Olin Lathrop wrote...

>Dave, I didn't say (and certainly didn't mean) that the soft start =
design
>was bad, only that I personally didn't like it.

I didn't take any offense, Olin; I'm just irked that this is one of
those instances where I've "winged" an answer to a question and possibly
missed a significant pitfall.  My gut sense is usually pretty good, but
it isn't infallible.

>...Others could legitimately argue I'm being silly...

You?  Silly?  Never-- even if you did once plug a carbon microphone into
a 120V outlet to show people what REAL sparks were like...

>...If the FETs were 1/2 on with cold bulbs (it won't happen like this, =
I'm just
>trying to pick something to see where it leads), that would be 35A x 60V=
 =3D
>2100 watts.  If this happened for 20mS, that would be 42W averaged over =
1
>second.

Ummm... they're 12 volt bulbs, not 120 volt bulbs.  The current's right,
but the voltage and the power dissipation are off by an order of
magnitude.

I've done some simulation of this MOSFET switch with soft-start, based
on a 50 millisecond turn-on period (from beginning of current flow to
the point where current is limited by Rds(on)), and modeling the lamp as
a conductance which decays exponentially from the "cold" value to
one-seventh that amount with a time constant of 50 milliseconds,
starting at the instant where current flow begins. =20

This model is biased very strongly toward the pessimistic side, since
the Motorola data (see my other post, in response to Dale, for details)
suggest filaments this size heat up much, much more quickly--at least
ten times as fast as the model I constructed.

I don't have a SPICE model for a 2SK2614, so I used a Fairchild BUZ11 in
my simulation.  It has very similar characteristics and it also has a
much more complete datasheet, including a chart showing Forward Bias
Safe Operating Area for different pulse widths.  According to this
chart, the BUZ11 can dissipate 120W in a 50 millisecond pulse over a
wide range of voltage/current combinations.

The simulations I did showed the MOSFET dissipating a little over 18
watts average over the three line cycles during which current limiting
was taking place; and this is with a current inrush whose duration is
something like an order of magnitude longer than might reasonably be
expected. =20

With no inrush (that is, the lamp starts out at its "hot" resistance
value), the MOSFET dissipation during the slow-start interval drops to 7
watts for 50 milliseconds.

Even if the 2SK2614 SOA curves aren't exactly like those of the BUZ11,
it appears the MOSFETs in this soft-start scheme will be experiencing
turn-on stresses roughly an order of magnitude below what they're
capable of withstanding-- and even lower if the soft-start interval is
shortened from 50 milliseconds down to 10 or 20.

A two-to-one safety margin would be cutting it close; but when the
margin goes up to ten-to-one, I start getting fairly confident of having
room enough for error.

Dave

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.