Is joule thief circuit gets overunity?

[TinselKoala]

http://touchstonesemi.com/products/current-sense-amplifiers

These aren't super fast but from looking at the data sheet they should be usable to well over 20kHz with much better accuracy than what you are using now and they draw only microamps of current themselves. And you might be able to get a free sample demo board, complete with the low-inductance path, very low resistance precision CVR.

[Void]

These aren't super fast but from looking at the data sheet they should be usable to well over 20kHz with much better accuracy than what you are using now and they draw only microamps of current themselves. And you might be able to get a free sample demo board, complete with the low-inductance path, very low resistance precision CVR.

Nice. Thanks for the info.  If I were measuring efficiencies close to 100% I would definitely look into those boards. I have saved the link just in case.

[ltseung888]

@Void,

Now that we are using the same connections for measurement, the results should be compariable.

See if Baord 118 or 119 show the same characteristics in having "noise".  Please adjust your DC Power Supply so that the LED just light up.  Reduce slightly and the LED should still be ON.  With Board 120, that occurs at CH1 Vrms = 344mV.  The Waveform for CH2 has spikes crossing the 0 ref line.

If a capacitor is connected in parallel with the DC Power Supply, there will be the longest lighting with the lowest drop of voltage around this range.  Poynt99 can use Board 33 to do the same.  He will be using the Tektronics.  Please display your waveform.

Many of you will comment that this is around the "noise" level.  Let us do the experiments first.....

[Void]

@Void,
Now that we are using the same connections for measurement, the results should be comparable.
See if Baord 118 or 119 show the same characteristics in having "noise".  Please adjust your DC Power Supply so that the LED just light up.  Reduce slightly and the LED should still be ON.  With Board 120, that occurs at CH1 Vrms = 344mV.  The Waveform for CH2 has spikes crossing the 0 ref line.
If a capacitor is connected in parallel with the DC Power Supply, there will be the longest lighting with the lowest drop of voltage around this range.  Poynt99 can use Board 33 to do the same.  He will be using the Tektronics.  Please display your waveform.
Many of you will comment that this is around the "noise" level.  Let us do the experiments first.....

@Lawrence:
As I have previously explained, as the amplitude of the waveforms being measured fall closer and closer into the noise level amplitude range, the degree of error in the measurements increases accordingly. Also, when making measurements close to the lower measurement limits of the measuring instrument, the degree of error is typically a lot higher than when making measurements that are at higher magnitudes and well within the normal measurement range of the measuring instrument. 

For the input voltage setting you are suggesting here, the input and output current magnitudes are well within the noise level range, especially the output current which is completely within the noise level, and also the current waveform magnitudes are down at the very lower measurement limit of my scope, which will also introduce even further error into the measurements. Regarding the noise, we can't assume that the noise waveforms are always perfectly symmetrical and therefore will average out to about 0 overall, as noise at any instance in time is the sum total of all noise sources affecting your circuit over the measurement period, and although some noise may be fairly symmetrical about the 0 voltage axis, other sources of noise may average out to be more positive or more negative overall, and the noise affecting a circuit can vary a fair bit over time depending on the sources of the noise. The result is that when the magnitude of waveforms being measured are well within the noise level, the noise could very potentially be skewing our measurements quite a bit, and it in all likelihood is. You just simply can't rely on such measurements being at all accurate under such conditions. As I have also mentioned before, you would in the very least need something like a high quality amplified differential probe to attempt such measurements, and you would also want to take steps to minimize noise in your circuit as much as possible. You would also want to be using a good high quality and very well calibrated scope for the measurements as well. Even then, you would still need to be aware of the measurement limitations of the equipment you are using to make your measurements, and be aware of how much error noise may still be introducing into your measurements.

With all the above being understood in regards to the measurement error being much too high to make meaningful measurements at the current waveform magnitudes which result when the input voltage is set very low, I made the following measurements just to show what kind of results you can end up with when attempting to make measurements under such conditions. If a person did not fully understand about measurement instrument limitations, and the effects of noise on measurements when waveform magnitudes are well within the noise level, then a person might make the mistake of thinking that measurements made under such conditions are meaningful. In reality however, it is very likely in the following measurement results that the measured values have a high degree of error due to the current waveforms being measured being well within the noise level. Anyone with an engineering or science background should be well aware of these sorts of measurement limitations and sources of error in measurements.

I used your scope probe connection method for these measurements.
I used a regulated DC power supply only (no super cap in parallel) for the input voltage source.
I made certain that the DC offset for Ch2 (used for current measurements) was calibrated as close to 0V as I was able to get it, before making the measurements.
(Again, these measurements are meaningless due to the current waveforms being measured being well into the noise level.)
Input Voltage: 328mV  (The output LED was lighting dimly at this input voltage setting.)
Input Power:   32.275uW
Output Power: 103.278uW
Efficiency:       319.994% 

Scope screen shots are attached.
Voltage traces are in yellow, current traces are in blue.
Input current waveform is inverted.

[Void]

@Lawrence: Just to try to help clarify what I was saying in my post above, I am attaching a scope screen shot of the Ch2 probe connected exactly as it was when I made the output current measurement shown in my previous post, but in this case the power supply is still connected to the circuit, but the power supply is switched off, to just show the noise level being picked up by the scope probe at this measurement point in the circuit. As you can see in the attached scope shot, the noise waveform with the power supply switched off looks very close to what the 'output current' waveform looked like in my measurements shown in my previous post. In other words, what is being measured as 'output current' under such conditions is actually mainly electrical noise that is being picked up by the circuit and probe. This will in all likelihood greatly skew the output current measurement and give you a meaningless measurement result. The actual output current that is passing through the output LED at this low input voltage setting on the joule thief circuit (for example at 328mV) is most likely in the very low microamps range, probably somewhere in the vicinity of 20uA or so. Also, such a low current is likely well below the minimum amplitude that my scope can measure with any degree of accuracy.
- void -