StakdTrax ENFR Download

How it works

How StakdTrax keeps several iPhones in step

Most “sync recording” apps try to start every phone at the same instant. That can't work, so StakdTrax doesn't try. Here is what it does instead, in plain terms.

Record is a timestamp, not a button press

An iPhone's audio engine takes an unpredictable moment to start, tens of milliseconds and different every time. Two phones told to start together will never actually start together.

So every phone starts listening when the session is ready, before anyone presses Record, and keeps listening. Every slice of audio it captures is stamped with the phone's clock. When the host presses Record, that becomes a marker on the shared timeline. Starting together turns from an impossible hardware problem into a timestamping problem, and timestamps can be corrected.

Three ways phones disagree

  1. Clock offset. Two phones' clocks don't read the same time. The phones exchange timing messages over the local link, keep the quickest round trips, and agree to within roughly half a millisecond to two milliseconds over Wi-Fi.
  2. Clock drift. Two phones don't count time at quite the same speed. A difference of 30 parts per million sounds tiny, but over a ten-minute take it adds up to 18 ms, enough to hear as a smear on a snare line. StakdTrax measures the drift across the take and corrects it when the stack is built.
  3. Audio-clock drift. The microphone's sample clock is separate again. A phone recording at a nominal 48 000 Hz might really be running at 47 999.7 Hz. StakdTrax compares samples recorded against time passed and resamples each track so it stays in step from the first bar to the last.

Huddle, calibrate, scatter

Phones held bottom to bottom while the host plays a sweep
The huddle: phones held close together while the host plays a short sweep.

Sub-millisecond alignment needs one more step. Sound takes about 3 ms to cross a metre of room, so a sweep heard by phones spread across a hall measures two things at once: how far apart their clocks are, and how far each phone is from the speaker. There's no way to separate the two without knowing the distances.

So the phones are held together first, within 30 cm, which keeps the travel time under a millisecond. The host plays a short sweep, every phone hears it at effectively the same moment, and the difference between their recordings is the clock offset alone, measured to a fraction of a sample. Then everyone spreads out to record.

It takes a few seconds. Calibrating, step by step.

After the take

When the phones send their tracks to the host, each track carries what's needed to place it: its clock readings, its measured sample rate and where the calibration sweep landed. That's why a phone that drops out and sends its take later still lines up exactly as it would have on the night.

The host then builds the stack: every track placed on one timeline, drift corrected, mixed, and also written out one file per phone with a CSV of the offsets. Collecting and stacking.

Download on the App Store