Mid/side EQ helps when a problem lives in one dimension of the stereo image: bass that has spread into the sides, reverb mud that sits only in the width, a centre vocal crowded by wide guitars. It breaks the mono when the side gets boosted further than the mix can afford, because a mono fold-down keeps the mid and discards the side. Every decibel added to the side is a decibel the mono listener never hears. With linear-phase filters and a correlation meter open next to the EQ, it reaches problems a left/right EQ cannot touch without changing the whole mix.
What are mid and side, in numbers?
Two signals built from left and right. Mid is the sum, M = (L + R) / 2. Side is the difference, S = (L - R) / 2. Decoding reverses it exactly: L = M + S and R = M - S. Alan Blumlein described the sum-and-difference idea in a patent filed on 14 December 1931 and granted in June 1933 (UK patent 394325), and the arithmetic has not changed since. The patent page gives both formulas with a factor of about 0.7 instead of 0.5; the scaling only matters if an encoder and decoder disagree.
Read the formulas as a map of a mix. A vocal panned dead centre is identical in both channels, so it lives entirely in the mid and contributes nothing to the side. A pair of double-tracked guitars panned hard left and right, two different takes, splits its energy evenly between mid and side. A reverb return with a wide, decorrelated tail puts about half its energy in the side, far more than any centred source. Solo the side and the lead vocal vanishes, which is the fastest way to hear what the width of a mix is actually made of (Sound On Sound).
What happens to the side in mono?
It cancels. A mono fold-down is (L + R) / 2, which is the mid and nothing else. A centre-panned source loses nothing. The hard-panned guitar pair drops 3 dB against the centre, because half its energy was in the side. Anything that exists only in the side, a polarity-flipped double or a widener's difference signal, disappears completely.
Now boost the side 3 dB with an M/S EQ to widen those guitars. In stereo the pair comes up about 1.8 dB. In mono nothing changes, so the gap between the stereo and mono versions of the guitars grows from 3 dB to about 4.8 dB. A 6 dB side boost gives the pair roughly 4 dB more in stereo and leaves it 7 dB down in mono. Those figures come straight from the formulas above, and they are the arithmetic behind a mix that sounds wide on monitors and thin on a single speaker.
Where does M/S EQ actually earn its place?
In four jobs, each with a number attached:
- Low end in the sides. A high-pass on the side at 100 to 150 Hz makes everything below it mono, keeps kick and bass anchored, and stops a wide pad from smearing the bottom octave. For a vinyl cut it stops being optional.
- Mud in the width. Room mics and reverb returns pile up between 200 and 500 Hz in the side, and a 2 to 3 dB cut there on the side alone cleans the image without thinning the vocal, bass or kick in the middle.
- A centre vocal crowded by wide guitars. A 1.5 to 2 dB dip in the side around 2 to 4 kHz gives the vocal room without touching the vocal's EQ or the guitars' own.
- Air. A shelf of 1 to 2 dB on the side above about 10 kHz opens the top of the image; Craig Anderton's Sound On Sound piece recommends a slight high-frequency side boost for the same reason. Beyond 2 dB the mono arithmetic starts to cost more than the width buys.
Each of those is a cut or a small shelf. The moves that go wrong are large side boosts below 8 kHz, where the instruments carry most of their level and the mono loss is easiest to hear.
Linear phase or minimum phase?
Linear phase, for any M/S move bigger than a decibel or so. A minimum-phase filter shifts phase around its corner frequency. Apply it to the mid and not the side, and after decoding the left and right channels no longer agree on timing at that frequency, which blurs the image in the band the EQ was meant to tidy. FabFilter's Pro-Q 4 help recommends linear-phase processing when filtering stereo channels to avoid unwanted phase changes (FabFilter), and Anderton warns against any phase shift or sample misalignment in M/S work for the same reason.
The cost is latency and pre-ringing. FabFilter's own figures for Pro-Q 3 at 44.1 kHz put the Medium linear-phase setting at 5,120 samples, about 116 ms, and High at 9,216 samples, about 209 ms (FabFilter). Pre-ringing grows with steep slopes, high Q and large gain changes, which is one more reason the M/S moves on a bus stay broad and gentle.
How do you check that it has not gone too far?
With a correlation meter and a mono button, before the move and after it. A healthy mix reads between 0 and +1 most of the time, and readings left of zero mean the side is outweighing the mid, so mono will lose energy (Production Expert). The guitar pair makes the threshold concrete. Untouched, it reads about 0. After a 3 dB side boost the same pair reads about -0.33, because correlation works out to the mid power minus the side power over their sum. One EQ move has pushed the meter into the zone it exists to warn about.
Then the mono button, at the same listening level as the stereo check. If a part disappears, or the vocal jumps forward by more than a decibel or two, the side is carrying something the mono listener needs.
When should you fix the source instead?
Whenever the problem belongs to one instrument. M/S on the bus treats every source that shares the side at once, so a 3 dB cut aimed at a boomy reverb also thins the guitars and the room mics. If the mud comes from one return, EQ the return. If one synth is wide in the low end, high-pass that synth's side at the source. The bus keeps the moves that belong to the whole picture, and on a master from another engineer those moves stay under a decibel until the mix engineer has heard them.
Frequently asked
Should M/S EQ go on every mix bus? No. It belongs where the image has a problem in one dimension. On a balanced mix a side boost buys width in stereo and costs level in mono, as the numbers above show.
What frequency should the bass go mono below? 100 to 150 Hz covers most pop and electronic mixes. For a vinyl cut the cutting engineer's limit decides, and 150 Hz is the usual starting point.
Linear phase or minimum phase for M/S work? Linear phase for any move bigger than about a decibel, because a phase difference between mid and side shifts the stereo image once it is decoded. Keep slopes and Q gentle to stay clear of pre-ringing.
If a mix sounds wide here and thin on a phone
Send me the stereo bounce. I will listen to the side alone and folded to mono, then send back the frequency ranges where the side carries more than a mono listener will ever hear, with the cut that fixes each one. Credits are at /credits, background at /about.