For this second run, I wanted to bring the frequency up to the range Lawrence is using.
I used a Dremmel tool to carefully cut a slice out of the core to bring down the core u. Now the Fo is about 52kHz.
05.png is showing an average INPUT POWER of 24.64mW (very close to the previous 6kHz test).
07.png shows that the OUTPUT POWER is at 51.27mW (again very close to the 6kHz test), but this must be divided by 10 to take into account the value of the CSR is 10 Ohms, not 1 Ohm. Therefore, the OUTPUT POWER is about 5.13mW.
.99
Poynt:
My congratulations again on this one. It's important to state that the traces all make sense now. The traces you show are indeed what they are supposed to look like, and what you expect them to look like. Nice sweeping curves showing the LTJT circuit charging and discharging the toroidal inductor, etc.
The math trace also looks like what it is supposed to look like. This shows convergence between the recorded data and what you expect that data to look like, indicating that you have valid data. Overall, an excellent job.
Note that if you were to increase the power buy using a larger toroid, and using higher voltages, the curves would look the same but the relative noise levels would start to decrease. The scope traces would become clean and noise-free. The operating frequency would also decrease. But the bottom line is that a "big brother" LTJT circuit would operate in the same manner, and show similar efficiencies. Scaling it up will NOT get you over unity.
Thank you so much for your efforts.
MileHigh
You're welcome of course, and thanks.
I hope that with the help of Lawrence and maybe the professor, we can determine why there is such a large discrepancy in my results so far, and those shown by him and the professor. I have an idea, but I am going to wait on revealing that for a little while.
It will also be interesting to see how the device I build from the parts Lawrence is sending me performs in comparison.
.99
When would these be needed?
http://www2.tek.com/cmswpt/psdetails.lotr?ct=PS&ci=13429&cs=psu&lc=EN (http://www2.tek.com/cmswpt/psdetails.lotr?ct=PS&ci=13429&cs=psu&lc=EN)
Poynt, stop being cheap. Get a good battery man. ;D
Poynt,
Just one question on the L3 secondary output graphs. The blue trace is the current and you can see it dips below zero. Is that due to the leakage current through the reverse-biased diode or is that possibly related to the calibration sequence for that channel? If I recall correctly the diode reverse-bias leakage current should be negligible. So it looks a bit fishy to me. You could double-check the reverse-bias current with a multimeter if need be. At the end of the toroid discharge cycle you can see how the voltage drops until the polarity-reversal happens (2.2 main divisions from the left on the TEK00006.PNG graph).
At the tail end of that discharge cycle the power goes negative and that doesn't make any sense. So it looks to me like the current waveform is negatively offset.
Assuming that I am correct and the current waveform needs to be offset upwards, it is likely going to marginally decrease the output power calculations for both trials. That's because you are showing positive power when the diode is reverse-biased, when in fact there is no power. This false-positive power is much larger in area than the false-negative power at the end of the discharge cycle. Of course the main power stroke will also be higher in level too.
MileHigh
MH,
The reverse breakdown of most LED's is on the order of 5V, so assuming that inductive kickback is present, this might well account for that negative current.
This is why it is important to either find a good LED model, or use a 5V zener when attempting to simulate this circuit in SPICE. I strung several 1N4148's together in my first attempt, but it will not work because this represents a diode with a VBr of about 500V.
In my simulation I will try a 5V zener, or perhaps 5x 1V zeners in series, then I will have roughly the 3V forward voltage of these super-bight LEDs, and the required 5V reverse breakdown parameter. Harvey, you may want to try this in Protel as well.
Note that when the power trace dips below zero (indicating some reactive power is present), that represents power going back out of the load, and this is why we can't use Pave=VRMS x IRMS to calculate the average power.
.99
Quote from: poynt99 on 2011.02.07, 13:24:07
The reverse breakdown of most LED's is on the order of 5V, so assuming that inductive kickback is present, this might well account for that negative current.
This is why it is important to either find a good LED model, or use a 5V zener when attempting to simulate this circuit in SPICE. I strung several 1N4148's together in my first attempt, but it will not work because this represents a diode with a VBr of about 500V.
In my simulation I will try a 5V zener, or perhaps 5x 1V zeners in series, then I will have roughly the 3V forward voltage of these super-bight LEDs, and the required 5V reverse breakdown parameter. Harvey, you may want to try this in Protel as well.
Note that when the power trace dips below zero (indicating some reactive power is present), that represents power going back out of the load, and this is why we can't use Pave=VRMS x IRMS to calculate the average power.
I assume that you agree that when the transistor is on, the LED in the L3 output loop is reverse-biased. If you scope probe is 1:1 then the reverse-bias voltage is only around 2 volts so no current should be flowing. I really can't account for the current going negative at the tail end of the core discharge cycle. You are showing about -24 mV across a 10-ohm resistor when the reverse-bias voltage is about -2 volts. So the reverse current is about 2.4 milliamps. That seems off to me. If you could double-check that with a multimeter that would be appreciated.
It's hard to be sure, but it looks like the discharge through the collector LED is during a brief 1 uS pulse just after the transistor switches off. You notice in the 04.PNG graphic that the current levels off to horizontal for about 1 uS at the end of the ramp-up. That's where the core is discharging through the collector LED just like in a conventional Joule Thief. The L3 loop is also discharging energy during this 1 uS period but most of the energy goes through the collector LED and one-ohm resistor because the impedance in that path is much lower.
After this 1 uS rapid discharge most of the energy stored in the core has been expended and the collector LED forward-bias voltage shuts off any further energy discharge through this path. Then the rest of the energy is discharged through the L3 loop because it can generate higher EMF than the L2 loop. So the discharge cycle shown in the 06.PNG graph represents the remainder of the core energy being dissipated through the L3 loop.
That's what I think is happening with the disclaimer that I am not on the bench and I am a bit rusty with all of this stuff.
MileHigh
Poynt:
It just occurred to me that there is something subtle going on with respect to your average input power calculation that probably has to be accounted for.
Let's assume for the sake of argument that I am correct, and that there is a big energy discharge through the collector LED for about one microsecond. You could easily confirm this by looking at the potential on the 1K base input resistor. If the potential snaps low about 1 microsecond before the current trace in the 04.PNG graphic stops, that would confirm my speculation in the previous posting.
So that means that during that one microsecond pulse, you are recording an energy discharge that comes from two sources; 1) the voltage times the current supplied by the battery, and 2) the voltage times the current supplied by the discharging L2 coil.
I also noticed that the current during this 1 uS discharge cycle is about 80 milliamps. I'm pretty sure that that's a huge amount of current for a typical LED and that high current times the voltage across the collector LED should (I think) represent a significant amount of the energy storage capacity of the toroid (not forgetting that some of that energy is coming from the battery itself).
The energy coming from the L2 coil is stored energy that was already accounted for during the current ramp-up. Therefore you have to subtract that energy from the input energy calculation. If you don't you are double-counting some of the energy supplied by the battery per cycle. That almost makes me queasy and reminds me of a luminary! lol
Of course with your uber awesome DSO you could do a high-bandwidth capture of the voltage pulse across the collector LED itself.
MileHigh
Quote from: MileHigh on 2011.02.07, 18:47:30
Poynt:
The energy coming from the L2 coil is stored energy that was already accounted for during the current ramp-up. Therefore you have to subtract that energy from the input energy calculation. If you don't you are double-counting some of the energy supplied by the battery per cycle. That almost makes me queasy and reminds me of a luminary! lol
Of course with your uber awesome DSO you could do a high-bandwidth capture of the voltage pulse across the collector LED itself.
MileHigh
MH,
I think it might be better to place the INPUT CSR directly in series with the battery, that way we can be certain that the current acquisition is correct. I will try that tonight...in the negative leg of the battery. The 1 Ohm in the emitter will be shorted for this test.
.99
Poynt:
I just noticed that in your computer graphic schematic for the circuit you show the collector LED being directly across the CE of the transistor. In your hand drawing of the circuit you show the collector LED being across the collector and ground. I was basing my thoughts on the computer graphic of the schematic.
Anyway, I hope that I gave you some food for thought!
MileHigh
Me again,
I think that I have to retract my statement about the double energy counting from the discharging L2 inductor.
I was looking at the graphical diagram of the circuit, "poynt99 schematic 1.jpg" that indeed shows show the collector LED being directly across the CE of the transistor. With this arrangement the one-ohm current sensing resistor does a correct job and is in series with the battery.
Where I screwed up was in the alleged "double counting" of the discharge from L2 after the transistor switches off. It's true that during the current ramp-up when the transistor is on energy is stored that eventually is discharged through L2. However, when L2 discharges after the transistor switches off, you are still only recording the voltage from the battery, and the current from the battery during this phase. I referred to this energy component two postings ago also.
This effectively means that the energy discharge from L2 after the transistor switches off is not seen a second time. Rather, it's more like the energy discharge from L2 is discharged across the collector LED only, and the current through the one-ohm CSR and the battery voltage accounts for the energy supplied by the battery. There is a voltage bump-up across the collector LED when L2 discharges, buy this is not seen because it is internal to the circuit.
In other words, indeed there is energy that is stored in the core and then released by the core via L2 in a full cycle, but the energy per cycle derived from recording the battery voltage and the current through the CSR only sees this once, not twice.
Sorry, I got lost there and simply had to go back to basics.
MileHigh
Good catch though MH, as I am pretty sure I built it as per my hand drawn schematic (which is incorrect). I will make the change when I get home tonight and see if the LED directly across C-E has any significant effect on the wave forms and results. In the simulation, it doesn't.
.99
Here are the results and scope shots with the circuit corrected (LED directly across transistor C-E). I've also included an updated drawing and notes.
09.png indicates an average INPUT power of 39.8mW (prior test was 24.64mW, ~61% increase in power).
11.png indicates an average OUTPUT power of 52.09mW/10 = 5.21mW (very close to the prior test).
These results compared to the previous run (when the LED was at ground), would indicate that the previous INPUT power measurement was incorrect.
Note the triangle wave shape of current now.
.99
Alright, I didn't like what resulted above, so it was time to clip that energy-sucking collector LED out of the circuit, and prove that the output will indeed go up significantly without it.
013.png is indicating an average INPUT power of 32.88mW (a decrease from the last run).
015.png is indicating an average OUTPUT power of 228.3mW/10 = 22.8mW (a more than 4-fold increase over previous tests!).
Now the input to output efficiency is at about 69%, and yes, the output LED became much brighter.
Who said that collector LED should not be considered part of the output? ^-^
.99
Hey Poynt!
QuoteHere are the results and scope shots with the circuit corrected (LED directly across transistor C-E). I've also included an updated drawing and notes.
09.png indicates an average INPUT power of 39.8mW (prior test was 24.64mW, ~61% increase in power).
11.png indicates an average OUTPUT power of 52.09mW/10 = 5.21mW (very close to the prior test).
These results compared to the previous run (when the LED was at ground), would indicate that the previous INPUT power measurement was incorrect.
Note the triangle wave shape of current now.
Ha! That looks better. You can see how my theories about the discharge of the collector LED were wrong because I was looking at the (formerly) wrong schematic and trying to fit the scope traces in with what I thought was the right schematic. Trying to put a square peg into a round hole.
Now we clearly see that both LEDs have current flowing through them for the full cycle of the energy discharge from the core. That feels a lot better!!! lol
Your new numbers reveal something else that's interesting. We can make a preliminary deduction that there is roughly (39 - 24) = 15 milliwatts of average power being dissipated though the collector LED. That's significantly higher than the 5.21 milliwatts of power being dissipated in the L3 output loop.
Obviously with a higher input power and the same L3-related output power, the efficiency of your original LTJT unmodified configuration is even lower than before, about 13.1%.
Really big shew!
MileHigh
Quote from: poynt99 on 2011.02.08, 02:17:35
Alright, I didn't like what resulted above, so it was time to clip that energy-sucking collector LED out of the circuit, and prove that the output will indeed go up significantly without it.
013.png is indicating an average INPUT power of 32.88mW (a decrease from the last run).
015.png is indicating an average OUTPUT power of 228.3mW/10 = 22.8mW (more than 4-fold increase over previous tests!).
Now the input to output efficiency is at about 69%.
Who said that collector LED should not be considered part of the output? ^-^
.99
It's a whole new ball game!!! That's starting to sound "about right" to me. Core losses and wire resistance losses coming into play along with the overhead associated with the CSR at the emitter, the base input resistor, the diode drop at the transistor base input, etc.
Now we just need some Lawrence Tseung magic to get you over the top!
MileHigh
Quote from: Harvey on 2011.02.07, 02:17:19
When would these be needed?
http://www2.tek.com/cmswpt/psdetails.lotr?ct=PS&ci=13429&cs=psu&lc=EN (http://www2.tek.com/cmswpt/psdetails.lotr?ct=PS&ci=13429&cs=psu&lc=EN)
Harvey,
Mostly in cases where a current and differential voltage probe need to me time-matched, but would also be desirable when making critical measurements using probes with differing cable lengths, or models.
As I am using 4 probes all the same make, model and cable lengths, it should not be a problem with these tests.
.99
Quote from: MileHigh on 2011.02.08, 02:19:26
Obviously with a higher input power and the same L3-related output power, the efficiency of your original LTJT unmodified configuration is even lower than before, about 13.1%.
After a long day, I came back and COP dropped to 13%. The road to OU just went from .8infinity to 1.1infinity. Thanks >:(
Anyhoo, after several...juxtaposing I found the LED on L3 may have mounted in opposite of the prof's. Therefore, the conclusion is...this juxtaposing is some good shi...sstuff. :D
Quote from: poynt99 on 2011.02.07, 20:07:31
Good catch though MH, as I am pretty sure I built it as per my hand drawn schematic (which is incorrect). I will make the change when I get home tonight and see if the LED directly across C-E has any significant effect on the wave forms and results. In the simulation, it doesn't.
.99
I must correct my post here.
The change in the simulation does make the same difference as observed with the real circuit. The CSR wave form becomes more triangular vs. sawtooth.
.99
Poynt:
It's the old cliche about how a programmer can only write about 10 lines of clean error-free code per day when you average things out over time. We both made mistakes and found them and recovered.
It's going to be interesting when the FedEx shipment arrives.
For Lawrence and Poynt: I believe that there are now two samples in transit, one by FedEx and one by regular mail. It would be very helpful if you can establish what measurements were made by Lawrence's group in Hong Kong (or China?) for each sample. Then when Poynt makes his measurements on each sample you will be able to compare the data.
Pictures of the setup and your handwritten notes are also appreciated.
MileHigh
MH,
The kit Lawrence shipped via regular mail is not an assembled unit, so there are no measurements of that one yet. As for this latest fully assembled unit, see Lawrence's data he already provided.
I'll post my test results of each one as they become available.
.99
Quote from: poynt99 on 2011.02.08, 02:17:35
Alright, I didn't like what resulted above, so it was time to clip that energy-sucking collector LED out of the circuit, and prove that the output will indeed go up significantly without it.
013.png is indicating an average INPUT power of 32.88mW (a decrease from the last run).
015.png is indicating an average OUTPUT power of 228.3mW/10 = 22.8mW (a more than 4-fold increase over previous tests!).
Now the input to output efficiency is at about 69%, and yes, the output LED became much brighter.
.99
So you finally cut out that transistor-LED as I suggested -- hurray! I did that weeks ago. O0
Quote"
.99: Who said that collector LED should not be considered part of the output? ^-^
NOT ME! At least, not if you understood me correctly (and I don't think you did).
My next suggestions --
1. Replace the 10ohm + 100 ohm resistors with a 1-ohm resistor, and take measurements. (Makes V*I more straightforward also.)
2. Replace the 1 Kohm resistor to the base with 500ohms; take measurements.
3. Then -- place a capacitor in series with the LED, that is, between the 1ohm R and the remaining LED -- and take measurements again. This is what I have done after some tinkering, with interesting results. I use a 100 uF capacitor most of the time.
I realize this is getting rather far from the Tseung circuit as he sent it, but my goal is to see what happens to the circuit when tweaked, to see whether the COP can be boosted up.
4. I have also wound my own toroid, the JT portion of it anyway, starting with a pre-wound 100 uH toroidal inductor from Jameco as the secondary winding -- again with interesting results. Jameco 386601, 0.037 ohms, 100 uH. I did this as a way to improve repeatability from one toroid to the next, hopefully, by starting with a pre-wound toroidal inductor. More on that later if anyone is interested.
Hi Professor.
Yes, I am interested in trying the 1 Ohm in the output to see if I get the same resulting phase shift you showed. Did you also get a better efficiency with the 1 Ohm in place of the 110 Ohms?
Will you post some input and output shots similar to the ones I posted...with MEAN values instead of RMS?
I'll try your other suggestions as well, as time permits. I suspect I will be receiving the kit from Lawrence very soon though. At the moment I'm more focused on testing a unit that Lawrence feels is worthy of testing, which is why I am looking forward to receiving and testing the fully-assembled unit.
Thanks,
.99
PhysicsProf:
I am going to make a few comments about your suggestions in advance of anyone trying them out. The purpose of commenting ahead is to hopefully shed some more light on the way the Joule Thief works.
Quote2. Replace the 1 Kohm resistor to the base with 500ohms; take measurements.
This can be expected to have a marginal effect on the operation of the circuit. Lowering the base input resistor will mean that the transistor switches on slightly faster. Because the JT does a "snap" when the transistor switches on and off, you are talking about changing the timing by a few microseconds at most. Also lowering the base resistor will increase the power dissipation slightly. I suspect that lowering the base input resistor will increase the operating frequency of the JT by a very small amount.
I should also state that I am assuming that the 1K resistor is low enough in value to fully switch on the transistor so changing the resistor to 500 ohms will not make much of a difference. In the unlikely case that the 1K resistor does not not in fact fully switch on the transistor, then things change. When you make a Joule Thief you want the transistor to act as an ON-OFF switch, you don't want the transistor to work in partial conduction mode.
Quote1. Replace the 10ohm + 100 ohm resistors with a 1-ohm resistor, and take measurements. (Makes V*I more straightforward also.)
In the past I made reference to the JT as a transformer and made reference to changing the load on the secondary affecting the impedance match and the power transfer. Those statements were incorrect because it's not really a transformer.
The real way of looking at the JT is that it is a circuit that operates in two cycles, first cycle charges the core up with magnetic energy and then the second cycle discharges that energy stored in the core, and then the process starts over again.
The following discussion assumes that the collector LED has been removed and the only output load is on L3.
The key point here is that once the core has been charged up with energy, then you discharge that finite amount of energy. So if you change the load on the L3 secondary coil, you will be able to affect the rate that the energy discharges, but not the amount of energy itself that's available to discharge into the load.
So with those points in mind, it looks like reducing the load to an LED in series with a one-ohm resistor will speed up the discharge cycle considerably. This shortened discharge cycle will therefore increase the operating frequency of the JT and as a result increase the average power consumption of the JT. However, the overall average-power-in to average-power-out efficiency is not likely to change considerably.
The most important thing to learn here is that the amount of energy per individual discharge cycle should not change substantially when you go from a (LED + 110 ohm) load to a (LED + 1 ohm) load. It all depends on the amount of current flowing through L2 the instant before the transistor switches off. If the amount of current is the same then the amount of energy stored in the core will be the same. Thus you are dealing with a fixed amount of energy that is available in the charged-up toroidal core of the Joule Thief.
MileHigh
Poynt:
Just a quickie. I am assuming that you are using vanilla 5% tolerance carbon resistors for your current sensing functions. Are you making actual measurements of the values of those resistors for the math calculations?
Thanks,
MileHigh
Quote from: MileHigh on 2011.02.08, 19:38:26
PhysicsProf:
I am going to make a few comments about your suggestions in advance of anyone trying them out. The purpose of commenting ahead is to hopefully shed some more light on the way the Joule Thief works.
This can be expected to have a marginal effect on the operation of the circuit. Lowering the base input resistor will mean that the transistor switches on slightly faster. Because the JT does a "snap" when the transistor switches on and off, you are talking about changing the timing by a few microseconds at most. Also lowering the base resistor will increase the power dissipation slightly. I suspect that lowering the base input resistor will increase the operating frequency of the JT by a very small amount.
[snip]
MileHigh
In fact, by direct observation of a JT circuit in my lab upstairs, lowering the base input resistor (from 1000 ohms to 500 ohms) resulted in a DECREASE of the operating frequency of the JT by a large amount -- 461 kHz to 361 kHz.
I say this not to embarrass you, MH, but rather to show (once again) that one's theoretical understanding of the JT circuit -- or the one I just tested in my home lab, a very simple JT circuit (without the collector LED) -- is in the OPPOSITE direction from what you predicted, and the change is large. I invite .99 to make the same test and check if this result is more general. Quick and easy test.
Also, MH, we see again the importance of testing with an actual apparatus -- experiment trumps theory... as previously noted throughout the history of science.
And again I encourage you to set up a JT to play with... not difficult.
.99 "Will you post some input and output shots similar to the ones I posted...with MEAN values instead of RMS?"
Yes of course, when I get time on the Tektronix 3032 (within the next week). I have not been up to the university to use the 3032 since I was there and posted results in this forum last time.
I've been using the ATTEN 1062C that I bought and arrived last Thursday. Lots of fun, and it has the channel 1 * channel 2 math function, which I am using routinely to display the Power = V * I waveform.
However, while this DSO readily calculates the MEAN (and RMS, etc) for channel 1 and channel 2 separately, I have not figured out how to get it to calculate the MEAN for the product V*I ... Not sure it has this capability -- I have written to the tech people at ATTEN. Also, have not been able (yet) to get the power waveform onto my MAC-Pro laptop where I could further analyze it... and where I could post the waveforms. I do have waveforms stored on a thumbdrive, if I could just read them-- the disc that came with this system was not helpful. If anyone knows the "tricks" of this new DSO, I'd appreciate hearing from you.
So for now, to get the MEAN of P = V*I, I have a long trip up to the university and arranging to use that scope. I am also trying to borrow a more expensive scope that will calculate the MEAN of the V*I waveform directly. I'm glad you're taking data .99 -- you will no doubt be the first to evaluate P-Mean (and the ratio Pmean-out/Pmean-in) for the Tseung circuit.
Quote from: PhysicsProf on 2011.02.08, 21:19:27
In fact, by direct observation of a JT circuit in my lab upstairs, lowering the base input resistor (from 1000 ohms to 500 ohms) resulted in a DECREASE of the operating frequency of the JT by a large amount -- 461 kHz to 361 kHz.
I say this not to embarrass you, MH, but rather to show (once again) that one's theoretical understanding of the JT circuit -- or the one I just test upstairs, a very simply JT circuit (without the collector LED) -- is in the OPPOSITE direction from what you predicted, and the change is large. I invite .99 to make the same test and check if this result is more general. Quick and easy test.
Also, MH, we see again the importance of testing with an actual apparatus -- experiment trumps theory... as previously noted throughout the history of science.
And again I encourage you to set up a JT to play with... not difficult.
Actually I am not embarrassed at all. There are limits to what can be done when you are just commenting and you haven't worked with one on the bench, nor have I seen your actual circuit. In most of my postings along these lines I make a disclaimer that I am not actually on the bench myself and therefore I can't be 100% certain of what I am stating. The process of exploring like this is part of the learning experience.
QuoteAlso, MH, we see again the importance of testing with an actual apparatus -- experiment trumps theory.
You are missing the point PhysicsProf. I was wrong with respect to your test. The empirical evidence on your particular Joule Thief shows large changes in frequency and a decrease in frequency when you lower the base resistance, the opposite of what I said. There are too many unknowns on both sides to draw any conclusions at all between what I stated and what you tested.
Do you know if your transistor was fully on or in partial conduction mode when you used a 1K base resistor? What about when you switched to a 500-ohm base resistor? My statement was based on the presumption that for both the 1K and 500-ohm case that the transistor would be functioning as a switch. Can you explain why the frequency dropped when you changed the base resistor? It would be fun to work together one day, time permitting, to figure out why the frequency changed.
There is a logical reason for the decrease in the operating frequency of your Joule Thief that is backed up by theory. So the challenge for those that are interested in pursuing the research is to explain your observations with sound theory. That's where the fun comes in as part of the intellectual journey.
Let's go back to the Wikipedia formula for the operating frequency of the Joule Thief:
F = ((V_batt x R_batt) / L_mutual). Note R_batt is the output impedance of the battery. This formula presumes that the transistor is acting like a switch.
On the the other hand, supposing the transistor is in partial conduction mode when the base input resistor is 1K ohm. That would effectively make R_batt become (R_batt + R_transistor) which is higher, which according to the formula would
increase the JT operating frequency.
Supposing the transistor is in full conduction mode and acting like a switch when the base input resistor is 500 ohms. That would effectively make R_batt lower, which according to the formula would
decrease the JT operating frequency.
So, based on the limited amount of information that I have on hand, it would appear to me that when you have a 1K base input resistor the transistor is operating in partial conduction mode. That would explain the decrease in frequency when you switch the base input resistor to 500 ohms.
As far as I understand, the desire is for the Joule Thief circuit to operate the transistor as a switching device because this then limits the power dissipated across the collector-emitter junction to a minimum.
This suggests an interesting experiment: As you lower the base input resistance, you should see the operating frequency of the Joule Thief decrease. Eventually you should get to a point where lowering the base input resistance even further does not change the operating frequency (i.e.; the basis for my original comments). That point where the frequency stops changing would be your ideal base input resistance. I am assuming that a standard Joule Thief circuit assumes that the transistor should be operating as a switching device.
MileHigh
Quote from: MileHigh on 2011.02.08, 19:45:00
Poynt:
Just a quickie. I am assuming that you are using vanilla 5% tolerance carbon resistors for your current sensing functions. Are you making actual measurements of the values of those resistors for the math calculations?
Thanks,
MileHigh
MH,
No I am not at the moment using the actual value of the "1 Ohm" CSR resistor, as I see no need until or unless the "n" begins to approach 100%. The resistor is 5% carbon I believe, and a value of 0.98 Ohms iirc.
.99
An open question for anyone who may wish to reply:
With reference to the scope shot below, what factor(s) determines the current reached before the transistor switches OFF?
In the shot below (TEK00012.PNG), the current (blue trace) rises to about 85mA.
(http://www.overunityresearch.com/poynt99/TEK00012.PNG)
.99
I'm finding you are raising some very good points, and questions. I certainly don't know all the answers. \
MH: "As far as I understand, the desire is for the Joule Thief circuit is to operate the transistor as a switching device because this then limits the power dissipated across the collector-emitter junction to a minimum." Agreed... how best to achieve this is an important question.
The circuit frequency is sensitive to just about any change I make -- resistances, LED's in or out, toroid windings, etc. It does seem to reach a natural resonance condition quickly in almost all cases.
Quote from: poynt99 on 2011.02.08, 23:57:10
An open question for anyone who may wish to reply:
With reference to the scope shot below, what factor(s) determines the current reached before the transistor switches OFF?
In the shot below (TEK00012.PNG), the current (blue trace) rises to about 85mA.
.99
This is the reason why I joked with you earlier to get a good battery. I noticed that the 6kHz on test 1 and the 50kHz on test 2 have the same amount of current. The slope of the current looks too linear so I suspected that current rise in L1 and L2 coil is controlled by the switching. I think the reason is that the battery has too high impedance.
Quote from: GibbsHelmholtz on 2011.02.09, 01:06:46
This is the reason why I joked with you earlier to get a good battery. I noticed that the 6kHz on test 1 and the 50kHz on test 2 have the same amount of current. The slope of the current looks too linear so I suspected that current rise in L1 and L2 coil is controlled by the switching. I think the reason is that the battery has too high impedance.
The supply voltage does have an effect on the oscillation frequency, true. However, the battery impedance does not have that much effect on things at the moment, and you even pointed out yourself that the current reached the same amount, even when the Fo was about a decade higher.
So, what are the factors that determine the maximum rise of the current? Why is the current slope too linear, what do you mean?
.99
Quote from: poynt99 on 2011.02.08, 23:57:10
An open question for anyone who may wish to reply:
With reference to the scope shot below, what factor(s) determines the current reached before the transistor switches OFF?
In the shot below (TEK00012.PNG), the current (blue trace) rises to about 85mA.
I'm pretty sure that it's the leveling off in the current rise that causes the transistor to switch off, not the absolute amount of current flow. It's the change in the rate of change of the current that determines when the transistor starts to switch off. So that's the second derivative (the "accelerated acceleration") of the current with respect to time that is the determining factor.
As the current climbs linearly as the core get's energized the positive EMF on the L1 coil keeps the transistor switched on with a constant voltage. Then the current starts to level off as the output impedance of the battery (plus the resistance of the transistor junction assuming it is in partial conduction mode) starts to become a factor.
This leveling off of rate of the current rise causes a slight drop in the positive EMF from L1. This triggers a positive feedback loop, (a 'cascade' effect) where the dropping EMF from L1 starts to switch off the transistor, and that in turn causes more leveling off of the rate of increasing current (eventually becoming decreasing current), which in turn reduces the EMF from L1 (eventually going negative), which turns the transistor off even more, etc.
This positive feedback loop causes the transistor to "snap" of very very quickly, much faster than a typical Bedini motor shuts off the transistor. You note that it's so fast that you can't even see it in your scope captures. They look like normal linear ramp functions but I have to assume that the at the tail end of the ramp the slope changes just slightly for a fraction of a microsecond and then all hell breaks loose.
A great analogy for a positive feedback loop that is somewhat applicable comes from trying to adjust the position of your car seat while driving. If your seat is free to move forwards and backwards and you reflexively touch the brake while you are doing 100 km/hour, then you will floor the breaks and will be unable to stop this from happening. The slightest touch on the brake pedal will slow the car down resulting in the seat sliding forwards which results in you pushing harder on the brake pedal, etc. It happens in a flash and you have no control over it.
MileHigh
OK, the feedback does sound plausible MH.
Are there any other factors involved with this Imax? What factors determine the Fo?
.99
Poynt:
Does your scope have low-pass filtering? It might be possible to see the slope change at the end of the ramp if you filtered out the noise.
As far as the Fo goes, we can see that for the charge cycle, it's related to R/L, and then I assume the discharge cycle is related to how fast you discharge the energy in the core. So I assume that part is load dependent, and depends on what is connected to the collector terminal and/or L3.
I just realized too that you can't forget the "backwards" thinking that you have to do for inductors. I slip there sometimes and I made some incorrect statements about the discharge of L3 recently.
So with the "inductor backwards" thinking cap on, the lower the load resistance the slower the core will discharge, and the higher the load resistance the faster the core will discharge.
MileHigh
Quote from: MileHigh on 2011.02.09, 02:51:17
Poynt:
Does your scope have low-pass filtering? It might be possible to see the slope change at the end of the ramp if you filtered out the noise.
As far as the Fo goes, we can see that for the charge cycle, it's related to R/L, and then I assume the discharge cycle is related to how fast you discharge the energy in the core. So I assume that part is load dependent, and depends on what is connected to the collector terminal and/or L3.
I just realized too that you can't forget the "backwards" thinking that you have to do for inductors. I slip there sometimes and I made some incorrect statements about the discharge of L3 recently.
So with the "inductor backwards" thinking cap on, the lower the load resistance the slower the core will discharge, and the higher the load resistance the faster the core will discharge.
MileHigh
Yes, the scope has an acquisition "averaging" function, and when utilized, it cleans up the traces quite a bit.
The battery voltage is one factor that affects Fo. One other factor that affects Fo is the time constant of the base drive, L/R, which is the reason changing the 1k to 500 Ohms decreases the frequency, i.e. smaller R yields longer tau. Incidentally, this also affects the Imax value. A smaller Rb and you have a larger Imax ("Imax" being the value at the top of the ramp).
So: decR, decFo, incTau, incImax,
What other parameter change might have a similar effect on Imax and Fo?
Why did Imax remain the same after I cut the core to increase Fo?
.99
Quote from: MileHigh on 2011.02.08, 21:59:12
...
Can you explain why the frequency dropped when you changed the base resistor? It would be fun to work together one day, time permitting, to figure out why the frequency changed.
There is a logical reason for the decrease in the operating frequency of your Joule Thief that is backed up by theory. So the challenge for those that are interested in pursuing the research is to explain your observations with sound theory. That's where the fun comes in as part of the intellectual journey.
...
MileHigh
Some variants of the Blocking Oscillator will show an "OFF
time" which is dependent upon an RC time constant.
This variant seems to be governed by the L/R time constant.
Was the slope and time of the inductor current increase relatively
constant in both cases?
Quote from: MileHigh on 2011.02.09, 02:35:23
I'm pretty sure that it's the leveling off in the current rise that causes the transistor to switch off, not the absolute amount of current flow.
...
MileHigh
Yes, that is the "trigger" which initiates the cutoff
cycle; inductor current flow ceases to increase.
1) current rise is linear with slope di/dt=V/L
2) Max current x number of turns produces core saturation. Double the turns and max current is cut in half.
The rest operates like MH pointed out.
PS. The horizontal segment before the current ramp is mainly governed by an RC time constant. C is the BE transistor junction capacitance. So, very some of these parameters and you vary Fo.
Quote from: EMdevices on 2011.02.09, 06:05:36
PS. The horizontal segment before the current ramp is mainly governed by an RC time constant. C is the BE transistor junction capacitance. So, very some of these parameters and you very Fo.
EMdevices,
Thanks for your comments but I will respond to the point above. Note that the transistor will also "snap" back on with another positive feedback loop. This will happen when the rate of decreasing flux in the core starts to level out when the core is running "out of gas." So the horizontal segment before the current ramp is associated with the toroidal core inductor discharging its stored energy through L3. So it's another L/R type of time constant at play.
MileHigh
MH, I'm assuming a sharp discharge cycle, and the graph seems to show the quick discharge at less than 1% of the time duration of the level current, if I'm not mistaken. During kickback, L/R will govern, true, but this is really small (occurs fast) like I said. The longer horizontal is really charge buildup in the transistor.
Quote from: EMdevices on 2011.02.09, 07:13:11
MH, I'm assuming a sharp discharge cycle, and the graph seems to show the quick discharge at less than 1% of the time duration of the level current, if I'm not mistaken. During kickback, L/R will govern, true, but this is really small (occurs fast) like I said. The longer horizontal is really charge buildup in the transistor.
By "scoping" the signal at the transistor
from base to emitter it is possible to
arrive at a clearer understanding of the
nature of the "cutoff" portion of the base-
emitter signal.
It should show a significant reverse bias
during that time; the result of some
'discharge' current.
All good comments, thanks. However, may I reiterate these questions:
1) What other parameter change might have a similar effect on Imax and Fo?
2) Why did Imax remain the same after I cut the core to increase Fo?
and I'll add one more since this has been mentioned a few times...
3) Is core saturation necessary for oscillation, and is it present with this LTJT?
Thanks,
.99
Poynt,
I'll take a stab at the questions at risk of repeating some points because I am not going to go back to reread both threads.
Quote1) What other parameter change might have a similar effect on Imax and Fo?
During the core charging cycle I'll go back and say it's the battery output impedance plus the transistor impedance that dominates. It's this resistance that determines when the slope of the core charging current starts to level off. So if you lower the effective resistance here you increase the Imax, and you decrease the Fo.
Also, if the source battery voltage is higher, then the initial slope of the current increase is higher. That means you will reach the the leveling off point in the slops faster, and this will result in increasing Fo. Note that the battery voltage will not have any affect on Imax.
If you increase the number of turns in L1 and L2 you increase the inductance with a risk of saturating the core. If we assume that the core does not get saturated then this will lengthen the charging and discharging time and decrease Fo. It should not affect Imax.
I am assuming that I might be missing something here and not really answering your question.
Quote2) Why did Imax remain the same after I cut the core to increase Fo?
Same answer as above, it's the resistance that determines Imax.
I am actually a little puzzled here about the frequency increase. If you have a core where the number of turns in L1 and L2 are constant, and then you make a cut in the core, you are reducing it's cross-sectional area. Won't that make the the core saturate sooner but not affect the value of the inductance associated with L1 and L2? But it is apparent that cutting the core is decreasing L and therefore increasing Fo. I am having some trouble visualizing this because the current waveforms show a linear rise (see below), indicating core saturation is not a factor here.
Quote3) Is core saturation necessary for oscillation, and is it present with this LTJT?
No, and in fact the linear rise in the current waveform is telling you that the core is not saturated. If the core got saturated you would see the current slope increase. One more time, it's the series resistance causing the slope to level off that triggers the discharge cycle.
That's my best crack and answering the questions!
MileHigh
1) What other parameter change might have a similar effect on Imax and Fo?
Imax is changed with changes in: number of turns, different core properties, radius of toroid
F0 is changed with changes above plus, change in output resistor value, higher R values increase F0.
2) Why did Imax remain the same after I cut the core to increase Fo?
given the same number of turns, same toroid radius, core saturation occurs at a certain H intensity, and that is proportional to N*I, or the number of turns times the current, and this quantity remained unchanged.
3) Is core saturation necessary for oscillation, and is it present with this LTJT?
Yes, and Yes, (this is the fundametnal quality of BOs, as opposed to regular oscillators.)
Some Discussion
This circuit is very simple and has been around for decades in one form or another. Core saturation defines the end of the ON time of the BJT and the length of the OFF state can be controled with RC time constants or power ouput impedance.
MH, you are absolutely right, I looked at the charts again, and I can see the OUTPUT waveform labled by p99. Disregard my RC and charge buildup in the transistor comments, not applicable here.
the figure I include shows the ON and OFF states.
During the ON state the current ramp is linear and dictated by di/dt = V/L, like I previously mentioned.
The end of the ON state occurs when saturation is reached and the coupling between the base coil and the collector coil begins to drop off sharply under positive feedback. A carefull analysis of these switching events involves magnetics and I can expand on this if needed.
During the OFF state the current is routed through the output resistors and the energy stored in the magnetic field is dissipated by the output resistors and the LED or diode and it follows a roughly L/R timeconstant decay. This off period/output phase seems to last about 3 or 4 time constants, and the exponential decay is clearly seen.
When the energy is pretty much disipated and the voltage has droped to a certain threshold, where the base circuit can now begin to function again, it will repeat the cycle.
Most Important parameters;
1) By far, the Core properties, number of turns and operating voltage dictate the duration of the ON time.
2) the Off time is mainly dictated by output impedance (in additon to the core properties, energy stored, etc...)
3) there are other parameters of course, but we don't need to get into them here since they don't have a big impact.
P.S. I should expand on core saturation, the complete formula is H = N*I / (2*pi*r), and H and B are governed by the material properties (see a typical B-H curve) So inspecting the equation, another thing that can change the performance is obviously the radius of the toroid, but not it's thickness (in the z-direction)
EMdevices:
It's fun to "look into the guts." Of course there is only so far I myself can go without being on the bench and poking around with a scope probe. I don't have a bench and I don't have a scope! I also downloaded pSpice and it's sitting somewhere in my download directory.
Honestly at this point I am content to see what the results of the real analysis are for the new LTJT sample due to arrive very soon, i.e.; the average output power vs. the average input power.
In the discussions so far for anyone that has been following, it should be pretty clear that the Joule Thief simply charges a toroidal magnetic core with energy supplied by the battery and then discharges that energy through some sort of a load, then repeats the cycle. This is an "astable multivibrator" pulse circuit and it "resonates" but it is not directly related to "resonance" in the way that we talk about an LC tank circuit. It's a kind of distant cousin to an LC resonant tank circuit.
Looking forward to the numbers!
MileHigh
P.S: Great composite graphic!
Absolutely MH,
I've worked on these things for years and they are realy simple and neat circuits, I love them. they can be used to do DC-DC conversion, trickle charge batteries, etc...,
But one thing is clear, they are not Over Unity devices, barring any external influences.
I'm going to share a measurement tip. If these devices are operating at almost 99% efficiency, it's very easy to make measuremnts errors and conclude they are over unity. I've made those mistakes in the past and I have a feeling I'm not the only one.
Here it is: if the scope probes are not properly calibrated, pulses voltage will overshoot more then reality and that will get you so excited when you calculate Pin vs. Pout by looking at the voltage spikes on the scope. I jumped up and down with joy many times, then I remembered the probe business and fell flat on my face. So, check those darn probes and use the same one for Pin and Pout calculations to minimize variability.
The FedEx package arrived today at about 13:00, so I already have the device at hand.
Hopefully I will be able to "sneak" some time away tonight to do a quick test of Pin and Pout.
.99
PS. I have a different view of those 3 questions (I fully agree with MH on #3 though), and I'll perhaps elucidate on those a little later on.
My answers in a non-technical manner, I would say are as follows;
1a) Add another secondary in parallel, or,
1b) let's say the secondary was 20 turns, then remove the secondary and wind four 5 turn secondaries connected in parallel.
2) I would say because the core transfer is not required to move more then a few degrees since the windings are overlapping. If the core had two 180 degree wound coils non-overlapping, then the core is required to really move the flux 180 degrees. If each of the winds is going all around the core, then the flux only needs to just nudge and the transfer is complete. Making a cut in the core will make not much difference under the later.
3) Here is where I differ. I doubt if a 1.5 volt driven core will achieve core saturation of more then 40% given the amount of voltage/amperage that core could really handle. So my answer is no and no.
wattsup
Quote from: MileHigh on 2011.02.09, 02:35:23
I'm pretty sure that it's the leveling off in the current rise that causes the transistor to switch off, not the absolute amount of current flow. It's the change in the rate of change of the current that determines when the transistor starts to switch off. So that's the second derivative (the "accelerated acceleration") of the current with respect to time that is the determining factor.
MileHigh
I concur. The only thing I would add is the "slam" condition begins when reasonable second derivative < 0.
No time to perform the tests properly tonight.
Unless something unexpected happens, I'll properly attend to these tests tomorrow night ;)
.99
no worries p99, life takes precedence.
In the mean time let me pontificate a bit more :P
Ok, if the information provided on the toroid is accurate, the following can be calculated approximatly:
N = sqrt( 830uH/1.48uH) = 23 turns in the primary coil (rounded down) (is this close p99? can you count the turns?)
H_max = (23 turns * 0.1 amps max) / (2 * pi * 0.02 m) = 18 A/m (these are rough estimates of max current from graph, and effective radius of toroid from picture)
Now look at the B-H curve below, for the 3C90 material composing the core, and notice they only show the nonlinear saturating region, and right where the chart begins, H is around 18 A/m at 25 deg temp, basicaly, this is where saturation starts. The curve below what's shown on the chart would look pretty much like vertical lines (or high u values) which I'm sure some of you have seen. So, what do you know, saturation begins at about the same number I calculated !
Guys, you can belive me or not, but the saturation characteristics of the core is what dictates the end of the ON cycle. Yes, a bit of this and that do play a part, but by far it's the saturation characteristics of the core.
EM
P.S. Don't be confused by the B (mT) values on the graph, the zero value corresponds to the onset of saturation and this is a relative value, referenced to B_sat, not specified explicitly.
Quote from: EMdevices on 2011.02.10, 04:08:39
no worries p99, life takes precedence.
In the mean time let me pontificate a bit more :P
...
Guys, you can believe me or not, but the saturation characteristics of the core is what dictates the end of the ON cycle. Yes, a bit of this and that do play a part, but by far it's the saturation characteristics of the core.
EM
...
This variation of the Blocking Oscillator is sometimes referred
to as a "flyback converter" or a "ringing choke converter."
"It is a blocking oscillator in which the recovery time is governed
by an L/R time constant rather than the more familiar RC time
constant used in circuits formally designated as blocking
oscillators. The transformer core of the flyback converter does
not saturate. However, in some designs core saturation does
occur. In either case the diode in the secondary circuit isolates
the converter from the load as the current ramp is developing
in the primary winding during the transistor on time. This current
ramp eventually terminates regeneratively when the transistor can
no longer supply the demanded current (or, in alternate designs,
when core saturation occurs.) The collapsing field then induces
secondary current that, because of its polarity, is delivered to the
load. A unique feature of this circuit is that the peak voltage exceeds
that corresponding to the transformer turns ratio - ordinarily by a
factor of three or four, but sometimes by as much as eight."
Irving M. Gottlieb
Power Supplies, Switching Regulators, Inverters & Converters
1976, 1977, 1984 Tab Books Inc.Whether or not the core saturates magnetically is dependent
upon circuit design. Core saturation is not necessary and is
most often not desired.
It can be argued that one can learn more about the Joule Thief around here than one can on the nine-hundred and eighty-six pages of Joule Thief discussion somewhere else! lol
EM,
Thanks for the pontification. ;D
I've listed two core types on my diagram; one blue, one white. These are two core types that I have on hand, and I used the blue one to construct my unit.
Unfortunately the blue core is not a Ferroxcube part, but I've seen elsewhere that the material is similar to 3E4 (as noted on my diagram "Like 3E4 material, probably Hi u"). This is a SANLIN part, and I was not able to find any good specs on it.
The white core I have is a Ferroxcube part, but I have not tried it yet. It is a smaller core.
Note also, that the inductance specs given at the top of my diagram were measured before I cut into the core to introduce a significant gap in order to lower the permeability. I succeeded and managed to increase the Fo by a factor of 9 or so.
Once I receive that part kit from Lawrence, I will try to determine an "AL" spec for his core and my two core types as well. I am certain that my cores are much higher u than his. The white TN16/9.6/6.3 3C90 cores are supposed to have an AL of about 1480nH, but as I've mentioned, I have not yet built one using this core.
.99
Hi .99,
I am aware of a very helpful program on calculating AL and u values of unknown toroidal cores. All what would be needed is to have an L meter and measure a 10 (or any preferred) turns of coil inductance wound onto the unknown core.
Here is the link, it is free and small: http://dl5swb.de/html/mini_ring_core_calculator.htm
Check its Tools in the upper Menu line, the first item in Tools is AL and permeability: these are calculated by the program from the measured L of known turns and the measured mechanical sizes (OD, ID, h).
I have found this very useful. In case you are already aware of this program, then sorry, perhaps others here may find also useful.
Gyula
Thanks gyula. :)
I was going to wind 10 turns on the core and measure the inductance, then divide by 10. That should give me "AL".
This looks like it could be a handy tool though. ;)
.99
Quote from: MileHigh on 2011.02.09, 15:04:07
I am actually a little puzzled here about the frequency increase. If you have a core where the number of turns in L1 and L2 are constant, and then you make a cut in the core, you are reducing it's cross-sectional area. Won't that make the the core saturate sooner but not affect the value of the inductance associated with L1 and L2? But it is apparent that cutting the core is decreasing L and therefore increasing Fo. I am having some trouble visualizing this because the current waveforms show a linear rise (see below), indicating core saturation is not a factor here.
Think core saturation at a certain absolute
B.
The
B in equation
B=u
H is relative
B.
L is reflected upon the relative
B.
When Poynt cut the core to reduced mu, he's altering relative
B.
Adding cuts or gaps in a core allows higher values of current before saturation is induced. It also helps to discharge remnant magnetism between charge cycles, allowing operation at higher duty cycles.
Then there is the subject of non-linear gapping, but that is for another chapter in the saga.
the AL value of the core is inductance per turns ^2 , so this is the applicable equation we need to use.
L = AL * N^2
Note: If the inductance is measured with a meter, who cares what the AL value is, this spec is given so we can approximatley calculate the inductance easily and quickly, but the measured value is the final word, unless of course we want to re-engineer and discover what materials we are dealing with.
I should say that cutting gaps into the toroid, as in breaking the magnetic path, will invalidate some calculations, but no matter what we do to the toroid will not affect the B_sat value, this is a material property. However, cutting up the toroid will affect what magnetic intensity we can generate given number of turns and current flow because this is dependent on the geometry, however reducing the height of the torroid does not affect the magnetic intensity like I've mentioned, but cutting gaps into the toroid, i.e. breaking up the flux path will most definitley affect the magnetic intensity since the reluctance has changed.
I'll be doing some testing hopefully over the next couple of hours after I solder some leads on to the assembled unit Lawrence was kind enough to send me via FedEx.
Other good news, the parts kit that Lawrence put in the mail finally arrived today! :) Thanks Lawrence.
It's going to be a fun weekend folks. O0
.99
OK, now having the OUTPUT scope issues taken car of, I wanted to retest my original circuit build. The first test is the same as where this circuit left off, with the collector LED clipped, and the only output taken from the secondary output LED.
secout_input_mean.PNG indicates an average INPUT power of 33.44mW.
secout_output_mean.PNG indicates an average OUTPUT power of 230.3mW/10 = 23.0mW.
n=68.8%
.99
Now let's look at the case where the secondary output circuit is cut open circuit, and the collector LED re-connected as the only load. Note that the LED goes through the 1 Ohm to ground.
colout_input_mean.PNG indicates an average INPUT power of 41.53mW.
colout_output_mean.PNG indicates an average OUTPUT power of 37.1mW.
n=89.3%
As you can see, it is far more efficient to load the JT from the collector than from a secondary winding with 110 Ohms series resistance.
Also of note, comparison of this unit with my P9901 air-core unit (http://www.overunityresearch.com/index.php?topic=729.msg10758#msg10758), you can see that a ferro-magnetic core is not used to make the device function (via core saturation), but rather to make it much more efficient, i.e. 90% vs. 50%. Q. How does it do that? A. By increasing the inductor Q.
.99
Good work POYNT
So to what do we attribute a turnoff mechanism with an air core?
In the ideal world, with an air core, there would be no turnoff mechanism as the inductor would draw greater current increasing drive and so on. In the practical world with a real battery with a internal resistance of 120 milliohms, there will be a limit to the current drawn by the inductor, allowing for a turnoff mechanism (when current goes steady state, drive disappears)
(edit: I'm also discussing an ideal circuit with no emitter resistor to limit current. If you are testing with the one ohm emitter resistor, that will also act as a current limiter)
Be interesting to simulate the air core version with zero battery impedance and zero winding resistance to check this.
Poynt:
Thanks for redoing the captures. If I can make a request it would be to do a set of captures where you also look at the L1 output voltage that drives the base input resistor. This is the critical signal that's associated with the Joule Thief feedback mechanism that shows when the transistor is switched on and off. Perhaps doing this for the "standard" LTJT configuration would be informative for your readers.
MileHigh
Quote from: poynt99 on 2011.02.08, 02:17:35
Alright, I didn't like what resulted above, so it was time to clip that energy-sucking collector LED out of the circuit, and prove that the output will indeed go up significantly without it.
013.png is indicating an average INPUT power of 32.88mW (a decrease from the last run).
015.png is indicating an average OUTPUT power of 228.3mW/10 = 22.8mW (a more than 4-fold increase over previous tests!).
Now the input to output efficiency is at about 69%, and yes, the output LED became much brighter.
.99
It just occurred to me to ask -- Did you account for the Power lost in MEASURING the current in the INPUT circuit, using (I presume) a low-ohm resistor?
That power dissipation in the measuring-resistor may be small, but should IMO be accounted for. I would use V*V/R to determine the power lost in the measuring-resistor, and subtract this from the "INPUT Power". Would you check this, please? (Perhaps you handled this already and I missed it?)
Quote from: PhysicsProf on 2011.02.23, 07:49:20
It just occurred to me to ask -- Did you account for the Power lost in MEASURING the current in the INPUT circuit, using (I presume) a low-ohm resistor?
That power dissipation in the measuring-resistor may be small, but should IMO be accounted for. I would use V*V/R to determine the power lost in the measuring-resistor, and subtract this from the "INPUT Power". Would you check this, please? (Perhaps you handled this already and I missed it?)
Hi Professor.
Yes, you are absolutely correct; the power in the current sensing resistor, and any other resistors in the power loop should have their power dissipation accounted for also.
In this case, the power in the 10 Ohm and 100 Ohm were not accounted for, and so in reality, the efficiency would be higher than the 69%. I would estimate it is actually between 80% to 90% if those two resistors are included. The best way to make the actual measurement, would be to replace the 110 Ohms with a single 1 Ohm as you have done with your tests.
I did not perform this measurement because it was not part of the claim Lawrence is making.
If you wish, I can do this test with a 1 Ohm installed. Let me know.
.99
Wait -- how did you measure the CURRENT coming from the battery, the current in the "input" or JT circuit -- what resistor did you use there?
Let's start with that.
Quote from: PhysicsProf on 2011.02.23, 15:21:41
Wait -- how did you measure the CURRENT coming from the battery, the current in the "input" or JT circuit -- what resistor did you use there?
Let's start with that.
The INPUT power was measured as per my diagram on the first page in this thread. Battery voltage and battery current (via the 1 Ohm in series with the transistor emitter) were measured and multiplied in the scope. Mean power was also shown.
This was performed with the collector LED clipped, and thus, not in the circuit. It was therefore not skewing the results.
.99
Quote from: poynt99 on 2011.02.23, 19:30:41
The INPUT power was measured as per my diagram on the first page in this thread. Battery voltage and battery current (via the 1 Ohm in series with the transistor emitter) were measured and multiplied in the scope. Mean power was also shown.
.99
Right -- that's what I thought. Now, did you calculate and subtract the power lost to the input-JT circuit due to power dissipation in this 1 ohm resistor?
I would like to know how much power is dissipated in the 1ohm measuring resistor, and how much power is consumed in the remainder of the input circuit. I don't know how to make my question any more clear.
Quote from: PhysicsProf on 2011.02.23, 20:35:14
Right -- that's what I thought. Now, did you calculate and subtract the power lost to the input-JT circuit due to power dissipation in this 1 ohm resistor? I would like to know how much power is dissipated in the 1ohm measuring resistor, and how much power is consumed in the remainder of the input circuit. I don't know how to make my question any more clear.
I'm trying my best here Professor. :)
Hopefully this will help. I didn't include the power in Rb as it is quite small, but it can be measured as well.
.99
Quote from: poynt99 on 2011.02.24, 02:04:26
I'm trying my best here Professor. :)
Hopefully this will help. I didn't include the power in Rb as it is quite small, but it can be measured as well.
.99
Yes, the power in Rb is quite small, but needs to be quantified since this resistor is intended only for measurement and is not a necessary part for the circuit to run. I find that it's presence (or absence) does affect the frequency at which the circuit operates, for example. And if there were a way around it, one would like to make measurements without perturbing the system.
At present, given the tiny currents in your DUT, it seems the best we can do is to be
quantitative in calculating the power dissipated in Rb, and that is what I have asked of you. I haven't seen any numbers yet, and request (again) that you provide the numbers --
QuoteI would like to know how much power is dissipated in the 1ohm measuring resistor, and how much power is consumed in the remainder of the input circuit.
"Quite small" is a start but is qualitative; a quantitative measurement is what I'm requesting, and you say this "can be measured." Please do. The numbers leading up to "Pcore" in your set of measurements would also be enlightening, if you would please share the numbers with us. Thanks!
Quote from: PhysicsProf on 2011.02.24, 14:24:14
Yes, the power in Rb is quite small, but needs to be quantified since this resistor is intended only for measurement and is not a necessary part for the circuit to run. I find that it's presence (or absence) does affect the frequency at which the circuit operates, for example. And if there were a way around it, one would like to make measurements without perturbing the system.
At present, given the tiny currents in your DUT, it seems the best we can do is to be quantitative in calculating the power dissipated in Rb, and that is what I have asked of you. I haven't seen any numbers yet, and request (again) that you provide the numbers --
"Quite small" is a start but is qualitative; a quantitative measurement is what I'm requesting, and you say this "can be measured." Please do. The numbers leading up to "Pcore" in your set of measurements would also be enlightening, if you would please share the numbers with us. Thanks!
Professor,
Rb, the 1k base drive resistor exhibits an average power dissipation of only about 0.5mW. This amounts to less than 1% of the total input power, so I have never really been concerned that not including it in the computations would make or break any claims, one way or the other.
It is possible to measure the power in Rb, but it is not all that straight forward. What one needs to do is use two scope probes to measure Vb and the input side of Rb so that we obtain the instantaneous voltage across Rb. We need to use the scope to take the difference between the two probe measurements to obtain VRb (voltage across Rb). The scope would need to square this math function to get VRb
2/Rb, and I am not certain many scopes will do this. For less than 1% of the total power, I do not feel this is worth all the effort making this measurement.
I thought you were more interested in the power dissipation in the current sense resistors, CSR1 and CSR2, and I have shown how to make those measurements. I have not made the measurement on these two resistors yet, and that will have to wait for the weekend.
.99
Quote from: poynt99 on 2011.02.25, 02:38:52
Professor,
I thought you were more interested in the power dissipation in the current sense resistors, CSR1 and CSR2, and I have shown how to make those measurements. I have not made the measurement on these two resistors yet, and that will have to wait for the weekend.
.99
This part is correct, I misunderstood your definition of Rb earlier -- I await your measurements this weekend.
My question stands:
QuoteNow, did you calculate and subtract the power lost to the input-JT circuit due to power dissipation in this 1 ohm resistor? I would like to know how much power is dissipated in the 1ohm measuring resistor, and how much power is consumed in the remainder of the input circuit. I don't know how to make my question any more clear.
Quote from: poynt99 on 2011.02.13, 18:48:58
Now let's look at the case where the secondary output circuit is cut open circuit, and the collector LED re-connected as the only load. Note that the LED goes through the 1 Ohm to ground.
colout_input_mean.PNG indicates an average INPUT power of 41.53mW.
colout_output_mean.PNG indicates an average OUTPUT power of 37.1mW.
n=89.3%
As you can see, it is far more efficient to load the JT from the collector than from a secondary winding with 110 Ohms series resistance.
Also of note, comparison of this unit with my P9901 air-core unit (http://www.overunityresearch.com/index.php?topic=729.msg10758#msg10758), you can see that a ferro-magnetic core is not used to make the device function (via core saturation), but rather to make it much more efficient, i.e. 90% vs. 50%. Q. How does it do that? A. By increasing the inductor Q.
.99
This is interesting. When you take measurements, please include a measurement of the power dissipated in the input-current-measuring-1-ohm-resistor [CSR1 -- did I get your definition correct this time?] for the above circuit.
That is, please re-calculate:
colout_input_mean.PNG indicates an average INPUT power of
41.53mW.
when you have subtracted the power dissipated in this current-measuring resistor; it is not essential to the operation of the circuit, but is useful for measuring the current.
And then re-calculate n, if you would.
Looking forward to your measurements.
I would add that I'm personally learning a lot from a study of this "simple" JT circuit, using my own oscilloscope now. The V*I math function on it is particularly useful.
And I'm learning a lot from the civil discussion we're having, and for that I thank you (particularly .99).
OK, so I did with my own circuit what I'm asking .99 to do.
I disconnected the "LT" part of the LTJT circuit, so just looking at the JT circuit. See attached photo of the set-up.
Connections across the "input" CSR we've been discussing (CSR1) and across the battery. The green waveform on the DSO is the total power input (Vbatt * I-csr1).
Next:
Red LED from the collector then across a 1-ohm measuring resistor (call it CSRb). Result is shown in the second photo.
It's a bit rough to calculate the energy into the circuit and the energy out to the LED-CSRb -- my DSO does not give the MEAN for the math function. I calculate energy as explained previously, taking the area under the green Power waveform, for one cycle. I hope .99 will do the measurement the other way this weekend... I may try to go to the University and use the Tek 3032 also...
Long story short, I gotta run! but I find that the NET input power to the circuit is a little less than the output into the LED-CSRb ...
I probably made some mistake. n is about 1.3 weird...
really gotta run
:o my mechanical model predicted a theoratical 1.5
I'll be re-testing this one as well, now that I've nailed down a good procedure.
Stay tuned...today.
.99
I've done a test using the above circuit, and I've realized a much better and simpler way to obtain the actual battery power and actual LED power, than what I proposed here (http://www.overunityresearch.com/index.php?topic=729.msg10831#msg10831). The power dissipation in the CSR resistors is also easily obtained. O0
QuoteNotes:
1) P11= probe, scope 1, CH1. P12= probe, scope 1, CH2. P21= probe, scope 2, CH1. P22= probe, scope 2, CH2.
2) INPUT power is obtained as follows:
2.1) Use the scope MATH to produce MEAN[(V1*V2)]. This is the total power of Vbat and CSR1 together. We will call it Pitotal.
2.2) Use the scope MATH to produce MEAN[(V2*V2)]. This is the power of CSR1 alone. We will call this Pcsr1.
2.3) Pvbat is computed by: Pitotal - Pcsr1.
3) OUTPUT power is obtained in much the same manner:
3.1) Use the scope MATH to produce MEAN[(V3*V4)]. This is the total power of LED and CSR2 together. We will call it Pototal.
3.2) Use the scope MATH to produce MEAN[(V4*V4)]. This is the power of CSR2 alone. We will call this Pcsr2.
3.3) Pled is computed by: Pototal - Pcsr2.
4) The efficiency n of the battery power delivered to the LED alone is then: n = 100(Pled / Pvbat).
5) If we were to insert a 1 Ohm CSR resistor in series with the transistor emitter, we could measure and calculate the transistor power dissipation in a similar fashion. We already have the voltage across the transistor, V3, so the other required voltage measurement would be that of a CSR3 resistor labeled "V5". Note, V5 is not the same as V2.
The test results for my LTJT with a core and no secondary circuit (normal JT) are as follows:
Pintotal = 29.05mW
Pcsr1 = 1.72mW
Pvbat = 27.33mW
Pototal = 31.3mW
Pcsr2 = 0.66mW
Pled = 30.64mW
n(vbat to led) = 30.64/27.33 =
112%So either I have a measurement anomaly, or..... ;D
.99
Does anyone see any flaw in the measurement method?
It seems correct to me.
.99
I see a possible flaw in the calculation for the power dissipation in the LED. The forward resistance is not constant as the LED is being pulsed. I think the the duty cycle needs to be factored into the calculation.
Hoppy
Hoppy,
Are you aware of the method being used for these tests? We are using the scope to multiply instantaneous voltage and current to get instantaneous power. Then we use a "measurement" function in the scope to provide the MEAN of that instantaneous MATH power trace.
All phase skewing and duty cycle allowances are taken care of by this method. There is no need to "do" anything, other than make the final calculation for efficiency.
Make sense?
.99
Yes, I understand the method being used but I don't understand the MEAN[(V3*V4)} expression as defining the LED power dissipation.
Hoppy
How are you accounting for the fact that CSR1 current flow
is both "input" (transistor on) current and also "output"
(transistor off) current?
Is there any way to fully isolate the two?
Quote from: Hoppy on 2011.02.27, 22:16:48
Yes, I understand the method being used but I don't understand the MEAN[(V3*V4)} expression as defining the LED power dissipation.
Hoppy
Hoppy, it isn't.
Check that line again:
Quote3.1) Use the scope MATH to produce MEAN[(V3*V4)]. This is the total power of LED and CSR2 together. We will call it Pototal.
.99
OK, I have found at least one problem. The INPUT power Pitotal must be ADDED to Pcsr1, not subtracted. This is because the power in the two are in opposite directions, and hence sign. So, let's try this again:
The corrected test results for my LTJT with a core and no secondary circuit (normal JT) are as follows:
Pitotal = 29.05mW
Pcsr1 = 1.72mW
Pvbat = 30.77mW
Pototal = 31.3mW
Pcsr2 = 0.66mW
Pled = 30.64mW
n(vbat to led) = 30.64/30.77 = 99.5%
This still seems rather high. Surely there is more than 0.5% of the power being dissipated in the transistor and base resistor...
.99
Yes thanks, I have mis-read this line.
Hoppy
Quote from: Dumped on 2011.02.27, 22:27:40
How are you accounting for the fact that CSR1 current flow
is both "input" (transistor on) current and also "output"
(transistor off) current?
Is there any way to fully isolate the two?
We are interested in the net power delivered by the battery, and the net power dissipated in the LED. Having a current sense resistor in series with each along with the voltage across each (including it's CSR) allows us to determine the power in each.
Does that make sense?
.99
Quote from: poynt99 on 2011.02.27, 22:45:34
We are interested in the net power delivered by the battery, and the net power dissipated in the LED. Having a current sense resistor in series with each along with the voltage across each (including it's CSR) allows us to determine the power in each.
Does that make sense?
.99
Yes, it makes sense if CSR1 is included in the "input"
inductor charge current path,
then
included with CSR2 for the "output" inductor discharge
path where the "source cell" and the "Inductor"
work series aiding to furnish "load" current which
puts the two resistors in series for that time.
Quote from: Dumped on 2011.02.27, 23:10:50
Yes, it makes sense if CSR1 is included in the "input"
inductor charge current path,
then
included with CSR2 for the "output" inductor discharge
path where the "source cell" and the "Inductor"
work series aiding to furnish "load" current which
puts the two resistors in series for that time.
So, what is your analysis...does it work correctly as I have it set up?
.99
Quote from: poynt99 on 2011.02.27, 20:28:32
I've done a test using the above circuit, and I've realized a much better and simpler way to obtain the actual battery power and actual LED power, than what I proposed here (http://www.overunityresearch.com/index.php?topic=729.msg10831#msg10831). The power dissipation in the CSR resistors is also easily obtained. O0
The test results for my LTJT with a core and no secondary circuit (normal JT) are as follows:
Pintotal = 29.05mW
Pcsr1 = 1.72mW
Pvbat = 27.33mW
Pototal = 31.3mW
Pcsr2 = 0.66mW
Pled = 30.64mW
n(vbat to led) = 30.64/27.33 = 112%
So either I have a measurement anomaly, or..... ;D
.99
Interesting, and in the same direction as my n=1.3 measurement on Friday.... Must be thoroughly checked!
QuoteQuote from: poynt99 on 2011-02-27, 23:51:41
OK, I've discovered that Pitotal must be added to Pcsr1, not subtracted. ...
.99
No, I think you were right the first time. You measure the total input, then SUBTRACT the power dissipated in the measuring resistor. The power dissipated in the measuring-resistor is not available for the rest of the circuit.
What would the power be in the input circuit if you reduced CSR1 to 1/2 ohm? then to 1/4 ohm, etc.
Perhaps the most important aspect of this exercise is getting the measurements right, subtracting when one should subtract, getting the ground connections right, etc. I'm not sure you/we are there yet. But I'm encouraged by the results (not ecstatic yet).
Professor,
I am quite certain that all currents must go through the CSR which is in series with the battery in order to capture all that is either leaving or entering the battery. That is why your transistor emitter must be tied to the circuit ground as shown.
The addition or subtraction of the power in the CSR is not so easy to see. The voltage drop across that resistor is actually reversed from what you might think. It caught me too, and Humbugger thankfully corrected me on that. So actually we are subtracting the two powers, but it so happens that one of them is a "negative" power so to speak.
You can think of the battery CSR as an internal resistance in the battery. In such a case, this resistance is part of the total power dissipation in the battery. It is the same case here.
Perhaps someone else can explain it better, I am quite properly burnt out at the moment. :D
If you are still in much doubt professor, I can show you with a simulation where it is quite clear.
.99
Yes, the power from CSR1 comes from the battery to acquires the output. However, if we reduced or cut that power out, we can save some energy and have the same output. The professor is right.
On the other hand, I think we have a problem with the method. CSR1 is subtracted from the whole cycle means it's being subtracted both the inductor charging and inductor discharge cycle. On the inductor charging cycle, the battery provide power to charge the coil and CSR1. On the discharge cycle, we have LED, CSR1, and CSR2 but... also there are two power source: inductor discharge and battery EMF. Therefore, we must add 1/4 CSR1 power to the output, 1/2 LED power to input, 1/2 CSR2 to input. How did I come up with this? This is just how I roll. ^-^ lol
OK guys, let's start with the basics.
In order to determine the average power dissipation in a circuit component, we must measure the voltage across, and current through that component. If we have a running instantaneous product of those two measurements, which is p(t), we can have the oscilloscope compute the MEAN of p(t) to indicate the average power.
Are we in agreement on that?
.99
Quote from: poynt99 on 2011.02.28, 01:03:43
OK guys, let's start with the basics.
In order to determine the average power dissipation in a circuit component, we must measure the voltage across, and current through that component. If we have a running instantaneous product of those two measurements, which is p(t), we can have the oscilloscope compute the MEAN of p(t) to indicate the average power.
Are we in agreement on that?
.99
That's one method, valid yes, but not the only method. Can we also agree that the resistor will HEAT UP and we can measure the rate of heating to determine the power dissipated in the resistor? Further, is this heat from the resistor CSR1 going to add to the light output of the LED? (Certainly not.)
Also, .99, please PREDICT what will happen to n when you replace CSR1 with a 1/2 ohm resistor... use your simulation if you wish. THEN make the measurement with a 1/2 ohm CSR1.
I should have been more specific, sorry. ;)
Using the oscilloscope method (so far your questions have only pertained to oscilloscope measurements), are we in agreement with the above?
.99
Quote from: poynt99 on 2011.02.28, 01:03:43
OK guys, let's start with the basics.
In order to determine the average power dissipation in a circuit component, we must measure the voltage across, and current through that component. If we have a running instantaneous product of those two measurements, which is p(t), we can have the oscilloscope compute the MEAN of p(t) to indicate the average power.
Are we in agreement on that?
.99
As I said, I am in agreement with this method, with the other questions I raised unanswered as yet.
Hopefully, Gibbs is in agreement as well.
I would hope that the efficiency will increase as the value of CSR decreases.
.99
Quote from: poynt99 on 2011.02.28, 02:19:17
Hopefully, Gibbs is in agreement as well.
I would hope that the efficiency will increase as the value of CSR decreases.
.99
what the heck, leave me out of this. :P You and prof do your thing. I have no problem. lol I will chime in and out like a magic man. ;D
OK, fine...be like that Gibbs. :P
Professor, et al:
It is very constructive to think of a battery as "a reservoir" that loses energy, and every other component in the circuit as "sinks" that gain energy. In this respect, think of the energy given up or lost from the battery as negative energy, and energy gained or dissipated by the circuit components as positive energy.
The CSR placed in series with the battery is a component of the circuit and can be thought of as a load actually. Therefore it GAINS energy. Meanwhile, our battery loses energy (most of the time).
Normally, we would have no issue conceiving of the power lost from the battery; it is simply the battery voltage times the battery current. If we used a current probe to measure the battery current, this would all be quite straight forward. However, we have resorted to using current sense resistors (CSR's), at least for the moment. The CSR placed in series with the battery (it matters not if the CSR is in the negative or positive leg of the battery) allows us to gauge the battery current, but in doing so it presents itself as a small load that eats up a bit of power. In the method I proposed to use for obtaining both the battery power and CSR1 power, the first measurement involves acquiring a power measurement that captures the voltage across both the battery AND CSR. I called this power Pitotal. This of course will not provide the true battery power figure we are looking for, so we have to subtract the CSR power from Pitotal in order to obtain the battery power figure. We know that the battery power is negative and it is much larger in amplitude than the CSR power, therefore we can safely assume that Pitotal will also be a negative value because the battery power figure dominates. We can therefore write:
Pbattery = (Pitotal) - (Pcsr1)
establishing the correct power polarities as discussed, we have:
-(Pbattery) = -(Pitotal) - (+Pcsr1)
if Pitotal measured to be 30mW and Pcsr1 measured to be 3mW, then we have the following:
-Pbattery = -30mW - 3mW
-Pbattery = -33mW
.99
We can look at it another way.
Let's assume we have a battery and 3 circuit components (C for component, not capacitor); C1, C2, and C3. The power of each is Pbat, PC1, PC2, and PC3. There are no other losses.
Let's assume that PC1=3mW, PC2=20mW, and PC3=10mW. We know that Pbat must therefore =33mW. So we can write the power balance equation as follows:
Pbat = PC1 + PC2 + PC3
33mW = 3mW + 20mW + 10mW
If we decide to capture the voltage across the battery AND one of the circuit components in the same measurement because they are in series, say PC1 for eg., then we have to rearrange the power balance equation;
Pbat - PC1 = PC2 + PC3
and substituting in the values:
33mW - 3mW = 20mW + 10mW
30mW = 30mW
The true battery power though is 33mW. ;)
So you see that although it may seem strange that we actually added the CSR1 power to the battery in the above example, mathematically, it is sound. This example works backwards from where we started in the actual test.
Hope that helps.
.99
I must be really slow or something tonight.
You write:
QuoteI would hope that the efficiency will increase as the value of CSR decreases.
.99
Now for me, I would hope that the calculated efficiency will STAY THE SAME as the value of the measuring resistor CSR decreases. That's the whole point of this exercise of subtracting out the effect of the CSR, in order to determine the COP / efficiency with the measuring resistor removed. Can you agree with that?Next,
You write,
QuoteLet's assume that PC1=3mW, PC2=20mW, and PC3=10mW. We know that Pbat must therefore =33mW. So we can write the power balance equation as follows:
Pbat = PC1 + PC2 + PC3
33mW = 3mW + 20mW + 10mW
Suppose we want to determine the input Power, Pin, in the case where there is no CSR1. That would be handy, wouldn't it? save us some trouble.
But with this method we need a measuring resistor to get the current. So next best thing, we reduce the ohms in stages, from say 1 ohm (giving 3mW dissipation) to 1/2 ohm (giving 1.5 mW lost) to 1/10 ohm (giving just 0.3 mW dissipated, essentially negligible).
Having done that, down to just 0.3 mW in CSR1, what is the input Power Pin?
And, for the sake of discussion, let's say the output Power stays constant somehow, at 20 mW.
So, what do you get for the COP = Pout/Pin for the two cases? Is your n the same as COP?
One of us is missing something, and I hope your answers will clarify -- starting with my first question --
Now for me, I would hope that the calculated efficiency will STAY THE SAME as the value of the measuring resistor CSR decreases. That's the whole point of this exercise of subtracting out the effect of the CSR, in order to determine the COP / efficiency with the measuring resistor removed. Can you agree with that?
Quote from: GibbsHelmholtz on 2011.02.28, 00:45:15
Yes, the power from CSR1 comes from the battery to acquires the output. However, if we reduced or cut that power out, we can save some energy and have the same output. The professor is right.
On the other hand, I think we have a problem with the method. CSR1 is subtracted from the whole cycle means it's being subtracted both the inductor charging and inductor discharge cycle. On the inductor charging cycle, the battery provide power to charge the coil and CSR1. On the discharge cycle, we have LED, CSR1, and CSR2 but... also there are two power source: inductor discharge and battery EMF. Therefore, we must add 1/4 CSR1 power to the output, 1/2 LED power to input, 1/2 CSR2 to input. How did I come up with this? This is just how I roll. ^-^ lol
Oops, I made a mistake. :-\ Sorry Poynt, this is why. I don't like to give people false hope.
I did my analysis based on 2 separate cycles (inductor charging and discharging), but the input power is already took account for both cycles. Therefore, the only thing left is add that 1/4 CRS1 to the ouput. Wow... I have to think about this.
Poynt et al,
As you include more and more miniscule dissipation components in your measurements and calculations, it is only natural that your measurements will converge on 100% efficiency. If you include the resistance of the inductor winds, the core loss, the transistor dissipation and all, you will certainly end up with 100% efficiency every time.
As you approach perfect unity COP, your measurement error will begin to dominate over the ever-decreasing unaccounted-for loss percentage. Remember, the scopes are 8-bit and, after doing scope-math, I'd be surprised if you had better than 2 or 3% accuracy, considering you are multiplying two numbers each with an unknown quantization error and offset error and probe cal errors, too.
For instance, to be absolutely proper, if you are going to include the shunt losses, you should equally consider the battery impedance. Just because the loss is internal to the battery itself, it still makes the measured "input power" appear to be less than it actually is (from the ideal voltage source within the battery) if you ignore the internal battery loss. Using worn-down penlight cells, you might be surprised at how high the internal resistance is.
I think we can all safely say, at this point, that the claims of Lawrence citing COPs (or FLEETs) of hundreds of percents are completely disproven. As we finagle over the last few percentage points above or below 100%, even the best scope measurements are going to be inadequate. With so much effort being put in by all parties involved, this casual onlooker is beginning to wonder what the point is.
Other than an exercise in finding the limits of accurate measurement using scopes and in accounting for and identifying every tiny loss factor, that is. Plus, of course, its kind of fun, iand a good exercise in measurement approach and communication. Soon you'll be breaking out the 0.01% tolerance Kelvin-sensing shunts and the NIST-traceable calibrated ten-digit accurate DVMs ;D
Humbugger
P.S. [edit] Please don't take my comments as being demeaning or dismissive. I think the group is making very useful solid progress in eliminating confusion and errors generally found in these kinds of exercises. It's a good thing, as Martha Stewart might say.
Quote from: PhysicsProf on 2011.02.28, 03:45:24
I must be really slow or something tonight.
Not at all, you are asking good questions.
Quote
Now for me, I would hope that the calculated efficiency will STAY THE SAME as the value of the measuring resistor CSR decreases. That's the whole point of this exercise of subtracting out the effect of the CSR, in order to determine the COP / efficiency with the measuring resistor removed. Can you agree with that?
We first need to establish what YOU mean by efficiency. As far as I know, we have only alluded to what the metric for
n is in our tests, and I have always assumed it is based on the LED power in relation to the INPUT power. If this is what you mean also, then we are on the same page, and we can continue with an answer to your question.
In the last or second last post, I mentioned that the CSR1 resistor that is in series with the battery is actually a small load on the battery, i.e. it dissipates real power. The amount of power it dissipates is related to the voltage across it and it's value of resistance. For a certain, fairly wide range, the larger it's resistance, the more power it will eat up. As it eats up more power, there is less power available for the intended load, i.e. the LED, therefore the efficiency n must decrease. Does that makes sense and answer your question as to why the efficiency will go up as the value of CSR1 goes down?
Quote
So next best thing, we reduce the ohms in stages, from say 1 ohm (giving 3mW dissipation) to 1/2 ohm (giving 1.5 mW lost) to 1/10 ohm (giving just 0.3 mW dissipated, essentially negligible).
Having done that, down to just 0.3 mW in CSR1, what is the input Power Pin?
And, for the sake of discussion, let's say the output Power stays constant somehow, at 20 mW.
The output power to the LED can not remain constant if other components in the circuit are changed to consume more power, as I described above. Remember what happens to the collector LED when you reconnect the LED load that is in the secondary circuit, which we have both disconnected in favour of the general JT connection?
I hope things are becoming clearer. 8)
.99
Hum,
At this point, we are interested only in determining the battery power and the LED power. The measurement I made today indicates an n of 99.5%. That figure does not include the CSR2 nor any of the other circuit components such as the transistor etc. I think our goal is to factor those out so we only have PLED/PBAT, and nothing more.
Any ideas why my n measurement came out so high?
Thanks,
.99
Quote from: poynt99 on 2011.02.28, 04:24:07
Hum,
At this point, we are interested only in determining the battery power and the LED power. The measurement I made today indicates an n of 99.5%. That figure does not include the CSR2 nor any of the other circuit components such as the transistor etc. I think our goal is to factor those out so we only have PLED/PBAT, and nothing more.
Any ideas why my n measurement came out so high?
Thanks,
.99
Zipons?
;D
I'll try the measurement again tomorrow if I have time, and this time I'll post the wave forms. In the mean time, do you see any flaws in my method for getting the power measurements?
.99
Good comments -- we're getting closer I think. Will digest and respond tomorrow.
Meanwhile --
Quote from: poynt99 on 2011.02.28, 04:40:05
;D
I'll try the measurement again tomorrow if I have time, and this time I'll post the wave forms. In the mean time, do you see any flaws in my method for getting the power measurements?
.99
Great -- now, could I convince you to replace the 1 ohm CSR1 with a 1/2 ohm, so we can see empirically
what happens to the calculated efficiency?
BTW, I would consider the 1 ohm CSR2 in your schematic to be part of the output load, and the interest to me would be COP = Pout/Pin, where Pout is Pled + Pcsr2, and Pin is still under discussion.
PS -- I still seek a way to calculate the COP that does not depend on our choice of resistance for CSR1, which after all is only there so that one can evaluate the input current.
Poynt:
I have only skimmed through this but I can make one recommendation. Since you briefly fluttered into over unity territory and are getting numbers like 99.9%, I think it's time to measure the values of your shunt resistors with your best multimeter if you haven't done so already.
You have Rosemary in a tizzy! :D
MileHigh
Quote from: PhysicsProf on 2011.02.28, 05:32:13
...
BTW, I would consider the 1 ohm CSR2 in your schematic to be part of the output load, and the interest to me would be COP = Pout/Pin, where Pout is Pled + Pcsr2, and Pin is still under discussion.
...
Since during the output pulse time interval CSR1 and CSR2 are series
connected within that discharge loop, would not the total output
power be:
Pout = Pled + Pcsr2 + Pcsr1
or, since CSR1 = CSR2,
Pout = Pled + 2(Pcsr2)
?
I would tend to agree with both of you, however, we can keep adding components on and on. At some point we have to draw the line.
CSR1 is a necessary evil at the moment, but soon I hope to show that it can be substantially reduced in size (while still obtaining an accurate measurement) so as to minimize these losses.
.99
Quote from: poynt99 on 2011.02.28, 13:21:56
I would tend to agree with both of you, however, we can keep adding components on and on. At some point we have to draw the line.
CSR1 is a necessary evil at the moment, but soon I hope to show that it can be substantially reduced in size (while still obtaining an accurate measurement) so as to minimize these losses.
.99
Ah, here we are getting somewhere -- reduce CSR1 as I have repeatedly requested so that it's effect on the system is finally insignificant.
That is the solution I've proposed and I think it is a good one.
After all, the measuring resistor should be a minor perturbation on the DUT so that we do not need to quibble over its effects on the measurement.
Humbugger wrote:
Quote
As you include more and more miniscule dissipation components in your measurements and calculations, it is only natural that your measurements will converge on 100% efficiency. If you include the resistance of the inductor winds, the core loss, the transistor dissipation and all, you will certainly end up with 100% efficiency every time.
I detect a certain bias here that is disconcerting. We are trying to determine EXPERIMENTALLY whether or not a simple system can demonstrate overunity, more power out than in. It is not a foregone conclusion that as we include "miniscule dissipation components in your measurements and calculations, it is only natural that your measurements will converge on 100% efficiency." That is what we are in process of determining, experimentally. Not by pre-determined conclusion without the need for experiments ("Ipse dixit" authoritarian style).
There is something else here that needs to be considered also -- we may vary the circuit components such as the toroid windings and get "better" results. That is, if there is a real effect here, we may expect the COP to move further above 1. I don't know whether .99 is willing to keep trying things; but I am.
Meanwhile, I have a card in hand I need to disclose... I took my little JT circuit which showed n over unity last week, to the University to test on the Tektronix 3032. My colleague allowed me to use his scope; unfortunately, I did not have access to his computer to allow me to upload records from the 3032 to the computer. But I did take a photo of the set-up (attached). The red waveform represents the POWER.
I took measurements much as did .99 later; however, as I said before, the output from the emitter I put into the circuit at the point he labels V2 rather than going directly to ground. I will test his exact set-up later hopefully this week as I travel back to the University.
You're not the only one taking measurements, .99... Here are the results in brief from last Friday, 25 Feb 2011. The input leg of my JT circuit showed mean Pin = 67.1 mW. This includes the power dissipated in the 1 ohm CSR1 which I estimated at 6%, or 4 mW.
The output leg of the JT, including the LED and a 1-ohm measuring resistor, showed mean Power Pout = 74.1 mW.
So the uncorrected
n = 74.1/67.1 = 1.10. If I subtract or add 4 mW from CSR1 to the Pin, still n >1.
But it is uncomfortably close to one. So I keep trying this and that, to see what the effect is on the COP. I have added a capacitor to the circuit which appears this weekend to increase the COP; results on that later as I can get back to the 3032.
IF the power in the output leg can be verified as greater than Pin, how is this possible? I still believe in conservation of mass-energy. So to me, this result would imply an input of energy from an unidentified source, rather than a violation of the laws of Physics.
And that would be interesting!
Indeed professor, it is strange to be hovering even around the 100% mark as in my second (corrected) calculation.
Hopefully, with the wave forms posted, and a little more time examining the computation method step by step, we can discover why the efficiency is coming out much higher than expected.
.99
Agreed. And I look forward to your next set of measurements with a lower-ohms CSR1 which should tell us much also.
Quote
Humbugger wrote:
"As you include more and more miniscule dissipation components in your measurements and calculations, it is only natural that your measurements will converge on 100% efficiency. If you include the resistance of the inductor winds, the core loss, the transistor dissipation and all, you will certainly end up with 100% efficiency every time."
I detect a certain bias here that is disconcerting. We are trying to determine EXPERIMENTALLY whether or not a simple system can demonstrate overunity, more power out than in. It is not a foregone conclusion that as we include "miniscule dissipation components in your measurements and calculations, it is only natural that your measurements will converge on 100% efficiency." That is what we are in process of determining, experimentally. Not by pre-determined conclusion without the need for experiments ("Ipse dixit" authoritarian style).
Yes, Professor, your "bias detector" is well calibrated. Please don't be too disconcerted, however. There are very good reasons for my "bias", as I hope you are aware. I make no pretense of having even a shred of doubt that the laws of energy conservation are entirely applicable here. I'm not engaged in the experiments except as an onlooker and occasional commenter.
I think that anyone educated in physics or electronics or mechanics probably, in their heart of hearts has a bias, as well. In general, to be biased toward a belief that the accepted "rules" of energy conversion and transfer apply to this circuit is quite healthy and normal from a science perspective. After all, the whole OU movement is looking for "exceptions" to the rules and, as far as I know, none have been demonstrated or applied in any practical device to date. If we weren't biased, why would the phrase "Exceptional claims require exceptional proof" be so prevalent? Do you not agree that claims of overunity performance are "exceptional"?
A healthy bias toward the enormous history and solid well-documented body of "conventional scientific knowledge" is the reason science today demands accurate investigation and well-designed experiment to vet exceptional claims, after all. Without that bias, the rigor of investigation and the practical advancements in technology that have resulted would definitely suffer, I believe.
Humbugger
P.S. I'm sure this response will win me no popularity contests here, but I'm not too worried about that. O0
My overall comment remains: I do not trust the accuracy of oscilloscope-based measurements using math on irregular waveforms to be much better than about 5% overall, and that's when immaculate bench practices are used. When all the error sources are added up, I'd much sooner trust a good multi-digit calibrated DVM looking at DC equivalent input and output currents and voltages derived by passive low-pass filters and precision shunts. Especially in the milliwat regions.
Hum,
I would agree. I also agree with the weariness of using 8-bit conversion on relatively low-level signals. The DC equivalent method is what I am shooting for next.
However, I am still surprised at the efficiency numbers both the professor and I are getting with this particular arrangement, and with different scopes. Flirting with the 100% area is quite unusual, isn't it?
Hopefully, additional careful testing will reveal what is going on.
.99
Yes, it does seem kind of strange getting so close to 100% n, but it wouldn't really surprise me much if the efficiency was as high as 95% at the low powers involved. I^2R losses are bound to be very minimal, the transistor base loss the same and I bet the Vsat of the transistor is pretty low under these conditions, too.
It might be interesting to take a close look at Vsat, which I'm guessing is the largest loss factor in the present circuit at the current levels involved. I mean just to get a ballpark as the the known losses there as a reference point. If you have 5 or 10% lost power in the transistor, you'll know to be surprised (spell that SUSPICIOUS) if you get n of 90-95% or more right off the bat.
You guys are doing a great job of whittling it down to the nitus gritonius by factoring in the shunts, etc. It is definitely a worthwhile exercise and I can see no glaring flaws anywhere. If the scope-based tests show n>95% or especially if they consistently show >100%, it's time to break out the RC filters, precision Kelvin shunts and a high-accuracy DVM.
Humbugger
Something to think about
All the parts in the circuit consume some power. Considering the input power vs output power is not a complete story.
If you can add up all the power used in the circuit by each component, those can be all added to the output figure, which is a good thing. ;]
Even a reed switch has a certain on resistance value that when it closes, voltage is present across that reed resistance, current flows, power dissipated.
The real figures may be better than realized. ;]
Mags
I've found something...
... but neither Kirchhoff or Faraday can handle this.
;D ;D ;D
Considered the JT pix I attached.
The charging and discharging cycle are as shown. We know that the current direction for both cycles are the same. All little minor details are omitted to see the bigger picture.
Now, let's start at point A and apply the "modify KVL" ( E + dB/dt coil = 0 ). Let's go counter clockwise for charging.
-dB/dt coil + Vbat = 0
Now for the discharging, let's go clockwise.
-Vbat +....woohoo, it's now positive dB/dt coil = 0 Phew...that was close call.
Let's try Faraday (E = dB/dt coil ). Let's go counter clockwise for charging.
Vbat = -dB/dt
Now for the discharging, let's go clockwise.
-Vbat = dB/dt ===> Vbat = -dB/dt !!
Good -- my digital power supply finally arrived today. I'm using a separate digital voltmeter to measure the output voltage more accurately; I'm wondering how reliable the built-in ammeter is for measuring the input current; it reads down to 1 mA. Perhaps as a starting point, it may give some information. More learning experience!
Ive been thinking more about the connection of the emitter direct to ground in .99's circuit, a concern I raised before. It seems to me that this represents output power which is not accounted for, just thrown away. I think we can do better. .99 at one point mentioned adding a 1 ohm resistor in series with the emitter -- and I think that is an important step to take. (Can you do it?) Call it CSR3. Would not Pcsr3 ADD to the output power and increase n?
Quote from: PhysicsProf on 2011.03.01, 05:37:20
Good -- my digital power supply finally arrived today. I'm using a separate digital voltmeter to measure the output voltage more accurately; I'm wondering how reliable the built-in ammeter is for measuring the input current; it reads down to 1 mA. Perhaps as a starting point, it may give some information. More learning experience!
It wouldn't hurt to double-check both the voltage and current with a known good DMM.
Quote
I've been thinking more about the connection of the emitter direct to ground in .99's circuit, a concern I raised before. It seems to me that this represents output power which is not accounted for, just thrown away. I think we can do better. .99 at one point mentioned adding a 1 ohm resistor in series with the emitter -- and I think that is an important step to take. (Can you do it?) Call it CSR3. Would not Pcsr3 ADD to the output power and increase n?
We need all the components' "return" paths to go through the CSR1, otherwise we will not obtain a true INPUT power figure. Perhaps redrawing the circuit slightly will shed some light on this. We don't need ground symbols in this circuit, so I have removed them. They only cause confusion it seems. You may now notice that CSR2 is in series with CSR1, but don't be concerned about that. Please also note that I have corrected the calculation for Pvbat in the diagram notes; i.e. Pitotal and Pcsr1 must be added together to obtain Pvbat.
Indeed I can use a 1 Ohm in the emitter to see what power our transistor is dissipating. I'll try this when I get to testing the circuit again, today or tomorrow.
.99
Quote.99:
Indeed I can use a 1 Ohm in the emitter to see what power our transistor is dissipating. I'll try this when I get to testing the circuit again, today or tomorrow.
Good. Call it CSR3. I have added this to my circuit and checked it briefly using my DSO at home. The power dissipated in this emitter 1-ohm resistor CSR3 is not negligible and should be included -- increasing n a small amount. My circuit now appears much like your schematic, except I've added CSR3.
Quote.99 Pitotal and Pcsr1 must be added together to obtain Pvbat.
I do not concede this point -- still awaiting data on what happens to Pin, Pout and n (COP) when a 1/2 ohm CSR1 is used instead of a 1 ohm resistor. I prefer data to theoretical discussions and I'm guided by the data.
In any case, as I've noted, as long as CSR1 is small (like 1/2 ohm), it matters little to the COP whether one adds or subtracts Pcsr1.
I have also observed an interesting change in the power waveform (across CSR1) when the voltage from my power supply (rather than battery) is decreased, down from 1.5V to about 0.6 V when the LED goes out. At about 1.1 volt (in my circuit, may vary with details of the toroid etc), the power curve goes from entirely positive to sometimes positive and sometimes negative. That is, the curve crosses the zero line and the waveform becomes more erratic (many more wiggles) as the voltage is decreased to about 1 volt. Clearly, the phase relationship between current and voltage changes, which I find interesting and perhaps significant.
It would be nice to have two oscilloscopes with math and also mean calculation on the math product (power)... Perhaps someday. ;)
I've been working with the circuit further. I find that the current flows through CSR3 then through CSR2, alternately -- they are both outputs yet they are out of phase (my DSO shows). Therefore, I connected the output of the emitter directly to the point labeled "p22t" (also "v4") in .99's latest circuit schematic. It makes some sense to measure the output current therefore in CSR2 alone, and CSR3 is not needed. Does this make sense?
CSR1 then gives the input current while CSR2 gives the output current (flowing from both the collector and the emitter), keeping things simplified.
A question for you, .99. In your latest circuit LTJT2 - schema02, you do not show any connection to ground -- which is a change from schema01. Why the change?
Hope you're finding interesting results in the lab, .99. O0 A couple more questions in the interim:
I've gone back through the thread trying to find details on your toroid and the windings you used... missed it somehow.
Can you repeat a few details on the toroid used and bifilar windings? size of wire, etc. -- for the ferrite-core toroid.
Did you measure the inductance for each winding? which I presume are close to the same.
(Or if someone has the data, please re-post them... )
This is my first post here and it is this thread that has got me here. I have some idea to share with you guys concerning the measurement. Please let me borrow the schematic attached to Reply #124 of this thread.
As professor already pointed out, csr1 and csr2 gives the input and output currents, very nice step of simplification (in math: Ibat = Icsr1 = V2, Iled = Icsr2 = V4, see attached schematic for definition of V). O0
And it is also great that .99 argued for Pbat = Pitotal + Pcsr1. I fully agree with him. Here is why: Pitotal = V2*V1, Pbat = Ibat * (V1-V2), Pcsr1 = Icsr1 * (0 - V2) = - Iscr1*V2 {note: V2 < 0}. Since Ibat = Icsr1 = Vcsr1, we got Pbat = Ibat*V1 - Ibat*V2 = Pitotal + Pcsr1.
However, the above relationship holds only if Rcsr1 = Rcsr2 = 1 ohm, but we all know such idea could never happen in real life. Therefore we must take the real values of Rcsr1 and Rcsr2 into the equation. By the Ohm's Law: V2 = - Icsr1 * Rcsr1, V4 = Icsr2 * Rcsr2. Take note that the currents Icsr1 and Icsr2 are not the same in quantity in general.
Therefore, Pbat = Pitotal + Pcsr1 = (|V2|/Rcsr1) * V1 + (|V2|/Rcsr1) * |V2| = (|V2|*V1+|V2|*|V2|) / Rcsr1 = Pbat_nominal / Rcsr1.
Here Pbat_nominal = |V2|*V1+|V2|*|V2|, which is what we have calculated assuming Rcsr1 = 1.
Similarly, Pled = Pled_nominal / Rcsr2.
And so, n = n_nominal * (Rcsr1/Rcsr2), where n_nominal = Pled_nominal / Pbat_nominal.
To obtain the real n, it seems that we also need to know (Rcsr1/Rcsr2). But a clever experiment design can help us find the real n without knowing the ratio: just repeat the experiments with the two resisters Rcsr1 and Rcsr2 swapped, and obtain n'_nominal. Now assume that by swapping, the real n does not change much. Then we have:
n = n_nominal * (Rcsr1/Rcsr2) and n = n'_nominal * (Rcsr2/Rcsr1).
Therefore, n*n = n_nominal * n'_nominal and n = sqrt(n_nominal * n'_nominal).
Note that the assumption of n being the same after swapping is based on the idea that Rcsr1 and Rcsr2 are insignificant components in the circuit and they are close in value.
Hopefully this rather long ranting can add something to our methodology.
lanenal
Basically, in my first post I proposed a method of more accurate measurement: just measure twice with the Rcsr1 and Rcsr2 swapped, then find the square root of the product of the two nominal n.
In the second post, I would like to propose a simplified efficiency calculation.
First of all, we group the components into three parts:
1. The bat and the Rcsr1 is grouped into the Power Source part. The idea is to view csr1 as an internal resistor of the Big-Bat.
2. The Led and the Rcsr2 is grouped into the Power Sink part. They both consume the energy from the coil.
3. The rest is grouped as the Joule Thief system. And it is the efficiency of this system that we should measure.
With that partition, the new COP measurement should be n = Pototal/Pitotal.
Note: Pototal = V3*V4, Pitotal=V1*V2, according to the schematics in my first post.
Note: I am using instantaneous measurements here (power, not the work), but it is clear that the results carries over to work as well, which is power integrated over time.
The same methodology proposed in my first post applies readily to the new measurement.
lanenal
P.S.: with this new n definition, .99's measurement will be 1.06 > 1. But since that's only a nominal measurement, and we don't know the ratio Rcsr1/Rcsr2, therefore the real n can't be determined yet. It could be less than 1, if the R values have +- 5% of error.
Quote from: PhysicsProf on 2011.03.01, 21:21:00
I've been working with the circuit further. I find that the current flows through CSR3 then through CSR2, alternately -- they are both outputs yet they are out of phase (my DSO shows). Therefore, I connected the output of the emitter directly to the point labeled "p22t" (also "v4") in .99's latest circuit schematic. It makes some sense to measure the output current therefore in CSR2 alone, and CSR3 is not needed. Does this make sense?
This may work in this case, but I would not recommend this practice. The purist and most scientific approach is to have a series CSR for every source and sink of power we are interested in getting a power measurement on. I would suggest we restrict our power measurements to INPUT, and the LED/CSR2 combination for the output. The LED (and it's associated CSR2) is the only real load in this circuit. The goal is to get as much of the used battery power transferred to the LED as possible. In theory, power in equals all dissipated power out. Our goal is to shift as much of that outpu power to the LED as we can. When you add up all the dissipations, of course they will add up to the battery power. We want to bias the distribution of power mostly to the LED.
So while it is useful to know how much power the transistor is eating up, it does not factor into the efficiency equation. Only the power source and the intended load factor into that.
Quote
CSR1 then gives the input current while CSR2 gives the output current (flowing from both the collector and the emitter), keeping things simplified.
Sometimes simplifications can cause us to overlook important things.
Quote
A question for you, .99. In your latest circuit LTJT2 - schema02, you do not show any connection to ground -- which is a change from schema01. Why the change?
As I said in my post, the ground symbols just add confusion to the matter when a grounded CSR for the battery is employed. I had hoped that re-drawing the circuit without the ground symbols would make things clearer...have I failed?
"Ground" is just a reference point. The probe reference points are still denoted on the schematic. Electrically, the two versions of the schematic are identical.
.99
Quote from: PhysicsProf on 2011.03.02, 05:17:08
Hope you're finding interesting results in the lab, .99. O0 A couple more questions in the interim:
I've gone back through the thread trying to find details on your toroid and the windings you used... missed it somehow.
Can you repeat a few details on the toroid used and bifilar windings? size of wire, etc. -- for the ferrite-core toroid.
Did you measure the inductance for each winding? which I presume are close to the same.
(Or if someone has the data, please re-post them... )
Here is a link to a post with some info on the cores.
http://www.overunityresearch.com/index.php?topic=717.msg10617#msg10617
Here is a link to the notes scan I mentioned in that post:
http://www.overunityresearch.com/index.php?action=dlattach;topic=717.0;attach=3603
Note that I physically cut the core I had, in order to decrease the core permeability and increase the frequency of operation. I made the inductance measurements BEFORE I cut it which of course changes all the parameters. But all the wire and turns information is there in my notes.
.99
Quote from: lanenal on 2011.03.02, 15:19:40
Basically, in my first post I proposed a method of more accurate measurement: just measure twice with the Rcsr1 and Rcsr2 swapped, then find the square root of the product of the two nominal n.
In the second post, I would like to propose a simplified efficiency calculation.
First of all, we group the components into three parts:
1. The bat and the Rcsr1 is grouped into the Power Source part. The idea is to view csr1 as an internal resistor of the Big-Bat.
2. The Led and the Rcsr2 is grouped into the Power Sink part. They both consume the energy from the coil.
3. The rest is grouped as the Joule Thief system. And it is the efficiency of this system that we should measure.
With that partition, the new COP measurement should be n = Pototal/Pitotal.
Note: Pototal = V3*V4, Pitotal=V1*V2, according to the schematics in my first post.
Note: I am using instantaneous measurements here (power, not the work), but it is clear that the results carries over to work as well, which is power integrated over time.
The same methodology proposed in my first post applies readily to the new measurement.
lanenal
P.S.: with this new n definition, .99's measurement will be 1.06 > 1. But since that's only a nominal measurement, and we don't know the ratio Rcsr1/Rcsr2, therefore the real n can't be determined yet. It could be less than 1, if the R values have +- 5% of error.
I think I agree with you here, if I understand correctly. However, as I pointed out above, I would not recommend that the transistor emitter be included in the Pototal calculation.
Also, I think it will be far easier to calibrate, i.e. tweak our resistors to be 1.00 Ohms rather than going through many complicated calculations. If we can make or buy very accurate 1 Ohm resistors, we are better off imo.
.99
Welcome, Lanenal, and thanks for your comments.
I have inserted a 1-ohm resistor in the emitter path and found the results interesting. [What I posted a bit earlier has been corrected at this point, in the following post.]
.99
QuoteAlso, I think it will be far easier to calibrate, i.e. tweak our resistors to be 1.00 Ohms rather than going through many complicated calculations. If we can make or buy very accurate 1 Ohm resistors, we are better off imo.]
I agree with that, but I've asked and hope you will provide us with data from your JT circuit when CSR1 at 1 ohm is replaced with a 1/2 ohm resistor, as I think you said you would do.
Note that in your circuit, you now have the collector path going through the LED and CSR2 AND through CSR1, back to the battery negative, and we'll be watching that you take this into consideration. Having looked at the oscilloscope output the way you have it connected, .99, I see that P-CSR1 is NEGATIVE on the scope -- I think this is what you were trying to say. Having seen the data, I understand now. That is, with CSR1 before the battery, then the absolute value of P-cSR1 needs to be added to determine the total Pinput.
By the same token, the current flow from the collector is such that the power from its current flowing through CRS1 must be added to the total output power. Correct me if I'm wrong.
\
All of this becomes simpler and less prone to error, IMO, when CSR1 is reduced to 1/2 ohm or even less (as long as the current is still measurable with acceptable accuracy), since the power dissipated in CSR1 becomes relatively small
Thanks for providing some details about the windings on your toroid, and how you cut it.
.99 wrote:
QuoteWhen you add up all the dissipations, of course they will add up to the battery power.
I detect a certain bias here that is disconcerting. We are trying to determine EXPERIMENTALLY whether or not a simple system can demonstrate overunity, more power out than in. It is not a foregone conclusion that "When you add up all the dissipations, of course they will add up to the battery power. " That is what we are in process of determining, experimentally. Not by pre-determined conclusion without the need for experiments ("Ipse dixit" authoritarian style).
Quote from: PhysicsProf on 2011.03.03, 05:19:08
I detect a certain bias here that is disconcerting. We are trying to determine EXPERIMENTALLY whether or not a simple system can demonstrate overunity, more power out than in.
The bias that disconcerting is your bias PhysicsProf. Yes you are trying to determine experimentally if a simple system can demonstrate over unity, but that shouldn't mean that you should go into the experiment with an expectation that you will measure over unity. That is what you are implying when you take issue with Poynt's statement that says, "When you add up all the dissipations, of course they will add up to the battery power."
It would be totally unscientific on your part and unreasonably biased to expect measurements that show over unity.
There is nothing wrong with being open minded but there are limits. It makes no sense to use inverse logic with respect to the conservation of energy.
MileHigh
Quote from: poynt99 on 2011.03.03, 01:24:14
This may work in this case, but I would not recommend this practice. The purist and most scientific approach is to have a series CSR for every source and sink of power we are interested in getting a power measurement on. I would suggest we restrict our power measurements to INPUT, and the LED/CSR2 combination for the output. The LED (and it's associated CSR2) is the only real load in this circuit. The goal is to get as much of the used battery power transferred to the LED as possible. In theory, power in equals all dissipated power out. Our goal is to shift as much of that outpu power to the LED as we can. When you add up all the dissipations, of course they will add up to the battery power. We want to bias the distribution of power mostly to the LED.
So while it is useful to know how much power the transistor is eating up, it does not factor into the efficiency equation. Only the power source and the intended load factor into that.
[snip]
.99
"Intended" by whom?
I have come back to this point because I think it is significant in our understanding and in our calculation of efficiency.
Let us consider the JT system as a "black box", that may or may not have more power out than power in. That is what we are trying to find out empirically.
We look at the current going into our "black box" as well as what comes out -- and the question here is -- what power is coming out? There are two currents coming out of the toroid/transistor combination which constitutes the black box -- one from the collector and the other FROM the emitter -- and both go from the black box to ground. Both can do useful work, such as heating a resistor. This output heat (from both paths) is not just for convenience in measuring -- it is OUTPUT POWER.
Yes, the original system has just one LED, connected to the collector. But why could we not hook a light to the output from the emitter also?-- the fact is, we could. It turns out that the voltage from the emitter is probably not high enough to light an LED (although I may try it), but some other form of bulb should work -- and in any case, we can certainly extract heat before the current is dumped to the ground connection simply by inserting a heating element called a resistor in the path from emitter to ground.
There is only ONE current coming out of the battery (or power supply) -- and for this we calculate the associated INPUT power, Pin.
There are TWO currents coming out of the black-box, and these must both be accounted for (not ignored) to accurately calculate the OUTPUT power, Pout.
I note that lanenal agreed, and ask if you also agree (dear reader, including .99) -- and if not, specifically why not? Just saying that original circuit by Lawrence or someone else did not have a heat or light-producing element in the path from emitter to ground is not a good reason to ignore this portion of the power emerging from the system.
Quote from: PhysicsProf on 2011.03.03, 05:19:08
Welcome, Lanenal, and thanks for your comments.
I have inserted a 1-ohm resistor in the emitter path and found the results interesting. [What I posted a bit earlier has been corrected at this point, in the following post.]
.99
I agree with that, but I've asked and hope you will provide us with data from your JT circuit when CSR1 at 1 ohm is replaced with a 1/2 ohm resistor, as I think you said you would do.
Yes, I will do that.
Quote
Note that in your circuit, you now have the collector path going through the LED and CSR2 AND through CSR1, back to the battery negative, and we'll be watching that you take this into consideration.
Yes, CSR1 should be considered as a load as I have mentioned. We will add it's power to the LED and CSR2 power.
Quote
By the same token, the current flow from the collector is such that the power from its current flowing through CRS1 must be added to the total output power. Correct me if I'm wrong.
Yes, we will consider CSR1 as an output dissipator of power.
Quote
All of this becomes simpler and less prone to error, IMO, when CSR1 is reduced to 1/2 ohm or even less (as long as the current is still measurable with acceptable accuracy), since the power dissipated in CSR1 becomes relatively small.
Agreed.
Quote
I detect a certain bias here that is disconcerting. We are trying to determine EXPERIMENTALLY whether or not a simple system can demonstrate overunity, more power out than in. It is not a foregone conclusion that "When you add up all the dissipations, of course they will add up to the battery power. " That is what we are in process of determining, experimentally. Not by pre-determined conclusion without the need for experiments ("Ipse dixit" authoritarian style).
I did precede that sentence with this one: "
In theory, power in equals all dissipated power out." Perhaps you might allow me some latitude in regards to being "jumpy" in response to anything that might suggest discussions about theory of operation and theoretical efficiencies. Thanks.
I am just as curious as to the unusually-high efficiency as anyone, which is why I'm going to re-do the tests and include the transistor dissipation and add CSR1 to the output power as well as CSR3. After that, the only remaining power dissipations are in the battery itself, the two coils, Rb, and the core. I do not know how to
measure the core dissipation directly though, so we may have to forgo that one.
.99
[quote ]
remaining power dissipations are in the battery itself
[/quote]
ehehe... kcits htiw POC
@.99
QuoteAlso, I think it will be far easier to calibrate, i.e. tweak our resistors to be 1.00 Ohms rather than going through many complicated calculations. If we can make or buy very accurate 1 Ohm resistors, we are better off imo.
Or if we can measure both Rcsr1 and Rcsr2 accurately, then n_real = n_nominal * (Rcsr1/Rcsr2), which should be easy enough.
@PhysicsProf
I have not really considered the emitter question carefully, yet it is undoubtedly a very interesting topic. It seems to me, a cycle in a basic JT circuit, has two phases:
phase 1. the energizing phase: the transistor is in the ON state and the current in the coil connected to the collector is increasing.
phase 2. the vanquishing phase: the transistor is in the OFF, and the the current in that same coil is going through the LED.
It seems to me that phase 1 is a pure input phase, and there is no active output (ignoring the wasted power in the transistor, the 1k transistor, and the coils). Then in phase 2, the battery AND the coil connected to the collector are powering the LED.
If the analysis above is fine, I tend to ignore the emitter output, not if we can have easier and more accurate measurements.
However, it is possible to calculate that output without inserting a third measuring transistor between the ground and the emitter. In fact, we should remove CSR2 and have the LED directly connect to the ground, then the new Pototal can be found by integrate [V3*V2] over time, and this time the Pototal includes the output energy spent on both the LED and the transistor. Moreover, this time the current is estimated by V2 on one single CSR1, so value errors in CSR1 won't be a factor anymore, the calculated n would be EXACT!!!
Thus, I would say yes for the sake of simplicity! Here is the proposed method:
1. take CSR2 out and connect the LED directly to the ground (or just leave as is and don't measure V4, measure V2 instead if you were using two scopes).
2. measure V1, V2, V3 as before.
3. calculate Pitotal by MEAN [V1*V2], which gives the time average input power.
4. calculate Pototal by MEAN [V3*V2] over time, similar to finding the Pitotal.
5. then n = Potatal/Pitotal, which is EXACT as the error in the value of CSR1 is balanced out in the way n is calculated. O0
This seems to be as simple as it can be. Also, it only leaves tiny energy leaks out of consideration, namely the leaks in the 1k resistor and the coils.
cheers,
lanenal
Quote from: poynt99 on 2011.03.03, 13:06:35
Yes, I will do that.
Yes, CSR1 should be considered as a load as I have mentioned. We will add it's power to the LED and CSR2 power.
Yes, we will consider CSR1 as an output dissipator of power.
Agreed.
I did precede that sentence with this one: "In theory, power in equals all dissipated power out." Perhaps you might allow me some latitude in regards to being "jumpy" in response to anything that might suggest discussions about theory of operation and theoretical efficiencies. Thanks.
I am just as curious as to the unusually-high efficiency as anyone, which is why I'm going to re-do the tests and include the transistor dissipation and add CSR1 to the output power as well as CSR3. After that, the only remaining power dissipations are in the battery itself, the two coils, Rb, and the core. I do not know how to measure the core dissipation directly though, so we may have to forgo that one.
.99
Thank you for your responses, .99. Glad to see that you are also "curious as to the unusually-high efficiency" We are remarkably in agreement at this point, and I'd like to emphasize your point:
QuoteI'm going to re-do the tests and include the transistor dissipation and add CSR1 to the output power as well as CSR3
.
I am looking forward to your measurements.
On my side, I measured the inductance of the 1" yellow toroid (from Jameco) that I wound bifilar that gives the better results -- both of my windings have L = 24 uH.
@laneal -- I'm puzzling over your post and will do some measurements using your method... thanks.
Quote from: lanenal on 2011.03.03, 15:10:41
[snip]
@PhysicsProf
[snip]
However, it is possible to calculate that output without inserting a third measuring transistor between the ground and the emitter. In fact, we should remove CSR2 and have the LED directly connect to the ground, then the new Pototal can be found by integrate [V3*V2] over time, and this time the Pototal includes the output energy spent on both the LED and the transistor. Moreover, this time the current is estimated by V2 on one single CSR1, so value errors in CSR1 won't be a factor anymore, the calculated n would be EXACT!!!
Thus, I would say yes for the sake of simplicity! Here is the proposed method:
1. take CSR2 out and connect the LED directly to the ground (or just leave as is and don't measure V4, measure V2 instead if you were using two scopes).
2. measure V1, V2, V3 as before.
3. calculate Pitotal by MEAN [V1*V2], which gives the time average input power.
4. calculate Pototal by MEAN [V3*V2] over time, similar to finding the Pitotal.
5. then n = Potatal/Pitotal, which is EXACT as the error in the value of CSR1 is balanced out in the way n is calculated. O0
This seems to be as simple as it can be. Also, it only leaves tiny energy leaks out of consideration, namely the leaks in the 1k resistor and the coils.
cheers,
lanenal
OK -- I see what you're doing and I like it! CSR1 is used as a current-measuring resistor for both the input and output legs of the circuit. As your wrote, and I wish to emphasize:
In fact, we should remove CSR2 and have the LED directly connect to the ground, then the new Pototal can be found by integrate [V3*V2] over time, and this time the Pototal includes the output energy spent on both the LED and the transistor. Moreover, this time the current is estimated by V2 on one single CSR1, so value errors in CSR1 won't be a factor anymore, the calculated n would be EXACT!!!Looks great to me -- I wonder if others agree, particularly .99 and humbugger. Any problems seen?
You have, I think, homed in on a simple yet elegant solution to measuring Pin, Pout, and n.
lanenal,
I tried your method in PSpice, and it comes very close to the actual summed power in the transistor and LED, but it is lower by about 1.5%. I can not account for that loss at the moment, but it would be handy to have only the one CSR in the circuit.
One other potential problem in combining all the output powers, is we are less likely able to identify where or which device is causing our efficiency to be so high.
At any rate, I'll leave it up to you and the professor to decide which method you want to use.
.99
Quote from: poynt99 on 2011.03.03, 19:12:32
lanenal,
I tried your method in PSpice, and it comes very close to the actual summed power in the transistor and LED, but it is lower by about 1.5%. I can not account for that loss at the moment, but it would be handy to have only the one CSR in the circuit.
One other potential problem in combining all the output powers, is we are less likely able to identify where or which device is causing our efficiency to be so high.
At any rate, I'll leave it up to you and the professor to decide which method you want to use.
.99
Great idea, Lanenal -- elegant. Thanks for checking this, .99.
I have done some tests now with Lanenal's method and I love the simplicity of it. The channel 2 probe stays put on CSR1, as do probe-ground leads. It is quick to move the channel 1 probe between the 2 points, for input and output.
So, I have done tests using my "best" toroid, one which I wound myself and call J1. It has looked the best using .99's circuit. I also extracted the small transformer from a Fuji AA flash throw0awat camera and wired that into the circuit for another test.
I'm using my ATTEN 1062 DSO, which has a MATH function and shows the Power waveform in green Using L's method, I plot the Pin and Pout and then I can straightforwardly compare the AREA under the waveform for one cycle, which gives me
Eout/Ein = Pout/Pin (since for the same time period, one cycle) = n.
Using L's method, I don't have to subtract or add P-csr1, which is great, I can simply calculate the areas under the Power curves (as a function of time, so this yields Energy) for In and Output, and Divide Eout/Ein to get n.
I have done this for toroid J1 and the Fuji transformer. The results --
Toroid J1: n = 1.4
Fuji transformer: n = 1.08
CAVEAT: Both results are approximate, given that I have done the area calculations by hand. I may put the CSV data from the DSO into Excel and see if I can refine the calculation; or go up to the University and use the Tek 3032 which calculates the Math Mean directly (giving Pmean with hardly any work) -- applied to the same circuit.
I will post photos of the set-ups soon... I haven't been able to extract the waveforms in a screen-shot; wish i could to show you directly. For J1, the area under Pout appears larger than the area under Pin, by eye. (Have I done something wrong, or is this something to cheer about? not real sure yet.)
The circuit is rather easy to put together on a circuit-board, and I encourage others to participate and check my and .99's results.
Well done guys! The research and the collaboration in this thread has been really really impressive -- especially if the result of COP>1 can be replicated. I will attempt my own replication in the next few weeks -- I'm moving my scope out of storage and into my office this weekend.
Lanenal , your ideas for power measurement were robust and brilliant.
PhysicsProf, thanks for taking the time to organize this and do all the hard work required to do the measurements. I know this is not easy, but you've gone ahead with it despite the criticism of some unnamed users on another boards , whose real purpose appears to discourage research rather than have open debate. I've built Joule Thief too and was mystified by the circuit. Measuring power on it is non-trivial question.
Poynt99, thanks for providing the board ,and providing the schematics and intellectual support for this research.
Let's let these COP>1 measurements keep us motivated to pursue multiple avenues of research in order to achieve what we all know is possible. In the meantime, we need to make sure that the methodology for Joule Thief power measurement is valid and will stand up to scrutiny -- but these results are very encouraging. Wikipedia knows nothing when it comes to suppressed knowledge. Wikipedia could get away with telling us the Romans invented the first 365-day solar calendar , unless people question the 'authorized' 'official' knowledge propagated by large institutional hierarchies.
Anyway great job guys, I'm very impressed with the research in this thread.
Thanks for the encouraging words, feynman.
More measurements and care, and I think the n=1.4 number is high. I have also concluded that my "area under the power curve" method, by hand, is a bit too crude to be reliable for numbers close to unity.
I plan a trip to the 3032 at the University as soon as I can get there -- and I look forward to .99's measurements.
Here's what Wikipedia says about Joule Thief in terms of waveforms:
(http://upload.wikimedia.org/wikipedia/en/thumb/6/6f/Joule_Thief_Waveforms2.PNG/800px-Joule_Thief_Waveforms2.PNG)
Good job guys, and thanks for sharing your results.
Quote from: poynt99
I tried your method in PSpice, and it comes very close to the actual summed power in the transistor and LED, but it is lower by about 1.5%.
It is certainly good to know that the measured Pototal is LOWER than the actual. I'd like to do an analysis to show that it is indeed the case IF we also include the power spent on the 1k resistor (our measuring method unchanged).
Pototal = P1k + Pbe + Pce + Pled.
P1k = power on the 1k resistor = I1k * V1k,
Pbe = power on the be knot of the transistor = Ibe * Vbe,
Pce = power on the ce knot of the transistor = Ice * Vbe,
Pled = power on the LED = Iled*Vled.
On the input side it is relatively simple:
Iin = Icsr1 = Ibat, Vin = Vbat - Vcsr1, so Pitotal = Iin*Vin = Icsr1 * V1
The measurement is exact (Note: Vin=V1, Iin=Icsr1=Vcsr1=V2, with Rcsr1=1).
So I will concentrate on the analysis of the output side, and consider the ON and OFF state of the transistor separately.
When the transistor is OFF, the analysis of the output is simple: Iin = Iled, Pototal = Pled = Vled * Iin. Thus our measurement is exact in this case.
Let's turn to the ON state:
Note that Ie = Ibe + Ice, I1k = Ibe, Iin=Ibat=Iscr1=Ibe+Ice+Iled, and Vled = Vce.
As the current Ice is climbing up, so the coil connected to the collector maintains a voltage drop, we have Vin > Vce. On the other hand, the time interval for Ibe to rise to its working level is tiny, so most of the time, Vbe+V1k = Vin + Vinduced (ignoring the resistance in the coil), where Vinduced is the induced voltage because the current in the other coil is increasing.
P1k + Pbe = I1k * V1k + Ibe * Vbe
= I1k * (V1k + Vbe)
> Ibe * Vin
> Ibe * Vce.
From which we can proceed as follows:
Pototal = P1k + Pbe + Pce + Pled
> Ibe*Vce + Ice * Vce + Iled * Vled
= (Ibe + Ice + Iled) * Vce
= Iin * Vce
Note that Iin*Vce is what we measured as Pototal (to see this, Iin = Iscr1 = Ibat, Vce = Vled = V3), so it is indeed less than the actual Pototal.
On the other hand, the measured Pototal should be very close, as Ibe is relatively small to Ice on the average. The bigger the beta, the better the approximation. The gap can be roughly estimated by integrating Ibe * (Vin - Vce) during the ON state of the transistor (ignoring Vinduced), now (Vin - Vce) is the voltage drop over the coil connected to the collector, Ice * (Vin-Vce) is the power stored on that coil, which will be released through the LED later. However, Ice is increasing almost linearly from zero up to (beta * Ibe) [Ibe is relatively constant, so is Vin-Vce, during the ON state], therefore MEAN[beta * Ibe * (Vin - Vce)] during the ON state = 2 * MEAN[ Ice * (Vin-Vce) ] during the ON state. The stored energy MEAN[ Ice * (Vin-Vce) ] during the ON state is released later on the LED, which can be estimated by Pled * ( 1 - Vin/Vled) during the OFF state of transistor. From our previous tests, Pled ~ Pototal, therefore the gap is estimated to be 2 * Pototal * (1 - Vin/Vled) / beta, which is about 200*(1-Vin/Vled)/beta percent of the Pototal. If beta = 50, Vin = 1V, Vled=2V, the gap is then estimated to be 2%. Which is close to your 1.5%, and that 2% also included the 1k resistor. If assume Vinduced = Vin-Vce, as the coils have the same # of windings, the gap becomes 400*(1-Vin/Vled)/beta; in this case if beta=100, Vin=1, Vled=2, the gap should be 2%.
In conclusion: the simplified method of measurement gives a good
underestimate of efficiency n.
@PhysicsProf:
It all Looks exciting -- look forward to your verified results. And your encouraging words are appreciated. I will do my best to join you [might take some time, I need to get a scope and learn how to use it].
@feynman:
Thank you for your encouragement, and it is great that you could join the force here.
cheers, lanenal
Quote from: poynt99 on 2011.03.03, 19:12:32
One other potential problem in combining all the output powers, is we are less likely able to identify where or which device is causing our efficiency to be so high.
Hi .99: very good question you have raised out there -- at this moment I can't say much about how to identify which component is causing the high efficiency, but let's keep thinking about it. However, it could be the case that the high efficiency comes as the synergy of a few components. Another possibility is that some energy comes from the ether (or, if you don't accept ether, the magnetic-electric field) instigated by the make-and-break of the transistor and picked up by the coils. I don't know, guys, just my wild educated guess for you guys to think about.
Later on, we might want to do some fine tuning of it, varying the base resistor, the coils, the transistor beta (by using different transistors) , and measure the efficiency and see how it changes with those parameters.
If we just want to ascertain n > 1, which seems to be our near objective, then the simplified measurement is what I would prefer. To reduce the errors caused by the inaccuracy of scopes, it is good to follow this procedure:
1. measure V1, V2, V3 and calculate n1.
2. swap the probes for V1 and V3, and calculate n2.
3. calculate the final n = sqrt(n1*n2).
This way, relative measuring errors of the probes are canceled out, and the final n should be very accurate. O0
cheers, lanenal
Good comments, and good adventure.
So, I will attach a photo-composite below, showing the JT circuit and Pout and Pin (labeled on a white card) for this circuit, using my ATTEN 1062C.
Conditions: Jameco 1-1/8" yellow toroid which I wound bifilar, 13 windings, 24 uH in each winding. Vbatt = 1.14 volts, actually from a power supply, which reads 58 mA current (input). This is approximately the point at which the Green Power curves just touch the zero line -- higher voltage, and the curves move away from the zero line indicating a steady draw of current in addition the sawtooth). Further, 1.14V is the Voltage for which I obtained the largest n with this particular toroid (according to my rough calculation). You may observe in the photo that I am using a red LED for these tests.
Note, feynman, that the curves are considerably more interesting than what you found in Wikipedia. Blue is the Voltage across CSR1, V2 in Lanenal's schematic. Yellow waveform with Pout represents V3 (including the LED); yellow with Pin gives the V1 waveform.
I will leave it to you to examine the green curves for Pout and Pin -- using Lanenal's circuit (which i also depicted in the photo, because I printed it out this morning and used it). Note that both Pin and Pout POWER curves come out Negative the way the circuit is hooked up per Lanenal's schematic.
You can see from the curves that calculating the area under Pin is rather easy -- a triangular sawtooth. Calculating the area under Pout is more challenging as it is an irregular spike -- and note that it reaches much larger peak power (and voltage) than the input. I did measurements and calculations by hand on these waveforms to get n -- and found n to be somewhat larger than unity in this case (and a few others), but I'm awaiting runs with the Tektronix 3032 in order to get more reliable and definitive numbers.
Suggest you compare Pout and Pin green waveforms for yourself.
That's very interesting O0. I did a little marking and rough measuring, see attached. My estimation is that it is anywhere between 90% to 110%. Look forward to the numerical results.
Absolutely PhysicsProf, the curves are definitely more interesting than Wikipedia shows. I realized that when I was experimenting with Joule Thief in the past. Even if it was only 90% efficiency , I think it's a fascinating circuit, and it's a great way to get people involved in experimenting with electronics. The evidence that it's COP>1 only adds to the mystery.
I will upload my scope traces over the weekend. Unfortunately I only have one old Tektronix analog scope , so I had a single channel trace of the voltage of the transistor waveform and the LED waveform, but it's accurate down to nanoseconds, which is nice.
Great work guys, looking forward to moving my gear this weekend.
http://cgi.ebay.com/Vishay-VPR5-1R0000-1-0000ohms-0-1-5W-resistors-7-/400200178193?pt=LH_DefaultDomain_0&hash=item5d2dca1a11 (http://cgi.ebay.com/Vishay-VPR5-1R0000-1-0000ohms-0-1-5W-resistors-7-/400200178193?pt=LH_DefaultDomain_0&hash=item5d2dca1a11)
For low power such as the Joule Thief and for those that can measure resistance accurately:
http://shop.ebay.ca/?_from=R40&_trksid=p3984.m570.l1313&_nkw=carbon+resistors&_sacat=See-All-Categories#
50 Value 1/4W Carbon Film Resistor (1R~10MR) 5% 2000pcs
$10, free shipping.
Use parallel combinations to achieve your desired resistance.
.99
Interesting results yesterday as I traveled to the University and used the Tektronix 3032 there, to derive Pin and Pout mean values, as we have been discussing.
I will start with a few "normal results", in part to get us all used to reading the waveforms and the calculated power.
Result 1: Using toroid "A" (from prototype A sent to me for testing by Lawrence Tseung; but just using the primary and feedback windings, not the secondary winding). Attached shows the resultant waveforms and calculated Power, with the input Power in red on the left and the output Power in red on the right. Channel 2 is always the voltage drop across the one-ohm resistor (see Lanenal's circuit diagram above in this thread), that is, the current waveform. Here, a nice sawtooth. Channel 1 is the voltage across the battery -- or, in these tests, the DC power supply -- for Pin, on the left. For Pout on the right, Channel 1 is the voltage drop across the transistor and (red) LED.
Notice that per the schematic, the voltage across the 1 ohm resistor is Negative, hence both Pin and Pout show up as negative -- for the discussion that follows, and to be consistent with what we know, I will set the y axis to be POSITIVE DOWNWARDS consistent with the power in and out being positive quantities. The shapes are interesting -- a saw=tooth for Pin and a single U-shape for the Power Out.
The 3032 calculates ch1*ch2 = V * I = Power and displays the result in RED waveforms, and it calculates the MEAN Power for us, given in the right- hand column. Per our previous discussion, to a reasonable accuracy over a dozen or so cycles, we get:
n = Pin / Pout = 36.11/ 40.15 = 0.9.
The estimated error, +- 2% based on .99's simulation and on the repeatablility I find for this ratio in these tests (for a given toroid and set-up).
High efficiency, but not surprising.
The second pair of scope-shots homes in on the input and output signals, showing detail. This juxtaposition will be useful for later comparisons.
A photo for the set-up is given in the attached. The circuit is straightforward, 2N2222 transistor, red LED, 1K ohm resistor to the base, 1 ohm resistor on input and output for measuring currents, bifilar wound toroid.
The photo shows the toroid which I wound, with the primary and feedback windings both having 24 uH... (data to follow for this toroid, called "eJ" ).
Data for this eJ toroid is shown in the attached, for input voltage from the power supply measured by an independent DVM as 0.993 volts.
Familiar sawtooth waveform for Pin, and U-waveform for Pout.
Result 2: n = Pout/pin = 39.64/44.08 = 0.90 = 90%
+- about 3%.
Ho-hum.... but there is more
More from the same toroid, this time at 1.04 volts from the PS:
n = 44.3/48.3 = 0.92
No surprises ... yet.
I want to show next that one can get variations in the waveforms by using different conditions -- which I demonstrate by looking at a transformer I and a friend extracted from a Fuji AA-battery-type disposable camera. Here we used the outer pins for the primary and feedback inductors (numbered 1,4 and 2,5 on another diagram... I'll see if I can dig it up...).
The waveforms become quite interesting, showing THREE distinct humps for both Pin and Pout -- see attached. I show a detailed shot for a few cycles in the first attachment so you can see the detail. The Fuji transformer likes sinusoidal-looking waveforms... ;) Why? an interesting question. Clearly the Pin and Pout waveforms can vary from what we see above.
Notice that sometimes I vary the vertical scale for the power waveform display -- the scale shows up in red at that bottom of the scope-shot so you can keep track of this parameter.
Result 4, Fuji: n = 24.2 / 29.2 = 0.83
+- about 3%. I suspect the lower value follows from the resistance in the very fine wire used in the Fuji transformer.
Without further ado, let me share with you -- inviting comment -- the most interesting result of yesterday's (4 March 2011) tests at the University, using a Tektronix 3032 scope.
Here I am using the eJ toroid that I wound. I began with a Jameco toroidal inductor (100 uH, Jameco part # 386601), and wound fifteen windings bifilar of 22-gauge insulated (one with plastic, the other enamael) copper wire. The inductance of EACH of my windings came out to 24 uH. The Jameco winding is not involved in this test run, and the wires are left unconnected from this winding (which would be the secondary winding for a Tseung-type system).
I have given results with this SAME eJ toroid above... where we observed Pin as a simple sawtooth and Pout a single hump or spike. Now -- please look closely at the waveforms for this run at 0.996 V (in from the power supply), in the attached.
First, we look at the detail for roughly 1.5 cycles in the first attached screen-shot. The red Pin waveform is approximately what we have seen before, a saw-tooth pattern. But the (red) Pout waveform is quite different!
Instead of a single hump or spike essentially bounded at zero-power as we have seen in the past, we now see a U-shape that significantly OVERSHOOTS ZERO, demonstrating current and voltage Out of Phase, as we also see comparing the voltage across the (LED+transistor) in yellow, and the current in the 1-ohm CSR shown in blue.
Further, notice the oscillations/wiggles in the waveforms for the output power and voltage waveforms. Very interesting. To me, this is striking and rarely-seen behavior, and I have been working with these types of circuits for months now.
To get the efficiency with some accuracy, we have the Tektronix 3032 calculate Mean Power for both Pin and Pout over numerous cycles -- with this interesting result:
n = 44.8 / 39.8 = 1.13 = 113%
+- about 3% as previously noted.
If you are visually comparing Pin and Pout waveforms in the attached, note that the scale in the screen display for Pin is 20 mVV whereas the scale for Pout is 50 mVV (which does not change the result above), which I did because the Pout curve was getting clipped on the 20mVV scale as i recall.
In order to encourage and facilitate comparisons, I include a final attachment which shows Pin and Pout for these data at 0.996V juxtaposed with the data for the same eJ toroid ran at nearly the same voltage (0.993 volts), but where the latter displays the "normal" pattern with just one spike. The two waveforms in the center of this attachment (trace 2 and 3 ) give Pout and Pin both on the same scale, 50 mVV, to make visual comparisons more straightforward for you. If you think the Tektronix 3032 may be making a mistake when calculating the Mean power in the case where the power is sometimes negative and sometimes positive, I invite you to evaluate the areas under the Pin and Pout curves for one cycle, subtracting the opposite-signed power region, and then evaluate n from this integral (that is, using n = Ein/Eout, as I have demonstrated before). It would be a worthwhile check. I have done this roughly, and the Tek 3032 calculation seems good.
You have a dozen questions? I have many also... ;)
After seeing this result, I returned twice to the same circuit and getting the voltage as close as I could, but could not generate the unusual "spiky" Pout waveform again in the limited time I had on the Tek 3032 at the University. And those results came out with COP = n of about 90%.
I recall having seen such a waveform at home before, but I do not know just how to reproduce this spiky waveform -- wish I did. And I am very open to suggestions.
I will say that I looked first at the many-cycle Pout (attached), then looked at many-cycle Pin (moving the channel-1 probe to do so), then moved the channel-1 probe back in order to re-measure Pout and record the approx detailed Pout waveform (attached). The Pout waveform remained the same over the several minutes required to do these measurements and record them on the computer, about 15-20 minutes. So however I got in this mode, with the unusual waveforms, zero-crossing Pout and COP evidently greater than one, the condition remained for a while.
When I saw that n was coming out larger than one, I was a bit excited and did go back and get the detailed waveforms recorded. However, I was thinking that I could simply come back later and get more data at this condition after running at other voltages, which was not the case yesterday afternoon -- I'm glad I got as much data as I did and recorded it for you all to see.
Now the question is -- what does this all mean? and how can one get back to this out-of-phase relationship between V and I in the output circuit?? (I have reported this condition before as you may recall, some time ago, but that was with the LT-type circuit and this is a simple JT.)
To me the result is striking and noteworthy and something of a breakthrough. But can it be repeated? Repeatability is a core requirement for solid science and progress.
Thank you PhysicsProf, these measurements are great and the result is fascinating! O0
Will study it more closely after come back from church.
Some tests coming up...
.99
Experimental data are the key.
It will be helpful to have these additional (V4 and V5 especially) measurements, .99. Please show the waveforms, especially the power waveforms, for comparison with the data I obtained with the 3032.
I'm not sure if I read the scope shots right -- the marks on the left margin of the shots, are they zero indicators? Could it be the case that there is an absolute bias of the scope measuring the voltage drop over CSR1? Also, from your reporting OU results:
Quote from: PhysicsProf on 2011.03.05, 23:41:59
To get the efficiency with some accuracy, we have the Tektronix 3032 calculate Mean Power for both Pin and Pout over numerous cycles -- with this interesting result:
n = 44.8 / 39.8 = 1.13 = 113%
+- about 3% as previously noted.
I found that it is even higher from the readings on the right margin of the first shot attached to that same post:
n = 55.08/45.20 = 1.21 = 121%. Am I getting it wrong? Sorry, I don't know much about scopes and I'm just hoping this result can stand scrutiny. Also, what is the sampling frequency of the scope shots? 100MHz? I don't know if I can afford a terribly high frequency scope.
lanenal
Quote from: lanenal on 2011.03.06, 06:51:44
n = 55.08/45.20 = 1.21 = 121%.
Not sure what you mean but I think n over many cycles to provide accurate calculation.
The Out of Phase result puzzled me quite a bit. It almost seems to me that that the current never passed through collector emitter. The only logic I can come up with for now is the JT turned from pulsing resonant into an LC circuit using the base emitter flow. ??? :-\
Quote from: GibbsHelmholtz on 2011.03.06, 14:40:43
Not sure what you mean but I think n over many cycles to provide accurate calculation.
OK, I see -- so that reading was just calculated from the cycles seen on the shot.
Quote
The Out of Phase result puzzled me quite a bit. It almost seems to me that that the current never passed through collector emitter. The only logic I can come up with for now is the JT turned from pulsing resonant into an LC circuit using the base emitter flow. ??? :-\
Me too -- it seems the current through the CSR1 is alternating? With the Diode and the transistor, I don't know how could the current flow in the other direction. Or am I reading the scope wrong again?
lanenal
P.S.: I saw a 200MHz two Channel Scope today, less than $100. But no Math capabilities. Nor can I download data onto a computer.
what I'm thinking is like this
http://www.youtube.com/watch?v=kQdcwDCBoNY
but... still thinking :-\
I've discovered the error in my test setup (scope grounds on wrong node), so here are the measurements again, as pertaining to "schema02", i.e. no CSR3 present.
Pitotal = -46.64mW
Pcsr1 = 2.79mW
Pvbat = 49.43mW [Note that you can obtain this value directly when measuring Pitotal just by inverting the CSR1 channel]
Pototal = 33.62mW (PLED and Pcsr2)
Pototal + Pcsr1 = 36.41mW
I did not measure PQ.
So, efficiency not including output power dissipated in Q is:
n = 36.41mW / 49.43mW
n = 73.66%
.99
Lanenal:
Quote
I'm not sure if I read the scope shots right -- the marks on the left margin of the shots, are they zero indicators?
That is correct.
Quote from: GibbsHelmholtz on 2011.03.06, 14:40:43
Not sure what you mean but I think n over many cycles to provide accurate calculation.
Right.
QuoteThe Out of Phase result puzzled me quite a bit. It almost seems to me that that the current never passed through collector emitter. The only logic I can come up with for now is the JT turned from pulsing resonant into an LC circuit using the base emitter flow. ??? :-\
Call it OOP... but not oops, at least not yet -- still checking various things. I tried adding an "antenna" to the system in one of my tests, but the noise produced is nothing like that seen in the evident-n>1 test. See my data from Friday (4 Mar 2011) repeated below -- compared with Pototal from .99's post today (6 March). The "input" power curves (on the left for each pair) look quite similar, as one might expect. Our "output" power curves are inverted relative to one another, but taking that into account, these waveforms differ rather dramatically. Mine shows the out-of-phase condition, evidently. As I say, I don't know how to get the JT into this Out of Phase (OOP) condition that I achieved on Friday, but I'm thinking about GibbsH's suggestion as a possibility..... but how, Gibbs?
@poynt99
Sorry to hear your measurements were incorrect. That's happened to me before too, and it's disappointing. Sometimes what I do is work on a few unrelated projects concurrently , so if one doesn't work out , I can take a break and come back to it.
@PhysicsProf
What you have accomplished -- if it is not due to errors in the setup -- is very significant. What appears to me to be happening is that 'regular' Joule Thief is not OU, but due to some sort of phase anomaly (phase, frequency, and resonance seems critical in this sort of research), the joule thief circuit can slip into OU operation.
What we have are the scope traces, which are the most important information. What I would suggest is that you attempt to replicate the experiment on a new circuit and toroid idential to the first (same windings, etc). That way you can make modifications and experiment without 'damaging' the original setup which produced OU. At the same time, I would suggest you just continually run the original circuit in the same place on the same scope and see if it gets 'pushed' back into OU.
Here's what I would consider as possible sources that could have 'pushed' Joule Thief out of phase into anomalous operation:
1) Short-circuit, whether through the human body, oscilloscope, or otherwise. This needed only to be temporary short circuit to get the circuit out of phase, such as brushing against two circuits leads with your fingers (human body is 100ohm - 100k resistor).
2) RF energy in proximity. For example, cell phone, radio waves, CD-R burner, etc.
3) Simple probability (0.5% probability of slipping into OU operation over 24 hour period. etc)
Those are my suggestions for the time being. Let's also make copies of your scope traces and work, and try to meticulously document the exact setup (schematic, scope probe points, probe types/impedence, photographs, etc). I will definitely attempt replication.
Thank you again for your efforts; I am following this with great interest. Furthermore, for anyone interested in OU research , there is an excellent Torrent containing Nikola Tesla documentaries and patents which is available here:
Nikola Tesla Books and Other Video:
http://thepiratebay.org/torrent/4755412
The above is a great source of inspiration if you are feeling down due to engineering problems. Tesla faced much worse (his lab with a lifetime of work burned up in a fire), and yet he still triumphed.
Poynt:
I am not really following the thread but congratulations on finding the error. 73% sounds right, 99% sounded wrong.
I don't know if Murphy's Law is to blame but for such a simple circuit it has really been a struggle, don't you think?
Feynman and PhysicsProf:
Quotedue to some sort of phase anomaly (phase, frequency, and resonance seems critical in this sort of research), the joule thief circuit can slip into OU operation
There is really no chance that Feynman's speculation is true. I admit I have not been reading or looking at the scope shots. What I gather from skimming through the thread is that PhysicsProf has been reporting over unity results for the past week or more.
Look at the lesson in Poynt's experience. With perseverance and more investigation you should figure this enigma out. I realize that I am just parachuting in here but mark my words, there is no over unity associated with PhysicsProf's testing. It's just a question of honing your skills and finding where you made an oversight.
MileHigh
MH,
It was those pesky grounds again. It's a good thing I re-drew the schematic without them, as it helped me find my error. Paying attention to the diagram in some detail is worth the time, I'm discovering. ;)
The numbers do seem to correspond to more commonly-achieved efficiencies now, agreed.
.99
Quote"
Look at the lesson in Poynt's experience. With perseverance and more investigation you should figure this enigma out. I realize that I am just parachuting in here but mark my words, there is no over unity associated with PhysicsProf's testing. It's just a question of honing your skills and finding where you made an oversight."
I welcome skepticism , but I don't take kindly to those involved in IO. Upon what basis are your categorical blanket statements? If it's non-experimental law of thermodynamics claims , please start a new thread. I do not want to discourage research in promising areas (sharp gradient DC in toroids).
@PhysicsProf
It just occurred to me that what you ought to do is add a variable component to the circuit ... perhaps a variable resistor, variable inductor, or variable capacitor. This allows tuning like in the video GibbsHemholtz posted.
Quote from: feynman on 2011.03.06, 23:12:37
I welcome skepticism , but I don't take kindly to those involved in IO. Upon what basis are your categorical blanket statements? If it's non-experimental law of thermodynamics claims , please start a new thread. I do not want to discourage research in promising areas (sharp gradient DC in toroids).
I don't know what "IO" is. The basis behind the statement is that an inductor can only put out as much energy as you put into it. It's a device that can store energy and then release that stored energy. It cannot create energy or harvest energy from "somewhere else."
You can easily do experiments to confirm these properties of inductors on the bench, and yes they do adhere to the laws of thermodynamics. It's tempting to think that "sharp gradient DC in toroids" can be considered a promising area of research. However, I think if you do do the research you will not find anything there. It's a cliche about sharp gradients accessing otherwise hidden vacuum energy.
Anyway, with perseverance PhysicsProf should start seeing some numbers that refute his current findings suggesting over unity.
MileHigh
Quote from: PhysicsProf on 2011.03.06, 21:15:28
Call it OOP... but not oops, at least not yet -- still checking various things. I tried adding an "antenna" to the system in one of my tests, but the noise produced is nothing like that seen in the evident-n>1 test. See my data from Friday (4 Mar 2011) repeated below -- compared with Pototal from .99's post today (6 March). The "input" power curves (on the left for each pair) look quite similar, as one might expect. Our "output" power curves are inverted relative to one another, but taking that into account, these waveforms differ rather dramatically. Mine shows the out-of-phase condition, evidently. As I say, I don't know how to get the JT into this Out of Phase (OOP) condition that I achieved on Friday, but I'm thinking about GibbsH's suggestion as a possibility..... but how, Gibbs?
Of course your data is differ from .99's data although the current inverted looked just like yours. lol and it's obvious that .99 is anti cop>1 ;D sorry Poynt, it's just a personal opinion. :P In your data, current rise from zero to max while collector emitter voltage is high (seems like LED is on) and dipped negative when current drop. That's the total opposite from other data. I still cannot comprehend it fully. I do have opinions that the mode of operation (psuedo resonance/OOP) is rather rare. I don't expect we can reproduce them easily and I don't plan to go that route if we don't have to. I do strongly believe it has something to do with the youtube vid I posted. I'm thinking we reproduce the circuit in the youtube vid and replace the cap with an LED, then insert a resistor to measure current from the input and output similar to the JT while pulse it with DC...
Edit: Sorry Poynt :-[ I know deep down inside you believe all the energy should add up. We're brothers on this one.
Quote from: GibbsHelmholtz on 2011.03.07, 01:02:37
Of course your data is differ from .99's data although the current inverted looked just like yours. lol and it's obvious that .99 is anti cop>1 ;D sorry Poynt, it's just a personal opinion. :P In your data, current rise from zero to max while collector emitter voltage is high (seems like LED is on) and dipped negative when current drop. That's the total opposite from other data. I still cannot comprehend it fully. I do have opinions that the mode of operation (psuedo resonance/OOP) is rather rare. I don't expect we can reproduce them easily and I don't plan to go that route if we don't have to. I do strongly believe it has something to do with the youtube vid I posted. I'm thinking we reproduce the circuit in the youtube vid and replace the cap with an LED, then insert a resistor to measure current from the input and output similar to the JT while pulse it with DC...
Interesting, thanks, Gibbs. I like that circuit also -- will build it, too. One should start a separate thread for it -- whoever replicates it first, OK?
QuoteI admit I have not been reading or looking at the scope shots. What I gather from skimming through the thread is that PhysicsProf has been reporting over unity results for the past week or more.
Look at the lesson in Poynt's experience. With perseverance and more investigation you should figure this enigma out. I realize that I am just parachuting in here but mark my words, there is no over unity associated with PhysicsProf's testing. It's just a question of honing your skills and finding where you made an oversight.
MileHigh
So long, MH. When you want to look at my scope-shots and data, you're welcome back.
There are some good ideas that evolve from discussing actual circuits and data, and I appreciate this feedback, including lanenal, feynman, .99 and Gibbs.
The little JT circuit is a self-resonator... seeks out a resonance rather quickly with major variables being the toroid windings, input voltage and resistance going into the base from the feedback-loop. All easy to change.
Getting at the OOP condition is difficult -- I've been playing with the circuit for several hours since Friday, looking at it with my 60 MHz Atten DSO, with no success-- on that circuit from Friday.
Besides, I think the point is that condition was reached and we -- I - am looking for a way to get there reliably and repeatably.
And I found something, inspired by comments on the LED being a diode -- it must be a leaky diode to permit the OOP condition (I think, may be wrong about that).
Anyway, I just replaced the LED with a RESISTOR, same circuit otherwise. The JT works fine, self-resonates readily with a resistor in place of the LED.
With a 10-ohm resistor replacing the LED, call it Rout, using my ATTEN, I find the input and output powers at about one based on areas under Pin and Pout, but I'll wait till I get back to the 3032 to get numbers. Furthermore -- with this Rout as a variable 1K-ohm resistor, I can see how the waveform varies as I vary Rout. Very interesting... and
I can get an Out-of-Phase condition quite easily with this Rout! (Although it does not have the same waveform as Friday's exactly...) So that is progress and as I say, the idea came out of this discussion. PS -- the Tek 3032 I use is a 300 MHz oscilloscope... glad I can use it, when I get up to the University!
This is getting more and more interesting...if we are willing to put aside our prejudice, this is a great learning experience.
@PhysicsProf
Replacing the joule thief LED with a variable resistor was an excellent idea. Does modulating this resistor value change the resonant frequency significantly?
Looking forward to replicating.
Quote from: feynman on 2011.03.07, 02:44:28
@PhysicsProf
Replacing the joule thief LED with a variable resistor was an excellent idea. Does modulating this resistor value change the resonant frequency significantly?
Looking forward to replicating.
Yes, the resonant frequency is sensitive to the resistance of Rout, and the toroid windings (and L) and the input voltage -- and the choice of transistor. The parameter space is quite large actually.
Glad you are joining in with real experiments, feynman.
For those serious about the fine points of energy measurement, efficiency, and switching power supply design, may I suggest you go to Power Integrations:
http://www.powerint.com/
http://www.powerint.com/en/pi-university
Sign up and go to their on line university. Read all of their application notes, especially on the design of inductor and transformers for high efficiency . They have some interesting chips with very clever internals. Study them. They are a long way from a JT design, however, some of the principles may apply, and there is a huge body of knowledge here that will help you get up the learning curve faster.
@PhysicsProf
Would you say this image is an accurate comparison of waveforms? I used your COP=0.90 circuit vs the COP=1.13 circuit and essentially photoshopped the images for comparison.
(http://feynmanslab.com/images/post4/PhysicsProf_comparison1_png.png)
http://feynmanslab.com/images/post4/PhysicsProf_comparison1_png.png (http://feynmanslab.com/images/post4/PhysicsProf_comparison1_png.png)
Thanks
Quote from: feynman on 2011.03.07, 19:11:45
@PhysicsProf
Would you say this image is an accurate comparison of waveforms? I used your COP=0.90 circuit vs the COP=1.13 circuit and essentially photoshopped the images for comparison.
(http://feynmanslab.com/images/post4/PhysicsProf_comparison1_png.png)
http://feynmanslab.com/images/post4/PhysicsProf_comparison1_png.png (http://feynmanslab.com/images/post4/PhysicsProf_comparison1_png.png)
Thanks
Yes, that's the right identification -- and useful as long as one is careful to note that the scale for the lower-left (input) power waveform is 20 mv*V, whereas the scale for the lower-right (output) power waveform is 50 mV*V, so that must be taken into account. Thanks.
Quote from: MileHigh on 2011.03.07, 00:52:10
I don't know what "IO" is. The basis behind the statement is that an inductor can only put out as much energy as you put into it. It's a device that can store energy and then release that stored energy. It cannot create energy or harvest energy from "somewhere else."
[snip]
MileHigh
Well, the inductor can 'harvest' energy inadvertently when it is being exposed to an external magnetic field. The result is that some energy is induced into the inductor and into the circuit it's in. It acts as a magnetic pickup. The toroid typically used in a JT is relatively immune because it has a closed tight magnetic path. If it was a bobbin inductor like a typical choke it could pick up much more energy from external magnetic fields.
Here is my comments on my blog about this thread. http://watsonseblog.blogspot.com/2011/03/2011-mar-8-joule-thief-efficiency-113.html (http://watsonseblog.blogspot.com/2011/03/2011-mar-8-joule-thief-efficiency-113.html) One thing I would really appreciate is a good closeup picture of the circuit. It would make it a lot easier to replicate not only the circuit but the results.
Quote from: acmefixer on 2011.03.09, 02:13:47
Well, the inductor can 'harvest' energy inadvertently when it is being exposed to an external magnetic field. The result is that some energy is induced into the inductor and into the circuit it's in. It acts as a magnetic pickup. The toroid typically used in a JT is relatively immune because it has a closed tight magnetic path. If it was a bobbin inductor like a typical choke it could pick up much more energy from external magnetic fields.
Here is my comments on my blog about this thread. http://watsonseblog.blogspot.com/2011/03/2011-mar-8-joule-thief-efficiency-113.html (http://watsonseblog.blogspot.com/2011/03/2011-mar-8-joule-thief-efficiency-113.html) One thing I would really appreciate is a good closeup picture of the circuit. It would make it a lot easier to replicate not only the circuit but the results.
Attached find a photo acquired at the University on 4 March 2011, the day the n = 1.13 value was obtained as explained in some detail above. The photo was taken with my HP digital camera, which is not great for close-up shots, but by comparing with the schematic (also in the attached), I think you can pick out the essential elements. For these tests, I used a power supply as described rather than a battery, and found that varying the input voltage was useful in changing the conditions of the experiment. The best results were obtained at 0.996V input voltage. Note that the power IN and power OUT are measured and compared, so the use of a power supply should not mitigate against the validity of the calculated N.
.99 wrote:
QuoteI've discovered the error in my test setup (scope grounds on wrong node), so here are the measurements again, as pertaining to "schema02", i.e. no CSR3 present.
Pitotal = -46.64mW
Pcsr1 = 2.79mW
Pvbat = 49.43mW [Note that you can obtain this value directly when measuring Pitotal just by inverting the CSR1 channel]
Pototal = 33.62mW (PLED and Pcsr2)
Pototal + Pcsr1 = 36.41mW
I did not measure PQ.
So, efficiency not including output power dissipated in Q is:
n = 36.41mW / 49.43mW
n = 73.66%
.99
Would you say that I made the same error? (See previous post showing connections used and photo showing probe connections to determine Pinput.) Thanks.
Where are your two scope grounds connected again?
It almost looks like one scope ground is attached to the same place as the other scope probe? Is that incorrect?
If you describe the photograph verbally in detail I will label the probes and grounds in Photoshop.
Quote from: feynman on 2011.03.09, 17:58:16
Where are your two scope grounds connected again?
It almost looks like one scope ground is attached to the same place as the other scope probe? Is that incorrect?
If you describe the photograph verbally in detail I will label the probes and grounds in Photoshop.
In fact -- the two scope grounds (one under the other, but connected together) are connected at the same location as shown in the schematic, "north" of the 1 ohm CSR.
The scope probes are then connected at points V1 and V2 shown in the schematic.
The larger cables coming in from the right are from the power supply, connected also as shown in the schematic.
Because of the location of the scope grounds the CSR1 voltage (V2) appears as negative, as expected from the schematic, whereas V1 and V3 appear as positive (generally -- the exception being the out-of-phase condition I have described). Therefore, Pin (V1*V2) and Pout (V3*V2) appear as negative (generally -- the exception being the out-of-phase condition I have described earlier in the thread).
Quote from: PhysicsProf on 2011.03.09, 17:36:27
.99 wrote:
Would you say that I made the same error? (See previous post showing connections used and photo showing probe connections to determine Pinput.) Thanks.
From what I can tell professor, your scope grounds are connected correctly, esp. if they are connected as per my schematic. A quick check is to take note that your scope grounds are NOT connected to the battery negative terminal directly. They need to be connected to the other side of CSR1, which I believe you actually have done.
.99
Quote from: PhysicsProf on 2011.03.09, 19:15:46
No -- the two scope grounds (one under the other, but connected together) are connected at the same location as shown in the schematic, "north" of the 1 ohm CSR.
The scope probes are then connected at points V1 and V2 shown in the schematic.
The larger cables coming in from the right are from the power supply, connected also as shown in the schematic.
Because of the location of the scope grounds, if a battery were used, the CSR1 voltage and the power appear as negative, as expected from the schematic.
Okay , sorry this is confusing. So, as I understand:
(http://feynmanslab.com/images/post4/4MarPhotoSchma_scaled.png)
-The red and the black leads on the right of the picture are the power supply.
-Both scope grounds are connected together . They are connected to CSR1 on the 'far' (away) side of the (-) of the power supply, labeled P11G and P12G in the schematic. P21G is also connected here. This ground is common across all calculated scope channels.
-This leaves us with three remaining scope probes (V1, V2, and V3). And V4 which is the same as V2.
-V1 (for measuring input power) is connected to the (+) of the power supply . [aka. P11T]
-V2 (for measuring input power) is connected to the (-) of the power supply, on the power supply side of CSR1. [aka. P12T]
-V3 (for measuring output power) is connected to the (+) of the LED, aka. the collector of the 2N2222 NPN transistor. [aka. P21T]
-V4 (for measuring output power) is the same waveform as probe V2.
Power_in = MEAN(V1*V2). Aka the total power of the battery and CSR1 together. Aka Pitotal.
Power_out = MEAN(V3*V2) Aka the total power of the LED and transistor together. Aka Pototal.
Is this correct? Does any one else have an opinion on these methods?
Also PhysicsProf , did you get a chance to check the COP range of the 1K variable resistor which replaced the LED?
Thanks
-Feynman
Quote from: poynt99 on 2011.03.09, 19:30:42
From what I can tell professor, your scope grounds are connected correctly, esp. if they are connected as per my schematic. A quick check is to take note that your scope grounds are NOT connected to the battery negative terminal directly. They need to be connected to the other side of CSR1, which I believe you actually have done.
.99
This is correct. Thank you, .99.
Quote from: feynman on 2011.03.09, 19:48:03
Okay , sorry this is confusing. So, as I understand:
(http://feynmanslab.com/images/post4/4MarPhotoSchma_scaled.png)
-The red and the black leads on the right of the picture are the power supply.
-Both scope grounds are connected together . They are connected to CSR1 on the 'far' (away) side of the (-) of the power supply, labeled P11G and P12G in the schematic. P21G is also connected here. This ground is common across all calculated scope channels.
-This leaves us with three remaining scope probes (V1, V2, and V3). And V4 which is the same as V2.
-V1 (for measuring input power) is connected to the (+) of the power supply . [aka. P11T]
-V2 (for measuring input power) is connected to the (-) of the power supply, on the power supply side of CSR1. [aka. P12T]
-V3 (for measuring output power) is connected to the (+) of the LED, aka. the collector of the 2N2222 NPN transistor. [aka. P21T]
-V4 (for measuring output power) is the same waveform as probe V2.
Power_in = MEAN(V1*V2). Aka the total power of the battery and CSR1 together. Aka Pitotal.
Power_out = MEAN(V3*V2) Aka the total power of the LED and transistor together. Aka Pototal.
Is this correct? Does any one else have an opinion on these methods?
Also PhysicsProf , did you get a chance to check the COP range of the 1K variable resistor which replaced the LED?
Thanks
-Feynman
Yes, you have stated it correctly, Feynman -- and .99.
Feynman:
QuoteAlso PhysicsProf , did you get a chance to check the COP range of the 1K variable resistor which replaced the LED?
Not yet -- I have not made the trip back to the University since the tests last Friday, March 4th, reported on above.
Hi poynt99
Is the output power measurement done using the product of rms voltage values, or by integrating over time the product of the instantaneous values V2(t)*V3(t)?
Only the second method can give correct results. The first one is false because the rms value of a product is not equal to the product of rms values.
May be I missed something. Could you clarify?
Quote from: exnihiloest on 2011.03.10, 11:26:55
Hi poynt99
Is the output power measurement done using the product of rms voltage values, or by integrating over time the product of the instantaneous values V2(t)*V3(t)?
Only the second method can give correct results. The first one is false because the rms value of a product is not equal to the product of rms values.
May be I missed something. Could you clarify?
We are using neither, but closer to the energy measurement you mention above.
We have been simply measuring input and output power using v(t) and i(t). It is the same result as the energy measurement, because the JT oscillation is periodic.
We've previously established the protocols for measurement, and all agreed that measuring the power is adequate and equivalent in this case. I hope you are not suggesting the contrary. It was also established long ago with Lawrence, that using RMS values is incorrect.
.99
@exnihiloest
Lanenal developed the energy test method, so I'm sure he can comment on details.
I use an old analog scope, so I don't have a scope with math functions. I think on a digital scope MEAN(V1*V2) is going to give the moving/instantaneous average of the instantaneous product of the two curves -- that is , the MA of the product of their two instantaneous values at time t. As such, it's definitely not RMS.
Quote from: poynt99 on 2011.03.10, 15:04:41
We are using neither, but closer to the energy measurement you mention above.
We have been simply measuring input and output power using v(t) and i(t). It is the same result as the energy measurement, because the JT oscillation is periodic.
We've previously established the protocols for measurement, and all agreed that measuring the power is adequate and equivalent in this case. I hope you are not suggesting the contrary. It was also established long ago with Lawrence, that using RMS values is incorrect.
.99
Correct. We calculate the input and output power using the math function for v(t) and i(t) [from the voltage drop across a 1 ohm resistor, typically) and then take the MEAN value of P(t) = v(t) * i(t) over numerous cycles. Not RMS values.
@PhysicsProf
I dug up my trifilar Joule Thief as well as my analog 2-channel tektronix scope. So I can't give power measurements, but I will be able to photograph waveforms. I'll post em as I get em.
I'm getting a digital 4 channel scope, but this will be a couple of weeks before I have this capability.
Quote from: feynman on 2011.03.10, 15:55:42
@PhysicsProf
I dug up my trifilar Joule Thief as well as my analog tektronix scope. So I can't give power measurements, but I will be able to photograph waveforms. I'll post em as I get em.
Looking forward to it, Feynman. Suggest we start a new thread as this one is far from its starting point.
H
ow does one set up a 'Bench'? I like this feature here and would like to record thoughts and progress from my electronics bench...
Quote from: poynt99 on 2011.03.10, 15:04:41
...
We have been simply measuring input and output power using v(t) and i(t). It is the same result as the energy measurement, because the JT oscillation is periodic.
We've previously established the protocols for measurement, and all agreed that measuring the power is adequate and equivalent in this case. I hope you are not suggesting the contrary. It was also established long ago with Lawrence, that using RMS values is incorrect.
...
Therefore I understand you calculate the power by summing (v(t)*i(t)) (not by making Vrms*Irms).
I perfectly agree, it is the right method. I was not suggesting any thing else. It was not clear for me what was used, and I just wanted to be sure (I don't master all subtetlies of English language), sorry if I disturbed the thread.
Thanks.
Quote from: exnihiloest on 2011.03.11, 10:44:57
Therefore I understand you calculate the power by summing (v(t)*i(t)) (not by making Vrms*Irms).
I perfectly agree, it is the right method. I was not suggesting any thing else. It was not clear for me what was used, and I just wanted to be sure (I don't master all subtetlies of English language), sorry if I disturbed the thread.
Thanks.
No problem at all exnih. Perhaps we are a bit touchy these days when anyone mentions using RMS for these measurements, as it seems we have to go to extraordinary lengths to change people's minds about the methods they were using. It required a lot of effort to make a little change, and I'd hate to have to go through all of that again.
Regards,
.99
Quote from: PhysicsProf on 2011.03.09, 17:34:01
Attached find a photo acquired at the University on 4 March 2011, the day the n = 1.13 value was obtained as explained in some detail above. The photo was taken with my HP digital camera, which is not great for close-up shots, but by comparing with the schematic (also in the attached), I think you can pick out the essential elements. For these tests, I used a power supply as described rather than a battery, and found that varying the input voltage was useful in changing the conditions of the experiment. The best results were obtained at 0.996V input voltage. Note that the power IN and power OUT are measured and compared, so the use of a power supply should not mitigate against the validity of the calculated N.
Hrm, some wires hanging loose -- another reason for a good closeup. One thing I can see, the black plastic transistor. A 2N2222 comes in a metal package, the PN2222A or MPS2222 come in the plastic package. One more unanswered question that needs to be answered before anyone can replicate the circuit.
Thank you for the followup.
Quote from: MileHigh on 2011.03.07, 00:52:10
...
Anyway, with perseverance PhysicsProf should start seeing some numbers that refute his current findings suggesting over unity.
...
I admire your optimism. ;)