Hi @Tim Edwards Can you please help me to extract the PEX (R+C) correctly for this gds!

@Ahmed Reda please check if this helps: https://openroad.readthedocs.io/_/downloads/en/latest/pdf/

Do the following:

gds read otam
select top cell
expand
flatten otam_flat
load otam_flat
select top cell
cellname delete otam
cellname rename otam_flat otam
extract all
ext2sim labels on
ext2sim
extresist tolerance 10
extresist
ext2spice lvs
ext2spice cthresh 0
ext2spice extresist on
ext2spice

@Tim Edwards Excuse me. I have one question, sometimes I have to use

cellname delete otam
cellname rename otam_flat otam

and sometime not. When I should use

cellname delete...

and

cellname rename...

?

It's just a convenience. Because when you use "flatten" you end up having to create a new cell with a different name, and when you extract that, you get a subcircuit with the new name, not the original name. You then either have to simulate or run LVS with the new name, or manually change it in the netlist. The commands in magic just remove the original cell and rename the flattened cell to the same name that the original one had, so that when it extracts, the subcircuit has the original cell name.

@Tim Edwards Also, should i use those steps for this gds !. I just have a little bit of conflict. Thank you for your patient.

This should work for any layout:

!#/usr/bin/env bash
magic -dnull -noconsole << EOF
gds read $1
select top cell
expand
flatten $1_flat
load $1_flat
select top cell
cellname delete $1
cellname rename $1_flat $1
extract all
ext2sim labels on
ext2sim
extresist tolerance 10
extresist
ext2spice lvs
ext2spice cthresh 0
ext2spice extresist on
ext2spice
quit
EOF

@Tim Edwards Thank you for your explanation.

This is a good layout for hitting all the known issues in extresist. Extraction around the bipolar transistor is particularly a problem, because it is causing extresist not to handle the nets connected to the bipolar base and emitter, which are important ones for simulating parasitics. I will need to delve into the bipolar extraction a bit and see what's going on. The other problems are minor; the contact issue, as I said, may produce a small error in resistance, but that's all. The "missing connection" errors (except maybe for the bipolars) are on source-drain tied dummy devices, so while they are not connected to the right points in the power and ground nets, it won't make any noticeable difference to the simulation.

@Tim Edwards Thanks. I got it.

@Ahmed Reda: Please compile magic version 8.3.372 (which I just pushed) when you get the chance, and test it on this bandgap. I worked through a list of problems, including (1) handling bipolars and similar devices where one or more terminals are in different planes from the device identifier type; (2) handling MOSFETs with source and drain tied together, and (3) properly handling the entire substrate as a resistor network. (1) and (2) were pretty clean fixes. (3) produces an ugly mess of the substrate which I assume is correct, but which may or may not give the SPICE simulator fits. Please give it a try in simulation. I am going to look into options for "extresist" that exclude or simplify the power rails, and check if there is a way to do more aggressive network simplification on the substrate network.