https://open-source-silicon.dev logo
Join Slack
Powered by
Hi All. Does anyone have an example of running MC ...
# xschem
d
Hi All. Does anyone have an example of running MC analysis in SKY130 using xschem / ngspice? I see MC analysis in the ngspice manual but they seem to be for generic models. No idea how to get statistical variation modelled into an instance from SKY130. Any help or examples much appreciated here! Thanks
l
a
Key change is use tt_mm library
d
oh great! this gives me something to get started with. Thanks very much!
@Luis Henrique Rodovalho: Just looking at your very helpful example there. So seems I don't need to manually edit any parameters in the netlist (e.g. manually add a gaussian distribution to vth). All this is self contained in the tt_mm file. I just simply need to run 'x' (=200 say) times and I will see a statistical spread in my results?
l
Yep, but you must save the measurements somewhere after each run.
👍 1
a
That's sound correct.
👍 1
d
Is there anywhere I can get documentation on the matching for this process, specifically Avt data for the various devices? As shown in the attached, when I perform MC analysis on pfet_01v8 (200 runs) I get an Avt of ~ 8.3mV.um. This seems reasonable as I have seen Avt's of ~ 8 in 180nm processes, However, it would be good to get some confirmation. Txs
@Luis Henrique Rodovalho: In your script why do you need to include "destroy all" at the end of your dowhile loop? Txs
l
The destroy command is used to release the saved signals from memory after each run. It can be a lot for larger circuits and/or simulations.
d
Thanks Luis. Good memory management tip so, will be useful when I start running MC runs across process.
11 Views
Open in Slack
PreviousNext
Powered by