jtroxas wrote:

> yah.. I thought I remember reading something about that along time ago.. I 
> just cant remember where I had seen it on the net... and so I put up this 
> post.. I also sometimes hear some scientist on TV saying something bout it..

It depends on what you look at. If you consider components only as "black
boxes" with average failure rates (say, a resistor has a MTBF of so many
hours), then, of course, the more components, the more likely a failure.

But as I wrote before, it's not that simple. For a decent failure
probability analysis for any given device, you need to first define
"failure" for the device and your purposes. Then you need to establish the
relevance of the different failure modes of each component for that device
failure (which depends on the other components). Then get the probabilities
of the different failure modes (which also depend on the other components).
Then add that all up (considering correlations). 

There's nothing that guarantees that a design with fewer components will be
more reliable. It may very well be that, for a given design, added
components increase the reliability rather than decrease it.

Look for example at PIC programmers. You want a really reliable piece? You
won't find them among the ones with the fewest components. 


> what about common household applience like our good old computers.. I
> dont know if they have several or secondary or failsafe units that will
> come alive  incase something fails...Unless were talking about something
> special like multiimillion dollar aircraft.. the space shuttle or
> something in that respect...

Of course, normal household appliances usually don't have built-in
redundant devices. But think of simpler examples... A power supply with a
resistor in the input part. The resistor every now and then fails open,
with a device failure as a consequence. Somebody thought that the resistor
might be driven too close to its maximum voltage and replaces the single
resistor with three in series: no failure anymore (of this kind). More
components and fewer (not more) failures.


Then there's the question what constitutes a "component". There are ICs
with millions of transistors. There are ICs that pack together different
devices on the same silicone -- are they as many components as they contain
integrated devices? There are modules that contain several components on
thick film; are these one component (they look like one from the outside)
or several? Furthermore, you'd have to have a significant average failure
rate for each component. I don't think these numbers exist for most of the
components you'll find in a typical design, and I don't think these numbers
are very relevant; the actual failure rate depends a lot on the individual
application, so average failure rates are not really useful. I don't think
the number of components is a very relevant reliability benchmark.
Stripping out the protection diodes from a PIC's input circuit doesn't
increase reliability.

Gerhard

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist