reply #368, hopefully in the Rigol scopes there are no "inherent surprises" to lead the user astray.
The message perhaps is that too good results measured by a scope need to be checked by other means too or find ways to circumvent the scope shortcomings. Often, this is not easy.
Gyula
OK, this is a re-visitation of the basic FWB version of the Parametric Charger that was first disclosed in posts #9 and #16 on the first page of this thread. From there, things sort of got off track and I have myself to blame due to my suggested "improvements"!
This inline version does not suffer from any capacitive coupling across transformer windings plus this test does not use a current probe but rather a 10 ohm Caddock precision film CSR with a RF probe connections on a 10M-4pf probe that is setup and calibrated for 100.0ma/volt.
Previously, attempts were made with this technology to run continuously with a given resistive load. Although this might yield slight gains, the best performance is to operate this type of device in an open rather than closed circuit. IOW, we allow the parametric capacitance to charge a fixed capacitor to a given level that requires a certain input energy. We would then discharge this capacitor into a suitable load to realize any gain. This process would be repeated for a useful OU device.
The schematic below shows the circuit used. The main difference from the original is the removal of C2 the series capacitor and the addition of the CSR. The circuit is basically a full wave bridge driven by a high frequency square wave pulse that charges the film capacitor C1. The top of C1 indicated by the plus sign is the most positive terminal during operation. The substrate diodes for each mosfet provide one means of charging C1 while the reverse bias parametric capacitance of the mosfet's drain-gate/source junctions provide the other means. At the correct frequency, L1 in conjunction with the parametric capacitance, provide a reactive input means that effectively reduces the input energy requirement. This is the gain mechanism for this type of device.
The first scope pix shows the Pin of 22.29mw over 44.1ms which allows us to calculate the input energy with 22.29e-3*44.1e-3 = 983uJ.
The second scope pix shows the scan for the voltage charging of C1. In this case we use the beginning voltage level of 25.54v and the ending voltage level of 44.81v to calculate the energy in C1 which is (44.81^2-25.54^2)*3.94e-6/2 = 2.67mJ. The apparent COP = 2.67e-3/.983e-3 = 2.71 .
These measurement are consistent and the next round will involve testing from a biased state in C1 although some testing has already been done in this area.
Regards,
Pm
Hi Partzman,
I've always found this line of thinking very intriguing. I think it was in about 1991 that I tried to get extra energy out of varactors but I just didn't know enough electronics to do it. (Don't know much more now!)
I can see that this version doesn't suffer from the capacitive coupling and offset issues that plagued your earlier tests.
Here's me being dumb-- what's the purpose of C1? Wouldn't the presence of fixed capacitance just lower the overall change in C and reduce the parametric amplification?
And for that matter, although the mosfets have obvious advantages over varactors, the varactors can have C ratios > 15, inherently giving more bang for the buck, I think.
It's good to be back!
Fred
Quote from: Orthofield on 2021.08.05, 18:32:00
Hi Partzman,
I've always found this line of thinking very intriguing. I think it was in about 1991 that I tried to get extra energy out of varactors but I just didn't know enough electronics to do it. (Don't know much more now!)
I can see that this version doesn't suffer from the capacitive coupling and offset issues that plagued your earlier tests.
Here's me being dumb-- what's the purpose of C1? Wouldn't the presence of fixed capacitance just lower the overall change in C and reduce the parametric amplification?
And for that matter, although the mosfets have obvious advantages over varactors, the varactors can have C ratios > 15, inherently giving more bang for the buck, I think.
It's good to be back!
Fred
Hi Fred,
Welcome back!
C1 serves as the storage load device for the parametric capacitors in the mosfets.
Yes, varactors would be interesting to examine.
Regards,
Pm
Hi Partzman,
Yes, I realized that C1 was the load after I wrote that. You had said as much in your description.
Fred
Hi partzman,
I'm surprised there hasn't been any comments about your research. I know from reading the thread about your previous tests that there were several errors that arose, but your present design takes those issues into account, and it's difficult to see how the results can be error. Of course it's always possible! But until an error is found, there's room for some excitement here!
Maybe it's not as glamorous and esoteric as some other projects, but that begs the question, are we looking for something that works, or something that's complicated and fraught with perplexities?
I searched long and hard and haven't ever found any detailed description of conservation of energy as applied to parametric processes. As far as it goes is the analogy to a parallel plate capacitor and the work needed to move the plates. But the diffusion layer in a semiconductor is a very different animal.
Fred
Hello partzman,
with MOSFETs instead of diodes please consider the following:
you increase new unknown parameters, that is 3 different capacitances and their dependence of the current, voltage, temperature and their mutual interactions which makes this setup more complex.
You like it that way eh ;) ?
https://www.rohm.com/electronics-basics/transistors/understanding-mosfet-characteristics (https://www.rohm.com/electronics-basics/transistors/understanding-mosfet-characteristics)
Miller-Capacitance here:
https://electronics.stackexchange.com/questions/83712/gate-capacitance-and-miller-capacitance-on-the-mosfet#83730 (https://electronics.stackexchange.com/questions/83712/gate-capacitance-and-miller-capacitance-on-the-mosfet#83730)
I am sure you know this but that is not the main point I wish to make, its the MOSFET-Driver which is my concern.
The MOSFET-Driver has been developed in order to accelerate the shutdown-time by removing the gate-charge as fast as possible. This charge is part of you power-in measurement and lost at the end of each single pulse.
What if we could recuperate this charge ? ...or we use another puls-methode which does not remove the gate-charge so at least this parameter and its interaction with the Miller capacitance is stable ?
And what would then be the value of your Input-Power ?
Mike
Hi Mike,
That's a very good point! I'd thought of this when fiddling with varactors, but never connected it here.
These gate charge recovery circuits do exist. A few minutes looking turned this up:
https://ieeexplore.ieee.org/document/885437
which is behind a paywall, but I'm pretty sure I have some circuits for charge recovery in my files somewhere.. looking now...
Fred
Hi,
Here are some gate charge energy recovery patents:
https://patentimages.storage.googleapis.com/17/26/cc/eaab36d23912cf/US7598792.pdf
https://patentimages.storage.googleapis.com/25/92/94/203187440eebc4/US6441652.pdf
https://patentimages.storage.googleapis.com/ea/ff/62/a795c6debce485/US7453292.pdf
https://patentimages.storage.googleapis.com/8b/f3/a4/8e12efcbcd0b71/US7285876B1.pdf
I was trying to find some non resonant circuits but so far they all rely on resonance and work at around 1 Mhz I think..
Fred
Hi Partzman,
This patent might be of interest. It's a Mosfet Parametric Amplifier with a charge pump type topology using two batteries, two switches and two mosfets. It doesn't include charge recovery.
https://patents.google.com/patent/US20050275026?oq=tsividis+mosfet+parametric#citedBy
Fred
Quote from: Kator01 on 2021.08.07, 21:00:48
Hello partzman,
with MOSFETs instead of diodes please consider the following:
you increase new unknown parameters, that is 3 different capacitances and their dependence of the current, voltage, temperature and their mutual interactions which makes this setup more complex.
You like it that way eh ;) ?
Hi Mike. Sorry for the late reply but we had business to take care of this weekend so I apologize. I totally agree that mosfet capacitances are complex so in this device I attempted to reduce that complexity to the basic drain capacitance with the gate and source shorted. There is however another reverse capacitance and that is bulk or substrate diode which varies with the particular mosfet geometry. Anyway, in themselves, these capacitances normally are all conservative with charge/discharge cycles that yield ~95% efficiency.
Quote
https://www.rohm.com/electronics-basics/transistors/understanding-mosfet-characteristics (https://www.rohm.com/electronics-basics/transistors/understanding-mosfet-characteristics)
Miller-Capacitance here:
https://electronics.stackexchange.com/questions/83712/gate-capacitance-and-miller-capacitance-on-the-mosfet#83730 (https://electronics.stackexchange.com/questions/83712/gate-capacitance-and-miller-capacitance-on-the-mosfet#83730)
I am sure you know this but that is not the main point I wish to make, its the MOSFET-Driver which is my concern.
The MOSFET-Driver has been developed in order to accelerate the shutdown-time by removing the gate-charge as fast as possible. This charge is part of you power-in measurement and lost at the end of each single pulse.
What if we could recuperate this charge ? ...or we use another puls-methode which does not remove the gate-charge so at least this parameter and its interaction with the Miller capacitance is stable ?
And what would then be the value of your Input-Power ?
Actually with the circuit as shown, L1 presents an inductive reactance of ~1600 ohms at 687kHz. The result is a series resonant circuit with the "C" being provided by the parametric drain capacitance. The operating frequency is higher than the normal resonance frequency resulting in a lagging current to voltage on the input. This is critical for the circuit to be able to produce OU and as I have found, the measurement of the slow sweeps require the highest sampling frequency available on the scope in order to be accurate and consistent.
You have offered some possible suggestions which would be interesting to test.
Regards,
Pm
Quote
Mike
Quote from: Orthofield on 2021.08.08, 17:58:06
Hi Partzman,
This patent might be of interest. It's a Mosfet Parametric Amplifier with a charge pump type topology using two batteries, two switches and two mosfets. It doesn't include charge recovery.
https://patents.google.com/patent/US20050275026?oq=tsividis+mosfet+parametric#citedBy
Fred
Hi Fred,
Thanks for posting the patent links and this one by Tsividis is most interesting. I attempted to replicate his circuitry in the past with no success and then I realized that the mosfet devices he used had a separate connection for the bulk or substrate diode and was not connected to the source. If you look at the ambiguity of the connections in Fig 6 and also the connections in Fig 8, it will become apparent then looking at all the Figs that the substrate diode is connected separately from the source. This is a critical requirement for his device to work but unfortunately, most all mosfets today have the substrate connected to the source and they will not work.
In fact at the time, I couldn't locate any such devices but they must exist.
Regards,
Pm
Hi Partzman,
Check these datasheets out, one of these may work for a Tsvidis rep:
http://www.aldinc.com/pdf/ALD1115.pdf
http://www.aldinc.com/pdf/ALD1105.pdf
http://www.aldinc.com/pdf/ALD1121E.pdf
http://ww1.microchip.com/downloads/en/DeviceDoc/mic94050.pdf
also ..30 ..31 .. 51 in this family.
https://datasheet.octopart.com/MC14007UBCPG-ON-Semiconductor-datasheet-531527.pdf
https://www.nxp.com.cn/docs/en/data-sheet/BSS83_N.pdf
Fred
Hi Partzman,
I didn't see your response to Mike above, but if I understand you correctly, your present circuit already has a sort of gate charge recovery?
Fred
Quote from: Orthofield on 2021.08.09, 18:53:26
Hi Partzman,
Check these datasheets out, one of these may work for a Tsvidis rep:
http://www.aldinc.com/pdf/ALD1115.pdf
http://www.aldinc.com/pdf/ALD1105.pdf
The above devices should work.
Quote
http://www.aldinc.com/pdf/ALD1121E.pdf
http://ww1.microchip.com/downloads/en/DeviceDoc/mic94050.pdf
also ..30 ..31 .. 51 in this family.
These appear to have the substrate connected to the source.
Quote
https://datasheet.octopart.com/MC14007UBCPG-ON-Semiconductor-datasheet-531527.pdf
This part appears to have the substrate connected to source.
Quote
https://www.nxp.com.cn/docs/en/data-sheet/BSS83_N.pdf
This part should work.
[/quote]
Fred
[/quote]
LtSpice appears to have seven different types of monolithic or 4-terminal mosfets. This may provide an easy way to test the concept if the models are accurate. We'll see!
Regards,
Pm
Quote from: Orthofield on 2021.08.09, 19:19:49
Hi Partzman,
I didn't see your response to Mike above, but if I understand you correctly, your present circuit already has a sort of gate charge recovery?
Fred
Hi Fred,
Actually, the present circuit is not a gate charge/recovery type. If this were the case, there would be no gain as can be shown by experiment and simulation as it would be conservative even though the elements are parametric.
This device works via off-resonance (above resonance) with the input current lagging the input voltage that produces a resultant lower input energy than output energy. IOW, it is best described as reactive to reactive converter where the output reactive energy is greater than the input reactive energy.
The following is a modification of the original and performs at a more consistent level. The schematic is shown first and it can be seen that effectively we have C1 being charged on the positive input pulse by the bulk or substrate diodes in M1-M4.
The parametric capacitance provides the resonant "C" with L1 and at the proper frequency, this resonant circuit produces the parametric increase in the voltage across C1. Without the parametric capacitance change in the mosfet drain to gate/source capacitance or the correct frequency, the voltage across C1 would simply reach a level equal to the supply voltage minus two diode drops and remain there. It is obvious the voltage increases in C1 above the supply voltage during the scan.
The 1st scope pix is the Pin scan which shows 8.117mw consumed over 64.76ms resulting in an input energy level of 8.117e-3*64.67e-3 = 525uJ .
The 2nd scope pix shows the start and finish voltage change in C1 which results in an output energy level of (36.01^2-26.42^2)*3.94e-6/2 = 1179uJ for an apparent COP = 1179/525 = 2.25 .
The 3rd scope pix is an expanded view of the Pin scan and shows the basic cyclic waveforms.
I would like to point out that this scan was biased as seen in the schematic and therefore represents a fairly accurate result one could expect from a continuously operating device. Also, the vertical resolution of these measurements is 16 bits and the horizontal sweep rate is 100 mega samples/second.
I will also add that when running continuously at the proper frequency wherein C1 reaches a stabilized voltage level, the Pin will be negative due to the reactive voltage/current phasing.
The drawback to this device at present is the overall low power.
Regards,
Pm
PM
QuoteThe parametric capacitance provides the resonant "C" with L1 and at the proper frequency, this resonant circuit produces the parametric increase in the voltage across C1. Without the parametric capacitance change in the mosfet drain to gate/source capacitance or the correct frequency, the voltage across C1 would simply reach a level equal to the supply voltage minus two diode drops and remain there. It is obvious the voltage increases in C1 above the supply voltage during the scan.
Interesting circuit, I see a series capacitor-inductor and mosfet switching on either side of the cap but I can't see a clear mechanism for gain.
However I can follow your line of reasoning that we could add/subtract series capacitance to vary a parameter ergo a parametric function. The reasoning behind the series parameter eludes me because the capacitance is in effect a storage mechanism relative to the action of the inductance. I can see the benefits of a charge pump configuration switching between series/parallel capacitance to modify the inductor I-V curve however what would the supposed change in a parameter act on?, it cannot act on itself in a series configuration.
This may help, the process in most of these FE devices relates more to energy or an energy state than a simple parametric function. If A happens we can expect B to follow as cause and effect plays out. However if we can hide B or part of it from A the normal flow of energy becomes interrupted. At this point "something" must intervene to conserve the energy in the system as energy must always be conserved in every case locally. Yes we could diverge on some tangent that energy is always conserved universally however that is not our concern.
In some sense it's like a game of hide and go seek, if action A happens then so must reaction B however if we could hide or transform part of B then energy cannot be conserved within the system and something internal/external must act to balance the equation.
My greatest insights came not from my R&D in electromagnetics or electrodynamics but more so critical thinking, reasoning, philosophy. The understanding that no matter how hard or fast I kick a dead horse it ain't going anywhere thus I should probably move on and try something different, something unique unlike what others do.
You will know it when you see it because it will be very much unlike anything you have done or thought in the past... something new.
Regards
AC
Quote from: Allcanadian on 2021.08.10, 02:23:59
PM
Interesting circuit, I see a series capacitor-inductor and mosfet switching on either side of the cap but I can't see a clear mechanism for gain.
Your description of the circuit is too simplistic as the mosfets do not switch but rather provide the series parametric resonance capacitance change. In order to help visualize the gain mechanism, ask yourself what causes the increase in voltage across C1 above the peak pulse voltage when the substrate diodes in M1-M4 are no longer forward conducting?
Quote
However I can follow your line of reasoning that we could add/subtract series capacitance to vary a parameter ergo a parametric function. The reasoning behind the series parameter eludes me because the capacitance is in effect a storage mechanism relative to the action of the inductance. I can see the benefits of a charge pump configuration switching between series/parallel capacitance to modify the inductor I-V curve however what would the supposed change in a parameter act on?, it cannot act on itself in a series configuration.
The parametric capacitance change may not even be needed as I'm reasoning at this point. This conclusion came from the observation that as the voltage across C1 increases, the drain to gate/source parametric capacitance decreases and causes a higher off-resonance frequency. When this state occurs, the input energy increases greatly and the gain drops to COP<1.
I would describe this circuit as a high frequency reactive to low frequency reactive converter requiring a driving frequency that is higher than the high frequency resonance during operation. If this is true, then a fixed series capacitance would work with a higher efficiency than the parametric capacitance. Testing will confirm this to be correct or not. If this is correct, then scaling is no problem.
Quote
This may help, the process in most of these FE devices relates more to energy or an energy state than a simple parametric function. If A happens we can expect B to follow as cause and effect plays out. However if we can hide B or part of it from A the normal flow of energy becomes interrupted. At this point "something" must intervene to conserve the energy in the system as energy must always be conserved in every case locally. Yes we could diverge on some tangent that energy is always conserved universally however that is not our concern.
In some sense it's like a game of hide and go seek, if action A happens then so must reaction B however if we could hide or transform part of B then energy cannot be conserved within the system and something internal/external must act to balance the equation.
My greatest insights came not from my R&D in electromagnetics or electrodynamics but more so critical thinking, reasoning, philosophy. The understanding that no matter how hard or fast I kick a dead horse it ain't going anywhere thus I should probably move on and try something different, something unique unlike what others do.
You will know it when you see it because it will be very much unlike anything you have done or thought in the past... something new.
Regards
AC
For those interested, here is a sim that shows the parametric drain capacitance change over time with a 20ma bias current up to the maximum 55v rating for the device. This is not the same device as used on the bench circuit which is the 14N05L or logic version which I'm sure is graded from the 1405 die. The error log displays the drain capacitance in nfd at the different time intervals.
Regards,
Pm
Hi Allcanadian,
I see Partzman has answered you far better than I can, but since I already wrote this I'll send it for what it's worth.
The gain is from parametric amplification.
https://en.wikipedia.org/wiki/Parametric_oscillator#Parametric_amplifiers
The Mosfets are not being used as switches but as a form of variable capacitor. Parametric amplification doesn't happen by switching reactive elements-- the C or L of the element has to change while holding a charge or flux.* Using a Mosfet as a parametric amplifier is completely conventional in itself, there are fiber optic amplifiers that do this now. The hypothesis is that the increase in stored energy can be more than the energy used to change the parameter. In the wiki article above, the typical example of a parallel plate capacitor is used, but the energy balance in mosfets and varactor diodes is much murkier, so there seems to be a possibility of excess energy there.
It looks like Partzman is not adhering to the idea of a parametric gain, but I don't yet understand where else any gain could be coming from... gotta read more...
*There does seem to be a possible counter example to that. A long time ago I posted a paper by W.L. Barrow where he used a mechanically switched capacitor to generate oscillations. It's near the start of the Mandelshtam and Papaleksi thread in this forum.
Fred
Quote from: Orthofield on 2021.08.10, 15:52:12
Hi Allcanadian,
I see Partzman has answered you far better than I can, but since I already wrote this I'll send it for what it's worth.
The gain is from parametric amplification.
https://en.wikipedia.org/wiki/Parametric_oscillator#Parametric_amplifiers
The Mosfets are not being used as switches but as a form of variable capacitor. Parametric amplification doesn't happen by switching reactive elements-- the C or L of the element has to change while holding a charge or flux.* Using a Mosfet as a parametric amplifier is completely conventional in itself, there are fiber optic amplifiers that do this now. The hypothesis is that the increase in stored energy can be more than the energy used to change the parameter. In the wiki article above, the typical example of a parallel plate capacitor is used, but the energy balance in mosfets and varactor diodes is much murkier, so there seems to be a possibility of excess energy there.
It looks like Partzman is not adhering to the idea of a parametric gain, but I don't yet understand where else any gain could be coming from... gotta read more...
*There does seem to be a possible counter example to that. A long time ago I posted a paper by W.L. Barrow where he used a mechanically switched capacitor to generate oscillations. It's near the start of the Mandelshtam and Papaleksi thread in this forum.
Fred
Hi Fred,
My conclusion that I stated in my last post which I repeat here,
The parametric capacitance change may not even be needed as I'm reasoning at this point. This conclusion came from the observation that as the voltage across C1 increases, the drain to gate/source parametric capacitance decreases and causes a higher off-resonance frequency. When this state occurs, the input energy increases greatly and the gain drops to COP<1.is totally incorrect!!!
Bench tests prove that I was wrong in that there is no increase in the voltage across C1 with a fixed capacitor of 3.3nf in parallel with an STPS 2150 Schottky diode.
However, logic did dictate that perhaps two parametric sources on either side of C1 was redundant and in fact that proved to be true. So, the right side of C1 is now connected to the CSR and single ended measurements can now be taken which simplifies things. So the circuit is now reduced to a pulse source, one parametric mosfet network, and C1.
I will be doing more testing on this arrangement.
Regards,
Pm
I would like to state that I have been having measurements of the various PC circuits that vary throughout the course of any given day. This problem reared it's ugly head in the earlier posts of this thread and I have to admit that I do not understand the problem.
There are times during the day that COPs are consistently in the 2 range and then after a period of time will change to .8 or so. This may be due to the long sweep times that push the limits of my Tek MDO scope therefore producing measurement errors. The waveforms when expanded look accurate but perhaps beat waveforms are being produced between the switching frequency and the sweep rates which confuse the measurement calculations. At the high sweep rate of 100 Msamples/sec, it takes about 15-20 seconds for the math calculations to complete and this is where and when I suspect the errors occur.
I have considered diurnal effects as well but the variations seem too broad over a short period of time IMO.
At any rate, if the devices worked at their best, the power levels are in the tens of milliwatts range which is not very useful unless large value parametric caps were found.
So, unless answers become apparent on the problem, I'll be leaving this research at the present time.
Regards,
Pm
Quote from: partzman on 2021.08.11, 13:43:42
...
So, unless answers become apparent on the problem, I'll be leaving this research at the present time.
...
Hi partzman,
After several experiments on parametric systems, I came to the conclusion that changing a parameter has a price, and that you can't get back more than that price.
However the way is not to be abandoned if we can make this price paid by a free natural phenomenon, for example the ambient heat. But I haven't got the right idea for that yet...
Hi F6FLT,
Yes, I tend to agree, although I still wonder about the hyperabrupt varactors....
In terms of ambient changes in light, heat, etc. Partzman and I tried some things along this line with the variable capacitance of solar panels. We actively varied the load to vary the C of the cells while in operation, to extract parametric power on top of the DC power. This didn't turn out to be very successful.
There are a lot of advantages to using solar panels as they are already installed everywhere.
Another way to use solar panels would be to switch them to parametric operation when the panel output goes below the controller's minimum voltage. Then, you simply charge the panels as a capacitor at irradiance peaks and discharge it when light intensity drops. Since the C rises with solar voltage, the charge/discharge cycle could be controlled by a solar cell. Since the output power is determined as much by the charging voltage as the parametric cycling, this might be a pretty good way to get considerably more power out of the panels from dusk to dawn.
Another thought I had along these lines would to use oscillations of the E field of the earth to bias a bank of varactors or mosfets and once again charge and discharge them to gather power.
There are also variable inductors designed so that a very small variation in gap causes a large change in L. These could draw power from ambient vibrations.
There have been numerous patents along these lines since the 50s, when ambient parametric generators were considered to power ocean monitoring buoys and other remote sensors.
Fred
Hi Fred,
Everything you said is very much in line with the kind of ideas I am exploring as well. I didn't know that there was a capacitive effect in solar panels, it is indeed an interesting track.
I came up with the idea of modifying a capacitance by heat, following some experiments with old capacitors from the 70s. When I used them in a high frequency oscillator, they were heating up, and the oscillation frequency was changing, which means that the capacitance was changing with the temperature. Unfortunately, the capacitance would have to decrease with temperature for the stored energy to increase, and the opposite was happening. I tested all kinds of capacitors, but either the capacitance remains stable or it varies in the wrong direction. Even if we had the right type, there would still be the problem that the thermal effects are slow, which makes effective practical applications difficult, as with the variable inductors you mentioned.
Using the oscillations of the electric field of the earth is also theoretically possible, but although the fields are high, the impedances under which they occur are also very high so that the recoverable powers are very low. The variations of the earth's magnetic field can also be used theoretically, but since they are very weak and slow, there again no practical application is possible.
The hyperabrupt varactors you mentioned seem to me to be the most interesting way, as well as all systems where very sharp nonlinearities appear. Systems out of their equilibrium state and in phases of rapid change seem to me to have been little studied in physics. Perhaps we could find new phenomena and solutions for our goal.
François
Hi F6FLT,
Two books from 1965 have complete analyses of a "ferroelectric converter" using BaTi and PZT capacitors doped in various ways to get a better time constant. These are "Direct Energy Conversion" by Angrist, and "Direct Generation Of Electricity" by Spring, both packed with interesting approaches to energy conversion. In order for the device to operate the cycling temperatures need to be from the Curie point, around 120 C, to above it. At high voltages the efficiency can approach Carnot but electrical breakdown occurs before that point. In most cases the efficiency seems to hover around 3 percent. And with the required rapid cycling of heat, I could only see them being used in, for instance, a device that rotates in sunlight.
Another concept for below the Curie point would be to pair the heat-exposed caps with an inductor and then work with an L drop rather than a C rise.
I thought of using the E field because these very abrupt varactors have a very high reverse bia impedance. And so do the FETs used for instance in atmospheric electrometers. I haven't explored the idea thoroughly.
A bit off topic, I did a patent evaluation of several thousand patents in the direct energy conversion field. Global, and as far back as I could get, which was usually around 1920 in most countries. Back to the beginning of the patent system in the US. This was how I noticed the C effect in solar cells, among hundreds of other interesting possibilities. The place where I could see the most likely possibility of vast improvements in efficiency and utility were in Seebeck generators. Before the development of ceramic materials, many clever inventors discovered and rediscovered the possibility of dramatically reducing heat transfer through the active portion of the thermocouple. They used point contacts, capacitive induction, arcs, and sparks to more or less completely decouple heat flow from current, and get around the Weidemann-Franz law that links heat and electrical conduction, and greatly limits the efficiency of these devices. The more sophisticated versions of this approach got efficiencies around 40%, which makes them competitive with other heat engines. I have a vast file on these various approaches but not being a hands on researcher but more of a paper pusher, I've failed to get anyone to seriously approach the subject.
Fred
@Orthofield
Efficiencies around 40% are not bad. I have also followed a lot of the improvements that have taken place in the last few years, especially with 2D materials or with the transverse thermoelectric effect. One has the impression that progress is important in the experimental field, but one does not see really interesting practical applications coming out of it.
So I'm going back to work... :)
François
Hi F6FLT,
Yes, given the current max efficiency of about 8%, 40% is way better than "not bad"! Like in photovoltaics, where there is stuff in the lab that can get maybe 30% (optical rectennas, metamaterials, etc.) these experiments will not be low cost alternatives in the next decade or so. If it isn't cheaper than other sources of power, it makes little difference that a lab has made it work.
The TEG tech I'm referring to doesn't require new materials. The problem is it requires AC or quasi-AC operation, which thermoelectric engineers can't seem to wrap their heads around...
Fred
Hi Orthofield
Quote from: Orthofield on 2021.08.14, 17:31:21
...
Yes, given the current max efficiency of about 8%, 40% is way better than "not bad"!
of course. It's just an ironic way of talking, like saying "that girl isn't ugly" when she's a charming model. :)
Quote
Like in photovoltaics, where there is stuff in the lab that can get maybe 30% (optical rectennas, metamaterials, etc.) these experiments will not be low cost alternatives in the next decade or so. If it isn't cheaper than other sources of power, it makes little difference that a lab has made it work.
It is rather reassuring to know that it is a problem of cost rather than technique, as there is often subsequent research into cheaper alternative materials using the newly discovered principle.
Quote
The TEG tech I'm referring to doesn't require new materials. The problem is it requires AC or quasi-AC operation, which thermoelectric engineers can't seem to wrap their heads around...
Fred
I have confidence in the engineers. They are no different from experimenters as us. They are very pragmatic and are always looking for ways to get around difficulties. I just think the issue is complex.
Quote from: F6
I have confidence in the engineers. They are no different from experimenters as us.
They are very pragmatic and are always looking for ways to get around difficulties. I just think the issue is complex.
You are describing the Ideal Science setting where the ability to conduct research freely without limitation would be the norm.
But, is that really the World of Science which is in power today?
Hi MuDped,
Yes, I agree with you on this last point. I posted an overview of the technology on the thermoelectric industry forum, and sent complete details to three different manufacturers of thermoelectric generators, including one in which I am an investor, and didn't hear a peep from any one of them. I have no interest other than getting the tech used, and there are no patent restrictions.
I do understand that they have to make decisions as to what to respond to. But still, not one answer? The technology is actually not at all complex, but it is counterintuitive for someone working within the field. So I must conclude that they can't see past their conceptual blinders. That's one reason why I hang out here-- much more free thinking.
The idea that scientists and engineers are searching for answers without any mental or corporate restraints is pretty naive...
Fred
Orthofield
QuoteThe idea that scientists and engineers are searching for answers without any mental or corporate restraints is pretty naive..
I had looked at parametric charging in the past and still am to some degree. My plan was to design a small lipo battery charger to power USB devices or trickle charge industrial/commercial batteries.
I chose this plan because it doesn't depend on what people think or believe with respect to science. If it works more people will buy it as word spreads and the company will grow. In my opinion the mistake most people make is trying to convince other people an unbelievable idea has merit. I would rather build a technology and sell it based on how useful it is rather than deal with people. As if to say... here it is, take it or leave it, I don't care what you believe.
It's better to build your own technology company than deal with people preoccupied with opinions and beliefs. Facts matter.
Regards
AC
Quote from: muDped on 2021.08.16, 17:07:48
You are describing the Ideal Science setting where the ability to conduct research freely without limitation would be the norm.
But, is that really the World of Science which is in power today?
I was not talking about science but about engineering, i.e. methods based on science, but practical, to obtain concrete results.
Engineers get concrete results that everyone can reproduce. We don't have any of that in FE. Engineering is never ideal, and FE even less so.
Quote from: Allcanadian on 2021.08.17, 16:21:31
Orthofield
I had looked at parametric charging in the past and still am to some degree. My plan was to design a small lipo battery charger to power USB devices or trickle charge industrial/commercial batteries.
This has been my goal from the outset of my FE research. Lead acid batteries can benefit from smart trickle charging as I can attest. I have a '39 Ford street rod with lots of electronics on board that had an Optima red top that would discharge below the LS2 computer limit in 2 weeks if the car was not driven. With a specially designed hi frequency trickle charger, that battery lasted 14 +years in the car before replacement. It was still working but I lost confidence in the battery.
Therefore, a self trickle charger that would be able to supply 200-500ma could save people lots on battery replacement.
Or, there could be a supercap LAB replacement with such a self charger, etc.
Quote
I chose this plan because it doesn't depend on what people think or believe with respect to science. If it works more people will buy it as word spreads and the company will grow. In my opinion the mistake most people make is trying to convince other people an unbelievable idea has merit. I would rather build a technology and sell it based on how useful it is rather than deal with people. As if to say... here it is, take it or leave it, I don't care what you believe.
It's better to build your own technology company than deal with people preoccupied with opinions and beliefs. Facts matter.
I could not agree with you more! This is exactly the business plan I've had in mind all along and basically, it is somewhat stealth because most would not recognize what they were purchasing or using. At this point I've fallen short on producing such a device!
Regards,
Pm
Quote
Regards
AC
Quote from: Allcanadian on 2021.08.17, 16:21:31
...In my opinion the mistake most people make is trying to convince other people an unbelievable idea has merit.
...
I agree with you almost 100%. Almost always, these ideas are completely vague, not expressed in operational terms, the results they would allow to obtain are not quantified, it is impossible to test them even by thought experiments.
But when the ideas are well formalized, expressed in a rigorous, even mathematical way, with a proposal for an experimental demo, it is interesting that those who have them present them, not to convince but to provoke analysis by others, to see possible flaws that they would not have seen, to have suggestions for improvement etc... before experimenting anything. This ties in with what we have talked about elsewhere, the notion of falsifiability. If the idea is falsifiable, it can be discussed. When I read the papers of Smudge or Distinti, I think they are quite right to spread their ideas.