[sdiy] Re-spin the classic ICs?

[Paul Schreiber]

> Or would Intersil consider introducing an improved 21st century
> version of the the CA3280, made in a modern fab?

Never going to happen.

Modern analog fabs, even on smaller 4" wafers (but these would run on 6" 
probably), you would get around 20,000 KGD (known good die per wafer). AND just 
like there a minimums for running pc boards in production, a typical minimum run 
(which is in turn, a function of reprogramming the diffusion ovens, reload mask 
reticules, setting up the test machines, sending the die out for packaging, blah 
blah blah) is 20 wafers. So we get 400K parts. Lord, on a 6" process over 1 
*million!*

BUT.....in order to 'respin' on modern fabs (and I've done this whole exercise 
regarding CEM/SSM), you need new mask sets. Well, in the "good old days" you 
could literally make IC masks *by hand* like a double-sided pc board. But now, 
you have to CAD them. BUT...then there is the new circuit design/HSPICE 
simulations/and internal lot testing (BTW: did I mention this process is at a 
*minimum* 26 weeks?).

How much money do you think all of that effort costs? Well, about $250K is a 
good point. And now our 400,000 ICs are in the $15/$18ea range *assuming* you 
can *sell* 400,00 at $15/$18ea. The "good" news is on a 6" fab we could hit 
probably $5.

I'll bet the WORLD TOTAL of CA3280E/AE (same die, this is a 'binning' for 
offset-vs-Iabc....oh CRAP that's another set of $$$) sales for the last *5 years 
total* is <100,000 parts. Ever semi company I've work for has a maximum 14-month 
ROI target (you have to show a profit after 14 months of sales). HAHAHAHAHA!

Paul S.
dasher of IC hopes