----- Original Message -----
From: Roger Morella <rjm@BUFFNET.NET>
To: <PICLIST@MITVMA.MIT.EDU>
Sent: Wednesday, October 20, 1999 10:46 PM
Subject: [OT] Survey... What is an acceptable failure rate?


> The following is a survey that I hope will benefit all of us regarding
> the prevailing consensus concerning failure rates and warrenty periods
> for industrial and consumer products.  The more people that respond, the
> more we will all benefit from the results.
>
> Preface: We have a situation where our customer is asking that we keep
> our failure rates at what seems to us to be an unacceptable limit.
> While we strive to keep our failure rates as low as possible, there are
> practical limits to how low you can go.  This is why we typically give a
> 1 year warrenty on most of our industrial products.
>
> Questions:
> In your opinion, what is an acceptable failure rate for electronic
> control assemblies in an industrial product?

Statistically, the MTBF *must* increase with system complexity.  Say for
example when manufacturing, a solder joint fails 1 in 100,000 times.
Is this acceptable? Most, including me, would be happy with this figure.
If your product contains 1,000 solder joints, 1 in every 100 units will be a
failure straight off the assembly line. Thats just the soldering now
consider
resistors,capacitors,transistors,ics,pcbs,assembly errors ....
These probabilities are cumulative, and in complex systems it is not unknown
for the system to be more likely *not* to work.

I beleive it was motorola who introduced the 6-sigma reliability concept,
which is the latest management buzz in my own company.  However, it has its
basis in hard mathematics, not bullshit - which is more often the case.

So, in summary, it depends on complexity and design.

> What about in consumer products?

I refer you to the answer above.

> What do you feel is a fair warrenty period for industrial products?

Failures once off the steep initial part of the 'bathtub' reliability curve
are rare.  Ask your customer if he would be prepared to pay for a 100%
burn-in
screening of the product prior to delivery.  The complexity and design will
determine the burn-in required.  You might want to ask searching questions
as to environmental & vibrational testing.  Mentioning money always makes
'em uncomfortable.

> What is fair for consumer products?

A 12 month warranty appear to be the norm.

> Has anyone considered an extended warrenty purchase program for their
> industrial customers, and if so, how did it work?

N/a

> Where can we find industry data that addresses these issues?

Try a reliability consultant.

> Thanks for responding
>