Hello all, I have problems chosing the best xsche...
# sky130
f
Hello all, I have problems chosing the best xschem/ngspice settlings to prevent numerical errors. I designed a comparator, but when I chose my 2 inputs close to each other ( 0.69996 and 0.7V) I experience noise while doing a normal transient analysis. My comparator output changes between runs. How I understand it now, is that this is due to numerical errors of ngspice? Which settings do you use to prevent this (I tried to change RELTOL, ABSTOL, METHOD,... but didn't find a succesfull setting yet. )? Or how do you do noise analysis without letting this influence the result? In attachment, the schematic and spicenetlist of my strong arm comparator, and a wantend and unwanted waveform. Have a nice day
1
m
ask in the #analog-design channel
👍 1
t
@Faedra Webers: You passed
tt_mm
to
.lib sky130.lib.spice
.
tt_mm
stands for "typical-typical mismatch". The "mismatch" part means that you're doing a mismatch analysis, and the circuit values will be different on every run, filling out the statistical distribution of device parameters. If you don't want that behavior, then use the
tt
corner, not
tt_mm
.
🙌 1