Monte Carlo example

[lhrodovalho]

Is there any example of any monte carlo simulation using the sky130A ngspice libs? The only example of monte carlo simulation I've seen for sky130A was the one which comes with the xschem examples. I've tried to use sky130.lib.spice mc and mc_mm_switch=1 and it doesn't seem to work.

[RTimothyEdwards]

You should not need to set mc_mm_switch, but using .lib <path_to>/sky130.lib.spice mc should work. What are you observing that leads you to conclude that this doesn't work? Don't forget to use .option SEED=random or else the "random variation" will be exactly repeatable (see the ngspice manual section "Analyses and Output Contro -> Simulator Variables -> General Options" and the section "Statistical circuit analysis -> Monte-Carlo Simulation").

[lhrodovalho]

It is working now.

Here is the code I'm using. It's just a cascode current mirror

* cascode mirror monte carlo testbench

* Include SkyWater sky130 device models

.lib "/usr/share/pdk/sky130A/libs.tech/ngspice/sky130.lib.spice" mc .param mc_mm_switch=1

.subckt n1_1 D G S B X0 D G S B sky130_fd_pr__nfet_01v8 ad=4e+11p pd=2.8e+06u as=4e+11p ps=2.8e+06u w=1.0 l=0.5 m=1 .ends

.subckt mirror ii io vssa x1b ii ii n1 vssa n1_1 x1a n1 n1 vssa vssa n1_1

x2b io ii n2 vssa n1_1 x2a n2 n1 vssa vssa n1_1 .ends

* supply voltages vdda vdda 0 1.8 vssa vssa 0 0.0

* input signals ii vdda ii 100n

* DUT xdut ii io vssa mirror vo io vssa 0.9

.option gmin=1e-15 .option seed=random .control

set appendwrite

let run = 1 dowhile run <= 100 reset op print -i(vo) wrdata cmirror.txt -i(vo) let run = run + 1 end

.endc

.end

The results are a little weird. It's repeating most of results.

0.00000000e+00 1.25723549e-07 0.00000000e+00 1.19438386e-07 0.00000000e+00 1.19438386e-07 0.00000000e+00 1.19438386e-07 0.00000000e+00 7.36299401e-08 0.00000000e+00 7.36299401e-08 0.00000000e+00 7.36299401e-08 0.00000000e+00 6.57856318e-08 0.00000000e+00 6.57856318e-08 0.00000000e+00 6.57856318e-08 0.00000000e+00 7.60332873e-08 0.00000000e+00 7.60332873e-08 0.00000000e+00 7.60332873e-08 0.00000000e+00 8.92142086e-08 0.00000000e+00 8.92142086e-08 0.00000000e+00 8.92142086e-08 0.00000000e+00 7.67733530e-08 0.00000000e+00 7.67733530e-08 0.00000000e+00 7.67733530e-08 0.00000000e+00 7.67733530e-08 0.00000000e+00 1.02408988e-07 0.00000000e+00 1.02408988e-07 0.00000000e+00 1.02408988e-07 0.00000000e+00 1.40886246e-07 0.00000000e+00 1.40886246e-07 0.00000000e+00 1.40886246e-07 0.00000000e+00 1.16252589e-07 0.00000000e+00 1.16252589e-07 0.00000000e+00 1.16252589e-07 0.00000000e+00 1.28616915e-07 0.00000000e+00 1.28616915e-07 0.00000000e+00 1.28616915e-07 0.00000000e+00 7.74502629e-08 0.00000000e+00 7.74502629e-08 0.00000000e+00 7.74502629e-08 0.00000000e+00 1.79339098e-07 0.00000000e+00 6.65927789e-08 0.00000000e+00 6.65927789e-08 0.00000000e+00 1.09944515e-07 0.00000000e+00 1.09944515e-07 0.00000000e+00 1.09944515e-07 0.00000000e+00 6.03082056e-08 0.00000000e+00 6.03082056e-08 0.00000000e+00 6.03082056e-08 0.00000000e+00 7.45341353e-08 0.00000000e+00 7.45341353e-08 0.00000000e+00 1.25462960e-07 0.00000000e+00 1.25462960e-07 0.00000000e+00 5.40797241e-08 0.00000000e+00 5.40797241e-08 0.00000000e+00 1.19822149e-07 0.00000000e+00 1.19822149e-07 0.00000000e+00 9.66296963e-08 0.00000000e+00 9.66296963e-08 0.00000000e+00 9.66296963e-08 0.00000000e+00 1.09885928e-07 0.00000000e+00 1.09885928e-07 0.00000000e+00 1.38002183e-07 0.00000000e+00 1.38002183e-07 0.00000000e+00 1.38002183e-07 0.00000000e+00 9.27004353e-08 0.00000000e+00 9.27004353e-08 0.00000000e+00 9.27004353e-08 0.00000000e+00 1.55837423e-07 0.00000000e+00 1.55837423e-07 0.00000000e+00 1.34809293e-07 0.00000000e+00 1.34809293e-07 0.00000000e+00 1.34809293e-07 0.00000000e+00 6.48953742e-08 0.00000000e+00 6.48953742e-08 0.00000000e+00 8.71935954e-08 0.00000000e+00 8.71935954e-08 0.00000000e+00 8.71935954e-08 0.00000000e+00 8.79232191e-08 0.00000000e+00 8.79232191e-08 0.00000000e+00 8.79232191e-08 0.00000000e+00 1.76363110e-07 0.00000000e+00 1.76363110e-07 0.00000000e+00 9.02236918e-08 0.00000000e+00 9.02236918e-08 0.00000000e+00 9.02236918e-08 0.00000000e+00 1.64945923e-07 0.00000000e+00 1.64945923e-07 0.00000000e+00 1.64945923e-07 0.00000000e+00 7.64482972e-08 0.00000000e+00 6.60547929e-08 0.00000000e+00 6.60547929e-08 0.00000000e+00 1.42333200e-07 0.00000000e+00 1.42333200e-07 0.00000000e+00 6.29370159e-08 0.00000000e+00 6.29370159e-08 0.00000000e+00 1.30169292e-07 0.00000000e+00 1.30169292e-07 0.00000000e+00 6.73920554e-08 0.00000000e+00 6.73920554e-08 0.00000000e+00 8.80050687e-08 0.00000000e+00 8.80050687e-08 0.00000000e+00 1.43119352e-07 0.00000000e+00 1.43119352e-07 0.00000000e+00 1.43119352e-07

[RTimothyEdwards]

Well, for one, generally you don't want to use mc with mc_mm_switch. The mc option is for process variation. The mc_mm_switch is for device mismatch. Generally speaking, one doesn't do both in the same simulation. Mismatch parameters are given at corners, and there are specific sections in the sky130.lib.spice for those.

But apart from that, I'm not sure it's valid to make multiple runs from inside an ngspice control block for monte carlo runs. I would at least try just making multiple ngspice runs and concatenating the results from all of them outside of ngspice.

[frbvianna]

@RTimothyEdwards Would you please provide more thoughts into why you think it's not valid to make multiple runs from inside the ngspice control block for the MC runs? I'm asking because that is how the ngspice manual exemplifies the MC runs - looping inside the control block. Though, I am seeing simulation results that are a bit odd in between iterations.