Peak detection is a matter of measuring the instantaneous amplitude. Set a zero‑crossing reference, calculate the derivative, and when it flips from positive to negative you’ve hit the crest. The exact instant is a phase fraction of a sample—usually a few hundredths of a millisecond. In practice I lock a trigger to that point, then buffer a few cycles to verify the shape. It’s the same as tuning a piano: you listen until the string’s vibration is crisp, then you record it exactly at that moment.

Right, the trigger should fire at the exact zero‑crossing with a margin smaller than one sample. Keep the buffer to a single cycle plus a safety margin; otherwise you’ll average out the peak. At 48 kHz a cycle is about 20 ms, so a 1‑cycle buffer is 20 ms—tight enough to preserve the crest but wide enough to capture context. That’s the sweet spot for matching a piano note.