Re: [avr-chat] newbie question

[Peter LaDow]

On Wed, Apr 29, 2009 at 9:33 AM, Vincent Trouilliez
<address@hidden> wrote:
> Another, more likely source of problems that could cause the
> hardware (CPU or whatever peripheral on the chip) is interferences from
> the environment. Like electronics used in satellites, which are
> sensitive to the sun's activity. When the sun sends a burst of
> radiation, the satellites must be tilted/turned around, to protect the
> electronics.

And it doesn't have to be quite so harsh as space, either.  I can
think of a project where code redesign was done to work around faults
related to environmental effects.  We had a product that would
experience random reboots in the field.  Turned out it was the
watchdog randomly expiring.  We spent months trying to capture the
cause of these events, certain it was a software fault of some sort.
We finally nailed it down to an alpha hit corrupting the memory in the
processor we were using.  The corrupted memory caused the processor to
occasionally "get lost" and fail to tickle the watchdog, causing the
hardware reset.

In the end, we had to move all the code out of the internal processor
memory and use external memory that was more robust with a lower FIT
rate.  I'm not sure what the AVR FIT rates are for their internal RAM.
 I did a quick perusal of their quality website
(http://atmel.com/quality/quality_reliability.asp), but it's not clear
to me whether the memory numbers are for flash only, or also RAM.  But
if they include RAM, their FIT rates are excellent.

Before this, I thought watchdogs were overkill.  Perhaps in consumer
applications it might be, but certainly in industrial, scientific, or
military applications I think a watchdog is invaluable.  Far better
for a critical device to reboot to a known state than to get lost and
act erratically.

Pete
-- 
--
"To love for the sake of being loved is human;  to love for the sake
of loving is Angelic."  -- Alphonse de Lamartine