Tools

Tempo to time

Note lengths in milliseconds at a given tempo. The 1/32 and 1/64 rows are the usual pre-delay values.

Delay feedback tail

How long the repeats keep going. The arithmetic behind why 90 percent lasts so much longer than 70.

Counted to −60 dB, the point the tail is usually considered over.

Note and frequency

Equal temperament, A4 = 440 Hz. The reverse field snaps to the nearest note and shows the offset in cents.

Samples and milliseconds

Converts both ways at the selected sample rate.

Decibels and linear gain

The conversion that explains why a copy 6 dB down does not cancel: it is still at 0.501, not 0.5 of nothing.

Adding two levels

Two identical, time-aligned copies sum to +6.02 dB. Two unrelated sources of equal level sum to +3.01 dB. This does the general case.

Pan law

What each channel receives as a source moves across the image. Explained in why a centred channel changes level.

Pan runs −100 hard left to +100 hard right. The law names are nominal: equal power is really −3.01 dB at centre, the compromise −4.52 and equal amplitude −6.02.

Compressor output

Where a peak lands once the ratio has worked on it. Hard knee. The full explanation is in what the ratio numbers mean.

Crest factor

The distance between peak and average, which is what decides how a track behaves even when two of them meter the same. See equal peaks, different behaviour.

A sine measures 3.01 dB, a square wave 0 dB. Unprocessed drums run high, a limited master low.

Filter slope

How much survives a given distance past the cutoff. Which slope to reach for is covered in what 12 dB and 24 dB per octave actually do.

The asymptotic slope, which holds an octave or so out. At the cutoff itself the response is already about 3 dB down.

Offset to comb filter

Where the cancellations land when two copies of one sound are offset. The arithmetic behind nudging tracks off the grid.

Where harmonics fold back

Distortion adds harmonics; the ones past Nyquist come back somewhere else. Oversampling gives them room. Background in aliasing and oversampling.

Bit depth and dynamic range

Each bit is worth about 6.02 dB. What to do at the bottom of that range is where dither goes and why.

Integer depths. 32-bit float stores an exponent as well, so it does not follow this arithmetic.