I'm running a transient sim for LSXO duty cycle that takes quite a while. I think I could create spice .meas statements for duty cycle, Vol, Voh, and rise/fall time with a single simulation. Is there a way to capture data for multiple electrical parameters in a single testbench with CACE? Maybe I just haven't found this in the documentation.

The

instramp

example project has an example of this. The

gbw.sch

testbench produces three results: gain-bandwidth, gain margin, and phase margin. The

simulate { ... }

block has a

format

line that shows how many values are in the output line, and there's a

null

for each one that should be ignored, and

result

for the one that is important to that parameter. Unfortunately, CACE is not yet smart enough to keep track of which testbenches have been run under which conditions, so it will run ngspice multiple times to get each result instead of using the same resulting data file multiple times. That's on my to-do list for CACE.

If I want to send an ideal sinusoid using a voltage source with a SINE() function, would CACE interrupt that? I have a testbench using these guys (image) that runs smoothly with xschem/ngspice, but is transformed into DC using CACE.

No, CACE shouldn't interrupt that. But it sounds like it is, anyway.

Well, I don't think it's realistic to find a bug fix in a few hours. I just pushed an update to https://github.com/b-etz/sky130_be_ip__lsxo if you're interested to see. The affected test is Vol and Voh.

If I set "keep simulation files" in Settings, I can see the testbenches, and the

SIN()

is present on the voltage sources. If I add nodes

net4

and

net5

to the

save

line, they can be plotted. So, nothing wrong with the sinusoidal voltages.

If I bypass the parasitics symbol, CACE runs with a sensible result. Interesting.

I think

lqfp_parasitics

is not being generated. I don't see it in the netlist. It appears to be an independent netlist you've defined in a schematic and are calling from the testbench. I think you probably will have to generate its netlist manually.

Does ".include {DUT_path}" include the whole directory? Or just the DUT and its dependencies?

It includes only the schematic

sky130_be_ip__lsxo

. I think that the issue may be that the

lqfp_parasitics

is defined in the xschem directory, but it is being used directly in the testbench which is in the

cace

directory, and

cace

specifically does not add the

xschem

directory to the path because it needs to avoid putting the netlist of the DUT directly into the testbench. You may need to duplicate

lqfp_parasitics.sch

in the

cace

directory so that it can be found when generating the testbench netlists.

Easy enough. Thank you!

Well, I'm not all that sure of my answer. I get the general problem, but haven't stared at enough to have a high confidence about the solution. There are probably multiple ways to work around it.

Can confirm that worked on my end, too. Glad you mentioned deleting the

.spice

generated files, because it would have taken me ages to try that.

Maybe I should give up checking for out-of-date files and just always regenerate netlists. There are just too many ways to subvert the out-of-date check.