(Oops, repost under EE)

List,

I have been running some tests with a n channel power mosfet used as a
linear current sink. It is controlled by an op amp with circuitry and a
sense resistor in series with the source to ground (basically.)

So, I intend to use it to discharge batteries ranging from a single cell
NimH (~1.2V) up to a 6S LiPo pack (~25.2V).

It is appropriately heatsinked with fan.

I have had no issues getting essentially 50W power dissipation as follows:

2.5A @ ~20.4V    (6S LiFe pack)

6.4A @ ~7.76V    (2S LiPo pack)

Now I run this test; discharge a 1S, 5000mAH LiPo cell (~4.2V unloaded)
and slowly increase current.

Everything is OK up to around 9A, then the MOSFET appears to turn all
the way on. Before protecting it with a fuse, the MOSFET would blow or a
PCB trace self destruct.

I put a 10A mini ATO fuse in the circuit and it repeatedly blows every
time I increase amperage from 8.9 amps to 9 amps.

replace the fuse and it will run again til you reach ~9A, then fail.

Go directly from 0A to a target amperage and it will fail at a bit lower
amperage; say 8.8 or so. (I think this supports the idea that it is heat
related; the die gets hot faster than the heatsink can remove the heat.)

I 'scoped the gate voltage; it is stable w/no visible spikes or noise.

Current feedback is subjectively (to my eye) nice and clean, AFAIKT with
my 100MHz o'scope.

No oscillation I can see in the control loop.

Power supplies seem fine.

At the time of failure a current meter in series with the LiPo cell
shows no deviation other than plummeting to 0, so I am assuming the
current spike is too fast for the meter to show.

Now. I suspect it is either

     * something causing the MOSFET to go into to "latch up", or
     * something to do with SOA.

I suspect the SOA problem; I'm not sure how latchup could occur with a
non-switching circuit, and everything relatively steady state.

Apparently, when running a MOSFET in the linear mode (not switching, as
most apps would) due to imperfections in the doping, die manufacturing,
internal lead attachment, whatever,  the die can develop a "hot spot"
which quickly hogs all the current, and causes thermal instability and
failure.

See

http://powerelectronics.com/704PET23.pdf

and

http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20100014777_2010016074=
..pdf

for more info.

Apparently this has become more of an issue as the mfg process is
optimized for switching uses and certain characteristics have changed
that make them less useful in linear apps.

IXYS has addressed this issue to some degree by modifying their die mfg
process and spec'ing and testing SOA at 95 deg C case temp and DC
continuous, for example:

http://ixdev.ixys.com/DataSheet/DS100200%28IXTH-TT75N10L2%29.pdf

The MOSFET I am using is a Fairchild HUF75345 in TO220 pkg. They do not
give a SOA curve at other than 25 deg C and they do not include a DC rating=
..

http://www.fairchildsemi.com/pf/HU/HUF75345P3.html

I have run these mosfets as part of discharge circuitry in the past, at
similar power ratings, but with slightly better heatsinking, as part of
higher amperage, multi-device circuits.

Ways I have thought of trying to test this w/o getting any more involved
than necessary include:

     * Buying some IXYS devices and seeing if they work in spec (does not
       prove the failure mechanism)
     * devising a way to cool the MOSFET and see if the magnitude of the
       power dissipation before failure increases. (same problem)

If this is the failure mode, then in order to prevent it from happening
I will have to limit max wattage based on temp and voltage, and it will
be subject to the device used where the failure point will be, right?

Also, I presume the device could be weakened by repeated use near the
failure conditions, so it will have to be identified and characterized
and then stayed away from.

     * Any idea how I can verify/rule out this failure mode?
     * It's something else that I haven't thought of?
     * ??


OK, let me have it, guys.

-- Mark Skeels Engineer Competition Electronics, Inc. TEL: 815-874-8001=20
FAX: 815-874-8181 www.competitionelectronics.com
-- http://www.piclist.com PIC/SX FAQ & list archive View/change your=20
membership options at http://mailman.mit.edu/mailman/listinfo/piclist

m
--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.