Am 18.12.2016 um 13:13 schrieb dick balaska:
> Am 2016-11-17 17:23, also sprach clipka:
> 
>> I once spent hours beside a climate chamber to hunt for a bug that the
>> hardware guys couldn't figure out.
>>
> 
> I got heat because a custom Altera part I was driving failed with my
> software. The part worked just fine with the old software (which was
> magnitudes too slow for which it was thrown out).  Clearly my problem.
> After a month of begging, I finally got an afternoon with the hardware
> guy, who took 10 minutes to figure out that the output of an 8 input
> nand gate was ringing into an adjacent line causing false triggers
> elsewhere.  Didn't show up when the part was running at 10% capacity.

Ah, that sounds like fun!

> And then, and *then*, I got big heat because the i82557 SIO chip would
> start sending 0's. No errors, hardware checksum was ok. This was my
> problem too.  Years later, I found out that shiny new Linux 1.0.3 won't
> support network cards with this chip (3C507) because of this known problem.
> Thanks for not mentioning that when you sold us on the buggy chip, intel.

That reminds me of the fun the hardware guys had when it turned out that
the designated MCU for the product didn't like to be started up too
shortly after losing power.

And I mean, it *really* didn't like that. Not in an "I'll throw a
tantrum" way, nor the "I'll go sulking in the corner for a while" kind.

No, I mean the "over my dead body" sort of dislike.

Literally.

Not good in a car, where engine start can occasionally cause power to
cut out on less-important pieces of equipment like ours.

Since a revised MCU was nowhere to be seen, and turning to a different
product was a no-go that late in the development cycle, hardware guys
had to add an extra circuit to the board that would absolutely
positively prevent the device from turning back on for a while after
powerdown.

Which of course would haunt us later down the line, when the car
manufacturer (who had approved the stopgap measure) complained that the
device didn't power up fast enough in some new test scenarios.

I miss that kind of bullcrap. Really.

Post a reply to this message