CUT-A-DOR 1.0 SOUND BANK AUDIT — SOFTWARE CHECKS
48 unique, additional source recordings; 48 unique source SHA-256 values; no factory source URL reused.
3 pinned CC0 repositories; Git blob hash and source size verified for every downloaded hit.
All source recordings: zero clipped samples. All prepared PCM: first/last frame zero, peak 26000.
Largest decoded peak: 29024/32768; minimum waveform SNR: 14.48 dB.
Loudest retained tail immediately before taper: -23.0 dB relative to source peak.
A heavily clipped retro kick, a two-sample clipped snare and a long noisy shaker were rejected and replaced.
Compact processing: no pitch-shift duplicates, no synthesized layers; antialiased mono resampling and shaped tails.
Bank output reproduced byte-for-byte in a second, separate output directory.
PASS 340 checks; 48 sounds validated, mapped and rendered; 24 all-16-pad FX stress runs; no sanitizer errors.
Native checks compiled against the actual firmware cad.c with AddressSanitizer + UndefinedBehaviorSanitizer.
Browser editor: exact encode match, pack/unpack roundtrip, decoding, category mapping and bad/truncated files checked.
No device flash, physical MIDI transfer, live CPU-load measurement or human listening approval was performed here.

BANK MEMORY (includes 512-byte header; limit 81920 each):
DUSTY ROOM: 71264 B; 16 hits; 10656 B remaining; SHA-256 548f200712b3294ec4d489bd133658b62e86a89d68753f407da5c1d1873bdd77
VELVET CUT: 77600 B; 16 hits; 4320 B remaining; SHA-256 e85ac66eb0bbb1e876384a359e2969a93aa717da55cbf4b5ca1875e13d6950ad
BREAK PIT: 75884 B; 16 hits; 6036 B remaining; SHA-256 6f6a2d98c5bd54f782b3e13fd86063d4e54209c3825df4adb22ebd67685efc26
Total stored: 224748 B; total spare across three slots: 21012 B.
All 48 zones are occupied; spare bytes allow longer/replacement sounds, not a 17th sound in any one bank.
No increase to firmware application image or fixed voice count is needed for the bank assets themselves.
Kit assignment uses both samples of every category, without falling back to factory sounds.
