# Audio comparison of retained CDs

All 14 audio tracks on each of the three retained CDs were compared as stereo 16-bit PCM at 44,100 Hz, allowing track offsets.

## December 1993 and September 1994

These contain effectively the same digital audio. Track lengths match exactly. September starts 109–127 stereo sample frames earlier (approximately 2.5–2.9 ms). After alignment, 99.8464–99.9981% of individual samples match exactly per track; interior waveform correlation exceeds 0.99999998 for every track. The remaining small sample differences do not justify a second FLAC collection. No gain correction was needed.

## March 1994

This is not the same digital transfer. All track lengths differ, and waveform differences persist within the music. A single sample offset does not reconcile the tracks, nor do the local comparisons show a simple constant gain change. This warrants retaining its audio separately. These measurements do **not** establish whether this is an alternate mix, an analogue transfer, lossy processing, or another alteration; that requires further investigation/listening. In particular, low whole-track correlation is not evidence of a different composition when timing varies.

| Track | December / September duration (s) | March duration (s) |
|---|---:|---:|
| 02 | 293.440 | 292.893 |
| 03 | 278.613 | 277.667 |
| 04 | 312.067 | 314.040 |
| 05 | 267.187 | 267.627 |
| 06 | 286.093 | 288.067 |
| 07 | 288.760 | 288.427 |
| 08 | 306.147 | 307.733 |
| 09 | 282.453 | 283.973 |
| 10 | 101.227 | 101.960 |
| 11 | 101.227 | 101.960 |
| 12 | 239.213 | 241.213 |
| 13 | 66.453 | 69.600 |
| 14 | 281.400 | 282.867 |
| 15 | 66.453 | 69.600 |

December FLACs are in `extracted/cd_1993_12/audio/`; March FLACs are in `extracted/cd_1994_03/audio/`. Both use FLAC level 8. March files are decoded and verified against the source PCM after encoding. September remains available losslessly in its original BIN without duplicate FLACs.

[Detailed measurements](retained-cd-audio-comparison.json). Reproduce with `python3 tools/compare_retained_cd_audio.py` (NumPy and SciPy). Local alignment scores and gains are exploratory measurements, especially for March; they are not authoritative speed or mastering estimates.

## Jurassic Park TFX slideshow preview

The Jurassic Park CD's TFX slideshow uses its audio track 12. This is the existing **TFX track 6**, with the September 1994 digital master: every overlapping sample is identical after a 774-frame alignment. December is also identical in the musical interior, with only very small final-boundary differences. The Jurassic Park track cut is about two seconds shorter, mostly omitting trailing silence.

The redundant preview FLAC and source Jurassic Park CD image set were discarded after verification. The slideshow gallery reuses our existing December track 6 recording.

[Existing TFX track 6 FLAC](../extracted/cd_1993_12/audio/track-06.flac) · [Detailed comparison and source hashes](jurassic-park-preview/audio-comparison.json) · [Retained preview and early cockpit screenshots](jurassic-park-preview/README.md)