Silicon mismatch data details.

[dinesh hegde]

Mismatch details

Where can we get the silicon mismatch data. I searched in process development kit (PDK) but it is not available. (I need data regarding variation of oxide thickness, implantation and doping across the Die)

 

[dick_freebird]

Mismatch details

If you need intradie -data- you'd probably have to pull it
yourself and extract these -process- parameters from
the device electrical structures.

You could back into it from PCM data if you had all-sites
data sets, such that you could get a site-site mismatch
picture and work back to a mismatch-per-millimeter value
(presuming there was any coherent result). But that is not
to be counted on, it's my belief that "local mismatch" dominates
even at die-scale distances and it's not Tox or implant
or activation at macro-scale that accounts for FET matching.
It's more about (especially as devices become very small)
things like short-range uniformity of dopant; if you have
on average 10 dopant per channel it's real easy to get
+/- 20% channel charge and a big VT mismatch, while
100 per channel that element will subside into the "noise".

It's something that is still better addressed by art than
trying to "overthink" the science, especially when there
is no systematic handle. Like, what would you possibly
do differently in layout, if you had a +/- 5% Tox gradient
across the dia cs +/-3%? Nothing, other than perhaps
specsmanship. You still draw it as clean and symmetric
as you can and then you take the yield.