What we tested
We generated three 22,050 Hz, 16-bit mono WAV files containing the same isolated C4, E4, G4, and C5 sequence. Every note lasted 0.900 seconds, every gap lasted 0.300 seconds, and each waveform measured 0.300 steady-note RMS.
Only waveform shape changed: sine, triangle, or square. Timing, pitch, RMS, envelopes, sample rate, converter controls, and converter code stayed fixed. Different peak levels were required to match RMS: 0.42426 for sine, 0.51961 for triangle, and 0.30000 for square.
| Waveform | MIDI events | Intended pitches retained | Unexpected result |
|---|---|---|---|
| Sine | 5 | C4, E4, G4, C5 | One short C4 tail |
| Triangle | 7 | C4, E4, G4, C5 | Three short same-pitch tails |
| Square | 19 | C4, E4, G4, C5 | Two simultaneous G2 notes |
The sine source stayed closest to four notes
The sine file produced the four intended pitches plus one 0.081-second C4 tail, for five events in a 97-byte MIDI file. Main-note durations ranged from 0.824 to 0.907 seconds.
The triangle source created more note boundaries
The triangle file retained all four intended pitches but produced seven events in 113 bytes. It added 0.093-second fragments after C4, G4, and C5. Its main-event velocities were also higher than the sine results even though steady RMS matched.
The square source fragmented and added lower notes
The square file produced 19 events in 213 bytes. E4 was divided into nine adjacent events, G4 into two, and C5 into five. During G4, the converter also wrote two G2 events lasting about 0.429 seconds each.
A square wave contains strong odd harmonics, unlike a sine wave. This test shows that matching pitch and RMS did not make the converter outputs equivalent; it does not prove which harmonic caused a specific event.
What to do with a bright or distorted source
- Review repeated same-pitch slices. Several adjacent events may represent one sustained source note.
- Check octave errors. Harmonic-rich audio can coincide with lower or higher detected notes that were not performed.
- Try a cleaner isolated stem. Keep the original file, then compare a less distorted version rather than assuming one treatment always wins.
- Keep the useful draft. Delete false notes and merge fragments in a piano roll when the intended pitch is clear.
Limits of this result
These are synthetic periodic waveforms, not piano, guitar, voice, or a mixed track. The test does not rank instruments, define an ideal tone, or predict every distortion effect. It records three exact outputs from the current local converter at default controls.