Hi, @Tim Edwards I'm having differences between discrete (

libs.tech/ngspice/*

) and continuous (

libs.tech/combined/*

) models. Which one should we trust? This is the reference oscillator used on the SDC Temperature Sensor. Discrete models predict a maximum frequency of near 200MHz, while continuous models near 150MHz.

Well, honestly I don't know because they should have been the same. But pick one of them and trust it, and we can use your temperature sensor to figure out whether it's right or not. Is the difference due entirely to FET models?

I think so, in the first image we use the

spice.{corner}.red

variations that reduces load time. while the latter imports the

combined

version. This simulations are of the oscillator, but we can't run the complete sensor with combined models, it throws this error

Netlist line no. 0:
Undefined parameter [temp]
Netlist line no. 0:

Do you know what could it be?

I am not sure what is in the

.red

files. I don't think they are necessary since ngspice version 42, which appears to have fixed all of the long load time issues. If you use the standard library (not .red), do you get a different result?

The

.red

files are generated with this script from osic-multitool. I'm going to update the simulations with the normal libraries.

@aquiles viza: I think for the error

Undefined parameter [temp]

, you have an outdated PDK source for `sky130_fd_pr`:

commit ba9a9cc7b38cf22d6c7f5bfe4cfdba06358ef552
Merge: c2bfb91 95f86d7
Author: R. Timothy Edwards <tim@opencircuitdesign.com>
Date:   Thu Jan 18 21:00:09 2024 -0500

    Merge pull request #13 from efabless/fix_temp_to_temper

    Changed "temp" to "temper" in the continuous pwell resistor model