Quote from: Groundloop on May 15, 2012, 12:00:06 AM
TK,

I think the reason that you did see the bias current increasing is that the RdsOn actually went UP in value
as the transistor heats up. That will fit the theoretical analyze. I can set up my test circuit and try it again, but
this time allow the circuit to run for a much longer time and see what I get for the bias current.

The RdsOn going UP in value is often the reason those critters blow. You generate a lot of heat in the transistor
and as it heats up the RdsOn climbs and you are generating more internal heat, etc. So at one point you get
too much heat and you heat sink can no longer cope with it and the transistor blows.

Back in 1/2 an hour or so.................................

GL.

@GL: How can the Rds go up at the same time the main current, Ids, goes up too, if the main supply voltage is constant? ETA: I see it now, it has to do with the rates of increase.

And...the power dissipated in the mosfet goes up linearly with its resistance, but as the square of its current. P=I^2R, right? So a small rise in current means much more power dissipation increase than a small rise in resistance does.

ETA: the graph shows that GL is right, though... the Rds does climb with the temperature. But I think the current goes up faster, so that the end result is the same: blown mosfet.

Quote from: TinselKoala on May 15, 2012, 12:12:34 AM
@GL: How can the Rds go up at the same time the main current, Ids, goes up too, if the main supply voltage is constant? ETA: I see it now, it has to do with the rates of increase.

And...the power dissipated in the mosfet goes up linearly with its resistance, but as the square of its current. P=I^2R, right? So a small rise in current means much more power dissipation increase than a small rise in resistance does.

ETA: the graph shows that GL is right, though... the Rds does climb with the temperature. But I think the current goes up faster, so that the end result is the same: blown mosfet.

TK,

Here is my bias current heating test over time.

Start:

My RLOAD = 10 Ohm. RdsOn = 1,6 Ohm. RSHUNT = 0,25 Ohm.

Main input 24 Volt @ 1,74 Ampere. BIAS was 0,09 Ampere @ 12,10 Volt.

15 Minutes later and running all the time:

Main input 24 Volt @ 1,92 Ampere. BIAS was 0,13 Ampere @ 12,05 Volt.

MOSFET is warm to the touch but not too warm to operate normally.
There is lot of heat coming from the RLOAD.

15 Minutes later on running all the time:

Main input 24 Volt @ 1,95 Ampere. BIAS was 0,13 Ampere @ 11,90 Volt.

MOSFET is much warmer now, almost so that it burns my hand.
So I will probably need a larger heat sink or put a fan on the existing heat sink. :-)

There is lot of heat coming from the RLOAD.
Seems that the circuit now is running stable and does not increasingly heating up more.

END TEST.

So yes, I agree that the bias current and the main current goes up as the MOSFET heats up.

GL.

Here is how she talks to someone who approached her with mildness and plain speech, and who is clearly much more expert in these topics than just about anybody I can think of. He never said a single harsh word to her, and this is how she responds.
And Ainslie thinks I am he ! That really cracks me up.  I bow in awe at cHeeseburger's knowledge and erudition. And his equanimity in the face of gratuitous insult and heavy-handed censorship. Clearly, I do not tolerate fools as readily as he does, neither do I have the knowledge that he does. Neither, apparently, do I have enough sense to just throw up my hands and walk away like he finally did. After all.... this will still be going on in another year, with no progress, whether I am here or not.

(Or is he out there, lurking and shaking his head, between facepalms.... or is he even closer than that?   Only Stefan knows, and I hope he doesn't tell.)

I note from that post that she doesn't know how to use the scope's zoom feature to show a time-amplified section of a waveform taken at a slower timebase.

Heck.... I can even do THAT on my old analog HP180a. Remember when I showed, simultaneously, the overall envelopes at 0.2 ms/div, and I expanded the timebase out from a portion of that and showed the oscillations at 1 microsecond/div, a zoom factor of 200? SIMULTANEOUSLY?  Is it POSSIBLE that my old analog garage-sale scope can do something that a modern, if low-end, DSO cannot?

Well, in all fairness it was a LLNL "garage sale". (That's Lawrence Livermore National Laboratory, for the acronym-challenged.)

SIGH.  Ainslie has a beautiful, if low-end, DSO that is used like an Etch-a-Sketch: to make art, rather than science.

@GL: thanks for running that test.

What do you think would happen if you used 72 volts instead of 24 volts, leaving everything else the same?