A quarter note in milliseconds is 60,000 divided by the tempo. At 120 BPM that is 500 ms, an eighth is 250, a dotted quarter 750 and a quarter triplet 333. Every tempo-synced setting in a session comes out of that one division, and a delay set to an arbitrary 380 ms against a 120 BPM track fights it rather than sitting in it.
The reverse works too: 60,000 divided by a measured millisecond value gives the tempo it corresponds to, which is how you find the tempo of a loop you only have the length of.
What are the note values?
At 120 BPM, where a quarter note is exactly 500 ms.
| Note | Multiplier | At 120 BPM |
|---|---|---|
| Whole | ×4 | 2,000 ms |
| Half | ×2 | 1,000 ms |
| Quarter | ×1 | 500 ms |
| Dotted eighth | ×0.75 | 375 ms |
| Eighth | ×0.5 | 250 ms |
| Eighth triplet | ×0.333 | 167 ms |
| Sixteenth | ×0.25 | 125 ms |
A dot adds half the note’s value, so a dotted eighth is three sixteenths. A triplet divides the beat into three rather than two, which is why the triplet figures are a third rather than a half.
Why is the dotted eighth the workhorse?
Because it is three sixteenths long, so its repeats land on the sixteenth grid but not on the beat. The echoes fill the space between beats instead of doubling them, which is what makes a delay sound like part of the arrangement rather than a smear on it.
A quarter-note delay reinforces the beat and quickly becomes muddy. A dotted-eighth delay creates the syncopated pattern that has been the default on guitar and vocal delays for decades, and it works because of that three-against-four relationship rather than because of the sound of any particular unit.
At 110 BPM the numbers are 545.45 ms for a quarter, 409.09 for a dotted eighth and 136.36 for a sixteenth. None of those is a round figure, which is precisely why the division is worth doing rather than eyeballing.
What else syncs to tempo?
More than delay. Reverb pre-delay of a sixteenth keeps the tail out of the way of the transient. Gate and compressor release times set to an eighth or a sixteenth breathe with the track instead of against it. Tremolo and auto-pan rates are the same arithmetic expressed as a frequency.
The general principle is that any time-based process either agrees with the grid or argues with it, and agreeing costs one division to arrange.
Swing complicates the grid deliberately. A swung eighth pair is not two equal halves of a beat but a long-short division, often close to a triplet feel, and a delay synced to straight eighths will sit against a swung part rather than with it. Where a session has swing applied, effect timings usually want to be set by ear rather than by division.
How does tapping go wrong?
In two ways that both produce a confident wrong answer. Tapping the hi-hat rather than the beat gives exactly double the tempo. Tapping every other beat gives exactly half.
Both are easy to catch because the error is a clean factor of two — if a tapped reading is 172 on something that feels mid-tempo, it is 86. A reading that averages the last eight taps settles within a bar or two and follows a tempo that drifts, which is what makes tapping usable on live material at all.
Tempo drift is a separate problem from tempo error. A live recording rarely holds one figure, so a fixed grid drifts against it over a few minutes even when the average is right. That is what tempo mapping exists for, and it is why a tapped average is a starting point rather than an answer on live material.
When do samples matter more than milliseconds?
Once you are trimming loops rather than setting effects. At 44.1 kHz one millisecond is 44.1 samples, so a 500 ms beat is 22,050 samples — a whole number, which is why 120 BPM at 44.1 kHz loops cleanly.
Tempos that do not divide evenly leave a fractional sample at the loop point, and that fraction accumulates into a drift over a long arrangement. It is inaudible on one bar and audible over a hundred, which is why sample-accurate loop lengths are worth checking on anything that repeats for minutes.
Questions people ask
Why 60,000? Because there are 60,000 milliseconds in a minute, and BPM is beats per minute. The division is unit conversion rather than music theory.
Does this work for odd time signatures? Yes — the beat length is unaffected by how many beats are in a bar. A bar of 7/8 at 120 BPM is seven eighth notes of 250 ms each.
What tempo should a track be? Whatever it is. Genre conventions exist and they are much looser than tempo charts suggest, and a track that sits at an unusual tempo is not doing anything wrong.
Why does my delay still sound off after syncing? Usually the feedback rather than the time. A synced delay with too much feedback fills every sixteenth and the grid stops being audible at all.
One division, then everything else is a multiplier. The BPM calculator and BPM to milliseconds give the note table, the delay time calculator covers the effects settings, and the BPM tapper finds the tempo when you only have the audio.