But, even with a large fan-cooled aluminum heat sink with a 1/2" thick
base and mounting the MOSFETs via a good phase-change thermal
interface compound (no pad)

and Aavid MAX clips (18lbs. pressure)

, the
MOSFETs blow. They'll run fine for minutes to hours but eventually
blow in 10-30 seconds after starting up with the maximum load, even
with a cold heat sink.

I'm not really sure how to use the linear derating factor for MOSFETs
but I'm wondering if this is the problem, i.e., that I'm exceeding the
rated power level for these FETs at the junction temperature I'm
running them at. The case and heat sink temperatures seem OK at this
high power level according to the rated theta-jc and theta-cs numbers
from the MOSFET datasheet.

After checking the datasheet, I found that these Absolute Maximum
Rating numbers:
Junction Temperature = 175 degrees-C.
Power Dissipation (at Tc = 25C.) = 470W
Linear Derating Factor = 3.1W/degree-C.

In parallel? If so, then what method are you using to force current
sharing?

If you're discharging batteries, why not use drain resistors as the
real loads and PWM the fets? That would save a ton of money all around
and be pretty much indestructable.

Good idea, I think I'll do the math again. I think I remember that
the current levels were too high to make that practical (over 100A at
0.9V min.) but it's worth revisiting.

-- remove SPAMMENOT for e-mail responses --

Are you sure nothing's oscillating?

No. For each degree that the case rises above 25C, you can dissipate
3.1W less than 470W. That's what the derating means.

No. For each degree that the case rises above 25C, you can dissipate
3.1W less than 470W. That's what the derating means.


It's not the temperature rise of the junction that's used?
I was calculating the temp rise in the junction (based on measured
case temp) instead. I have the case temperature from earlier on,
before the FET blew. Here goes...

Max power rating for FET = 470W
Linear derating factor = 3.1W/degree-C
Ambient temperature = 25C
Measured case temperature = 100C (actually, it's 97C but I rounded up)
Case temp rise = 100C - 25C = 75C
Derating = (3.1W/degree-C) x (75C) = 232.5W
Power available to FET as a load = 470W - 232.5W = 237.5W

This seems to indicate that the 125W load the FET was dissipating at a
case temperature of 100C wasn't a problem (by itself). But, the
junction temperature was a lot higher (140C, calculated). Shouldn't
the temperature rise of the junction be the basis for the derating?

Thanks!

-- remove SPAMMENOT for e-mail responses --

So, for instance, if you have a 175C maximum junction temperature, a
3.1W/K thermal resistance, and a case temperature of 174C, you'd better
hold the dissipation to 3.1W!

What? You're confusing yourself (and me). All the derating stuff is
there for people who don't understand thermal resistance (I'm opening
myself up for comments here, I know it).

So you've calculated a junction temperature of 140C? I assume your
devices are rated for 125C, or something less than 140C, then.

With the linear derating factor calculation results I got, I'm
concerned that I might not have the proper thermal resistance data
though. Calculating the junction temperature from the measured case
temperature gives me this:

T_junction = (T_case) + (P_dissipated) x (Theta_junction-case)
T_junction = (100C) + (125) x (0.32)
T_junction = 140C

Are you sure nothing's oscillating?

I can't find anything more than 5mV of noise or voltage variation
anywhere with the scope. I tried grounding the probe to both the main
supply ground and at op-amp gnd (and measuring around). Even when
that noise on the input to the op-amp, the cap on its output seems to
slow everything down enough to not let the noise affect the MOSET's
gate voltage.

My (possibly) exceeding the total dissipation rating of the FET isn't
a possible cause for the popping of FETs too? Still don't know if my
calculations and assumptions about linear derating factors are right.

John

Viewing the schematic, I can see no fault in the sharing method,
however you should be wary driving these parts from a 5V source, as
some may not be capable of the full intended load at this gate
voltage, particularly if the source resistors have a signifigant
voltage drop.

This would mean that those that can, will pass more current than their
compatriots. Easy to check with a voltmeter.

I wouldn't trust clips as far as I could throw them, in a high
current/power/temperature application.

You can get oscillations at pretty high freqs (>50 MHz) that will
cause problems. A ferrite bead or small valued resistor in series with
the gate lead can help some oscillation problems. Make sure you use a
scope that can see above 100 MHz.


Hmm...I have a 1K resistor from the op-amp output to the gate and
0.01uF cap from op-amp output to GND. I *think* that slows down the
amp enough to prevent oscillations at that high a freq., but it's easy
to add a bead too. Thanks!

John

the MOSFET can oscillate, all by its self. There is a nice Siliconix app
note (in the mospower apps manual) that does a Routh-Hurwitz stability
analysis on a MOSFET, showing why it oscillates, and how to stop it.

I use decent bolts (not cheese) and apply rated torque.

Just use gruntier spring clips

unless you have enough compliance in your spring, those reflow-type
thermal interfaces require you to re-torque the fastener. I've done that
with 600V 1200A IGBTs - heat to 80C for a while, let cool then re-torque.

whats the surface finish like on your heatsink? have you checked the
failed untis, to ensure an even spread of goop (a lack thereof indicates
a bent heatsink)