PER_PART_TILING.md

music/complete_tracks/PER_PART_TILING.md

THE PER-PART TILING TABLE — every layer measured alone, worst layer reported

CANON SUBORDINATION — this document is PROPOSAL-TIER: it serves canon and never outranks it.
The canon: the CVD · the T1 foundation docs · the T0 registries (registries/) · the spine
(docs/spine/CH_*.md) · the region pages (_source/02_Tier_2_Region_Pages/). Authority order: docs/DOC_MAP.md § 0.
Canon served: T0_Theme_Registry · T0_Motif_Registry.
If this document disagrees with canon, CANON WINS and this document is the defect.

Emitted by harness/music_gen/anti_tiling.py, measured on the SHIPPED MIDI in

build/audio/complete_tracks/tracks/. Nothing here is read off a generation record.

0. Why this ruler exists

The lane's three earlier verbatim-repeat rulers all reduce a bar to an ENSEMBLE signature.

An ensemble signature says two bars differ the moment any single part differs, so a pad that

tiles four bars across half a track reads CLEAN as long as the flute above it moves. A

listener does not hear the sum; a listener hears the layer. **The sameness mechanism was

broken by measurement, not by composition** — and the ensemble_concealment proof below

reproduces the concealment on demand.

Refusal threshold: 8 bars, per part, no role exempt. That is the § 1.3

number the lane already refuses on the ensemble, applied where the ensemble was hiding it.

1. The mutation proofs

mutationexpectedmeasuredpasswhat it proves
tiled_8x288PASSthe § 1.3 defect (bars 9-16 verbatim repeat bars 1-8) is measured at its own length
transposed00PASSLEIT-03 development is not tiling; a transposed restatement must read CLEAN
louder_copy88PASSvelocity cannot launder a repeat; the same bar played louder is the same bar
repeated_bar_in_silence11PASSrests are not tiles; a tooth that counts silence CONCEALS the runs that matter
clean_control00PASSthe positive control: a genuinely non-tiled part reads 0
ensemble_concealment88PASSTHE BROKEN MECHANISM, reproduced: a pad tiling one bar sixteen times under a flute that never repeats reads 0 on the ensemble ruler and 8 on this one

2. The twenty tracks, worst part first

worst run is the longest verbatim run of bars inside a single part. worst cov is the

share of that part's sounding bars sitting inside a repeated run of four bars or more —

the proportion a listener means by "it is all the same". ens is what the OLD ruler

reported, carried here only so the gap between the two columns is on the record.

trackarmbarspartsworst partworst runworst covens (old ruler)verdict
HF_MT_01_TITLEA9612counter :: cello_ens.sus_vib, trombone.sus625.8%0GREEN
HF_MT_02_EXPLORATION_TRAVERSALB9610frame:cello [cello_ens.sus_quiet]652.2%0GREEN
HF_MT_03_BATTLE_ORDINARYA10011perc_hi :: clash_cymbals30.0%0GREEN
HF_MT_04_BOSS_CHAPTERB9610perc:bass_drum [bass_drum]1875.6%2REFUSED
HF_MT_05_BOSS_APEXA17612perc_mid :: bass_drum749.1%0GREEN
HF_MT_06_SANCTUARY_RESTB884harp:left_hand [harp]1676.9%4REFUSED
HF_MT_07_TRADE_AND_FESTIVALA9610perc_low :: bass_drum527.1%1GREEN
HF_MT_08_SORROW_LOSSB568strings:cello_divisi_1 [cello_ens.sus_vib]644.4%0GREEN
HF_MT_09_WONDER_DISCOVERYA727harmony0 :: violin_ens.sus_quiet30.0%0GREEN
HF_MT_10_CREDITSB10819perc:timpani [timpani.hit]983.3%2REFUSED
HF_MT_11_REGION_FLORESA12811perc_low :: bass_drum, toms.low633.6%0GREEN
HF_MT_12_REGION_BALIB13610drum:answer [toms]1987.8%4REFUSED
HF_MT_13_REGION_SUMATRA_JAVAA808harmony0 :: cello_ens.sus_quiet30.0%0GREEN
HF_MT_14_REGION_SRI_LANKAB766strings:contrabass [contrabass.sus_quiet]834.9%0REFUSED
HF_MT_15_REGION_SOUTH_INDIAA14010perc_low :: bass_drum, toms.low636.4%0GREEN
HF_MT_16_REGION_SWAHILI_COASTB686perc:frame_drum_implied [frame_drum]877.8%0REFUSED
HF_MT_17_REGION_ETHIOPIAA12411bass :: cello_ens.sus_quiet519.0%0GREEN
HF_MT_18_REGION_CENTRAL_AFRICAB10010perc:clapped_sticks [claves]1895.1%5REFUSED
HF_MT_19_REGION_WEST_AFRICAA12811perc_low :: bass_drum, toms.low59.4%0GREEN
HF_MT_20_REGION_SOUTH_AFRICAB10010perc:hand_clapping [claps]41100.0%2REFUSED

12 of 20 green on the per-part ruler; 8 refused.

3. Every failing part, named

HF_MT_04_BOSS_CHAPTER

partnotessounding barsdistinct bar sigsruncoveragefirst instance
perc:bass_drum [bass_drum]804581875.6%bars 68-85 restated at 69-86

HF_MT_06_SANCTUARY_REST

partnotessounding barsdistinct bar sigsruncoveragefirst instance
harp:left_hand [harp]403991676.9%bars 54-69 restated at 70-85

HF_MT_10_CREDITS

partnotessounding barsdistinct bar sigsruncoveragefirst instance
perc:timpani [timpani.hit]12122983.3%bars 77-85 restated at 78-86
brass:trombone [trombone.sus]272210831.8%bars 38-45 restated at 44-51

HF_MT_12_REGION_BALI

partnotessounding barsdistinct bar sigsruncoveragefirst instance
drum:answer [toms]1325471985.2%bars 12-30 restated at 13-31
drum:lead [frame_drum]342118181884.7%bars 12-29 restated at 13-30

HF_MT_14_REGION_SRI_LANKA

partnotessounding barsdistinct bar sigsruncoveragefirst instance
strings:contrabass [contrabass.sus_quiet]605325828.3%bars 6-13 restated at 12-19

HF_MT_16_REGION_SWAHILI_COAST

partnotessounding barsdistinct bar sigsruncoveragefirst instance
perc:frame_drum_implied [frame_drum]42185877.8%bars 50-57 restated at 51-58

HF_MT_18_REGION_CENTRAL_AFRICA

partnotessounding barsdistinct bar sigsruncoveragefirst instance
perc:clapped_sticks [claves]1124151895.1%bars 73-90 restated at 75-92
perc:drum_sequence [toms]602141790.5%bars 74-90 restated at 76-92

HF_MT_20_REGION_SOUTH_AFRICA

partnotessounding barsdistinct bar sigsruncoveragefirst instance
perc:hand_clapping [claps]1176244198.4%bars 48-88 restated at 50-90
perc:leg_rattles [shaker.large]25634334100.0%bars 36-69 restated at 64-97
perc:skin_drum [frame_drum]321731488.2%bars 76-89 restated at 77-90
voice:2_stand_in [clarinet.sus]10255261378.2%bars 19-31 restated at 31-43
voice:1_stand_in [oboe.sus_vib]604020865.0%bars 21-28 restated at 29-36

Generated by harness/site/structure_site.py — the URL path is the repo path. review root