Hear a Small Timing Offset

Production guides · Listening library · Meme Plugins · 10 November 2026

A short offset between a sound and a copy changes how their samples combine. This experiment keeps the source fixed and compares equal-gain blends with exact delays of 24 and 96 samples at 48 kHz.

The source is a new synthetic four-note phrase. The two processed files contain 0.5 times the dry source plus 0.5 times a delayed copy. They have not been normalised to hide the level changes created by the combination.

Listen to the files

Start at a low listening volume. Play one example at a time. Nothing plays automatically. You can pause with each player or use the button to stop and reset them all.

Original phrase
This is the unchanged mono source.

Download this WAV · 4 seconds · 48 kHz / 16-bit PCM

Dry plus a 0.5 ms copy
The copy starts 24 samples later. Both paths use a gain of 0.5.

Download this WAV · 4 seconds · 48 kHz / 16-bit PCM

Dry plus a 2 ms copy
The copy starts 96 samples later. The blend gains stay the same.

Download this WAV · 4 seconds · 48 kHz / 16-bit PCM

Follow the comparison

Compare the original with each blend, paying attention to tone as well as level. The notes change pitch while the sample offset stays fixed. Describe the particular note or moment that changes rather than assuming the whole file becomes uniformly quieter.

Reproduce the offset

At 48,000 samples per second, 24 / 48,000 × 1000 = 0.5 milliseconds; 96 / 48,000 × 1000 = 2 milliseconds. The generator adds leading zeros to the delayed path and blends the arrays before writing 16-bit PCM. The phrase has silence after the last note, so these short offsets do not remove its final event.

What was verified

The source, offsets and blend gains are explicit in the downloadable recipe. All three WAV files were independently checked as four-second, mono, 48 kHz, 16-bit PCM. Their actual sample peaks are stored in the build manifest, and every encoded sample stays inside the integer range.

Keep the limit clear

Combining related signals with a timing offset can create frequency-dependent reinforcement and cancellation. This is a controlled mono blend, not a measured microphone setup. Moving real drum tracks to one visual reference can change other microphone relationships; this demonstration does not choose an alignment for you.

Source and reproducibility

All audio on this page was newly generated from mathematical oscillators for this batch. No recordings, loops or client audio were used. The complete Python generation recipe (.txt) recreates the four listening experiments with the Python standard library and checks their file metadata with FFprobe. The recipe specifies fade times, note frequencies, transformations and matching gains. These checks verify the stated digital construction; they do not constitute a listening study.

Technical background: Audacity: Mixing audio tracks.


Phase and polarity guide · All production guides