11 Jun
2006
11 Jun
'06
9:05 p.m.
You're dealing with cheap (inexpensive) consumer
products.
Well, actually, I was on the expensive Total Control side of USR, with
$35,000 rack modem systems and $155,000 integrated AOL cabinets.
How did they handle in warranty repairs -- toss the
part
in a box and donate the box to a local tech school?
Scrap, mostly. It was up to the techs. I think for a while they watched
the clock, and if it looked like they were going to overspend time on a
debug, they would fling the board into the bin.
But you are looking at it from a consumer products
viewpoint.
High volume, drive the cost and quality as low as the market
will tolerate. And, if something fails, just replace it
(since the labor cost of *repairing* -- not just "retesting" -- a
"defective" product exceeds the DM+DL for that product).
See above.
Imagine the device you are selling is used in a
manufacturing
environment. I.e. someone relies on your device to make
*their* products. Being "down" for an hour, day, *week* can
cost more than your product -- even if your product was a few
kilobucks.
Yes, when an AOL node went down, big or small, costs were staggering. We
did not measure using hours, days, or weeks, but seconds and lost
packets. Sidenote, when the pick-n-place robots went down at USR, it was
about $1500 per hour per line.
OTOH, if your design supports repair in the field,
then you
have more options available.
In todays environment, especially networking, often the cost one would
spend replacing the component (including the entire years salary for the
tech) would get dwarfed by the lost revenue due to the downtime. I am not
exagerrating. I pretty much can not think of anyone of importance that
does onsite board level repair in a working environment. Those days are
long gone. Even the military pretty much went to depot repairs in the
1970s.
They are just as
guilty of monovision as the design engineer who never walks the floor.
Definitely - and the trick is to get everyone working together.
Caps are as likely to fail as a socket -- do you keep
caps
out of a design? *Or*, do you design them in WITH PLENTY OF
MARGIN?
Remember, we are not talking about reliability (yes, sockets are a problem
there as well), but yeild. My point is that socket use takes a big toll on
yeild and manufacturing issues.
Anyway, yes, one tries to minimize the cap count. There is no black and
white, of course.
Muntz was not a dumb man.
William Donzelli
aw288 at osfn.org