Rosemary Ainslie Quantum Magazine Circuit COP > 17 Claims

[MarkE]

The raw data from those runs, along with the graphs:

http://www.youtube.com/watch?v=ZN-wzEGXuqg

Using the switching circuit it looks like you are getting about 13W equivalent from 17W in, or roughly 76% efficiency.  A clean capture of the drain voltage at a good resolution like 5V/div and drain (source) current will allow estimation of switching and conduction losses which should come out to about the 4W difference. 

[TinselKoala]

Yes, that's right. The efficiency figures don't include the power dissipated at the mosfet of course; to get the full power dissipation of the circuit I'd have to immerse the whole thing in oil and do the same kind of runs that way. I'm sure that the goalposts will be moved out that far, eventually, by the Great Scientist, but until then I'm not going to show any of that working.

Here's how I'll be condensing and displaying the data from experimental runs as they accumulate. For the performance to be "overunity" or better than straight DC power the data point from an individual run has to fall above the DC calibration line. I've got two data points so far. I can put the results of any circuit, any input power, any operating parameters on this graph as long as it uses the same fancooled load cell in the same condition and reaches a stable or nearly stable temperature in under 60 minutes.

[MarkE]

It seems that so far:  Your passive devices act passively. 

[MarkE]

I meant to ask:  How have you set-up the gray box input power measurements?   I think that you want to show that the error bars are small enough that they cannot change any comparative conclusions. 

[TinselKoala]

I meant to ask:  How have you set-up the gray box input power measurements?   I think that you want to show that the error bars are small enough that they cannot change any comparative conclusions.

Right you are. The primary data comes from the camera datalogging: The two DMMs measuring Voltage across the supply filter cap and in-line current in the positive lead to the DUT. Oscilloscope calibrations show that the voltage is essentially constant and the DMM's indicated current is within 5 percent of the value read from the scope using Vdrop across the non-inductive CSR arrangements (using the Link DSO, so I have real numbers, not just fingerpointings.) I've placed +/- 5% error bars on the DUT datapoints.... the bars are about the same size as the markers ! I've also made the error bars symmetrical, but I think the DMM's error is usually mostly on the low side. If pedantry and precision require it, I could take multiple data runs and use statistical procedures to get more accurate estimates of the error, but I think that this would yield even smaller error ranges on the plot.

I've added a run with the SWeir frequency-compensated Shifting Paradigms test board at about the same parameters as the two Grey Box trials, and I have another one yet to plot at near the original Quantum spec of 2.4 kHz and very short duty cycle. I couldn't get below 5.6 % using the F43, I'll have to resort to the DP101 to get a 3.7 percent duty cycle.

On the chart below, I reversed the axes so that Y error bars would correctly reflect the uncertainty in the power measurement, not the temperature measurement. So, on this chart, for a device to be doing better than the straight DC power, the data point must be _below_ the blue line of the DC calibration. That is, if a device needs more power to get to the same temperature as the DC powered load, it is less efficient and its point will be above the blue line. If the DUT needs _less_ power than the DC case to get to the same temperature, its point will fall below the blue line.