Hi all, I'm attempting to run PEX on a 64bit width...
# xschem
j
Hi all, I'm attempting to run PEX on a 64bit width circuit using xschem (v3.1.0) + ngspice(v40+). The test bench is calling on a 37.6MB spice file generated by Magic(
ext2spice cthresh 0, extresist tolerance 10
). Loading the test bench is taking an excruciating long time, which it is still loading (start time: 02:57pm EST) the circuit. I have set within the
.spiceinit
file with
ngbehaviour=hsa
. Is there a way to speed up this process or a different way to go about this?
h
Why are you going with
cthresh 0
? I would think that something like
ext2spice cthresh 0.01
would be more than sufficient.
Also, I do this
ext2spice rthresh 100
in my PEX script.
j
Thank you for the advice! I was following a recipe posted on slack and do not have good insight to what values to use when doing PEX since I'm clueless. To my understanding
cthresh 0
within Magic equals to all possible parasitic capacitor, is that accurate? Is there a method to figure out what cthresh or rthresh values to use; if its trial and error, do you change the values based on circuit/simulation run time?
This is the recipe I have used for PEX. Aside from
cthresh 0.01
and
extresist tolerance 10
, are there any settings that I should be looking out for?
Copy code
flatten cellname_flat
load cellname_flat
extract all
ext2sim labels on
ext2sim
extresist tolerance 10
extresist 
ext2spice lvs
ext2spice cthresh 0
ext2spice extresist on
ext2spice
The PEX spice file size is 35.4MB, should I be worried or is there a target file size to aim for?
h
There is no “proper” size for a PEX netlist. There is a “good enough” which you need to decide per application. See here a PEX script that we usually use, and it seems to work fine. https://github.com/iic-jku/osic-multitool/blob/main/iic-pex.sh
j
Thank you for sharing the script. I will download and install the tools. Edit: Do I need to run
iic-osic-setup.sh
to be able to use
iic-pex.sh
properly?
h
No, just run
iic-pex.sh
j
Hi @Harald Pretl, I finally got around to properly use
iic-pex.sh
. I would like to ask whether I need to have the flat version of the cellname included or I can just use the hierarchal one? Plus, if full R-C mode produces a .spice file of 35MB, is this a point where I should give up on having the parasitic resistances included?