Interactive experiment

Reproducible fault campaign

A fixed synthetic campaign fixture separates a fault-free control, attempts, unique target-class/time-bin pairs, and overlapping outcomes. The denominator is 12 pairs: four target classes (ROM check_done, debug permission, first fetch, key release) × three time bins (t1, t2, t3). The A-edge to D-edge labels in the row data identify an edge/location, not a target class.

Controls

Run Campaign attempts to run Maximum fault events per attempt

The fixture is not a measured campaign or probability estimate. Outcome categories may overlap; keep the denominator and untouched bins. This is a deterministic teaching fixture: no RTL or simulator ran. The fixture key identifies a sample row; it is not a simulator replay command. The synthetic fixture is not hardware evidence.

Static worksheetStatic denominator: 12 target-class/time-bin pairs = 4 target classes (ROM check_done, debug permission, first fetch, key release) × 3 time bins (t1, t2, t3). In a campaign row, the A-edge, B-edge, C-edge or D-edge part identifies an edge/location, not a target class; the time bin is separate. “Campaign not run” is the zero-attempt, injection-off mode; it is not a completed campaign, and the 12 bins it lists as untouched are unevaluated because no campaign ran; that is not a completed coverage report. The fault-free authorized acceptance example is a separate option and a one-row control record, in which an authorized commit is accepted, an independent reference model (outside the injection targets and their influence cone) marks it authorized, no fault is injected, the implication passes, and the cover is hit; it is a teaching row, not an RTL, simulator, formal, or physical result, and it is counted separately from the 12 campaign rows. Campaign-not-run exports no rows; the control example exports exactly one control record. Campaign rows are a fixed synthetic fixture; count attempted rows, selected unique target-class/time-bin pairs, activated-and-observable unique pairs, untouched pairs (no activated, observable attempt), unauthorized commits, availability failures, and late alerts separately. The attempts slider selects records; the separate per-attempt event budget never changes the attempt count or the 12-pair denominator. When the budget is below a row's required events, the row applies no fault: it shows no effect or trace, its assertion is “not run”, and its planned effect stays recorded only under the planned fields. Repeated pairs are intentional. The synthetic fixture is not hardware evidence. Outcome classes overlap; no RTL or simulator ran.