Hi, i saw some post on gm/Id method, an IEEE paper...
# analog-design
p
Hi, i saw some post on gm/Id method, an IEEE paper for the same https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=535416&tag=1 However, to plot graphs, 1. gm/Id Vs Vgs --> for gate Voltage 2. gm/Id Vs Vdsat --> i.e. VDD (also called as gate overdrive voltage) 3. gm/Id Vs gm/gds --> intresic gain 4. gm/Id Vs Vth ---> threshold voltage 5. gm/Id Vs Id/W ---> Current Density i'm searching for ngspice commands to be used with sky130, from some time. But, with few forms not able to find help. So, can anyone having example for such. It would be great help.
a
@Pritesh Ps You could find a large dataset generated here for nfet_01v8 for different W, L, Ldiff, we didn't vary nf but we plan to have another dataset for that: https://drive.google.com/drive/folders/1cbCYK9wuP3NY63LCNjdvb834iI2N1R5F?usp=sharing Also, we extracted all the Cs to be able to follow along the lecture that prof. @Boris Murmann has published ealier: https://open-source-silicon.slack.com/archives/C016HUV935L/p1682635431536879?thread_ts=1682627330.227489&cid=C016HUV935L
@Pritesh Ps Please don't use the data above. Model that derived this data are not working correctly.
p
@Boris Murmann thank you professor.! @Amro Tork I have been trying for the same as professor with provided script from several days! Since i'm not familiar with GoogleLab much, I thought only i'm having have issues to recreate the data. Thanks for conforming though. I'm back to working with xschem, ngspice, sky130. and trying to figure out how to do it there.
a
@Pritesh Ps The models are not reliable. I advise to use GF180MCU not skywaters
Here are the models
p
@Amro Tork thanks for the suggestions. But, i have already designed some blocks by calculation with sky130. and switching to GF180 in middle of masters project will become problem for me.
do you know any other possible way to get those graphs from sky130?
a
@Pritesh Ps yes, you will need the same setup provided by Dr. Murmann above. But the problem is that the models are not valid for all W & L combinations. You are limited by the W & L values in the model.
p
yeah, i tried to make the same setup but don't know much of setting up the anaconda (python) environment. Also, google colab thing is limited run and needs to be purchase for more runs. and both are not feasible options.
๐ŸŒ 1