Jinx wrote :

> How intensive is batch testing ? Say you have a bunch of micros
> made from several wafers. Presumably, ideally, each wafer and
> subsequent layers would be made of homogeneous material. So
> 
> (1) if you're working with known, controlled materials why does a
> batch or part fail to meet spec in the first place
> 
> (2) if one piece from a batch or base wafer fails, can you assume
> that the rest of that batch/wafer also will fail. Surely it can't be
> necessary to test every piece in a high-volume product
> 
> (3) therefore is a selected sample rigorously tested under 
> all conditions
> and that assumption made, for whatever reason, for the rest of them

I can't see why this would be any different from producing just
about anything. Statistical inspection was developed during
WW-II when the US military found out that they didn't had the
resources (mostly time) to test and inspect *every* bomb made.

The methods developed are today known as "MIL-STD-105D".
Google for that, and the first hit is :
http://www.itl.nist.gov/div898/handbook/pmc/section2/pmc231.htm

In short, what you do is to first define an AQL for your product.
AQL = "Acceptable Quality Level", or the acceptable %-ige of
defective products leaving the production. Say an AQL = 0.01
says that one defective out of 10.000 is "OK".

Then, using your AQL, you look up in the MIL-STD 105D tables
(today computerized, but first on paper) to get your sampling
lot size and the number of accepted rejects in the lot.

Let's say that you have 100.000 parts batch. The tables could then
tell you to sample 100 parts, and that 1 error is OK, but if you
get 2 in the sample lot, the whole 100.00 batch is rejected.

One thing that is *very* important is that you use a as
random sampling from the whole lot as possible. Special
random number tables was printed that you used to both
decide from which box (or whatever) you should pick your
samples, and also at what time of day you should make
the sample.

I can't see any reason why you can't use statistical sampling
inspection on chips also.

Jan-Erik.



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