I hope this reply answers the PIC's wide temperature operation
questions.

Generally all ICs are designed for full temperature operation of -55¡C
to 125¡C except for special applications (like down-hole oil well
operation). How well the design makes it over the temperature range
depends on the skills of the design and process fab engineers and the
quality of fab run. Wafer level tests mark the die for various grades
testing at 25¡C, but the goal is still to get the highest yield for the
tightest specs over temperature. Some IC companies don't temperature
grade at the wafer level at all. Some companies do wafer testing at the
temperature extremes, mostly for military or automotive engine
compartment ICs.

Final test of packages parts is also generally only done at room
temperature because of the extra time and cost of temperature testing.
So packaged parts are graded and labeled for various temperature
operating ranges biased on key parameters measured at only room
temperature.

So what does this have to do with operating a PIC at temperature
extremes?
Well it is most likely that Microchip does only room temperature
testing. Therefore there is a certain probably that any given lower
temperature labeled part will, in fact, operate over a wider temperature
range.

What will happen if you operate the PIC over a higher temperature range?

The lower temperature is not critical because there are no common
failure elements that cause permanent damage. A PIC will just stop
normal operation, if a transistor runs out of drive or a resistor bias
point changes because of a parameter change with temperature.

The higher temperature is critical. There are two major failure
mechanisms that will cause permanent IC failure: current leakage and
metal migration.

Current leakage can just stop a cells operation or cause SCR parasitic
latchup with total cell destruction. Destruction can be prevented by
monitoring and limiting supply current while raising the IC's
temperature. Do not use parts that show an abnormal increase in supply
current.

Metal (aluminum) and polysilicon trace migration can not be avoided.
This is a basic wearout mechanism for ICs. The higher the chip
temperature the sooner it may fail. The silicon IC process has been
designed around the intrinsic maximum junction temperature of 150¡C for
1000 hours of operation (minimum). There is a temperature function for
the IC process used by Microchip that will predict the failure rate but
I don't know if they publish it. Semiconductor-industry field history
gives an average of twice the failures at 160¡C compared to failures at
125¡C. However the PIC's lifetime should be tens of years or more if
>operated at no more than 125¡C. If a PIC (or other silicon IC) is
operated at 200¡C, the lifetime will be measured in tens of hours!

So the bottom line is "What is the PIC's temperature?" - not what
frequency is it operating at, even though there is some self-heating
>from higher clocking frequencies.     

- Wayne Foletta
BMI - Saratoga, CA

> ----------
> From:         wsiemens@ryan-tech.com[SMTP:wsiemens@ryan-tech.com] on
> behalf of Wendall Siemens[SMTP:wsiemens@RYAN-TECH.COM]
> Reply To:     pic microcontroller discussion list
> Sent:         Wednesday, January 14, 1998 8:19 AM
> To:   'pic microcontroller discussion list'
> Subject:      RE: Temperature specs
> 
> I am currently using windowed packages in high temperature and low
> temperature applications. I have taken a PIC16c73A up to 185 C and had
> it run fine at that temperature (only 455khz though). We spec all our
> tools using this device at 160 C.
> 
> My present project using a PIC17c44 has run from  -55C to 160C at 5MHz
> using the windowed ceramic package also.
> 
> I don't know how they will run at 20MHz, but for seven more degrees I
> don't think you will have a problem.
> 
>       -----Original Message-----
>       From:   Kolesnikoff, Paul [SMTP:pkolesni@BALL.COM]
>       Sent:   Wednesday, January 14, 1998 8:38 AM
>       To:     PICLIST@MITVMA.MIT.EDU
>       Subject:        Temperature specs
> 
>       I have to deliver a unit with several 16C63 PICs and 24C32
> EEPROMs.
>       After I built a prototype, a temp spec was added of +77deg C.
> The
>       commercial temp range goes to +70 deg C.
> 
>       My question is - are the industrial and automotive grade PICs a
>       different design, or are they a select in test kind of part?
> 
>       If they are the former, where do I get the high temp parts? If
> the
>       latter, can I just heat up the PICs I have to 80 deg C and see
> which
>       ones break?
> 
>       Also, does the clock speed affect this? I'm running at 20MHz.
> 
>       I guess I should also mention that these are windowed parts.
> 
>       Thanks,
>       Paul Kolesnikoff
> 
>