GM East Burst Heater Circuit

[MarkE]

Ms. Ainslie objects that I have suggested additional tests to Greg.  Greg acknowledges the possibility that what he has observed will turn out to be a difference in battery discharge behavior using pulses rather than DC.  Which BTW would still be a very useful thing.  If that is the case, versus over unity, and I believe it to be so, then one could either research to see if optimum frequency and/or duty-cycle is known or conduct experiments to find out what they are.  Perhaps more than 25% improvement is possible using the right wave shape and frequency.

Another possibility is that the differences that Greg sees are due to extra power coming from somewhere.  That would mean that at long last a self-powered over unity device could be engineered.  While I discount the likelihood of that being so, this is why we conduct experiments:  to find out the truth.  The thermal tests that I have suggested will make quick work of determining how much power the arrangement delivers.  If there is extra power, and Greg continues to be as meticulous as he has, then thermal tests will provide solid evidence.  If that is the case, then I know of at least one person who would willingly finance the capacitor tests.

Ms. Ainslie seems to think that I am out to derail Greg somehow.  She is welcome to her suspicions.  She is also welcome to try and point to anywhere that I have derailed a legitimate idea by any means.  I submit that the thermal tests that I have suggested can be conducted far more quickly at much lower cost than the capacitor phalanx that Greg has been pursuing.  I submit as I have above that thermal tests that reliably indicate OU would attract many offers of assistance.

[TinselKoala]

I think it's interesting that the "cheapo" undamped probe seems to respond as well as, or perhaps even a bit better than, the much more expensive undamped probe. Less overshoot, smoother ringing.

This makes me feel a bit more confident, since I have a couple of the 12 dollar probes too. (Along with some of the much more expensive ones.) My scopes don't have the bandwidth necessary to see any differences in the probes; it's nice to see the test done on a 200 MHz scope. Thanks!

[MarkE]

I think it's interesting that the "cheapo" undamped probe seems to respond as well as, or perhaps even a bit better than, the much more expensive undamped probe. Less overshoot, smoother ringing.

This makes me feel a bit more confident, since I have a couple of the 12 dollar probes too. (Along with some of the much more expensive ones.) My scopes don't have the bandwidth necessary to see any differences in the probes; it's nice to see the test done on a 200 MHz scope. Thanks!

The el-cheapo probes are a pretty good value.  They are no match when performing coaxial probing.  They do OK for just looking at microcontroller I/O.  If you use a damping resistor then you can look at switching circuits and actually tell what you are looking at.  The better practice is to use a coaxial probing connection. That's just not always convenient.  Steve has collected all kinds of data on probes under different circumstances.

[MarkE]

So here are a couple of data points that should at least be intriguing to anyone following Greg's experiments:

He has a temperature difference set-up with one TC probe stuck inside the ceramic heater resistor and one in ambient air.  Using the switching circuit, he records an average current of 5.6mV/0.0499 Ohms, at ostensibly 24% duty-cycle leading to a 34.54C temperature rise.  The rms value of that is: 5.6E-3/.0499*1/(0.24)^0.5 = 229mA.   The DC current that yields the same result he measured as 363mA, 58% larger.  This is a temperature measurement comparison and so has nothing to do with battery effects.

So:

1) Is the reported duty cycle accurate? 
2) Were the battery voltage, power supply voltage, and current sense resistor readings accurate?
3) Were the thermocouple readings accurate proxies for like power dissipation by the heater resistor?

Any of these can be checked to see that they are true or false. 

1) Is a matter of taking a good scope shot of the MOSFET drain voltage.  Measuring duty-cycle is not difficult.  It is unlikely that Greg confused a 10% duty-cycle for 24%.
2) Can be verified with an additional DMM to double check the readings.  They are probably close.
3) This is trickier.  There are lots of ways to go wrong with temperature measurements as proxies for heating power.  The known reliable way to obtain heating power and energy readings is to measure the temperature rise of a known liquid thermal mass.

From a sanity check standpoint, it seems unreasonable that a simple pulse circuit should yield 58% excess energy over its DC rms equivalent for the simple reason that such circuits are used in billions of products everyday.  Someone else should have noticed.  Still, "should have" is not a scientific answer.   The primary unknown can be resolved with a pair of heat transfer experiments.

[TinselKoala]

If reliable measurements of temperature that yield heating power are actually desired, a brief literature search will reveal how it is generally done. The same load should be used for both the control and experimental trials. The imperfectly insulated load will of course take some time to reach its equilibrium temperature with the surroundings; at this point it will be dissipating the same amount of power it is receiving from the supply. (Or from wherever it is getting its power, like its zipons being agitated or whatever.) The temperature-time curves are plotted for both DC supply from a voltage regulated supply, and the Burst Heater circuit. The power supplied in each condition can be measured as DC power in, and the power dissipated at the load can be determined by finding the equilibrium temperature and comparing it to that obtained at the various non-extraordinary DC power levels used for calibration. This might take some time, as in a good load cell it might take 30 minutes or more for it to reach equilibrium temperature. The load's thermal mass and insulation should be chosen so this thermal settling time is not too long and not too short, and of course the load must be allowed to cool back to (regulated, constant) ambient temperature before beginning each trial run. Several trials should be made at each DC power level to allow for statistical accuracy checks. If variations are made in the operating parameters of the Burst Heater circuit, several trials should be performed at each set of parameters as well. The same load cell should be used because small differences that may not be under the experimenter's control could affect results if different load cells are used. A schedule of trials can and should be prepared beforehand and the order of trials can be randomized so that load cell ageing effects, if they occur, can be distributed evenly or randomly over the data set. Automated data logging, or at least a chart recorder record of time and temperature, would be a big help, but it is actually possible for a determined individual to record the necessary datapoints with paper and pencil.
This process will yield a series of curves in temperature-time plots. The equilibrium temperature reached by the Burst Heater trials can be compared to the temperatures reached at the various DC power levels and interpolated to yield the actual power dissipation level of the load cell when it is powered by the Burst Heater circuit.
Presumably the conventional power input to the Burst Heater circuit can be accurately measured... it could be powered by the same DC power supply at the same power levels as used for the DC calibrations, for example.
This process will also yield the data necessary for actual input and output energy comparisons as well, if anyone cares to do a bit more math.