> The point I am making is they were all tested and sold cheap because they
> were junk not because they left the testing up to the buyer.

Thunks:

-  That may well be true when process yield is low.
For mature parts and processes it may well be difficult getting enough
junk to go round ;-).
When results get good enough the bottom spec parts may be within an
acceptable specification in the vast majority of areas or even well
above spec across the board. I have no chapter and verse to quote but
this sort of occurrence has been often enough mentioned for decades.
Modern metal film 5% resistors tend to cluster around their nominal
values rather than being normally-spread as was the case in ancient
times. So much so that I've occasionally been caught out when
prototyping by assuming an accuracy better than the stated one - as
this is often but not always true.

-  I'd suspect that there would be some cutoff point below which a
part would be rejected - eg hard RAM faults in the defined addressing
range. Some ICs that provide mass repeated facilities (eg RAMS, CCD
imagers and similar) either have spare 'rows' that are used to make up
for occasional faults or may have say 50% of the part disabled if a
lower spec part exists at 50% capacity. eg on windowed EPROMS of yore
it was common enough to find that a 2KB part and a 1KB part looked
identical. Row or bank selection can be by fuse burning or LASER
trimming or presumably even by a late-in-process metalisation stage.
[The inexact language here is a clear indication that this is an area
I've heard much on but have not investigated in any detail :-) ].

-  Failure to meet temperature spec for all relevant specs could well
be acceptable if the failure still allowed a "sensible" full spec
temperature range.

-  Ability to sell a rebadged full spec part with no guarantee of
performance being met also allows a manufacturer to undercut lower
cost competition while not destroying their own profit margins in eg
western markets.


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