Configure and interpret an Optimum™ margin study

Start with a nominal circuit that really uses the parameter

Extract the complete resistor-study ZIP into a new writable ASCII directory and use a matching current JoSIM®-Pro 2.9 Optimum™ installation with a valid JoSIM entitlement. Run from that extracted directory. These commands qualify the scripted CLI route; the maintained 2.9 GUI also has Run > Optimum™ studies, whose complete native walkthrough and images remain separate.

The fixture ramps a current from zero to 1 mA between 10 and 20 ps. R1 uses RVAL, initially 45 ohms. The SP reference expects node N to reach 50 mV and allows a 20% voltage deviation, giving an independently known passing resistance interval of 40 to 60 ohms. The initial 45-ohm circuit passes.

Keep the SP header as the bare node name N. This SP route generates its voltage print from that name; replacing the header by V(N) produced a missing-label failure. CSV and EVENT input conventions are different. The circuit uses fixed-step PHASE/BDF inside Optimum™.

Run verification and the selected margin search

1. Check that resistor.cir, resistor.sp, verify.toml and margin.toml are together. Read the complete inputs below before changing a physical model.
2. Run the commands below. A failed nominal verification stops the study; fix its first diagnostic before widening search ranges or changing tolerances.
3. Read RVAL's negative and positive sides, the critical side and the current margin_report.txt. The supplied search covers scale factors 0.5 to 1.5. Preserve the inputs and record the actual payload and settings with the result.

optimum-cli -r 1234 verify.toml
if ($LASTEXITCODE -ne 0) { throw "Nominal verification failed" }
optimum-cli -r 1234 margin.toml
if ($LASTEXITCODE -ne 0) { throw "Margin study failed" }
Get-Content .\margin_report.txt

Interpret the percentages and search resolution

Margin varies each selected parameter while the others stay nominal. It does not show that all parameters may move to their worst individual limits simultaneously. A search limit reached while still passing is a censored boundary, not a discovered failure; use the structured receipt to distinguish bracketed failures from search limits.

Binary search assumes a connected passing interval near nominal and uses a directional bracket scan. The checked exhaustive variant, with accuracy = 0.005, gives 11.00% and 33.00%, consistent with its coarser grid. A pocket-search variant writes a pass/fail map. Neither grid establishes that a narrower disconnected pocket cannot exist.

Choose the right parameter and tolerance controls

The parameter must be consumed by the circuit. Replacing R1 N 0 RVAL by a literal R1 N 0 45 leaves RVAL unused; the checked search then reports the 50% search limits on both sides. That is not evidence of a physically robust resistance range.

If RVAL is derived from RBASE and RBASE is searched, set margin.parameter_dependence = true when dependents should be re-evaluated. The checked frozen-dependent case reaches only the search limits, while re-evaluating the dependent recovers the expected resistance margins.

For this ordinary SP voltage trace, amplitude_threshold controls the acceptance fraction. Setting it to 0.25 changes the analytic interval to 37.5–62.5 ohms and widens the reported margins accordingly. Changing the phase threshold alone leaves this voltage margin unchanged. Do not treat one tolerance field as universal across voltage, current and phase checks.

Recover an empty selection or failed nominal result

1. Check the actual .PARAM names and their uses before applying a selection regex. Start with the exact anchored ^RVAL$ pattern supplied here.
2. If selection fails, replace the search_string entry by the explicit entry RVAL = {} under [parameters], then run again to a new output name or fresh folder. The explicit selection reproduces the expected margins.
3. If the process instead exits abnormally, retain the configuration, diagnostic and actual payload identity. An earlier installed review binary crashed on an empty selection; the current checked source build rejects it with a diagnostic. A common version number alone does not identify that behaviour.
4. Keep the same acceptance rule and meaningful numerical/physical model while investigating a margin discrepancy. Nominal RVAL = 30 fails verification in this fixture and correctly prevents every study mode from exporting a result.

Complete resistor.cir

* Public one-resistor Optimum teaching circuit; RVAL is actually consumed
.PARAM RVAL=45
I1 0 N PWL(0 0 10p 0 20p 1m 100p 1m)
R1 N 0 RVAL
.TRAN 1p 100p
.PRINT V(N)
.END

Complete resistor.sp

time N
0 0
1e-11 0
2e-11 0.05
5e-11 0.05
1e-10 0.05

Complete verify.toml

mode = "verify"

[verify]
method = "spec_file"
circuit = "resistor.cir"
file = "resistor.sp"
amplitude_threshold = 0.20

Complete margin.toml

mode = "margin"

[verify]
method = "spec_file"
circuit = "resistor.cir"
file = "resistor.sp"
amplitude_threshold = 0.20

[parameters]
search_string = "^RVAL$"

[margin]
accuracy = 0.001
binary_search_steps = 12
min_search = 0.5
max_search = 1.5
output = "margin_report.txt"