Stereo Width Management (on the Cheap)
Home producers working with stereo samples may have good reason to reign in the stereo-ness of their sounds. Here’s an easy, and free, way to accomplish this.
by David Baer, Jan. 2015
This discussion probably won’t be of interest to those who record in the studio and mix from recorded tracks. However, home producers who build up productions in the box using sample players and synths can be faced with a problem in mixing. If your samples or synth presets are blatantly stereo and you do nothing but throw them in the mix that way, you may be getting a very unsatisfactory result with respect to a realistic sound stage.
For example, you might be using a piano sample set that distributes the panorama of the sounds the way the pianist would hear them: low notes on the left and high ones on the right. But even with a less hokey situation, if there’s stereo content, it may be at cross purposes with how you want your mix to sound. If the sound has a reverb insert, then you probably want to turn that off and defer reverb treatment to a bus so that all the instruments/vocalists in the ensemble share the same acoustic space. But even with reverb disabled, you still may have considerable width in the sample or preset and need another way to realistically fit that into the mix’s soundstage.
An easy solution is to collapse the stereo track to mono and use the pan control to place it in the mix. But in doing so, you will have lost all the width. Maybe you’d like to keep some of it, but restrict it to a narrower wedge in the panorama. This is easy to do, and it can be done with one of several excellent free plug-ins. Read on to discover how.
The Majesty of Mid, the Mystery of Side
Before we go further, let’s take time for a brief discussion of Mid/Side (hereafter “MS”) signal processing, because that’s the key to how we can control stereo width. Even if you know what MS is all about, you may find the following brief discussion of interest. I gained several insights into this subject when playing around with how to put these concepts into words.
Now, there’s really no mystery in Mid or Side – I just couldn’t resist riffing on my favorite Spinal Tap lyric. As you’ll see, this boils down to the most simple of arithmetic – there’s not a trig function or logarithm in sight.
Stereo signals consist of a Left/Right (hereafter “LR”) pair of signals – this we all know. The LR data can be converted to MS data and back again in lossless fashion. There’s a fascinating history to LR/MS signal representation (it’s the key to how stereo LPs work, for example), but we won’t explore that here. Just the facts, Ma’am!
There are several common pairs of formulas for MS encoding/decoding. It doesn’t matter if the audio signals are continuous voltages or discreet pairs of pulse-code sample values. At any time there are two LR values and two MS values. With either, we can easily find the other with the pairs of formulas. The only key is that the paired formulas must be used in going both directions. Here they are:
Note that the second version of Option 2 is mathematically the same as the first version, but in using the second we’ve cleverly avoided needing any logarithms implied by the inclusion of dB amounts.
There are no major surprises here. We’d expect the M signal to be the same one we’d get when collapsing a stereo signal to mono in our DAW, and we can see that’s exactly what’s happening. Likewise, when the S signal is continuously zero, we have no stereo; converting such to LR results in the same signal in both L and R. So, here’s our first (and only!) strategy: to make a stereo signal less wide, we just reduce the S amplitude.
But before we leave these formulas, let’s look at a few more interesting aspects of this subject. First of all, notice that with option 2, we could use the same converter for LR to MS and for MS to LR. No matter what the input was, the output would be guaranteed to be the correct values. With options 1 and 3, we have to be careful to use the right conversion mode depending on whether it’s MS to LR or LR to MS.
Another interesting bit of trivia: note that if we invert the S signal (essentially multiplying S by -1), the result is that the L and R signals are swapped when we convert back to LR. In other words, we’ve swapped the channels.
Now, if we have L and R signals that are identical, we convert to MS and end up with a zero S, just as stated earlier. If, on the other hand, L = -R (they are identical except that one is an inverted copy of the other), then it’s M that ends up zero – meaning that a stereo to mono conversion in our DAW will result in silence. If you did not appreciate the role of coherent phase in mono-compatibility of stereo signals, maybe this will drive home the point.
Narrowing Things Down
But let’s get down to business. There are several stereo imaging plug-ins available that can do the job of narrowing your stereo. I happen to own two, not because I needed them, but because they came as part of bundles that I purchased for other reasons. One of these is the Nomad Factory Stereo Imager ST2S.
This is a simple one-trick pony. It can used to narrow or widen a stereo image. We’ll talk about the consequence of widening images later. However, we can easily guess at exactly what’s happening under the covers. The LR stereo signal is converted to MS, the S is adjusted accordingly (reduced for narrowing, increased for widening) and the signal is converted back to LR. If you narrow the image and want to place it in a specific location in the panorama, you do it with your DAW pan control.
A more sophisticated tool is the Waves S1 Stereo Imager. This can do the narrowing/widening but it also has a few other tricks. Specifically, it can also pan the output using the control labelled Rotation.
You can further finesse the output with the Asymmetry control, which further dictates the level of the sound within its L/R orientation. A picture is way better than text in this case to show the possibilities. In the three images below, we see a signal that’s narrowed to 50% of its original width, then placed to the right, and finally adjusted so the (now) center extreme is attenuated.
Now, if you happen to have a stereo width plug-in, then great – you’re all set. But if you do not, be of good cheer. There’s absolutely no need to spend your hard-earned money on one.
DIY Stereo Width Control
There are several of free plug-ins available for use in LR/MS conversion scenarios. We’ll look at two of them here: Blue Cat’s dual-channel Gain control and Voxengo’s MSED plug-in. Both are notable for being PC and Mac compatible and for being available in 32 and 64-bit versions, and the Voxengo plug-in even is available as VST 3, which is pretty impressive for a freebie. They can be downloaded from the following links:
Let’s start with the Blue Cat Gain control. There are actually three of these that come in a suite: a mono gain control, a linked-stereo gain and a dual-channel gain. We are interested in the third of these, which have independent adjustment knobs for the left and right channel of the stereo signal. What’s of particular interest is the Mid-Side button. When pressed, the stereo signal is converted to MS on the way in and back to LR on the way out. The two level controls become M and S level controls. So, with a single instance of this control we can do everything the Nomad Factory ST2S does.
If we want to narrow the signal, we simple dial in a lower value on the Side control and that’s it. The question becomes: how much? You can probably guess this correctly just using common sense, which would suggest that to narrow the stereo image to become 50% of the original perceived width, reduce the Side knob by 6dB. To go down to 25%, reduce Side another 6 dB (for a total of 12dB reduction), and so on. Well, just to make sure, I ran some tests (using just the Waves S1, however, not the ST2S). Those tests bore out the common sense guesses precisely.
Now, with the Blue Cat solution, we cannot reduce Side beyond 24dB, so we cannot achieve complete narrowness (i.e., mono). But with a 24dB reduction, we get things down to 6.25% of the full width, which should be plenty.
To pan the resultant narrowed signal, we can use the DAW’s pan control. If we want to further finesse the result to have something akin to using the Asymmetry control in the Wave’s S1, then we can simply put another instance of the Blue Cat gain plug-in behind the one used for narrowing. This one we leave in normal (non-Mid/Side) mode and adjust the left and right gains to taste. We don’t have the advantage of a handy little graphic to help us visualize what we’re after, but just adjusting by ear will handily get the job done.
<Voxengo MSED graphic – right justified>
The Voxengo MSED plug-in is another great option. It operates in three modes: Encode (LR to MS), Inline, and Decode (MS to LR). As you can see, there are independent level adjusters for Mid and Side. To narrow the image, put MSED in Inline mode, reduce Side level, and you’re done. But in the case of MSED, we also have the Mid and Side Pan controls. They do largely what you’d expect, but the manual doesn’t really say what’s going on under the covers. Imagining what happens to pan Mid is easy. What does it mean to pan Side? The conclusion I came to was to just not worry about it. Adjust by ear and move on to more important undertakings.
We’ve concentrated on how we can make a stereo image narrower so far, but the opposite is obviously also available. If we can narrow by reducing the level of Side, the increasing the level should produce more wideness, right? The answer is a qualified “yes”, but be careful. Clearly, either of our freebie solutions can be used in this fashion, since the Side levels can be increased as well as decreased in either. So just try it. Adjust by ear and see if the results work. Depending on the material, this trick can sometimes work extremely effectively, but other times not so much.
The danger is in totally annihilating mono-compatibility. If keeping your mix mono-compatible is important, then checking that you’re not degrading the mono playback is particularly important when using the widening trick. By now, the reason should be obvious: if you overcook the S signal, you may be introducing pronounced out-of-phase-ness in the LR domain. If your DAW has a phase meter, you can use it for this purpose. Otherwise, just listen and your ears will tell you everything you need to know – as usual.