Fixture 001: Radial GRIN Baseline¶

Purpose¶

Fixture 001 is the compact radial GRIN baseline used to validate that the current curved-ray transport path can launch, render, capture, and report a stable result through the dedicated harness.

This note records the current harness contract and the local run history that supports promoting A2 as the official Fixture 001 baseline.

It also records the first verified repeat-sweep result for Fixture 001 so the project has a concrete example of how to characterize a fixture control surface using only hardened, fully verified runs.

Post-Field-Fix Baseline Refresh (2026-03-26)¶

The r_outer / r_inner field-region correction changed the accepted visual output for Fixture 001 by enforcing zero field acceleration outside the active region. The March 26, 2026 rerun is therefore accepted as the refreshed post-fix comparison baseline even though run_verified = false against the older pre-fix capture.

Accepted post-fix baseline:

Timestamp: 2026-03-26T21-14-34
Output path: output/fixture_runs/fixture_001/2026-03-26T21-14-34
Status: ok
guard_progress exits: 0
forcedAdvance events: 0
processed_rows: 164
traced_pixels: 48864
source_hits: 2364
miss_hits: 46500
turnThreshold: 4

Promotion rationale:

scheduler remained clean
capture completed with full expected row coverage
compare mismatch was against the stale pre-fix baseline, not against the current field-region logic
resulting output is consistent with the corrected active-region cutoff

Current harness¶

Harness script: scripts/run_fixture_001.sh
Reporter: tools/fixture_001_report.py
Scene: res://test-grin-basic-visual-minimal.tscn
Fixture identity expected by launch audit: grin_basic_visual_minimal
Fixture id: fixture_001
Launcher token configured by the harness: run_fixture_001

Default thresholds and capture settings¶

FIXTURE_001_SETTLE_FRAMES=12
FIXTURE_001_MIN_RH_STEP=20
FIXTURE_001_MIN_PROCESSED_ROWS=64
FIXTURE_001_CAPTURE_FILM_OPACITY=1.0
FIXTURE_001_COMPARE_GRID=1
FIXTURE_001_COMPARE_CROSSHAIR=1

What the harness records¶

The report step builds params.json, metrics.json, and summary.txt from the Godot log and capture artifact. The primary baseline signals are:

launch audit status
render transport selection
traced pixel count
processed row count
capture success
scheduler guard exits and forced advances, as seen in run.log

Verified-control contract¶

Treat a Fixture 001 run as usable for control-surface interpretation only when all of these verification checks are true:

runtime_fingerprint_present = true
assembly_timestamp_present = true
effective_step_matches_requested = true
row_diagnostics_present = true
scheduler_clean = true

This is the current baseline procedure for validating Fixture 001 controls:

Launch only through scripts/run_fixture_001.sh.
Rebuild against the hardened Windows runtime mirror before each run.
Generate params.json, metrics.json, summary.json, and ledger append through the existing report flow.
Interpret only runs that remain fully verified after report generation.

Verified characterization update (2026-03-22)¶

A verified top-config repeat sweep was run on March 22, 2026 through the hardened Windows runtime mirror at /mnt/c/godot/godot_xPRIMEray.

The repeat set used three verified runs each for:

stepLength=0.040, errorTolerance=0.012, turnThreshold=2.0
stepLength=0.040, errorTolerance=0.010, turnThreshold=2.4
stepLength=0.040, errorTolerance=0.010, turnThreshold=3.2

For the turnThreshold=3.2 branch, errorTolerance=0.010 was used explicitly as the nearest baseline-compatible path already established in prior Fixture 001 work.

Repeatability summary¶

Config	mean hit_rate	median hit_rate	std dev hit_rate	mean traced_rate	std dev traced_rate	robust_score	verified runs
`0.040 / 0.012 / 2.0`	`109.654/s`	`110.781/s`	`1.814/s`	`2361.923/s`	`22.902/s`	`109.874/s`	3
`0.040 / 0.010 / 2.4`	`109.540/s`	`108.830/s`	`1.159/s`	`2373.456/s`	`4.672/s`	`108.251/s`	3
`0.040 / 0.010 / 3.2`	`112.289/s`	`112.484/s`	`3.249/s`	`2404.632/s`	`50.686/s`	`110.860/s`	3

Topology call¶

Fixture 001 verified characterization indicates a stable local basin centered near stepLength=0.040, errorTolerance=0.010, turnThreshold≈3.2.

The best verified operating point in current testing is 0.040 / 0.010 / 3.2, with 0.040 / 0.010 / 2.4 as a lower-variance fallback.

This is best described as a stable basin with a mild ridge toward turnThreshold=3.2, not a fragile single-run spike.

Why this matters¶

The top verified point was not isolated; neighboring verified settings stayed in the same performance band.
All nine repeat runs stayed fully verified, so the result is not tied to a build/runtime mismatch or scheduler regression.
This gives Fixture 001 a usable state-of-the-art reference for future fixture-note writeups: record the winner, the fallback, and the basin shape.

Radial profile interpretation¶

Fixture 001 optimization improves the same outer-mid annular hit band rather than shifting or broadening it. The best verified config, 0.040 / 0.012 / 2.0, increases localized hit conversion in bins 5 and 6 without changing rendered coverage.

Historical run history¶

The current local Fixture 001 history under output/fixture_runs/fixture_001 contains six recorded runs from March 19-20, 2026.

Label	Timestamp	StepsPerRay	Status	guard_progress exits	forcedAdvance	processed_rows	traced_pixels	Notes
A0	`2026-03-19T22-08-00`	500	ok	18	18	164	39200	First functional captured baseline, but heavily scheduler-limited
A1	`2026-03-19T22-49-38`	500	ok	7	7	164	24320	Improved over A0, still scheduler-limited
A1.1	`2026-03-19T22-55-23`	500	ok	10	10	164	27616	Additional 500-step run, still scheduler-limited
A2	`2026-03-19T22-57-53`	350	ok	0	0	164	56288	First scheduler-clean capture in the current run history
A3	`2026-03-20T18-26-33`	300	ok	1	1	172	71744	Single-parameter variant from A2; higher coverage and traced pixels, but scheduler guard regressed
A4	`2026-03-20T18-29-51`	400	ok	0	0	164	40608	Single-parameter variant from A2; scheduler-clean, but lower traced pixel count than A2

StepsPerRay Sweep Comparison¶

The completed single-parameter StepsPerRay sweep around the current baseline uses the following runs:

Label	StepsPerRay	guard_progress	forcedAdvance	traced_pixels	processed_rows	runtime
A3	300	1	1	71744	172	29.458s
A2	350	0	0	56288	164	28.825s
A4	400	0	0	40608	164	29.777s

Baseline Decision¶

A2 is now the official Fixture 001 baseline.

It remains the strongest balance of stability and output quality in the current local run history:

300 produced more traced pixels, but it regressed scheduler cleanliness with one guard_progress exit and one forcedAdvance=1 event.
400 stayed scheduler-clean, but it reduced traced_pixels from 56288 to 40608 and increased runtime from 28.825s to 29.777s.
350 preserves scheduler cleanliness while keeping stronger output quality than the stable 400 variant, making it the best current balance.

A3 variant result¶

Label: A3
Timestamp: 2026-03-20T18-26-33
Scene: res://test-grin-basic-visual-minimal.tscn
Fixture: grin_basic_visual_minimal
Launcher observed in run log: manual
Transport: GRIN_Optical
StepsPerRay: 300
Status: ok
Capture succeeded: true
guard_progress exits: 1
forcedAdvance events: 1
ready_frames: 12
render_health_step: 41
processed_rows: 172
traced_pixels: 71744
source_hits: 3361
miss_hits: 68383

Compared with A2, A3 only changes StepsPerRay from 350 to 300. The run still captures successfully, but it is no longer scheduler-clean because one guard_progress exit and one forcedAdvance=1 yield abort appear in run.log.

A4 variant result¶

Label: A4
Timestamp: 2026-03-20T18-29-51
Scene: res://test-grin-basic-visual-minimal.tscn
Fixture: grin_basic_visual_minimal
Launcher observed in run log: manual
Transport: GRIN_Optical
StepsPerRay: 400
Status: ok
Capture succeeded: true
guard_progress exits: 0
forcedAdvance events: 0
ready_frames: 12
render_health_step: 41
processed_rows: 164
traced_pixels: 40608
source_hits: 1928
background_hits: 6411
miss_hits: 32269

Compared with A2, A4 only changes StepsPerRay from 350 to 400. The run remains scheduler-clean, but traced pixel count drops from 56288 to 40608, so A2 remains the stronger baseline in the current local history.

Official Baseline¶

Label: A2
Timestamp: 2026-03-19T22-57-53
Scene: res://test-grin-basic-visual-minimal.tscn
Fixture: grin_basic_visual_minimal
Launcher observed in run log: manual
Transport: GRIN_Optical
StepsPerRay: 350
Status: ok
Capture succeeded: true
guard_progress exits: 0
forcedAdvance events: 0
ready_frames: 12
render_health_step: 41
processed_rows: 164
traced_pixels: 56288
source_hits: 2717
miss_hits: 53571

Accepted Post-Fix Comparison Baseline¶

Label: A2-post-field-fix
Timestamp: 2026-03-26T21-14-34
Scene: res://test-grin-basic-visual-minimal.tscn
Fixture: grin_basic_visual_minimal
Transport: GRIN_Optical
Status: ok
Capture succeeded: true
guard_progress exits: 0
forcedAdvance events: 0
processed_rows: 164
traced_pixels: 48864
source_hits: 2364
miss_hits: 46500

This accepted comparison baseline supersedes the older pre-fix capture for future post-r_outer / r_inner validation work.

Next Study Direction¶

Hold A2 fixed as the reference configuration and run the next single-parameter micro-sweep on StepLength.

Recommended rationale:

StepsPerRay has now been bracketed at 300, 350, and 400.
StepLength is the next most direct ray-march control that can change both scheduler behavior and output coverage while leaving the established A2 baseline intact.
A focused StepLength sweep should make it easier to see whether A2 can gain more traced coverage without giving back scheduler cleanliness.

Acceptance criteria for baseline reuse¶

Treat Fixture 001 as passing the baseline when the run preserves these checks:

status = ok
capture_succeeded = true
launch audit status = ok
guard_progress exits = 0
forcedAdvance events = 0
processed_rows >= 164
traced_pixels > 0

Notes¶

The recorded run history shows that reducing stepsPerRay from 500 to 350 is what coincides with the first scheduler-clean capture in the current local evidence set.
The current history was gathered from existing local artifacts, not by launching a fresh run during this doc update.