Testing the TK Tar Baby

[TinselKoala]

TK:

No, because what you are interested in knowing is what the power consumption is for a constant voltage, and not a variable voltage.

If you can imagine using a much larger capacitor so that the voltage is much steadier, then the slope of the cap discharge at say, 72 volts, would be much easier to measure.  So once you have the (nearly constant) slope of the change in voltage at 72 volts, and you know the capacitance, you can derive the current, i = C dv/dt.   So you could then get the power consumption at 72 volts.

I believe that you can also simply drop a power probe on the capacitor itself.  I am not sure if you can low-pass-filter the power probe data (i.e.; some sort of a running average), but I assume that you get the idea.

MileHigh

Ah....no. But thank you for confusing the issue further.

What I am interested in is the correct computation of average power dissipated during a time interval, and that is the number of Joules divided by the number of seconds. The waveshape is irrelevant to that computation, isn't it? Steady, wiggly, positive/negative....... as long as you count up the net Joules, adding all the positive ones and subtracting all the negative ones, you get the net ENERGY that passed your measurement point. And if you know the time interval from when you started counting (the full cap at t=0) to when you stopped counting Joules (oscillations stop and cap voltage flatlines) you know the average Joules Per Second "over" that interval. The average power in Watts, in other words.

And I am still asking: For the AVERAGE POWER levels cited by Ainslie, like -40 Watts, when there is a DC component that averages, say, 20 Watts .... what is the negative power level required during the oscillations to bring the average to the claim?

For example, say we have 100 Watts  positive inst. power during the DC phase, gate HI, and it is 18 percent of the total period. This means we have an AVERAGE of 18 Watts positive over the whole period. But the average power computed somehow by the NERDS is -40 Watts. So whatever is happening in the oscillations is bringing down the DC average of 18 Watts down to -40 Watts.

I think the way to find this value is like this:
(100 W x 0.18) + ( OscPower x 0.82) = P_avg  for the whole period.

Solving for OscPower,
(OscPower x 0.82) = Pavg - (100 x 0.18)
OscPower = (Pavg-(100x0.18))/0.82

Inserting the claimed Pavg value of -40 W we have
OscPower = (-40 - (100 x 0.18))/0.82
OscPower == about -71 Watts.

Right?

So now I am asking, where on the scope traces is this high negative power level to be found?  I know that it is a spurious measurement caused by the probe locations and excess wiring; that is not the issue. I want to know where the numbers _come from_ that lead to the claims by the NERDS of the high negative power levels. If they claim a -40 average for a trial that includes substantial DC power during the non-oscs...  then the oscs need to be even more powerful than the stated value, and I want to know where the numbers are coming from.

[MileHigh]

Shades of the Mandelbrot Set!   

[TinselKoala]

With the Watt probe directly on the 8uF capacitor, we get the TRUE instantaneous and average power over the 3ms period.

The average is correctly negative, because sources give up power, while loads dissipate power.

And it looks like about the same value that I got using the energy and the time interval-- something around | 8 Watts | .

[poynt99]

And the Joules of energy delivered by the capacitor.

[MileHigh]

TK:

Yes on your calculation of -71 watts.

So now I am asking, where on the scope traces is this high negative power level to be found?  I know that it is a spurious measurement caused by the probe locations and excess wiring; that is not the issue. I want to know where the numbers _come from_ that lead to the claims by the NERDS of the high negative power levels.

Sounds like you would need hands-on with the DSO (or the spreadsheet dumps if they exist).  Naturally you are expecting -71 watts measured power during the negative oscillation phase.

The big caveat for me is that I don't know the DSO.  What does it do when it's on a very slow timebase and is measuring minutes worth of data.  You are likely undersampled during the negative oscillation phase, but do Monte Carlo methods give you good data anyways?   When your display window on the actual screen is much smaller than the buffer of recorded data, does the DSO average the entire buffer anyways or does it just average what is in the display window, etc.

I suspect that all of these issues were over the heads of the NERD team.

I agree that the negative averaged power data is suspect when you have long on times for Q1 with no oscillation and heavy positive power dissipation in the load.

The DSO "doesn't lie" or does it?

MileHigh