Comb filtering, and how to spot it

Comb filtering is what phase cancellation looks like when one signal is a delayed copy of another. Instead of a single dip, you get a row of evenly spaced notches marching up the spectrum — the shape that gives it the name, since it looks like the teeth of a comb on an analyser.

The delay sets the spacing

The delay between the two copies decides where the notches fall. A delay puts the first notch at the frequency whose half-wavelength matches it, then repeats the dip at odd multiples of that frequency.

The rule of thumb is worth remembering: a delay of t seconds puts the first notch near 1/(2t) Hz. A 1 ms delay drops the first notch around 500 Hz, with more at roughly 1.5 kHz, 2.5 kHz, and so on up the range. Shorten the delay and the notches spread apart and climb higher. Lengthen it and they pack in tighter and lower. That relationship is the tell: if you move a mic and the whole notch pattern slides, you’re looking at comb filtering, not EQ.

Where the delay comes from

Anything that makes the same sound arrive twice, slightly apart, will do it:

• Two mics at different distances on one source — the workhorse case in multi-mic recording.
• A mic and a reflection — direct sound plus the same sound bouncing off a wall, the floor, a desk, or a music stand a foot away.
• A DI blended with a mic’d amp, where the amp path is a few feet and a little processing later than the DI.

Live sound gets its own version when the same source reaches a listener from two speakers, or from a speaker and a hard surface.

What it sounds like

Static comb filtering reads as hollow, thin, or slightly metallic — a coloration rather than an obvious effect. When the delay moves, it turns into the swept, swirly sound of a flanger, because the notches are sliding through the spectrum in real time. That’s the same physics a flanger pedal uses on purpose.

Spotting and fixing it

On an analyser the pattern is unmistakable once you know it: regular, evenly spaced dips rather than the random peaks and valleys of a normal instrument. Folding to mono usually makes it worse and easier to hear.

You have three ways out, and which one fits depends on the source:

• Change the delay. Move the mic, or time-align the tracks so the copies land together.
• Kill the reflection. More distance between mic and surface, or some absorption, removes the second copy at the source. The 3:1 rule — put the second mic at least three times as far from the first as the first is from the source — is a useful starting point here, not a law.
• Correct the phase relationship when the copies can’t simply be lined up in time.

As with any phase effect, some of it is just the character of a source in a real space. Fix the comb filtering that’s robbing you of low end and body; leave the mild coloration that’s part of the sound.