Power Mosfet forensics

[KlausST]

Hi,

...but he wants a textual description....

I think I'm not the only one here that prefers a schematic.
Text is very often unprecise and leaves a lot of room for interpretation.

Didn't you already recognize this in this thread?

But maybe this is the cause why you refuse to post a schenatic and a PCB layout.
For sure you are free to describe your schematic and PCB layout with words.

Klaus

 

[kathmandu]

@Klaus:

No offence, I just don't have time for your fancy replies.. one day, I'll google for "h-bridge" and I'll post a schematics for your convenience. Anyway, it has nothing to do with my question though..


@schmitt trigger:

Thanks for posting that pictures. Anyway, as far as I read, that parallel arm it's just a component of the primary side (named "the excitation branch", where "I0" is the no-load current - usually negligible). The other primary branch is "X1" from the diagram you posted. That's it, if V2 is zero (short-circuit), you've got X1 + X`2 (and R1 + R2) in series with V1 (transformer input voltage).

To anyone who doesn't bother using some words, could I get a further explanation, please?

 

[schmitt trigger]

The total series impedance, the vector sum of R1, R2, X1, X2 is much lower than the shunt impedance.
If you short the secondary, that series impedance will be in parallel with the shunt impedance.

The resulting parallel impedance is always lower than the lowest of its components.

In other words........... a very, very low short circuit impedance.

 

[kathmandu]

By "shunt impedance" you mean that parallel arm? If that's a very low impedance, how come the current flowing through it is usually neglected (the no-load current)?

Anyway, there's the load current flowing through X1+X'2+.. prior to output short-circuit. That current can't change abruptly, right?

So here's my original question: how to calculate the interval between t0 (the short-circuit moment) and an arbitrary t1, when the current flowing through X1 it's three times bigger than the rated current? I need that interval to calculate the propagation delay of a short-circuit protection circuit.

 

[KlausST]

Hi,

Honestely, I want to help you. I want to save your time and I want that the given response safely removes the problem.

I'll google for "h-bridge" and I'll post a schematics for your convenience.

There is no need to post any schematic / layout. I know how a h-bridge works.
It only makes sense if you post your schematic / layout. But for sure you are free to do it your own way.

My personal opinion to the Mosfet fails:
* The root of the fails is somewhere in your design.
* The fail after a long time of operation to me sounds like a long time degradation of Mosfet inside isolation barriers. The isolation becomes weaker and weaker...then the leakage current rises continously ... then you get continously increasing power dissipation.
Either the increased temperature kills the Mosfet or the weakened isolation almost immediately gives up.
I want to exclude sources of failure to avoid that you do a redesign yourcircuit, but the true error still exists...and in onother 1 or 2 years you get the next fail....

But I assume you are not interested in further postings from me, therefore I will keep in back..

Klaus

 

[FvM]

You finally confirmed that you need to determine X1+X2 to know the current rise time.

 

[kathmandu]

@Klaus:

I have always appreciated your advices in the past but with your last replies you were just making fun of one of my assumptions or the other.

Like I have previously stated, the Mosfet died because of a short-circuit at the inverter output (230V). I only had a (slow) fuse as a bare protection so no real protection at all. There's no problem with the PCB, as I changed its layout few times before I got an optimized version (and I checked all the signals with a scope during a heavy load: motor starting, thyristor phase angle control and so on).


@FvM:

X1 represents the primary reactance and X2 (actually, is X`2 = X2/N^2) is the secondary one, right? Unfortunately, I have no means of measuring the transformer windings inductances. Moreover, this is my main (off-grid) inverter thus I can't afford to take it off line for further (empirical) measurements.

Do you know, from your experience, the usual values of those reactances for a 4-5KVA low frequency (toroidal) transformer? I'll be grateful at the most if you could even specify a real-life value for that rise time though..

 

[FvM]

Series reactance of mains transformers can be expected in a range of a few percent of the rated load impedance, e.g. 2 to 10 percent. With toroidal transformers, it's usually at the lower end.

Generally, I agree with your consideration that a fast electronic overcurrent detection circuit can fully protect the H bridge against output short circuit and overload.

2% Xs for a 50 Hz transformer refers to about 65 µs rise time to rated current, or 200 µs to the threefold.

 

[kathmandu]

2% Xs for a 50 Hz transformer refers to about 65 µs rise time to rated current, or 200 µs to the threefold.

Thanks a million, that's the best piece of information I could ever get!

I already have a current transducer in my circuit (for energy measurements only) but the MCU polling time is about 10us (the transducer itself has a response time of 5us).

I always thought that such a delay (10us) is inappropriate for a short-circuit detection. Now that I have few tens of microseconds available, I'll modify the software routine to check for overcurrent conditions and to inhibit the PWM signals accordingly.

Once again, thank you very much for your kind support.