What is mid/side processing?

Mid/side processing in plain English

A stereo file stores two channels, one for the left speaker and one for the right. That's how it plays, but you can describe the same sound another way: as what the two speakers have in common, called the mid, and what's different between them, called the side. Mid and side hold exactly the same information as left and right, arranged differently.

The centre of a typical mix is in the mid: the kick, the bass, the snare and the lead vocal, all panned to the middle so both speakers play them equally. The side holds the width, things like a stereo pad, doubled guitars panned apart, the room sound on the drum overheads and the tail of a stereo reverb.

Once a sound is split this way you can give each half its own effects, say a compressor on the mid and an EQ on the side. Then the two halves are turned back into left and right, and the result plays on any speakers like any other stereo file. That round trip, from left and right to mid and side and back, is what mid/side processing means. You'll often see it written as M/S.

Encoding and decoding mid and side

The conversion is simple arithmetic, done on every sample:

• Mid = (L + R) ÷ 2
• Side = (L − R) ÷ 2

Converting back is just as short: L = mid + side and R = mid − side. Turning left and right into mid and side is called encoding, and turning them back is decoding. If you encode and decode with nothing in between, every sample comes out exactly as it went in, so the conversion itself costs nothing in quality.

Suppose a vocal is panned dead centre and peaks at 0.5 in both channels. The mid is (0.5 + 0.5) ÷ 2 = 0.5 and the side is (0.5 − 0.5) ÷ 2 = 0, so a centred mono part lives entirely in the mid. Now take a guitar panned hard left at 0.6, with nothing in the right channel. Its mid is 0.3 and its side is 0.3, which means a hard-panned part sits in the mid and the side equally.

Flip the polarity of one channel of a centred sound, so that L = 0.5 and R = −0.5, and the mid becomes 0 while the side is 0.5. Everything has moved to the side, which is why a polarity problem makes the centre vanish in mono. The phase lesson covers that in detail.

Not every plug-in uses the same scaling. Some write the mid as L + R and halve it on the way back, and some multiply by 0.707 so the level stays the same in both forms. The sound is identical either way; only the numbers on the meters move.

The halfway row assumes an equal-power pan, which sends about 92% of the level to the near speaker and 38% to the far one. Other pan laws give slightly different numbers, and the pan laws lesson explains why.

The mid is what mono plays

When a mono speaker, like a small Bluetooth speaker or a mono PA in a bar, plays a stereo file, it adds the two channels together and turns the sum down so it doesn't overload, often by half: (L + R) ÷ 2. Apart from that level, it's the formula for the mid, so Mid only, in the lab or in a plug-in, is a mono check.

Whatever you do to the side is never heard in mono. Turn the side up 6 dB and a mono listener hears no change at all. That makes side processing safer than the widening tricks that use short delays, which cut notches into the tone when the channels are summed (the comb filtering lesson shows how). It also means that the more of a mix you move into the side, the more a mono listener misses, because the side is the part they lose.

Whatever you do to the mid is heard by everybody. A cut, boost or compressor on the mid changes the mono version by exactly as much as it changes the stereo one.

Reading the lab

The two long screens at the top show the whole loop, split into its halves. The pale shape is the mid or side before any processing, and the coloured shape is after it, teal for the mid and violet for the side. Both lanes use the same scale, so you can see straight away how much smaller the side of the full mix is than its mid. The amber strip under each lane shows the gain reduction of that half's compressor.

The round screen is a goniometer. Each dot is one moment of the sound you're hearing, with the mid running straight up and down and the side running across. A mono sound draws a vertical line, Side only draws a horizontal one, and a normal stereo mix draws a ball that's taller than it is wide. Next to it, the width screen shows how much of the energy is in the side, slice by slice through the loop, with the original as a dashed line.

Each half has its own Level, Low cut and Comp. Low cut is a 24 dB per octave high-pass filter, and Comp sets the threshold of a 4:1 compressor with a 10 ms attack and a 120 ms release. The Listen buttons pick between the full stereo result, the mid alone and the side alone, and A/B switches all six controls off at matched loudness. The Right now panel explains what your settings are doing, using numbers measured on the loop as you change them.

What is mid/side processing used for?

Making a mix wider or narrower

Changing the balance between mid and side is how most width controls work. Turn the side up and the edges of the mix come forward; turn it down and everything pulls in towards the centre, until at zero side you have mono. Ableton Live's Utility device has a Width control that does this, where 0% is mono and 100% leaves the sound as it was. The stereo width lesson goes further into how wide is too wide.

On a full mix, 1 to 2 dB more side is usually enough to hear, and 6 dB is a lot. Narrowing is just as handy: a stereo synth pad with its side turned down 4 dB takes up less room and leaves a clearer gap in the middle for the vocal.

Putting the low end in mono

A low cut on the side makes everything below its cutoff mono, while the width above it stays. On a full mix or a master, 100 to 150 Hz is the usual place to start. The bass and kick then hit both speakers equally, and nothing changes in mono because the side was never in the mono version.

Vinyl is one reason mastering engineers do this. A stereo groove moves side to side for the mid and up and down for the side, so heavy bass in the side makes the groove rise and fall in depth, and too much of it can throw the stylus out of the groove. Digital releases don't have that limit, but wide bass rarely helps a mix and tends to sound vague on big systems.

EQ on the mid or the side

Any EQ move can go on one half only. On a music bus without the vocal on it, cutting 2 to 4 kHz in the mid makes space for the lead vocal without dulling the edges, and a gentle high shelf on the side brightens the width while the centre keeps its tone. This is a big subject with a lab of its own in the mid/side EQ lesson.

Compressing the mid or the side

A compressor on the mid alone holds the centre steady: the kick, snare, bass and vocal all get turned down on the loudest moments while the stereo spread underneath is left as it was. A compressor on the side alone evens out the width, so loud wide moments like a crash cymbal or a big stereo chord come down without the vocal being touched.

If you squeeze the mid hard, each kick pulls the centre down and the side suddenly sounds louder by comparison, so the mix gets wider on every hit and narrower between them. In the lab you can watch the violet line in the width screen jump with the beat. A little is fine, and on some records it's part of the sound; a lot sounds like the stereo picture pumping. On a mix bus, keep each half's compressor to 1 to 3 dB of gain reduction, the same as you would with an ordinary bus compressor.

Other effects on one half

Anything that runs in a mid/side mode works the same way. A little saturation on the side adds harmonics to the edges and brings them forward a touch without changing the vocal. Put a reverb on the side only and its whole tail sits in the edges with the centre left dry; both moves vanish in mono, like every other side change.

Mid/side recording

The idea started with microphones. Alan Blumlein's work on stereo in the early 1930s included it, and mid/side recording is still used on acoustic instruments, choirs and room sound. One microphone points straight at the source and records the mid. A second, with a figure-of-eight pattern, faces sideways: it picks up the left and right of the room with opposite polarity and records the side.

The two recordings are decoded with the same arithmetic as above, L = mid + side and R = mid − side. Because the side is its own track until it's decoded, you can choose the width after the session by turning that track up or down, and the mid microphone on its own gives a mono version that can't cancel. Logic Pro's Direction Mixer decodes a mid/side pair. In any DAW you can do it by hand: pan the mid track centre, duplicate the side track, pan the two copies hard left and hard right and flip the polarity of the right one.

How to use mid/side processing, step by step

This order works on a mix bus or a master. The lesson in the lab covers the same moves.

1. Listen to the mid and the side aloneBefore you change anything, solo each half and notice what's in the side, how loud it is and whether there's low end in it. If your plug-in has no solo, a mono utility after it lets you hear the mid.
2. Mono the low endPut a low cut on the side and raise it while you listen to the side alone. Stop where the rumble goes and the bass notes of any wide parts still come through, usually somewhere between 80 and 150 Hz.
3. Set the width with the side levelMove it a decibel at a time and compare with the original at the same loudness. If the vocal starts to feel further away, you've gone too far.
4. Compress one half only if it needs itStart with 1 to 2 dB of gain reduction on the mid and listen for the width moving with the beat.
5. Check it in monoSwitch to the mid alone, or your DAW's mono button, and make sure the mix still works. Everything you did to the side will have gone, which is what a mono listener hears.

Plenty of tools have a mid/side mode built in. Ableton Live's EQ Eight has one, and many of iZotope Ozone's modules can run in mid/side. For anything else, the free Voxengo MSED plug-in can encode a track so the mid is on the left channel and the side on the right. Any effect that processes its two channels separately then works on mid and side, and a second MSED decodes the result.

Starting settings for common jobs

These are places to begin, chosen so the change is easy to hear. Compare each one with the original at matched loudness and in mono before you keep it.

Common mid/side mistakes

Turning the side up too far

More side always sounds impressive for a few seconds. Past 3 or 4 dB on a full mix the centre starts to feel weak, and the mono version loses more of the mix than it did before. The less obvious problem shows up in the numbers. Take the hard-left guitar from earlier, with a mid and side of 0.3 each, and turn the side up 6 dB to 0.6: the left channel becomes 0.9 and the right becomes −0.3. The guitar now leaks into the right speaker upside down, which is the hollow, phasey sound of an over-widened mix.

Low-cutting the mid

The mid is where the kick and bass live and it's what mono plays, so a low cut on the mid thins out the whole track for every listener. Low cuts almost always belong on the side.

Judging it by volume

Turning the side up also makes the whole file louder, and louder tends to sound better. Match the level before you compare, as the lab's A/B does, or you'll end up keeping every widening move you try.

Forgetting that panned parts are in both halves

The side isn't a separate track of wide instruments. Anything panned away from the centre is partly in the mid and partly in the side, so a mid compressor reacts to hard-panned guitars too, and a mid cut changes them as well as the vocal.

Train your ears to hear width

Small changes in width are hard to hear at first, because nothing gets brighter or louder. The quiz under the lab turns the side up or down by a hidden amount, covers the screens and matches the loudness, then asks whether the result is wider or narrower than the original on A/B. Eight rounds take about three minutes. Headphones make it easier to begin with, and once you can pass it on headphones, try it on speakers.

Questions people ask

Keep going