What we tested
We generated three 22,050 Hz, 16-bit mono WAV files. Each contained the same 2.400-second A4 sine tone at 440 Hz and 0.55 peak amplitude, with 0.250 seconds of leading silence and 0.500 seconds of trailing silence.
The reference tone was uninterrupted. The other files replaced exactly 40 ms or 160 ms at the center of the tone with digital silence. Phase, pitch, sample rate, level outside the gap, converter controls, and every other input setting stayed fixed.
| Condition | MIDI events | Main A4 layout | MIDI size |
|---|---|---|---|
| No dropout | 2 | 0.221 s for 2.358 s | 67 bytes |
| 40 ms dropout | 3 | 0.244 s for 1.219 s; 1.462 s for 1.127 s | 76 bytes |
| 160 ms dropout | 4 | 0.255 s for 1.115 s; 1.660 s for 0.930 s | 86 bytes |
A 40 ms gap was enough to split the held note
The uninterrupted source produced one 2.358-second main A4. Its second event was a 0.070-second A4 tail fragment at 2.591 seconds.
With 40 ms of center silence, the converter ended the first A4 at about 1.463 seconds and began the second at 1.462 seconds. The same 0.070-second tail fragment remained, bringing the event count from two to three.
The longer gap added another short fragment
The 160 ms condition produced a 1.115-second A4 before the dropout and a 0.930-second A4 after it. It also produced a 0.140-second A4 starting at 1.521 seconds, between those two main sections, plus the same 0.070-second tail fragment.
All detected events stayed on A4. The dropout changed segmentation, duration, velocity, and file size rather than producing a different pitch in this run.
What to check before conversion
- Listen for edits and mutes. A tiny silent cut inside a sustained note may become a new note boundary.
- Inspect the waveform. Zoom into suspicious breaks instead of assuming they are rests in the performance.
- Review repeated pitches in the piano roll. Adjacent notes at the same pitch may be one interrupted source note.
- Repair only when appropriate. Close an accidental gap in the audio or merge the MIDI notes later; keep intentional articulation separate.
How this differs from repeated notes
A repeated-note test begins with separate performed notes and asks whether the converter keeps them separate. This test begins with one sustained oscillator and inserts silence into its middle. The output may look similar, but the source and the cleanup decision are different.
Limits of this result
This test does not establish a universal dropout threshold. It does not cover compressed streaming audio, clicks at edit points, changing background noise, real instruments, chords, mixed music, or other transcription systems. It records three exact outputs from the current local converter.