[net.micro] diagnostics for new computers, fresh out of box

LIN@Mit-Mc@sri-unix (11/20/82)

From: Herb Lin <LIN@Mit-Mc>
can anyone tell me about what i should do in the way of diagnostics
when i actually open the box that a new computer comes in?  it seems
there should be diagnostics that could give the system a workout
BEFORE marginal components fail when out-of-warranty.  any suggestions
for (a) home-programmed tests, (b) diagnostic packages, (c) hardware
diddling, (d) anything else?  how long should these things be run?

also, what is the "recommended" factory burn-in time for components in
testing?  clearly 200 hours is better than 48, but how much better?

i'll post answers to the list.

thanks.

LIN@Mit-Mc@sri-unix (11/22/82)

From: Herb Lin <LIN@Mit-Mc>
Date: 20 November 1982 08:31-EST
From: Charlie Strom <CSTROM>
To:   LIN
Re:   diagnostics for new computers, fresh out of box

Sudden infant mortality is surely a very common occurrence in the
microcomputer field. Memoery tests are very useful; the best I have
seen is called the Rasmussen Memory test (MEMR2) from Lifeboat and is
most likely available from them for a modest fee. There are also a
plethora of public domain memory tests available in the public domain
and they are generally useful for "hard" errors only; I have seen
several instances of bad memory where the mory tests insist that all
is well. Those that exercise the memory at full speed are the best.
Look at one called WORM and another called UMPIRE (both in the public
domain .) I did just have ocassion to repair a CCS 16K static board on
which a 2114 went bad after about 18 months of almost daily use. This
is the exception.
The best way to burn in the system is to use it in normal operation
intensively for a while. I find that the real life conditions will
show up any faults much more effectively than the so-called test
programs. If your machine runs WordStar for example, chances are that
there are no problems.
I will leave comments re disk drives and other components to others
who are more knowledgeable about them.
						Regards,
						Charlie

LIN@Mit-Mc@sri-unix (11/22/82)

From: Herb Lin <LIN@Mit-Mc>
Date: 20 Nov 1982 at 2046-CST
From: mknox at UTEXAS-11
To:   LIN
Re:   Reply to: diagnostics for new computers, fresh out of box

To answer your question about burn-in;

  48 hours will catch most of the 'infant mortality', the fact that new
parts are much more likely to fail than ones that have been in use for
some time.  I would say that 200 hours would buy you only a 5% to 10%
maximum increase in reliability.  The military only requires 160 hours
of burn-in (at rated temperature) on full 885B screened parts for applications
up to and including deep space probes where service calls are VERY expensive.