RSpice

RSPICE / DOCS / VERIFICATION

Verification.

The Verify workspace judges runs against the plan's specifications: yield over Monte Carlo trials, outputs across corners, tuned and optimized variables, device safety, and regression against a baseline run.

WORKSPACE
Verify, Alt+5
LIMITS
Verify › Specification matrix…
DATA
Read from the run itself, never interpolated
EXPORT
Regression as JUnit or TAP

01 / THE VERIFY WORKSPACE

One run, one table, one verdict per row.

Open the Verify workspace (Alt+5) and pick a page from the Verify menu: Monte Carlo yield, Process corners, Parameter tuning, Optimization, Safe operating area or Regression. Every page works the same way: a toolbar names the run it reads and sums it up in one line, a table holds that run's own values, and selecting a row shows its detail. When the run holds nothing for the page, one line says what would produce the data.

02 / SPECIFICATIONS

The limits every page judges against.

A specification is a limit on a named measurement, held by the simulation plan. Verify › Specification matrix… opens the editor for those limits, a view in the Results workspace; Simulate › Setup › Requirements & specifications shows the same limits and flags any whose measurement no analysis produces.

03 / MONTE CARLO YIELD

How many trials meet every limit.

Add a Monte Carlo analysis to the plan and run it. The yield page has one row per output the trials measured, with the plan's limits beside its statistics.

  • Type a limit into its cell: it becomes the plan's requirement, and the kept trials are judged against it again at once, without a new run.
  • A trial that did not converge counts as failing, and a guard band narrows what passes. Cpk against a one-sided limit is the one-sided index.
  • The toolbar shows the joint yield: a trial passes only when it meets every limit, so the joint yield can be lower than any one row's.
  • Statistics come from the exact samples the run kept; an output whose samples were not kept shows a dash rather than a number.

04 / PROCESS CORNERS

Each output at each corner.

Enable corners in the run set (Simulate › Setup › PVT, sweeps & variation), then run the plan. Rows are outputs and columns are corners; the nominal column is the run's own reference corner. A cell is coloured only when it breaks its limit or sits close to it, and a corner with no value for a row shows a dash. Every value is read from the run: nothing is interpolated. Compare nominal shows each corner as its signed difference from the nominal value, and Export CSV writes the table. A corner run with no outputs needs measurements added to the plan's outputs.

05 / PARAMETER TUNING

Adjust design variables against the last run.

Tuning needs at least one design variable in the active plan. The page puts the variable editors on the left and the response they produce on the right, with the limits judged against both the baseline and the candidate run. It tracks when the displayed result no longer matches the tuned values. Commit tuned parameters writes the values to the plan and queues the candidate runs; results are stale until you rerun.

06 / OPTIMIZATION

The optimizer's iterations.

Configure optimization opens the plan's Optimization analysis, where the bounded variables, the objective and the algorithm are set, and Run optimization runs it. The table has one row per iteration: the cost first, then each design variable the optimizer moved. The toolbar names the best iteration and whether the optimizer converged.

07 / SAFE OPERATING AREA

Every device check one run evaluated.

Add SOA checks to the plan, then run them (Edit SOA checks and Run SOA checks are on the page's toolbar). The Electrical SOA table has one row per evaluated check, worst first: violations, then checks near their limit, then the rest by how little margin they have. Clicking a row selects that device on the schematic. A check the run did not evaluate is not listed.

08 / REGRESSION

A run against a locked baseline.

  1. Choose Select baseline to pick the baseline run of the same plan.
  2. Choose Run regression. The toolbar reads, for example, Run 12 vs baseline Run 9 · 14 / 15 pass.
  3. Select a target to see its tolerance rule and, for a waveform, its comparison plot.
  4. Export JUnit / TAP writes the verdicts for a CI system.

Each check is anchored to a content digest of the baseline run and the tolerance rule in force when it was compared. A target whose rule was deleted, whose waveform is unusable, or whose record no longer matches is reported as a coverage gap, never counted as a pass; Remove orphaned rules clears rules whose target has gone.