Hi all,
some updates and questions regarding my cell/mask editor:
- The pixel-based editor works quite well now. With some practice, a
few hours were enough for me to draw 15+ simple cells (inverter,
buffer, nand, nor, and, or, tri-state inverter; each with different
drive strengths). Especially since many of those are simple variants
of each other.
- I defined the standard cell height to be 140 lambda, which is way
too much, even when you are inexperienced and use a very conservative
DRC. Much space is wasted. I will eventually restart at 100 or 70
lambda.
- The DRC I'm using is *very* conservative, even more so than MOSIS
SCMOS. I will mostly keep it that way unless/until you know more about
the actual constraints of your process.
The most constraining open questions regarding the process are about
wells. I will post them here in case you already have some information
in this regard. Note that I am thinking of purely digital circuits for
now, ignoring analog and clock network related stuff.
- Should I assume n-wells only, p-wells only, or both? If both, Do I
have to keep a minimum spacing between them? (MOSIS SCMOS says "no
spacing, just don't make them overlap", probably because in digital if
they touch they make a reverse-biased diode). The spacing is the
crucial point here since it affects the cell layout.
- Can the n-wells (or p-wells) of different cells be merged without
getting issues with electrical noise? Some cell libraries seem to do
this, i.e. they use a single long well across a whole row of cells
instead of one well per cell (or even one well per transistor). This
is super important because separate wells means spacing between wells
and even more spacing between source/drain and well boundaries, so it
constrains the layout a lot.
Greetings,
Martin