Rosemary Ainslie circuit demonstration on Saturday March 12th 2011

[Rosemary Ainslie]

I really hate to butt into this pissing match but I agree with you Poynt99. These are the actual waveforms that need to be seen. However, I really don't think you'll ever get Rosemary to see it that way.

It's like turning on a light switch for 2.08 minutes. I can look at the whole on/off cycle on my scope but it doesn't really tell me whats going on in the circuit. I really need to see only 2 to 4 cycles on my scope to see whats going on.

And the irony here MrMag is that I ALSO agree with you.  It is - indeed - the only way to look up close.  But it is, nonetheless NOT the full cycle.  Which also makes it undersampled.  IF they required this then ask for it.  Don't accuse us of UNDERSAMPLING.

Rosemary

[cHeeseburger]

Before anyone gets too serious here about analyzing waveforms, I'd like to remind folks about the three really big problems I have long ago pointed out and yet another one which is likely new.

1)  The shunt inductance has a reactance of several times the value of the shunt resistance, so the signal being fed to the scope that is supposed to represent the instantaneous current value is indeed not representative of the instantaneous current value but far more representative of the rate of change of the current...a quantity that is unrelated to input power.

2)  The wiring inductance inbetween the batteries appears to be about equal to the wiring inductance from the batteries to the benchtop (the effects of the latter which I presume have been eliminated by probing at the batteruy end terminals).  So about half of that huge error signal has been eliminated.  The AC voltage that is still being fed into the scope as "battery voltage" is indeed not representative of the battery voltage but rather represents the battery voltage PLUS a huge amount of rate of change of current information (di/dt) that again, has no bearing on input power.

3)  Rosemary has reported that about 5 Watts of circulating power at 1.5MHz is present in the gate circuit flowing in the wiring between the signal generator and the MOSFET gates.  This shows up in the shunt and is inseperably mixed in with the battery current due to the shunt's position in the circuit (inside the gate drive loop).

4)  If indeed the scope probe for the "battery voltage" is now placed at the battery terminal end of the wiring over to the bench, the I assume the scope ground clip is also placed over there at the battery negative terminal and no longer at the "holy point" ground on the experiment board.  The ground clip for the shunt probe, on the other hand, is still located at the "holy point".  What this means is that there are now two current paths for the negative battery wiring from the battery to the bench and one of those paths is right through the scope grounds.

The large distortion this causes in the current measurement can be seen easily in Rosemary's post #549 http://www.overunity.com/index.php?topic=10407.msg280993#msg280993.  I noticed this long ago but didn't mention it before now.

So, in conclusion, whether over-sampled or undersampled and regardless of whether we are viewing a few cycles of 1.5MHz or five minutes worth across the screen, the signals being fed into the scope are NOT representative of the input battery voltage and current in the first place.  They are hopelessly polluted with other information that has no bearing on the power into or out of the batteries.  Unless and until these problems are resolved, the entire oscilloscope exercise is meaningless.

Humbugger

[poynt99]

No Poynty.  They're NOT right.  They're undersampled.  They are not the whole of the cycle.  I'm well aware of the fact that they give a very favourable result - especially in terms of our argument.  But they do not give the full picture.  They are UNDERSAMPLED.  TRULY UNDERSAMPLED.  Is that you you meant?  You WANT us to undersample?  In which case - that would certainly put paid to any acceptance of any of the results that we present.  VERY strange you are Poynty.  I'd be inclined to think that you want us to disgrace ourselves with the wrong presentation.  You really do take us all for fools. 

Rosemary

The area of interest is for the most part when the oscillation is present. When the oscillation is not present there is essentially no power being dissipated in the load (according to the wave forms), and the ratio in time between these two phases represents the power duty-cycle.

You seem quite confused as to the what the terms "sample rate" and "undersampling" mean. I would suggest you educate yourself about this before trying to educate anyone else on this important issue. Reading the links I provided should help you understand.

.99

[poynt99]

There are two separate "cycles" at play here, and this is part of your confusion.

The first cycle is that determined by the function generator. Let's call it "cycle1". The period and duty cycle of the function generator determines the period of time that the circuit is oscillating, and the period of time that it is not.

The second cycle is that determined by the period or frequency of the oscillation itself. Let's call it "cycle2".

You are concentrating on capturing multiple cycles of cycle1. This is not 100% correct and not necessary in this case. If in fact there is no power dissipation during the "dead" times, then you should try to achieve either constant oscillation, or adjust the cycle1 to be much much shorter in time, perhaps 10 or 100 times longer than the cycle2 cycle time. This way you are able to adjust the scope to capture sufficient cycles of both cycles (or just the one in the case of constant oscillation), and adjust for sufficient sampling to avoid aliasing errors.

When the oscilloscope time base is set to capture a number of cycles of cycle1, when cycle1 is on the order of 100's of seconds, the sample rate is then insufficient to properly capture the much higher frequency cycle2 wave form. This is undersampling, i.e. information is missing.

Imagine a standard movie film that runs at about 25 frames per second. Now imagine that the projector bulb only turns on for one frame out of every 5000 frames that run across the lens. What do you suppose the movie will look like in such a case?

The mechanism turning the bulb on and off is analogous to your sample rate when the time base is set for 100's of seconds, and the running 25 frames per second is analogous to the 1.5MHz oscillation signal.

.99

[Rosemary Ainslie]

Poynty - I'm going to give this one last effort.  Thereafter you can post whatever it is that you require to salvage some of that bruised ego.

Under USUAL conditions it is enough to extrapolate any part of any waveform and magnify it to determine its value.  Under such circumstances then LESS IS MORE.  10 samples preferred or 4 preferred - it makes no difference.  But those waveforms are then REQUIRED TO BE PRECISELY PERIODIC.  And then, that reduction in the number taken as a representative sample range cannot be considered UNDERSAMPLED as each waveform is precisely the same as the next.

THEREFORE UNDERSAMPLING first and foremost relates to ANY ATTEMPT to isolate any part of a complex waveform and then INFER OR DETERMINE that it can be ENTIRELY REPRESENTATIVE OF ALL THE WAVEFORMS OVER AN ENTIRE CYCLE.   

Where you Stefan and All ERRED - was to assume that you can isolate a few waveform samples on a COMPLEX CYCLE.  Correctly speaking the first 20% or so, of each cycle of oscillation - on this particular waveform shows RADICAL APERIODICITY.  It ramps UP.  And then it ramps DOWN.  NOW.  And remember that representative 20% also lasts for a commensurate 20% period of time over 2.08 minutes.

1 - Now.  Let's - for the purposes of this argument - say that the switch is ON for 20% of each cycle.  FOR SOME REASON THERE IS NO DELIVERY OF ENERGY FROM THE BATTERY DURING THIS TIME EVIDDENT ACROSS THE SHUNT.  But let's ignore that question for now. (I've done some serious deletions and editing here.  This point is very interesting and I'll get back here one day IF I can get past this obsessive interest in UNDERSAMPLING)

2 - THEN.  FOR the next 20% of 2.08 minutes of that cycle - there is a variation to the waveform that EXCEEDS that required periodicity - BECAUSE THE WAVEFORM FIRST RAMPS UP AND THEN RAMPS DOWN.  And that variation is evident over the battery and the shunt.  Correctly we now need each and every sample over that entire period - else any attempt to isolate a mere 4 waveforms absolutely WILL NOT BE REPRESENTATIVE.  IT WILL BE UNDERSAMPLED.  Here we would need to look at each and every one of thoe 500 000 points of data capture related to these samples over 20% of that ENTIRE cycle. 

3 - THEN only do we get back to that 60% of the time when the oscillation becomes TRULY PERIODIC.  Then is the ONLY time that we can infer any kind of power analysis from studying an isolated 4 or 5 detailed waveforms.  And having done that study we would still need to factor it in to a precise quotient of time during which this oscillation was evident.

4 - NOW - to get back to the point.  Your's, Humbugger's and Stefan's CLAIM was that we were DELIBERATELY UNDERSAMPLING TO ADD TO THE ARGUMENT.  I hope I've FINALLY EXPLAINED THIS.  I know why you claimed this.  Your ASSUMPTION - shared by your cronies - was that if you actually LOOKED at a few of those samples rather than at MANY - then - not only would you see a positive mean and cycle mean average across the shunt but that the math trace would also default to positive.  Herein lay your hopes.  But look back at my posts.  I AM ON RECORD.  I assured you all that a detailed analysis of those waveforms indicates an EVER LARGER GAIN to the system delivered to the battery.  A full sample across an entire cycle is the MOST CONSERVATIVE VALUE that we can manage.  For Heaven's sake POYNTY.  Just take a look at those values.  IF you think I isolated them then I'll take a series of consecutive downloads.  NEVER DOES THE CYCLE MEAN - THE MEAN OR THE MATH TRACE DEFAULT TO ANY POSTIVE VOLTAGES EVER - on that particular waveform.  WAKE UP AND SMELL THE COFFEE.  AT it's VERY LEAST - this is ANOMOLOUS.

So.  Just to conclude this ridiculous debate.  IF you require isolated waveform samples - then I asure you -  there is every indication that our efficiency is elevated to values that are entirely inappropriate to what is also actually evident.  I am, therefore, not prepared to evaluate this power analysis on isolated samples.  Because if we did, or had we done so, then you would and could have all accused us of UNDERSAMPLING.  And ONLY under those circumstances would you have been RIGHT.

Kindest regards,
Rosemary

Added the word 'us'.