I wish! Actually, what made me think of this as a solution to making
very large chips is that IBM employed a similar solution to improve
yield on 64K memory chips back in the 70's.
IBM designed the 64K chip in quadrants of 16K each. When the chip was
manufactured each quadrant was tested and circuitry deleted (via laser I
think) if necessary to make 16K, 32K or 64K chips depending on how many
of the 4 quadrants had defects. I don't recall there being a 48K chip
but I suppose that could have been possible. It also added a new twist
to production planning since you started out trying to manufacture one
thing but what you got out the other end might be usable but different
than what you tried to make in the first place.
This design to improve yield had a dramatic effect on manufacturing.
The memory plant in Burlington, Vermont had a new, major building under
construction at the time based on capacity requirements stemming from
the old yield model. The exterior of the building was finished but the
building was turned into a warehouse instead of a new production line.
The new design was so successful in improving yield that the new
building wasn't required.
Does anyone know if the current sensors are pixel mapped to be able to
fill in the oddball pixel that might have been defective from day 1?
Chuck Norcutt
Moose wrote:
> Chuck Norcutt wrote:
>
>>Scalability is one thing. Production yield is another. I'm solving the
>>yield problem.
>>
>
> Good for you ! Can we buy stock now?
>
> Moose
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|