Measuring INPUT Power Accurately and with no Oscilloscope

[poynt99]

This weekend ended up being spent mostly with family, so I was not able to perform all the testing I wanted to.

I did however get everything set up, and to the point I was able to obtain some preliminary results.

Using a ~3kHz square wave (~50% duty) drive from a 555 circuit (more or less the original Ainslie circuit built using Groundloop's PCB), I was able to confirm what the Pin measurement was with ~15VDC supply input through the switching circuitry on a 10 Ohm inductive resistor.

The 100W load resistor got up to about 64ºC, and the measured Pin's using 4 different methods were as follows:

1) PSU meters: 0.68A x 15.2V = 10.3W
2) Scope:  MEAN[vbat(t) x 4icsr(t)] (4i to account for the 0.25 Ohm CSR) = 9.88W
3) Scope: MEAN[vbat(t)] x MEAN[4icsr(t)] = 9.91W
4) DMM: 15.18VDC x 158.2mVDC(Vcsr) x 4 = 9.6W

No external filtering was used for the DMM measurement, and the signals were fairly "clean" with little ringing. So far, so good.

Next step is to obtain some ringing in the 1MHz frequency range, and perform the same tests again. Stay tuned for more this week.

.99

[poynt99]

Here is a preview of the Burst Oscillator Circuit I developed in order to bench-test/prove the power measurement methods I have previously described and simulated (summarized on schema01 below). Also shown is an example wave form output in BURST mode. The oscillator can be operated in CONTINUOUS mode (constant oscillation), or for a more interesting and perhaps challenging measurement, BURST mode, depending on SW1 (yet to be shown, but shorts across RDUTY for continuous).

Note the included equation for obtaining the Pout measurement. 

The parts should arrive Thursday. Then the build and tweaking.

Any questions or doubts etc., feel free to raise them; I'd be happy to address all.

Regards,
.99

PS. The TC4426 is a MOSFET driver chip, modified to provide BURST control as well as drive to the IRF840 switch.

[poynt99]

I've been inspired to update this thread and move this burst oscillator circuit forward; to the point it can be built and used.

I've refined the design to make it more practical. It now features a few improvements as well.

This somewhat unique oscillator is largely based on the circuit that the RATS (Rosemary Ainslie's Team of Scientists) came up with a while back. The circuit retains similar wave forms to my past simulations of the RAT circuit, but this implementation is much more eloquent than the original RAT'S design.

I've now isolated the "bias" circuit from the oscillation path, which has been replaced with a large capacitor (one electrolytic, and one film). This is beneficial because there are no longer currents of any significant degree flowing in the DC bias network. The DC bias is best implemented by using a pack of 4 'AA' batteries, which also provide the supply (VDD) for the MOSFET driver. This arrangement also puts to rest the notion that the bias supply (Function Generator in the RAT circuit) supplies any appreciable energy to the circuit. It can't as shown, and it doesn't.

Again, 'M4' in my schematic is equivalent to the 'Q2' MOSFET in the RAT circuit, and my diode 'D1" is equivalent to the RAT circuit 'Q1'. The diode and 'Q1' MOSFET are interchangeable, and I've chosen the diode for now.

The 'BURST/CONT' MOSFET driver (green outline) can be replaced with any function generator, 555 circuit, or similar. All that is required is a minimum 0V to 5V pulse train (no offset required) with any desired duty cycle. The switch 'S1' as shown allows you to quickly switch between BURST mode (frequency and duty cycle set by the potentiometers) and CONTINUOUS OSCILLATION mode. BURST mode should be the most challenging to measure, so that is the mode I'm using.

THE CSR was moved over to the 'proto-board' circuit end (vs. the battery end), and the wiring lengths to the battery array are denoted, now as per the RAT circuit.

The 'LWire1' is approximately 10 inches of say #14 or #16 jacketed wire, coiled or straight. This may require some experimentation.

Disregard the small green doughnut near S1, it's an error message from PSPICE. See "burst_osc_schematic01.png" for the schematic.

After refining the design I decided to run some similar tests to those performed in my "detailed_analysis06.pdf" document. These follow.

The first test was performed as per the 'RAT" method, whereby the battery voltage scope probe is placed at the top of the load resistor (as shown in "burst_osc_schematic01.png") rather than on the battery terminal itself. As well, this "battery" voltage is multiplied with the CSR voltage in real time to produce a p(t) wave form. Then this wave form is averaged to get the final result for battery power, or INPUT power.

You can see from the "rat02.png" graph, that the INPUT power measured is -74W. This is in line with the results the RATS obtained.

Next, we move the battery voltage probe over to the battery as shown in "burst_osc_schematic02.png" (the green probes). This I have named a "TRUE" measurement of battery voltage and hence power.  As seen in "true02.png", the INPUT power measured is now +21.3W.

So, not only has the polarity changed, but the value is quite different as well. Let's move on.

The next test shows the "averaging method" I have been promoting. This requires only two DC voltage meters and a simple calculation. See the red "Pin(AVG)" equation on the schematic.

From the graph "avg02.png" we see that the power measured is +21.57W. Again, note it has a positive polarity, and that it is very close to the "true02.png" measurement above. NB. I am using the same probe locations as those used by the RATS, as shown in "burst_osc_schematic01.png"! The parasitic inductances and oscillation throughout the circuit have no significant effect on the measurement. This is in stark contrast to the errors exposed when using a scope and computing p(t) from v(t) x i(t) at these probe locations.

In PSPICE we have available a "Wattage Probe" that gives us instantaneous power p(t) directly. Thereby, we can obtain the actual power delivered by the battery simply by averaging the measurement on this probe. The graph "actual02.png" shows the result of -21.54W as the actual average INPUT power used in this circuit. Notice the polarity? All the polarities above (EXCEPT the -21.54W) actually need to be inverted because the voltage probe across the battery is reversed. Remember, power sources lose energy ('-' sign), and loads gain/dissipate energy ('+' sign). Easy to remember.

So, we have three INPUT power measurements that correlate quite closely, and one that is completely "out to lunch", that obtained using the RATS measurement method which produced a -74W result.

These results are consistent with those results found in the "detailed_analysis06.pdf" document.

Folks, which method are YOU going to AVOID using?

Stay tuned. I'm hoping to soon build the Burst Oscillator circuit and demonstrate these tests for you.

.99

[poynt99]

Cool stuff, this method work with a modified sine-wave inverter ?
(I need to measure the input power from the DC side for my resonant amplification experiment: all measure will be in DC to avoid error...)
Can I use it for a rectified unfiltered DC OUPUT ?


Edit: I have also a scope (DSO 2090) to get REAL power including AC (distorted dephased sine wave of course), I can use the Math function ChannelA mean * ChannelB mean ?

Schubert,

I simulated this method on an unfiltered 60Hz full wave rectified output to a pure resistive load, and it does not produce the correct INPUT power result. Why? Because we are not starting out with a pure DC source like we would have with a battery.

One option to overcome this issue, would be to incorporate a large capacitor on the supply output, but for a 1500W load, you would need something on the order of a 1F capacitor, which is not necessarily practical.

However, IF you know for certain that there is no significant phase shift between the output voltage and current (check it with your scope), you can use a good quality true RMS voltage meter to measure both the voltage and current. Then the results are multiplied together (remember to factor in the value of the CSR resistor) to give you the INPUT power result.

If there IS a significant phase shift, then the best method will be to use the oscilloscope as per normal methods.

.99