PopularFX
Home Help Search Login Register
Welcome,Guest. Please login or register.
2024-03-29, 12:50:34
News: Registration with the OUR forum is by admin approval.

Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 [14] 15 16 17
Author Topic: Parametric Charging  (Read 60083 times)
Group: Elite Experimentalist
Hero Member
*****

Posts: 1399
... .-.. .. -.. . .-.
Ah ok, was just a wondering about that circuit section.
Thanks for having tried it  O0


---------------------------
ʎɐqǝ from pɹɐoqʎǝʞ a ʎnq ɹǝʌǝu
   
Group: Professor
Hero Member
*****

Posts: 2992
  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?
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
  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
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 4056
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
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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

   
Group: Professor
Hero Member
*****

Posts: 2992
 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?
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 4056
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
   

Group: Tinkerer
Hero Member
*****

Posts: 1693
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
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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
   
Group: Experimentalist
Hero Member
*****

Posts: 1665

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
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 4056
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
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 4056

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
   

Group: Tinkerer
Hero Member
*****

Posts: 1693
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.
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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

 
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 4056
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
« Last Edit: 2018-10-11, 10:42:30 by Itsu »
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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!
« Last Edit: 2018-10-11, 21:49:14 by partzman »
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
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
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 4056
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

   

Group: Elite Experimentalist
Hero Member
*****

Posts: 4056
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
   
Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 [14] 15 16 17
« previous next »


 

Home Help Search Login Register
Theme © PopularFX | Based on PFX Ideas! | Scripts from iScript4u 2024-03-29, 12:50:34