fix(pc): Phase 4.4 — virtual-clock HIL sims, real physics fixes, HSMS close sentinel - #172
Merged
Merged
Conversation
… close sentinel The soak suite's wall-clock coupling had been masking a stack of real defects; giving the sims a virtual clock surfaced and fixed them all. Virtual clock (both HIL drivers, time_acceleration param, default 1.0): - rtp_hil: _update_simulation now accumulates a residual instead of int-truncating elapsed/dt and resetting the clock — the old code silently DISCARDED time on any poll faster than one timestep, so sim progress depended on caller cadence (the source of the flakiness). Recipe/segment elapsed times run on the same accelerated clock. - ion_implant_hil: dose integration and beam jitter use accelerated elapsed; at 1.0 production behavior is unchanged. Physics defects exposed and fixed once the sim actually ran 12 hours: - Power budget: losses were computed over the FULL wafer area in every zone (4x double-count) while each zone got 1/4 of a 10 kW bank — the plant saturated ~450C and could never reach a 1000C setpoint. Zone loss area is now area/num_zones and the bank is 10 kW/zone. - Zone integrator applied a dimensionally-wrong dt^2/tau double-lag (response scaled with the timestep, ~25x too sluggish). Temperature now integrates directly; the zone time constant models the LAMP's first-order response, where the lag physically belongs. - PID anti-windup: integral clamped to actuator authority. - set_target_temperature honors ramp_rate_C_per_s via setpoint profiling (was silently ignored). - set_target_temperature no longer clobbers RUNNING_RECIPE status — is_running was False from the first poll of every recipe. - get_recipe_progress preserves the last completed run's summary — natural completion cleared _run_id, so finished recipes reported progress_pct 0.0 and the stress test scored 0% completion. - 72h test: per-recipe timeout now counts ramp AND dwell durations (dwell-only meant ramp-dominated recipes always "timed out") with a 300-sim-second cool-to-ambient floor; poll tightened (0.5s real was 500 simulated seconds). HSMS transport (the Phase 4.4 equipment_manager skip): - Reader-task EOF and close() now push a None sentinel through the data queue; recv_data_message wakes deterministically and raises HsmsConnectionError instead of hanging until cancelled. Queued messages still drain first; the sentinel re-arms for later getters. - test_equipment_handler_exits_when_connection_closes un-skipped: 5/5 consecutive passes. ci.yml: corrected the stale Phase 4.1 comment claiming physics files were excluded (no exclusion mechanism ever existed; all 441 run green). Proof: soak suite 6/6 twice consecutively (was 1 pass, 4 xfail, 1 xpass); RTP 12h holds 1000C with 1.5C pyro std, 0.04C mean control error; recipe completion 3/3; pc unit+integration 460 passed. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
Phase 4.4 was the tracked "physics reconciliation" debt. Investigation showed the soak tests' wall-clock coupling had been masking a stack of real defects — in the simulators, the recipe engine, and the HSMS transport. This PR gives the HIL sims a true virtual clock and fixes everything that surfaced once 12 simulated hours actually ran.
Root cause chain
_update_simulationint-truncatedelapsed/dtand reset the clock, so any poll faster than one timestep advanced the sim zero steps while consuming the time. Sim progress depended on caller cadence → the pass/fail flips between identical runs.dt²/τdouble-lag (physics ~25× too sluggish, timestep-dependent). The lag now lives on the lamp (actuator), temperature integrates directly.set_target_temperatureignored the requested ramp rate (now honored via setpoint profiling) and clobberedRUNNING_RECIPEstatus — every recipe reportedis_running=Falsefrom the first poll._run_id, making finished recipes report 0% progress — the stress test scored 0/3 on recipes that had completed.recv_data_message— the handler hung until cancelled (the skipped equipment_manager test). A close sentinel through the data queue makes shutdown deterministic.Plus: PID anti-windup, and the stale CI comment claiming physics files were "excluded" (they run — all 441 — and pass) corrected.
Proof
time_accelerationdefaults to 1.0 — production behavior unchanged🤖 Generated with Claude Code