Brandon, listening to the tab against the record: “it’s like the AI Songsterr puts a cymbal or hi-hat hit on every beat even when there’s not.” This page measures that against the tab’s own note data. The verdict on the notation is that the description is accurate. The verdict on the recording is still open, and the last section says exactly why.
| MIDI | Lane | Notes | Share | Family |
|---|---|---|---|---|
| 36 | Kick | 203 | 31.9% | |
| 51 | Ride | 201 | 31.6% | cymbal / hat |
| 38 | Snare | 119 | 18.7% | |
| 45 | Tom low | 41 | 6.4% | |
| 49 | Crash | 39 | 6.1% | cymbal / hat |
| 43 | Floor tom hi | 14 | 2.2% | |
| 42 | Closed hat | 11 | 1.7% | cymbal / hat |
| 48 | Tom hi-mid | 5 | 0.8% | |
| 47 | Tom low-mid | 3 | 0.5% |
Ride 51 carries 201 notes. Closed hat 42 carries 11. The generator picked one cymbal and parked on it for the length of the piece.
One tick per bar, left to right, bar 1 to bar 99. Hover any tick for its numbers.
Zappa marked this piece with capital-letter sections in his own scores, reproduced in the Hal Leonard One Size Fits All guitar book and catalogued by Kasper Sloots. Reading those bar ranges and timestamps as 3/4 returns 116.5, 114.5, 117.2, 122.7 and 116.8 BPM per section against the tab’s stated 116. Two independent sources land on the same meter and tempo, which is what licenses every number on this page.
| Section | Bars | Starts | Beats | With cymbal | Saturation | Full bars | Ride | Hat | Crash |
|---|---|---|---|---|---|---|---|---|---|
| A main theme | 1–22 | 0:00 | 66 | 61 | 92.4% | 20/22 | 55 | 0 | 6 |
| B F-Em-Dm-G | 23–36 | 0:34 | 42 | 37 | 88.1% | 12/14 | 37 | 0 | 0 |
| C variation on the main theme | 37–64 | 0:56 | 84 | 77 | 91.7% | 25/28 | 54 | 11 | 14 |
| D steady G pedal, G Mixolydian | 65–79 | 1:39 | 45 | 39 | 86.7% | 11/15 | 36 | 0 | 3 |
| E outro, ends on a sustained A | 80–103 | 2:01 | 69 | 34 | 49.3% | 7/20 | 19 | 0 | 16 |
Positions are in quarter notes from the barline. K kick, S snare, RD ride, CR crash, hh closed hat.
| Count | Bar content | Note |
|---|---|---|
| 21× | 0 K+RD | 1 RD | 2 S+RD | 3/2 K | 5/2 K | cymbal on every beat |
| 10× | 0 K+CR | 1 RD | 2 S+RD | 5/2 K | cymbal on every beat |
| 9× | 0 K+RD | 1 RD | 2 S+RD | 5/2 K | cymbal on every beat |
| 5× | 0 K+RD | 1 RD | 2 S+RD | 3/2 K | cymbal on every beat |
| 4× | 0 K+CR | 1 hh | 2 S+hh | 5/2 K | cymbal on every beat |
| 3× | 0 K+RD | 1 RD | 2 K+RD | cymbal on every beat |
The five most repeated bars are the same idea five ways: a cymbal on beat 1, beat 2 and beat 3. Between them they account for 49 of the 99 bars. The single most common bar, kick+ride, ride, snare+ride, appears 21 times. Across the whole track there are 46 distinct bar patterns for 99 bars.
The gap between one cymbal onset and the next, over all 251 of them.
| Gap, quarter notes | Count | Share | |
|---|---|---|---|
| 1 | 230 | 92.0% | |
| 4 | 4 | 1.6% | |
| 3/2 | 4 | 1.6% | |
| 2 | 4 | 1.6% | |
| 3 | 2 | 0.8% | |
| 6 | 2 | 0.8% | |
| 5 | 1 | 0.4% |
230 of 250 gaps are exactly one quarter note. A drummer playing a slow waltz leaves gaps of several sizes across a piece. This one holds a single size for 92% of it.
Songsterr writes signature into a bar only when the meter changes.
Reading each bar independently and defaulting to 4/4 reports this piece as 93 bars of 4/4
and gives 63.4% saturation with 1 fully saturated bar. Carrying the signature
forward gives 89 bars of 3/4, 81.0% saturation and 75 fully saturated bars.
Same file, same script, opposite verdict, no error raised.
Two checks catch it. Every bar’s durations must sum to its bar length, and the summed bar lengths at the stated tempo must land on the record’s runtime. The correct reading gives 2:38.3 against a 2:39 track. The wrong one gives 3:22.
The source the tab names, mgIK23Karm4, is pulled and separated. The audio runs 159.02 s and the tab computes to 158.3 s, a 0.45% difference, so the 3/4-at-116 reading is now confirmed against the record itself rather than only against Zappa’s score.
| Threshold | Onsets found | Per second |
|---|---|---|
| 0.04 | 353 | 2.22 |
| 0.06 | 302 | 1.90 |
| 0.08 | 234 | 1.47 |
| 0.10 | 193 | 1.21 |
| 0.14 | 151 | 0.95 |
| Section | Window | Hi-band RMS | Against track average | Tab saturation |
|---|---|---|---|---|
| A main theme | 0-34 s | 0.12646 | 1.02 | 92.4% |
| B | 34-56 s | 0.12266 | 0.99 | 88.1% |
| C variation | 56-99 s | 0.11438 | 0.93 | 91.7% |
| D G pedal | 99-121 s | 0.07278 | 0.59 | 86.7% |
| E outro | 121-158 s | 0.15254 | 1.24 | 49.3% |
Every number here measures the notation. None of them measures Chester Thompson. Proving that a written cymbal is absent from the record needs the recording split into separate cymbal lanes. A mixed Demucs drums stem holds kick, snare, toms, hat, ride and crash together and cannot say which one was struck; on Cosmik Debris that same shortcut produced two wrong verdicts that the six-lane split overturned.
Status. The pull and the Demucs drums stem are done. The six-lane DrumSep split is running and checkpoints per 60-second block. The last step, scoring all 248 written cymbal beats against ride.wav and hh.wav, waits on it. Until then no claim is made about which cymbal Chester Thompson struck.
| What | Where |
|---|---|
| Tab metadata and per-track notes | songsterr.com/api/meta/6822181 → dqsljvtekg760.cloudfront.net/6822181/8906238/<hash>/3.json |
| Live editor export | frank zappa-Sofa No. 1-09-05-2026.gp |
| Section letters, bar ranges, timestamps | Hal Leonard One Size Fits All guitar book, via Kasper Sloots, zappa-analysis/one-size-fits-all.htm |
| Drum credit for the album take | MusicBrainz recording |
| Analysis scripts | ~/Projects/_outputs/zappa-sofa-no1-cymbal-audit/analysis/ |
| Related audit | Cosmik Debris hi-hat audit, Zappa drum transcription inventory |