Matt Bonner wrote:

> > I would do tests on time related items, such as USARTS (as you have said),
> > I2C, pwm etc. Also, running the outputs at their rated currents on several
> > pins to led's.
> >
> Not possible on the last point: Ohm's Law has to come into play at some
> point.  At 25mA per pin that's 1/8 W - pretty hard to dissipate at 150
> deg C.  Anyway, the only way you can get high temp out of a part is to
> run it at pretty close to "storage temperature" currents (i.e.: zero).
> High loads should be buffered with a VFET (VN10KE) or whatever.  BTW, at
> what temperatures do LEDs melt? <g>

Why not possible ? It's more easy (and cheaper too) to glue PIC to some
 radiator and not apply any additional switches like VFET or so.
You should use well temperature conducting glue and according radiator
 and you will dissipate your 1/8 W at 150*C without any troubles.
Another way - to make a tight contact PIC with radiator through
 termo-conducting paste .

BTW there are another way to eliminate overheating by using surface
 layers of copper (usually GND or Ucc) in multilayer PCB as radiator
 but 150*C is extremly high temperature for this method.

WBR Dmitry .