• Previous message: Steven M Brown: "RE: pseudo-resistor layers"
• In reply to: Steven M Brown: "RE: pseudo-resistor layers"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

 > following: the pseudo-layer calculation is easily fooled and extresist,

hi Steve, i think your 'the calculation is ... fooled', refers to
you're not getting the L/W you want, given the area/perimeter of the
'rpoly' to poly/prp boundaries you draw, but those are in your
control, right?

consider a fet. if you have a W/L = 5/2 fet, but instead of drawing it
like this:

 *****
 *****

you draw it offset like this:

 *****
  *****

then the L/W derived from the sum of area/perimeters for each tile
(10/5+5+2+2 in the first case, 10/5+5+3+3 in the second case) is not
going to give you the W/L you expect (5/2).  this is a limitation of
magic, that it won't extract L/W correctly for non-rectangular
non-convex regions (since it double counts some perimeter tiles).

by the way, a separate issue is that even with rectangular regions,
the quadratic equation solution for L/W can sometimes surprise you
with large diode-connected fets. it's "right", that is, the area and
perimeter it yields matches what you drew, but surprising, and "wrong"
from the standpoint of spice, which needs to have accurate L/W for
proper parasitics, altho this usually isn't a problem with diodes,
and that's the only case i've seen it.

/jeff

• Next message: Craig Stevenson: "Cadence interface to Magic?"
• Previous message: Steven M Brown: "RE: pseudo-resistor layers"
• In reply to: Steven M Brown: "RE: pseudo-resistor layers"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]