Now I've got my LVS working, I'm trying to do post layout simulation

Whoa, whoa, back up. . .

extract do resistance

doesn't do what you think it does. In fact, it doesn't do anything because it's related to .sim files, not .spice files.

<screeetch>

There is a recipe for doing full R-C extraction floating around, but I have it memorized, to I'll add it here:

load r2r
flatten r2r_flat
load r2r_flat
select top cell
cellname delete r2r
cellname rename r2r_flat r2r
extract all
ext2sim labels on
ext2sim
extresist tolerance 10
extresist
ext2spice lvs
ext2spice cthresh 0
ext2spice extresist on
ext2spice

Beware that this script flattens the cell and then renames the cell to the flattened version, so don't go save the flattened version back to disk, just quit after running this script.

thanks!

For simulation, you do not need to set

.param mc_mm_switch

and

.param mc_pr_switch

. Those are taken care of by the

.lib

statement. The 3rd argument to

.lib

is a section of the

.lib

file (

sky130.lib.spice

) to read and execute. That section defines the parameters. For the record,

mc_mm_switch

is a mismatch parameter (enables mismatch analysis by applying random variation to each component) and

mc_pr_switch

is a process variation parameter (enables monte carlo analysis by applying random variation to properties of the entire process).

thanks!

Actually, your extraction was more or less correct but hierarchical parasitic extraction is very unreliable so you need to do the flattning part.

it's commented out in my PDK.

What version of the PDK does volare claim to be using? That particular problem was fixed in open_pdks early on, years ago.

volare/sky130/versions/78b7bc32ddb4b6f14f76883c2e2dc5b5de9d1cbc/sky130A

That version is half a year and 94 commits out of date

hm

Even then, it should have been run through open_pdks and that resistor line should have been uncommented.

how would you recommend me install the pdk?

Since a lot of people are using volare, I'd like to work out what went wrong with that, and why.

so you know, I templated caravel_user_project, then ran make setup

And the answer to your general question: You already looked at how to do co-simulation with iverilog running verilog code for stimulus generation. The other method to use when you have an internal digital block is xspice. There are two methods to xspice: One is to run yosys and synthesize to the small base set of predefined digital blocks understood by xspice. I don't particularly like that method, since it involves re-synthesizing and potentially introducing differences in the digital interface. The other method is to use my script

spi2xspice.py

from the "qflow" repository (opencircuitdesign.com) which uses a "generic" logic block I created for xspice which Holger Vogt integrated into the ngspice source. That creates an xspice block that is the equivalent of the low-level SPICE. Then you replace the SPICE subcircuit with the xspice one in the netlist.

could I remove the digital block in the spice file and drive the analog part with ngspice cosimulation?

Well, that's where the problem of the parasitic extraction not working hierarchically will become a problem for efficient simulation. I might have to do some development on the parasitic extraction code so that it can handle extraction on a single hierarchical level and ignore subcells, or do that when subcells are abstracted.

Looks like the two of you (and maybe @Stefan Schippers) are the guys to watch among all the "looks ok but isn't, for some subtle reasons" flood of opinions (as opposed to fact) in the discussion forums. For the chipalooza, I want to go for the building block (Ibias for everyone else) that requires parasitic extraction and PVT/matching simulation first, i.e. goes exactly contrary to the schedule, and the scheduled selection funnel. What's the best way to come up to your speed, without incessantly pestering you with questions that distract you from your cutting-edge work, but not having to sift through a flood of opinion in the discussion forums? Any links not subject to discussion where I could start looking, while staying out of your way, would be best. Lots of stuff to see here. Still, move along. Don't mind the Fly on the Wall. Thanks.