On Fri, 15 Mar 2019 at 01:36, David Van Horn <
david.vanhorn@backcountryaccess.com> wrote:

> Does anyone know if transistor gain variations for the same part, unit to
> unit, are distributed flat across the min-max or gaussian?
>
> Related, but not pertinent :-)

Exact detail of the "what I did" fades with time (about 11 years ago) but
the 'story' is true.

I was in a factory in China where a large number of NimH AA cells were in
the store.
All the boxes looked the same and while the factory said that supply had
been from several batches they did not know how to distinguish between
batches or know any details about quantities etc.
Some batteries behaved in an acceptable manner and some were substandard.
It appeared that this may be batch based but this was not certain. The
differences appeared to be on a per box basis but this was not 100% certain
and testing 100% or even sampling per box was unattractive.

Memory fades, but I found a way to ID boxes by batch externally but this
was not enough to weed out good and bad boxes.

I asked if they had any 'scales' and they produced a digital office scale
with a resolution of (I think) 1 gram.
I started to weigh boxes and it became obvious that there were noticeable
differences.
I grouped batteries by mass in 1 gram increments on a box by box basis
(Each cell weighed in the order of 25-30 grams so not many groups would
have been expected.
One set of boxes had a single peak weight distribution with a semi gaussian
distribution.
The other set had a double peak weight distribution with one somewhat
higher than the other, significant separation and a noticeable dip between.

Low spec batteries came from the double dip boxes and (probably) low mass
samples.

Absolute weighing would have probably produced a similar result but the low
tail of the 'good' batteries probably overlapped the low tail bad batteries
which would have hidden both the different-batch cause and given a mix of
good and bad lower mass batteries.

The factory were suitably impressed. Such somewhat fortunate results help
one's image and credibility :-).



            Russell
--=20
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.