|
|
Magic Mailing List |
|
From: Steven M Brown (smbrown AT cse DOT psu.edu) Date: Thu Oct 23 2003 - 11:57:29 EDT
On Wed, 22 Oct 2003, Jeff Sondeen wrote:
> hi Steve, you need to achive the same area/perimeter of the 'rpoly'
> part of each resistor, if you want them to extract the same value,
> since magic ultimately determines L/W by solving an equation using the
> total area/perimeter it stores per node, so results vary from the
> small area/perimeter differences of the 'rpoly' areas that you drew.
>
> if you draw the same area, you get the same "L/W" values:
>
> m1001 ext_in a_19_24 int_in Gnd polyResistor w=2 l=136
> + ad=0 pd=0 as=0 ps=0
> m1002 feedback a_19_13 int_in Gnd polyResistor w=2 l=136
> + ad=0 pd=0 as=0 ps=0
Here are some results from my morning calculations and experiments of the
resistance extraction features when used on nuisance_resolved.mag
and nuisance.mag (resistor terminated at int_in):
ext_in feedback
------ --------
psuedo 1309 1976 <-- extracted as a 3x151 and a 2x152 pseudo-fet
extres N/A 1980 <-- partial results as described below
by hand 1950 1976
The label on the left is the manner of resistance extraction, the column
is the termination of the other end of the resistor.
The result of these calculations and observations would seem to be the
following: the pseudo-layer calculation is easily fooled and extresist,
though accurate and thorough, does not spot resistors that are terminated
externally (a situation which comes up when testing incomplete designs).
> this is less kludgey when you use 'silicide_block' layer to define the
> resistor (which when available changes ohm/sq from like 5 to like
> 125), which has to be drawn explicitely 'distinct' due to spacing
> rules anyway.
I will look into incorporating this into my design. Increasing the
resistance by a factor of 25 (or any factor > 1, really), would save
space. This still does not resolve the resistance extraction difficulties.
Steve
|
|
|
|