Claimed OU circuit of Rosemary Ainslie

[Harvey]

Hi Poynt,

I've been looking at the spice subcircuit for the IRFPG50 to see what the internal inductance is relative to the specs:

.SUBCKT IRFPG50  10 20 40 40
*     TERMINALS:  D  G  S
M1   1  2  3  3  DMOS L=1U W=1U
RD  10  1  92.4M
RS  40  3  5.92M
RG  20  2  211
CGS  2  3  176P
EGD 12  0  2  1  1
VFB 14  0  0
FFB  2  1  VFB  1
CGD 13 14  223P
R1  13  0  1
D1  12 13  DLIM
DDG 15 14  DCGD
R2  12 15  1
D2  15  0  DLIM
DSD  3 10  DSUB
LS  30 40  7.5N
.MODEL DMOS NMOS (LEVEL=1 LAMBDA=2M VTO=3.1 KP=1.37)
.MODEL DCGD D (CJO=223P VJ=0.6 M=0.68)
.MODEL DSUB D (IS=25.3N N=1.5 RS=0.139 BV=1K CJO=341P VJ=0.8 M=0.42 TT=243N)
.MODEL DLIM D (IS=100U)
.ENDS IRFPG50

Quite frankly, I need to rely on your expertise here to tell us which of those acronyms is related to the inductance of the device. The specifications call out nanohenries - specifically 5.0 on the Drain and 13 on the source. I see an LS up there, with 7.5N which may be for the Source - but I am uncertain.

You really are our resident expert on all things Spice and we could use your input here.

I have looked at your possible explanation for the pulse shift and find it intriguing. You surely do show a change in the voltage at the shunt by adding the inductance in there. Is that before, either side, or after the shunt? Is it included as part of the shunt? - IOW is the voltage across the entire value of impedance, or is it just the effective result on the resistive section?

I am asking, because I see some large differences between the simulation and the reality. The gate spike is not present, or is early in the sim by comparison. Also, the small single ring on the shunt falling slope seems drastically modified by added inductance. And symmetry seen in the reality appears greatly skewed in the sim. I am trying to re-evaluate my own perspective and see it from your point of view here, but I am having a bit of trouble aligning the sim with the reality and could use your input to get a grasp of it.

Also, I would like to know where the probe was placed in the reality shots for your system. I need to check with Glen, but I imagined that he had put it right on the shunt lead as close to the body as possible. Wouldn't this preclude the inductive forbearance present in the leads between the source pin and resistor?

I like that your at least posting the possible solution to our question, and really hope if we have any readers with qualified degrees that they can weigh in on this as well.

Cheers!

[poynt99]

I checked both models for the IRFPG50 that come with PSpice (one in the IRF and one in the pwrMOS libraries), and neither include any inductance at all, even though there are differences between the two models in general.

I see the Protel model is different yet again, and does at least include a value for the Source (LS) inductance.

However, the purpose of including a small inductance (in series with the shunt resistor), was to account for that which is added due to the presence of the shunt resistor only. This inductance is magnitudes higher in value than the parasitic Source inductance, and running an analysis with a LS of 13nH resulted in little change in the wave form with a LShunt of 1uH present.

The shunt probe was placed on the MOSFET Source pin, just as it was in reality.

Added comments: Analyzing down to this detail to see what is happening in and around the major events, it is not surprising that the sim results are not exactly as the real ones. A much higher degree of modeling (MOSFET and wiring) would be necessary I think to encompass all the subtle effects seen in the real scope shots. The shunt inductance was added (from a real measured value) and shown to be a possible cause of the significant shift observed in the shunt wave form, which is fairly evident in the simulation results, even with as little as 0.3uH for LShunt. Granted the simulation wave form does not look exact, but it does possibly demonstrate how and why the negative shunt voltage pulse coincides perfectly with the Drain spike. For an instant, that 0.25 Ohm shunt becomes about a 5.5 Ohm shunt.

.99

[Rosemary Ainslie]

Hi Harvey

I get it that you and Poynt are exploring whether this effect is due to phase shift - and that the waveform captured by the TDS3054C is somehow misrepresenting the true value of the current flow from the battery?

I would buy into this but it does appear that the battery depletes at the rate of current measured at this shunt.  Curious?

Also.  I am not sure that Poynt is quite as necessary to the cause as you seem to require.  No offense Poynt.  It's just that I need some open admission that the DC average measured across the shunt at source - has some relevance to the exercise.  And more to the Poynt - that it is within the measuring capabalities of the DSO.  Personally I prefer to work with the more open minded. 

[WilbyInebriated]

Added comments: Analyzing down to this detail to see what is happening in and around the major events, it is not surprising that the sim results are not exactly as the real ones. A much higher degree of modeling (MOSFET and wiring) would be necessary I think to encompass all the subtle effects seen in the real scope shots.

i told you this a long time ago over at energeticforum... it took a page of you dancing around my increasingly specific questions to finally admit it. then you put me on your ignore list

we would have to get our resident 'brownie point' expert milehigh here to weigh in on what the amounts are, but i think that's 15 brownie points for me and 50 demerits for you.


edit: speaking of milehigh,

I don't think so because you already set up your "Escape!  Escape!" modus operandum in your last posting.

it's modus operandi not modus operandum. ie: 'being stupid and pompous is milehigh's modus operandi'. operandum... lol dum indeed.

[poynt99]

Here is the overall picture.

First, the VDrain and Vshunt wave forms with Rshunt = 0.25 Ohm + 1uH

Second, a plot of VDrain again, but this time against the instantaneous shunt impedance.

In the center of the Drain spike, we see that the shunt is actually at a value of 5.4 Ohms, and spikes quite high on the falling edge of the Drain spike.

Clearly from this, we can see that using a constant value of 0.25 Ohms throughout to calculate the shunt current in the spreadsheet (as I and others have been doing) will lead to a significant error in the power calculations, in particular during the transients.

The solution is to use only a pure resistive shunt, or a current probe, otherwise the power calculation results will be quite skewed as we've seen. This is most likely only one of the problems being encountered in making these power calculations, as I still recommend differential probes for the voltage measurements.

.99