Claimed OU circuit of Rosemary Ainslie

[poynt99]

Did a test tonight with the shunt resistor at the high side of the load instead of at the source. Still got wonky results:

POS = -6.72W
PIL = 1.31W
PIM = -8.09W
PIS = 0.068W

A run I did last night with the shunt still in the source was as follows:

POS =1.56W
PIL = 1.99W
PIM = -0.46W
PIS = 0.52W

The true POS (as measured with the add-on filtered shunt) is exactly 2.0W in each case.

Looking back at all 8 prior runs I've done so far, it would appear as though the POS and PIL values are swapped, just as they do for the second set of results above. Also, all PIL values were very close to 2W, and POS close to 1.5W.

.99

[MileHigh]

Hey Poynt,

It is late now and I wanted to look at some DSO scope captures again but instead let me just wing it here.  You asked about getting better results with the DSO and I think that it can be done.

I made reference to this before but let me try fleshing out the concept.  I think that you want to be selective about what parts of the waveform that you want to look at and how you process different parts.  I hark back to my posting the other day that showed the first part of the cycle before the MOSFET switches off actually records the dissipative and reactive power, which may imply that you are recording more power than you really should be recording.

Let's split the waveform up into two parts, and use the point where the current makes the zero cross and starts reversing as the division point.  It's probably a bit more complicated than that, but if you can account for the energy +/-10% that would be very convincing, much better than what you are seeing now.

So from the MOSFET switch-on to the current making the zero cross is what we will discuss first.

At the end of this phase you can get the total energy dissipated in resistive part of the load resistor, a no brainer.  Just please make a real measurement of the actual resistance to make the calculations more accurate.  Then the stored reactive energy would simply be the total measured energy less the dissipated energy.

It would be "fun" to use a spreadsheet column to show the "phantom voltage" for the resistive component of the load resistor and another column to show the "phantom voltage" for the inductive component in the load resistor.  You could do the energy integration on the inductor current times the phantom inductor voltage to get the stored reactive energy, which should be the same as calculated above.

The MOSFET dissipative energy during this part of the cycle is a no brainer also, just voltage times current times time.  I just checked a timing diagram - IMPORTANT - For the MOSFET dissipative energy, stop integrating about 150 nanoseconds before the zero cross event.  The reason is that you don't want to record any of the rising potential associated with the big delayed spike.

The shunt resistor dissipated energy is also a no brainer.

So, at this point we have three rock solid dissipative energy recordings, and we know how much reactive energy is stored up in the system.

Now here come the magic (I hope!)....  You can ignore the massive delayed spike for your energy calculations, ignore it!

There is a rational reason for this.  First, we know how much reactive energy is available to create that delayed spike, we just recorded it with the DSO.  We also know that the MOSFET is switched off and is just acting like a small capacitor and there is no power dissipation in the forward current direction.

So, the trick is to simply switch back to dissipative energy recording mode, but this time for the reverse current.

You record the total energy dissipated through the load resistor and the shunt resistor, and you completely ignore the MOSFET "output energy."   However, you do record the energy dissipated through the MOSFET body diode.  You also record the energy pumped back into the power supply (or battery).

Again, the rational is to completely ignore the high voltage spike, and only record the dissipative events that are the result of that high voltage spike.  Does that make sense to you?  The key factor being that we previously recorded the reactive energy in the first part of the cycle.  The spike is just the very same reactive energy creating a reverse current flow, and we can simply record the energy events that result of that reverse current flow.

Before the zero cross you have recorded the dissipative energy and the reactive energy.  This should be extremely accurate, there are no funnies going on.

Then, for the second part of the cycle, all that you want to do is a reality check:  Does my recorded reactive energy from the first part of the cycle equal all of the dissipative events in the second part of the cycle?

i.e.:  Reactive energy(first cycle) = (load resistor + shunt + MOSFET body diode dissipative energy in second cycle) + Power supply returned energy second cycle + "other energy."

If you do a good job recording the data with the DSO you should be able to get real numbers with no kooky stuff going on.

For example:  What is the power dissipated in the MOSFET?  It's just the dissipative energy associated with the switching process in the first cycle plus the dissipative energy associated with the body diode in the second cycle.  That's it!  No crazy spike to deal with - you simply ignored it because you understand the energy flow and have decided to do a "work around" - only look at the resulting events associated with the expenditure of the stored reactive energy.  There is no need to factor the reactive energy in your power calculations because you know that you only want to look at the "burn off" of the reactive energy, and not the reactive energy itself.

This may all look like a ton of work, but there is a trick for that also.   In your spreadsheet you want to make a "super tab" with all of the columns of recorded DSO data, derived variables (like the phantom voltages), your relevant averaging functions and summation functions, and columns that give you the "real time" power dissipation, etc, etc.

Then you just copy/paste your original super tab over to a few clone super tabs.  Then you go into your clone supertabs and start deleting ranges of rows of data that you know that you want to "chop out" because you want to selectively ignore the spike or whatever, etc.

Going back to the MOSFET average power calculation again, you could chop up the data on one of the cloned supertabs and trim it down so that you only are left with the total dissipative power for the MOSFET.

It is a bit of work for sure, but it's the real deal.  There is no way you should get negative power for the MOSFET in this scenario.  You don't even look at the high voltage discharge of the MOSFET capacitor, you ignore it because you actually are only interested in the body diode energy dissipation in this case!

I apologize because I know that I am making my points two or three times.  I am tired and will not do a clean up.  Plus I think it helps to drill the concepts in sometimes.

I hope this makes sense!

MileHigh

[Hoppy]

It makes sense to me MH.

Hoppy

[Hoppy]

Could a split parallel connected shunt be used with one half being diode blocked for forward current and the other for negative? This would show an accurate timing representation between the load charging and discharging current at a particular point in the circuit.

Hoppy

[poynt99]

MH,

Yes I like the whole idea and I do agree with it. However, it is based on the premise that there is no possibility for COP>1, and that the spike dos not need to be recorded. This would be a valid test perhaps in a mainstream forum, but you and I both know that this idea is not going to fly with Rose and the crew.

In fact it seems to fly in the face of her very claim. We could do the measurement that way, but it would only validate what mainstream expects, so what is the point really? It's not going to wash with the new-age.

I'm afraid the only thing that will wash, is if we measure and compute the entire cycle, with no manipulation of the data except for cycle trimming. The whole point of dong this exercise, is to prove things out one way or another, to their satisfaction, within reason of course. I know that chopping up the data and eliminating chunks of it will be quite unsatisfactory to Rose et al.

Hoppy, an interesting idea with the two unidirectional shunts. I fear that may add some complexity to the computations and wave forms however. I think we need to simplify if possible.

I've come to the conclusion that measuring this apparatus correctly is not possible with the equipment I presently have. To measure this circuit properly, two differential probes (one for the load resistor and one for the MOSFET) and one current probe is required.

Grounding must be eliminated completely, and even though I have no electrical ground loops, I believe there exists an electrostatic one, which may only be eliminated if we can eliminate ground all together. The differential probes do this nicely.

So, it's time for me to design a four channel differential probe, good to at least 25MHz. The current probe is a more difficult one to tackle however, so it may not happen. But with a differential probe on a shunt, this should work just as well I hope.

.99