Phase cancellation: causes, comb filtering and fixes
Last updated 7 min readTags:MixingPhaseRecording
Learn why phase cancellation makes combined tracks sound thin, how comb filtering works, and when to change polarity, timing or microphone placement.
Phase cancellation occurs when related signals combine and parts of their waveforms oppose one another. The level falls at the affected frequencies. A snare may lose body when you add its second microphone; a bass may sound weaker when you blend its DI and amp recordings.
With a short, fixed delay between a sound and a copy of itself, some frequencies reinforce while others cancel. The repeated peaks and dips are called comb filtering. The name comes from the comb-like shape of the frequency response.
If a sound becomes thinner when you add another track, compare each track alone with the pair together. A loss of body in the combined sound is a reason to check polarity and timing before adding EQ.
Polarity and timing cause different problems
A polarity inversion reverses the sign of every sample: positive becomes negative and negative becomes positive. Two identical, time-aligned signals at equal level cancel completely if one has inverted polarity.
A delay moves one signal later. Its effect depends on frequency. A 1 ms delay is half a cycle at 500 Hz, so equal copies cancel there. At 1,000 Hz it is a full cycle, so they reinforce.
This distinction matters when you use the Ø button on a mixer or plug-in. It inverts polarity; it does not remove a delay. On a delayed pair, the button changes which frequencies cancel. One setting may sound better, but neither necessarily solves the timing problem. See polarity versus phase for the full distinction. [1]
How to check phase cancellation in a mix
1. Listen to the tracks together in mono
Solo the suspected pair and route them to the same mono bus, or pan both to the centre. Listen to each separately, then together. Check whether the combination loses low end, attack or clarity.
For a stereo effect, compare the stereo output with its mono fold-down. A delayed copy panned opposite the original may sound wide in stereo yet become hollow when the channels are summed.
2. Compare both polarity settings
Invert the polarity of one track and compare. Keep the setting that gives the more useful tone, then check it in the full mix.
A clear improvement tells you that the relationship between the tracks matters. It does not prove that the original problem was solely a polarity reversal: a delayed pair can also improve after a flip.
3. Test a small timing adjustment
If neither polarity setting gives a satisfactory result, inspect the waveforms for a recognisable shared transient. Use its arrival time as a starting point, then adjust one track in small steps while listening. You can move the later recording earlier or delay the earlier signal.
The largest peaks need not represent the same part of the sound, especially with different microphones or an amp recording. Visual alignment is a starting point, rather than proof that the tracks now combine well.
For layered kicks or other different samples, check several hits and the whole decay. An adjustment that strengthens the attack may weaken the sustained low end.
4. Check the result in context
Repeat the polarity comparison after changing timing. Then restore the rest of the mix and listen at a comparable output level. Judge the tone and attack as well as the increase in level.
With a multi-microphone drum recording, check the edited track against the overheads and room microphones too. A change that improves one pair may worsen another. [1]
Common recording and processing cases
Top and bottom snare microphones
Top and bottom microphones capture different aspects of the drum, so they are not identical copies. A polarity inversion on one channel often improves their combined sound, but the best choice depends on placement and the recording.
Compare both settings with the overheads playing. If the snare still sounds thin, test a small position change during recording or a timing adjustment afterwards. Perfectly matched waveform peaks are not the aim; a convincing snare sound is.
DI bass and an amp microphone
The microphone signal usually arrives later than the DI because sound must travel from the speaker to the microphone. The amp, cabinet and any digital processing also affect the relationship between the recordings.
At about 20°C, a 0.5 m acoustic path takes roughly 1.46 ms. For an otherwise identical delayed copy, that would put the first cancellation notch near 343 Hz. A real amp recording has a different frequency response, so treat this calculation as an illustration, not a prediction of its exact notches.
Try delaying the DI or moving the microphone recording earlier. Compare both polarity settings, and listen across different notes. If the blend remains uneven, reduce the level of one recording or use it mainly for a particular frequency range.
Parallel processing
A parallel chain can cause comb filtering if its output arrives late relative to the dry track. Check the DAW's plug-in delay compensation and any external hardware routing.
Correct latency compensation does not make every parallel blend phase-neutral. Filters and other processing can change phase differently across the spectrum. Bypass processors one at a time to find which change affects the blend. [1]
How delay sets the comb-filter frequencies
For a signal added to a delayed copy with the same polarity, the first cancellation minimum is:
First notch in Hz = 1 ÷ (2 × delay in seconds)
Further notches occur at three, five and seven times that frequency. Their spacing is:
Notch spacing in Hz = 1 ÷ delay in seconds
| Delay | First notch | Next two notches | Spacing |
|---|---|---|---|
| 1 ms | 500 Hz | 1,500 Hz; 2,500 Hz | 1,000 Hz |
| 2 ms | 250 Hz | 750 Hz; 1,250 Hz | 500 Hz |
| 10 ms | 50 Hz | 150 Hz; 250 Hz | 100 Hz |
These are calculated values for an ideal delayed copy. Complete nulls require equal levels and matching signals. Different recordings generally produce less regular results. [2]
For microphone recordings, the relevant distance is the difference in sound-path lengths from the source, not simply the distance between the microphones. Two microphones can be far apart yet equally distant from the source.
Do not infer a delay from a description such as “thin” or “hollow” alone. Many recording and EQ choices produce those qualities.
Why the quieter copy causes less cancellation
In the ideal two-copy case, similar levels produce deeper notches. If the delayed copy has half the amplitude of the original, the sum falls to half the original amplitude at the minima, about −6 dB. At the maxima it rises to 1.5 times the original, about +3.5 dB.
Lowering the second track therefore reduces the depth of the comb filtering. This is often sufficient when that track contributes only a little ambience or texture.
When the problem comes from the room
A microphone can capture a direct sound and a delayed reflection from a desk, floor or wall. Those signals have already combined in the recording, so shifting the recorded track cannot separate them. Change the microphone or source position, reduce the reflection, or record again. [3]
Comb filtering can also occur during monitoring as direct and reflected sound meet at your ears. If the tonal change varies as you move your head, investigate the room and speaker placement before changing the mix. Check on headphones or another monitoring system too.
The 3:1 rule concerns microphone spacing to reduce unwanted pickup between microphones serving different sources. It is not a general instruction to separate two microphones deliberately recording the same instrument. [4]
What meters and EQ can tell you
A spectrum analyser may reveal repeated dips, but the source's own spectrum can obscure them. A single broad dip does not establish the cause. Masking can make a sound harder to hear without creating a cancellation notch.
A stereo correlation meter compares the left and right channels. It cannot identify every cancellation problem between tracks already combined on a mono bus. Use it alongside listening and the mono compatibility check.
EQ after the sum cannot recover a frequency that has cancelled completely. Where you still have the separate tracks, first test their polarity, timing and relative levels.
Some comb filtering is intentional, as in flanging. If the combined sound has the tone you want and holds up in mono, there is no need to remove the effect merely because it contains cancellation.
Sources
- iZotope: What is phase in audio? Five ways to address phase issues
- SynAudCon: The delay comb filter, useful equations, Joseph G. Mitchell, pp. 21–22. Table values above are calculated from the delay equations.
- DPA Microphones: The basics about comb filtering and how to avoid it
- Shure: Audio Systems Guide for Houses of Worship, section on interference effects and the 3:1 rule.