"François Yves Le Gal" wrote in message
...
On Mon, 27 Mar 2006 19:06:26 -0500, "Robert Morein"
wrote:

Designers who used the original Hitachi devices
boasted that the outputs could be freely shorted.

I doubt that engineers boasted this. Sounds more like yet another
marketing
gimmick.

You should understand that a Mosfet isn't a single transistor: it's die is
made of multiple elementary cells connected in parallel (hence the highish
input capacitance among other characteristics).

Yes, I know this.

If you go outside the SOA,
thermal regulation won't be instantaneous - temperature has to rise on the
whole die, which displays a non negligible thermal inertia - and some or
all
cells can thus be destroyed.

I understand this; your references are very informative, but I do not yet
reach the conclusion that all MOSFETs are practically susceptible to this
effect with respect to audio applications. Your references do make it clear
that they should not be deliberately shorted.

Perhaps the negative region is less significant
to device destruction than with newer devices.

It's nearly the same.
Your reference contradicts that. Quoting, "Use of a very?low?on?resistance
device offers low losses for steady?state operation but may cause the device
to fail during a short circuit or an overload."
This remark suggests that the older devices are, indeed, more resistant.

When the operating point is set at low currents, negative tempco is very
real. See for instance fig. 4 in
http://www.eetasia.com/ARTICLES/2005...N23_EMD_AN.PDF, which
deals
with Mosfets used in hot-swap controllers, but the underlying theory also
applies to an audio context.

It's a good reference, but it was obviously authored with reference to
current devices, not antiques. Also, Sander was trying to bias a MOSFET,
which, presumably had a standard audio load, or open circuit output. The
paper specifically states that the sign of the temperature coefficient is
dependent on the gate to source voltage. I do not see an obvious answer to
the question of whether it is possible to bias an older Hitachi device into
thermal runaway, as related to Sander's activity.

BTW, thanks for the references. They are useful, and I will take note of
them.