What we tested
We generated a 22,050 Hz, 16-bit mono WAV containing four isolated sine tones: C3, C4, C5, and C6. Each tone lasted 0.900 seconds, with 0.300-second gaps and the same 0.68 peak amplitude. We kept one file unfiltered and processed two copies with second-order low-pass filters at 600 Hz and 300 Hz.
We ran all three files through the current local converter at its default controls, downloaded each MIDI file, and parsed the note events. The converter logic was not changed.
| Source condition | MIDI events | Intended pitches retained | MIDI size |
|---|---|---|---|
| Unfiltered | 7 | C3, C4, C5, C6 | 114 bytes |
| 600 Hz low-pass | 6 | C3, C4, C5, C6 | 105 bytes |
| 300 Hz low-pass | 8 | C3, C4, C5, C6 | 122 bytes |
The cutoff did not become a pitch ceiling
A low-pass filter reduces frequencies above its cutoff gradually. It is not a MIDI rule that rejects every higher note. C5 has a 523.251 Hz fundamental and C6 has a 1046.502 Hz fundamental, yet both notes remained in the 300 Hz output.
The filtering did change their levels. C6 measured 0.48080 RMS before filtering, 0.14879 after the 600 Hz filter, and 0.03885 after the 300 Hz filter. Its MIDI velocity fell from 76 to 61 and then 54, while its detected pitch stayed C6.
Event counts moved in both directions
The 600 Hz result merged the unfiltered C4 fragments into fewer events, reducing the total from seven to six. The stronger 300 Hz filter did not continue that trend: C3 split into three events and C5 split into two, raising the total to eight.
That is why fewer high frequencies do not automatically mean a simpler or more accurate MIDI draft. A filter changes note level, attack shape, and the evidence the pitch model receives. The result still needs to be checked against the source.
How to compare a low-pass edit
- Keep the original. Render a separate filtered copy so the source remains available.
- Use one short passage. Compare identical notes with only the cutoff changed.
- Check quiet high notes. Look for missing notes, lower velocities, shortened ends, or extra fragments.
- Prefer the closer draft. Do not choose a result only because it contains fewer events.
What this test does not prove
It does not define an ideal cutoff, test a real instrument or full mix, or predict other slopes, equalizers, video encodes, or converters. It records this tool's current output for three controlled files and shows that low-pass filtering can change MIDI structure without creating a hard upper note limit.