A stereo file arrives as two channels, left and right. That’s how it gets to the speakers, and it’s a poor description of what’s on it. Almost nothing in a mix is “a left thing” or “a right thing.” The kick is in both. The vocal is in both. The room on the drums is in both, slightly differently, and that slight difference is the whole reason it sounds like a room.

Mid/side takes the same two channels and asks a better question: what do these two signals have in common, and where do they disagree?

The arithmetic

Two sums, and that’s the entire encoding:

The mid is everything both speakers play the same. The side is everything they don’t. To get back to left and right you add and subtract again:

Run a file through the encode and straight back out the decode and you have the original file, sample for sample. Nothing is approximated and nothing is thrown away. It’s the same audio described on a different pair of axes, the way you can give a point on a map as north/east or as a bearing and a distance.

That reversibility is what makes the technique safe to build on. A mid/side plugin that you leave at its defaults has done nothing to your mix, because the decode undoes the encode. Everything interesting happens in the gap between the two, where you process one half and leave the other alone.

What ends up where

On a conventionally balanced mix, the mid holds the things that were panned to the centre: kick, bass, snare, lead vocal. The side holds the reverb tails, the doubled guitars, the stereo synth pads, the room mics, and the top end of anything that was recorded with two microphones.

You can hear this directly. Solo the side channel on a finished record and you get an odd, hollowed-out version of the song with the lead vocal and the bass mostly gone: the karaoke effect, and the same trick behind every vocal-removal plugin. Solo the mid and you get something that sounds like an old AM radio broadcast: centred, dry, and narrow.

Neither of those is a sound you’d use. They’re a diagnostic. Once you’ve heard which parts of your mix live in which half, the reason to process them separately becomes obvious, because the problems live in different places too.

What that buys you

The useful cases all have the same shape: something you want to fix is in one half, and the thing you’d damage by fixing it is in the other.

Low end is the standard example. Bass energy belongs in the middle, where both speakers reinforce it, and stray low frequencies in the sides make a mix feel loose without ever showing up as a level problem. A high-pass on the side channel only tightens the bottom without touching the bass guitar at all.

Compression is the case people underestimate. A stereo compressor keyed off the whole mix ducks the reverb every time the kick lands, because the kick is loud and the detector doesn’t know it should ignore it. Compress the mid and the side on separate detectors and the centre gets controlled while the ambience keeps its shape.

Then there’s saturation, where the argument is about harmonics rather than level. Drive applied to a full stereo bus adds harmonic content to the vocal and the room at the same time, and the vocal is usually the one that can’t take it.

Where it stops working

Two honest limits, and both catch people out.

The first is mono. Sum left and right together and the side channel cancels completely. The arithmetic: L + R = (M + S) + (M − S) = 2M. Whatever you put only in the sides is gone the moment someone plays your track on a phone speaker, a club PA fed in mono, or a laptop. This isn’t an argument against using the sides. It’s an argument for checking in mono before you commit to anything dramatic there.

If a chorus falls apart when you hit the mono button, the width you added was doing more work than you thought, and the mix underneath it needs to stand on its own first.

The second is that boosting the side channel is not the same as making a mix wider in any musical sense. Turn the sides up and you’ve changed the ratio between centre and edges, which reads as width and also as a quieter vocal and a smaller-sounding kick. Past a certain point it stops sounding big and starts sounding like the middle fell out.

Watching it happen

A goniometer makes the whole thing visible, which is why every plugin here has one. It plots the two channels against each other, rotated so the centre runs vertically. A mono signal draws a straight vertical line. Content that exists only in the sides lies flat across the horizontal. A normal mix sits somewhere between: a vertical-ish cloud that leans and breathes with the music.

Watch it while you move a width control and the relationship stops being abstract. The cloud stretches sideways as you add side content, and flattens toward horizontal as you push past what the mix can carry, which is the point where you should check mono, because that flat shape is what cancellation looks like on the way in.