IHP-Open-PDK
IHP-Open-PDK copied to clipboard
IHP-GmbH
missing model for `sg13_lv_nmos`
Trying to run the following ngspice simulation:
.title sg13g2_inv_1
.lib /content/IHP-Open-PDK/ihp-sg13g2/libs.tech/ngspice/models/SG13G2_hbt_woStatistics.hsp.lib
.include /content/IHP-Open-PDK/ihp-sg13g2/libs.ref/sg13g2_stdcell/spice/sg13g2_stdcell.spice
X1 GATE DRAIN VPWR 0 0 VGND sg13g2_inv_1
Vgnd VGND 0 0
Vdd VPWR VGND 5.0
R VPWR DRAIN 50k
Vin GATE VGND DC 0V
.op
.option post nomod
.end
.control
save all GATE DRAIN
dc Vin 0 5.0 0.01
display
wrdata output.txt GATE DRAIN
.endc
will result in the following error:
Error: unknown subckt: x1.xx1 vpwr gate 0 vpwr x1.sg13_lv_nmos x1.w=740.00n x1.l=130.00n x1.ng=1 x1.ad=0 x1.as=0 x1.pd=0 x1.ps=0 m=1
Simulation interrupted due to error!
See the following notebook which reproduce the issue: https://colab.research.google.com/gist/proppy/5582eeae8d2e2eccb48cd591b904185f/ihp-pdk-playground.ipynb
Is that expected? (i.e: the model for nmos/pmos will only get developed/released later)
Hi! Yes, this is expected.. For now, we have only models for HBT (BJT) devices, for the MOSFET models we're still discussing the Spectre to SPICE format translation with 3rd parties. I hope they will be released in summer timeframe. Also we're considering to put nmos/pmos Spectre models to the GitHub (but this will certainly help you only in case you have access to the simulator that supports them)
maybe @RTimothyEdwards and @atorkmabrains can share some insight on how the conversion was done respectively for SKY130 and GF180MCU.
@sergeiandreyev @proppy we actually use home grown scripts for this. But we also go into debugging the model card parameters and tweak them to match the desired outcome. That's why we always request a golden measurements data that we could use as a reference. It's very lengthy process if we faced issues related to not matching behaviors.
@sergeiandreyev Please contact me privately at linkedin if you want us to collaborate: https://www.linkedin.com/in/amro-t-24701458/
@sergeiandreyev You don't need to release spectre Model cards for that BTW.
https://semimod.de/ -- who is based in Europe might also be able to offer assistance, particularly if you are doing recharacterization stuff or regenerating from raw data.
The topic of spice model generation and validation comes up at lot in the CHIPS Alliance Analog work group (https://www.chipsalliance.org/workgroups/) and on the SKY130 raw data repository (https://github.com/google/skywater-pdk-sky130-raw-data).
Thanks a lot for the suggestions and offerings we will have a closer look at them in the next days. As @sergeiandreyev already mentioned we are about to start the process of transfering MOS models into Spice format and plan to have it done by late summer/early autum.
Markus from SemiMod: Thanks for your efforts @noherbrferurtth @proppy We are looking forward to helping with the model conversion and will join the IHP meeting that is coming up.
Also, I am delighted to share that we recently got funding for developing an open-source MOS Model extraction tool, so we are currently looking for (i) reference models and (ii) a test chip for measurement. I think the IHP technology would be perfectly suited.
edit: The spectre model cards would be helpful for us.
- we have now MOS Spectre models available in the repo:
libs.tech/spectre - and SemiMod @metroid120 did an outstanding job to convert these models for use w/ ngspice & Xyce (also HBT models were updated) The data is located in their GitLab repository https://gitlab.com/dmt-development/ihp_sg13g2_compact_models
update: we now have device simulation models available in the repo: lib.tech/ngspice/models
for usage please consult respective readme files and documentation on SemiMod GitLab repo
Hi @proppy , I guess this issue is resolved now.. Can we proceed and close this item as fixed?
