Splitting out your stereo signals into separate Mid and Sides tracks can open up a world of creative options in your mix.
In this article I'll take you through some cool but simple Mid-Sides techniques that add space, colour and interest to your tracks or overcome common mixing challenges. Now, I realise that some of you regard Mid-Sides (M‑S) processing as powerful voodoo that's best avoided, so I'll make everything really easy to follow. While some simple maths can explain the conversion between left-right (L‑R) stereo and M‑S stereo, you won't need it here (though if you do want to know more, check out Hugh Robjohns' explanation: https://sosm.ag/mid-sides-explainer). All you really need to understand is that the Mid signal is what you hear when you fold your stereo mix down to mono, and the Sides is everything else — any frequencies in the left or right channels that disappear in mono.
Converting L‑R to M‑S and back to L‑R again is a breeze: all you need is an M‑S encoding/decoding plug-in. Your DAW may include one, but cross-platform options are available from Blue Cat Audio, Melda Audio, Goodhertz, and DDMF, amongst others. I've used Voxengo's excellent freeware MSED here.
You'll find a ZIP file of hi-res WAV audio clips for each of the step-by-step exercises in the right-hand Media sidebar on this web page; feel free to download these source files and work through the exercises in your own DAW, or just to listen to the clips of the results. [The SoundCloud MP3 examples embedded in this article are purely for instant auditioning purposes.]
Download | 52 MB
For most creative M‑S processing I like to split the original L‑R source out to two tracks in my DAW (I've used Cubase here but it works in any DAW). This allows me to process the M and S components of the signal independently, and a quick glance at the mixer makes it obvious what's going on. I can throw whatever processing I want at either the Mid or the Sides only, and can send the results anywhere else in the DAW. As you'll see, this opens up some wonderfully creative options. I suggest you save my routing matrix as a template in your DAW:
- Open a blank project and create a stereo track, which will host the original stereo file you want to process.
- Disable this track's master-bus routing (eg. in Cubase, change it from 'Stereo Out' to 'No Bus'); you don't want to hear this track in your mix.
- Create two stereo effects tracks/busses, naming one track 'Mid' and the other 'Sides', and use your source track's aux send facility (at unity gain) to route the signal to both of these tracks.
- Insert MSED as the first effect in both the Mid and Sides tracks.
- In the Mid track's MSED, mute the Sides, so that only the Mid component passes through.
- In the Sides track's MSED, mute the Mid, so only the Sides signal passes through.
And that's it: the track for the original stereo source is now split into separate Mid and Sides tracks. You can apply any effect or process to the Mid or Sides tracks, or to both, and it's all summed back to L‑R at the master stereo bus so you can hear the result normally. You can also balance the M and S components to taste by altering the Mid or Sides track's level in various ways: using the DAW's track gain or fader controls; using the gain knob in the relevant MSED instance; or changing the original stereo track's send levels. To me it makes the most sense to use the gains on MSED, as you can then do all your M‑S work (muting, balancing etc) in one place, and it's easier to use the matrix as a template for other sources you want to process.
Our first technique provides the basis for several that follow — by processing only the Sides component of a stereo drum loop, we'll create a cool shuffle effect to add interest. Import the stereo file 'drum-beat-104.WAV' to the first stereo track (the source track in your matrix) and name the track accordingly. If you want to use a different file that's fine, but note that for M‑S techniques you must start with a genuinely stereo file featuring different L and R signals — a dual–mono signal saved as a stereo file won't work.
Insert a delay plug-in after MSED on the Sides track (I've used Waves H-Delay, but any stereo delay is fine). Set the DAW tempo to match the file (104 bpm), and the delay to 100 percent wet and a straight 1/8 tempo-sync'ed delay, with very little feedback. On playback, this should combine with the dry Mid signal to create a lovely shuffling type of effect, and give the drum beat a whole new feel but, importantly, without compromising the solidity of the loop's bottom end. The following clips are the results:
drum-beat-104: The original, unprocessed beat.
drum-beat-104-sides-solo: The processed Sides component only (the delay effect in isolation).
drum-beat-104-ms-delay: The full stereo result, with the unprocessed Mid and processed Sides balanced to taste.
A similar approach can inject ambience and motion into a staid piano line, by creating a delay tail that echoes and hangs in the air. Again, we'll add an effect only to the Sides component, leaving the Mid to provide a sense of focus and clarity.
Start with the same delay you used for the drums, and after this insert a reverb. I've used IK Multimedia's excellent Classik Studio Reverb but, again, any will do. The key to success is not to overuse either the delay or the reverb; they need to be blended by ear in such a way that they combine to form a single texture. To do this, set the wet/dry mix of each effect to taste. (Instead of using the wet/dry controls, you could instead send the Sides track to a delay on a separate track, and send that on to the reverb, but there's little to be gained by that in this case, and it's easier and tidier to use them as inserts.) To hear the results check out these clips.
amb-piano-120: The original, unprocessed piano part.
amb-piano-120-sides-solo: The processes Sides signal — nicely balanced, the delay and reverb effects form a lovely ambient backdrop.
amb-piano-120-delay-rev: The full effect, with dry Mid and processed Sides — the piano sounds deeper and wider than the original, but note the lack of an obvious 'hollow centre' that often results from some other stereo widening techniques.
Traditionally, beat producers 'lift' drum sounds from vinyl records and layer these samples to combine their different characteristics into a unique new sound. But simply slapping one sample on top of another usually means losing some clarity and impact. Why? Well, if two layers include similar frequencies at different amplitudes, the louder sound masks the quieter one; and the sounds will also be smeared in time where two layers play the same frequencies at slightly different times.
To get around this, the most effective layering technique is to choose one sample layer as the foundation, and use elements of additional samples to enhance and support it. Conventional processing like EQ and compression can be used to separate and define each layer, but these often have undesirable side-effects. M‑S processing offers us another, less messy way to layer things.
In this example we'll layer a mono snare sample with a stereo clap. Mono signals have no Sides component, so you could process and balance any mono sound on its own track, but I still prefer to route them to the same MSED M‑S tracks as I do stereo samples. That way, I can audition mono or stereo samples in a consistent way; it's drag-and-drop simple. Furthermore, I might later decide to apply stereo effects to the mono sample, and this way I'll be able to decide how best to represent these effects in an M‑S layering scenario. These two clips form our starting point.
mono-snare: The original mono snare sample.
stereo-clap: The original stereo clap sample.
The processing is similar to previous examples: you mute the Mid of the clap sample, to leave a clear space for the mono (ie. Mid-only) snare. It's oh-so simple, but by combining the Mid of one sample with the Sides of another, you can meld them together tidily without any processing whatsoever — just balance to taste. You can hear the results in these clips.
mono-snare-mid: The dry snare sample bounced from the Mid track.
stereo-clap-sides: The stereo clap with the Mid muted, and no effects or processing.
clap-snare-msed: A balance of the mono snare and the Sides component of the stereo clap.
It already works — but the real beauty of this way of working is that you can apply effects and dynamics processing to either component, and mix and match the levels of the components until the desired result is achieved, without either sound treading on the other.
A reverb that's fed only from the ambient Sides component... spreads its output across the sound stage [so] will give you reverb in both mono and stereo.
When mixing, I run my reverb and delays effects as aux sends rather than inserts, so I can route multiple sources to them. In a typical project, I'll have three distinct reverb tracks: one (invariably a plate type) for vocals; one for drums (usually a plate, room or booth, depending on the sound I want); and a more general-purpose one, for creating a common space in which all the various sounds can sit together. (I'll also have a few instrument-specific effects inserted directly on the track of the sound I want to process.) These stereo 'effects return' tracks provide another opportunity to play with M‑S muting and processing.
Start with the piano clip from the earlier example, routed to the Mid and Sides tracks as before. Then add a new stereo reverb track, as in the screenshot (I used a Valhalla one for the examples; they make great reverbs for a very reasonable price), and send a signal to it from the piano source track.
Now let's look at what happens according to where you place MSED in the signal chain. Place MSED, with the Mid channel muted, after the reverb. Press play, and you'll hear the original piano and a reverb effect derived from the whole piano sound. But because MSED follows the reverb, you can only hear the Sides of the output of the reverb, as indicated by the horizontal line on MSED's goniometer (the graph on the right of its GUI). Note that if you fold this mix down to mono, the reverb sound will disappear entirely — which may or may not be a good thing: it can be a useful way of adding 'fairy dust' to a stereo mix without cluttering things up in mono, but it will change the character of your mix on mono playback systems.
Now, try it the other way around: drag MSED before the reverb, again keeping MSED's Mid channel muted. Now you have a reverb that's fed only from the ambient Sides component of the piano, but as the stereo reverb spreads its output across the sound stage this will give you reverb in both mono and stereo. But in both cases, the dry sound remains nice and focused. It's worth listening to see which approach works best on any given song. And note that if you really want to go down the rabbit hole, you can go to town with multiple stages of Sides-fed stereo delay and reverb in this way...
These next two clips demonstrate the difference.
amb-piano-valhalla-sides: The reverb effect when the reverb is fed with the whole stereo piano signal, but the Mid of the reverb return is muted.
amb-piano-sides-valhalla: The reverb effect fed from the Sides-only piano signal, returning a mono-compatible reverb effect.
The next technique solves a common problem, but I need to explain the nature of the problem in some detail before we get underway. Something that seems to trouble a lot of inexperienced producers is management of their mixes' bottom end. It's not exclusive to the bass and kick drum — any sound that has elements residing in the low-frequency range needs careful consideration — but I'll focus on the kick and bass here, since it makes it easy to demonstrate the problem and the solution.
The interaction between the bass and the kick in modern EDM requires careful handling. Low frequencies have lots of energy — a higher amplitude for the same perceived level as higher frequencies. So when two or more sources with lots of low end overlap, they eat into your mix headroom and can also trigger dynamics processors (eg. a bus compressor or limiter) in an ugly way. Also worth noting is that where two sounds contain the same frequencies at the same time, the louder one will mask the quieter one, and despite the apparently high level according to your stereo bus' meters, this will leave things sounding wimpy and congested. Furthermore, where parts have only some notes/beats playing together and others in isolation, the sounds will lack consistency through the song, as the masking effect kicks in and out.
The trick is to remove from one sound frequencies that clash with those in another. But while that prevents unwanted triggering of compressors and leaves more headroom, simply EQ'ing the whole signal all of the time can compromise one sound or the other. A popular technique to address this is to 'sidechain' a compressor to duck (attenuate) one sound every time the other plays — for example, sending the kick channel to the external sidechain of a compressor placed on the bass channel, so that every time the kick rises above the compressor's threshold, gain reduction is applied to the bass. Each sound when playing on its own sounds as it normally would, but when they play together, the bass 'ducks' out of the way of the kick.
But while simple ducking can be just the ticket, sometimes there are unwanted side–effects — often, for example, it's really noticeable if you duck a bass part that has significant stereo spread (eg. chorused bass parts, or sounds treated with delays), or which is providing an important contribution to the mix higher up the frequency spectrum.
Thankfully, used in a similar way, dynamic EQ allows us to duck only the problem frequencies. And, taking it a step further, if we deploy a side-chained dynamic EQ only on the Mid component of the bass, we can often achieve the same thing even more transparently.
In this example, a drum beat and bass part combine to form a fairly busy pattern, and we'll use the kick to duck some clashing frequencies on the bass' Mid channel. These are the dry source sounds.
You can do this with any side-chainable dynamic EQ using the MSED matrix we used for the previous techniques, but it's simpler if you use a dynamic EQ which allows you to process the Mid and Sides channels separately. FabFilter Pro-Q 3 is one such tool, and I've used that here. Having inserted Pro-Q 3 on the bass track, activate its external side-chain (different DAWs handle side-chaining in different ways, so if in doubt check the manual, or Google!). Then created an EQ node, set the node's mode to Dynamic EQ, and set it to operate on the Mid channel only (you can access all these functions in Pro-Q 3 by right-clicking on the node). It's not usually difficult to find the clashing frequencies by ear, but Pro-Q 3 has a natty little feature whereby it highlights thee clashing frequencies for you — the stronger the overlap between the two signals, the redder the display. (This 'External Spectrum Visualiser' is accessed through the Analyser tab at the bottom.)
So far, so simple... but part of the joy of having a dynamic EQ is that you can treat different parts of the signal with different nodes. So, just for kicks, also try creating a second node: again, set it to dynamic, and this time set it to operate on the Sides only, and target the 'twang' frequencies in the bassline. Pro‑Q 3 has a band solo function which greatly helps in isolating the exact frequencies you want to process.
Now to tell each node how to react to the side-chain signal. We want the low-frequency Mid-only node to cut, and to do that you drag the Gain knob's outer ring anticlockwise. (Don't turn the inner part of the Gain knob or you'll apply conventional 'static' EQ.) Then set the Mid node's Q (bandwidth) so that the node ducks the shared frequencies but not more than that.
We'll use the second (Sides) node not to duck that twang but to accentuate it, just enough so it isn't obscured by the kick. All you need do differently is drag this node's dynamic EQ control clockwise. The combination of using ducking to attenuate the low frequencies and expansion to pronounce the mid-range frequencies in this way results in a controlled yet nicely dynamic response.
In the previous example, we were only concerned with managing the bass and kick in a mix, and they're generally fairly percussive sounds. This next one sheds light on how you might deploy a similar M‑S dynamic EQ tactic to manage deeper evolving sounds, whose frequencies overlap massively and together occupy a huge chunk of the spectrum. The sounds I've chosen are a cello line and a huge 'soundscape cluster' pad sound.
Insert FabFilter Pro‑Q 3 on the soundscape cluster track — it's this sound that needs to be ducked to allow the cello plucks to peep through, and also expanded so the soundscape lifts when the cello plucks hit the higher registers. Create two EQ nodes — one set to Mid, which we'll use to duck a frequency range where there's overlap, and the other for the Sides, which will expand a higher frequency range. Again, make both nodes dynamic, and this time select a pretty wide bandwidth. In the bass/kick example we were only concerned with fairly narrow ranges because both parts were short in duration, but with evolving and pattern-based sounds wider bandwidths tend to work better because they yield smoother responses. These clips demonstrate the effect.
Listen to the processed version, and notice how little congestion there is, and how the cello plucks burst through nicely thanks to the frequency conscious ducking. Once the cello plucks start to die away, the cluster takes over and a wash of lovely evolving frequencies hits you, thanks to the expansion process on the Sides — and by boosting the Sides you also get a subtle and engaging shift in the stereo width, which serves to enhance the pad sound even more.
There are countless ways to use M‑S in sound design, mixing and mastering — for instance, I've not even touched on the more conventional techniques such as M‑S mastering EQ and compression, and it would take more space than a Tolstoy novel to cover it all in detail. But I'll leave you with a few more ideas that you can look into if you want to take things further:
- In mixing and mastering, engineers often turn to M‑S equalisation to target only sounds that are panned wide (eg. a wall of electric guitars or a stereo synth pad), whilst leaving important elements like kick, snare and vocals in the middle untouched — or vice versa.
- M‑S EQ or compression can be really helpful for managing a stereo vocal group, as it allows you to process the lead vocal (usually largely in the Mid) or backing vocals (often more in the Sides) independently.
- When mixing drums, try compressing only the Mid — it usually allows you to compress the kick and snare without screwing up the cymbal or other percussion sounds that lie either side.
- A high-pass filter on the Sides combined with a low-pass filter on the Mid can help to give more focus to sounds panned centrally while tightening up the bottom end.
- Try using the Mid signal to key (side-chain) expansion on the Sides of a mix, to gently lift the air band every time the Mid rises.
- For a really different effect, use your DAW's automation to modulate a reverb between the Mid and Sides. It can create amazing motion effects. Or try auto–panning (or pan automation), place MSED post-pan, and mute the Mid. As with other techniques described here, everything in the middle remains clear and focused, but you still get that lovely panning effect.
The potential uses for M‑S are limited only by your imagination, and I implore you to get busy exploring the techniques that incorporate M‑S magic!
Eddie Bazil has been mixing and sound designing for well-known artists and synth/sampler companies since the 1980s. He is the brains behind audio-production video and eBook tutorial company Samplecraze. If this article has inspired you, check out the Samplecraze website: www.samplecraze.com
You'll sometimes see Mid and Sides called different things. For example, the Mid and Sides are often described as the 'Sum' and 'Difference', respectively, a description that's rooted in the maths used to encode L‑R channels to M‑S. Occasionally, including on the Fairchild 670 vari-mu compressor, you'll also encounter the archaic term 'Lat-Vert', which stems from vinyl mastering.
You'll find audio clips for each of the step-by-step exercises in a ZIP file of hi–res WAV audio examples in the righthand Media sidebar or use the link below. Feel free to download the source files and work through the exercises in your own DAW, or just to listen to the clips of the results.
Download | 52 MB