From Mastering

The Beginner’s Guide to Mid/Side Processing

Mid/side processing is a powerful technique that can be extremely useful in the right circumstance. But it can be a little difficult to get your head around and understand exactly what it’s for. Here we look at what mid/side processing is, how to achieve it and what to use it for.

What is mid/side?

Mid/side – sometimes called sum/difference – is a technique whereby a stereo recording is changed from the usual left and right to mid and side. The ‘mid’ or ‘sum’ is the mono portion of the recording (everything the left and right signals have in common) and the side is the stereo portion (everything else).

Using mid/side, we can more easily zone in on different parts of the stereo recording for EQ or compression say, or even manipuate the stereo image itself.

Manipulating M/S with free plugins

The easiest way of constructing an M/S matrix is to use a free plug-in like Voxengo’s MSED. This let’s you chain the relative gain of the mid and side signals to effect stereo width (another free plugin for this is included in BlueCat’s free gain package).

MSED will also act as an M/S encoder and decoder. Set it to encode and the left portion of the signal will become the M and the right, the S. You can then use any plugins that allow you to manipulate the L and R channels separately (but that don’t include an M/S matrix themselves) to manipulate the M/S signal, before sticking in another instance of MSED set to decode to convert it back to stereo.

How to construct a mid/side matrix without plugins

As easy as using MSED is, for additional flexibility (and understanding of what’s happening) it’s worth constructing your own M/S matrix at least once or twice.

The principle is simple – the M signal is everything common to the L and R signals, and the S is everything that is not. So, to construct your own M/S matrix:

  1. Split your stereo recording into 2 mono tracks – 1&2. Pan centre.
  2. Make duplicates of each one – 3&4. Pan centre.
  3. Send each pair to a separate group – buss 1&buss 2
  4. Invert the phase on track 4.
  5. Create 2 new buses – M&S
  6. Send buss 1 to M and buss 2 to S
  7. Pan M hard left and S hard right
  8. Invert the phase on buss S

This way you can now use any plugin that works in mono to manipulate the M&S signals separately.

Common uses of M/S processing

M/S processing is useful whenever you need to really zone in on a particular part of a stereo recording and the stems aren’t available. For example, if there’s an edgy sounding guitar at the edge of the field you can use an EQ cut on the S channel only to smooth it out.

How successful this sort of thing is depends on where instruments are panned, sometimes you just have to leave it be. Some common uses of M/S processing though are:

  • Stereo width
  • Increasing/decreasing stereo width by adjusting the relative gain (more S = wider). This has fewer artifacts than using ‘stereo widener’ plug-ins that often make use of comb filtering. Another approach is to use a high boost (8-16kHz) on the S channel to increase top end width.
  • Mono-ing bass
  • Tighten up bass by using a high pass filter at around 100Hz on the S channel.
  • De-essing
  • Use a de-esser on only the M channel (for lead vocals) or S (for backing vocals) in a complete mix.

As you can see, the sorts of problems that M/S processing can address are usually better dealt with earlier in the chain – but when that’s not an option, mid/side can be tremendously powerful.

Low Frequency Management in Mixing and Mastering Part 1 – Filters and EQ

Low frequency content presents one of the biggest problems in mixing and mastering. Getting it right is a fine balancing act: too little and your track lacks weight, too much and everything else is overwhelmed.

For many people, if you sort out the low end the rest will follow. Assuming you’ve got a monitoring chain that lets you hear the low end with some degree of confidence (more on that next time), how do you do it?

High pass filters in mixing

Most instruments carry some low frequency content in the signal, but other than the kick and the bass you don’t really need anything else down there.

One of the most useful mixing tips you will ever learn is to high pass filter everything except the bass and the kick.

  • Find the cleanest EQ you’ve got – I use Fabfilter’s Pro-Q but your DAW’s stock EQ will probably do fine.
  • Sweep up on each instrument until you hear a difference.
  • Scale it back until you don’t.

This gets rid of all the ‘junk’ bass that’s cluttering up your low end, as well as freeing up energy for higher frequencies. As a rule of thumb, 100Hz is pretty safe, but you can often go higher on some instruments depending on how dense the mix is.

Clearing space for the kick

Once you’ve got everything else out the way, you might still find your bass masking your kick.

  • Use a frequency analyser (try BlueCat’s free FreqAnalyst if you don’t already have one) to see what frequency your bass drum is at.
  • Make a 2-3dB cut in the bass instrument at that frequency. Keep the Q fairly gentle (you’re at the deep end).
  • You may need to make a compensatory boost higher up.

Then of course you can always use a sidechain compressor on the bass using the kick as the key. But that’s another post for another time.

High pass filters in mastering

You might have heard of the infamous ’30Hz rumble cut’. Often advocated in dance circles, it is common practice to put a high pass filter at 30Hz across the master buss for testing in clubs, to get rid of any unsightly low frequencies that will only show up on the biggest of subs.

This is certainly a good idea when you can’t hear that far down on your monitors. But as with any master processing, you have to be more careful with high pass filters across the master buss than on individual tracks.

For starters, if you’ve been filtering conscientiously while mixing, you shouldn’t have too much unwanted noise down there anyway, and the extra cut might take too much out.

Additionally, any filter you use will affect frequencies higher up in the spectrum by virtue of the phase shift inherent in any (non-linear) EQ process. The alternative is to use a linear phase filter, but again, this comes with its own baggage – I often find linear phase processors make things sound ‘smaller’ and make transients softer.

For what it’s worth, I find the filters in the Sonnox EQ to be pretty good. Steinberg’s PostFilter (in Nuendo and Wavelab) is great – particularly for very steep slopes. Fabfilter Pro-Q can do the job sometimes, and is very useful when linear phase really is needed.

But I nearly always use the butterworth or bessel curve in the TC Electronic MD3. The difference is startling. The whole track becomes more open and the bass very often sounds stronger as a result of the cut. The same can’t always be said for the others.

Bottom line? Don’t use any old EQ you have lying around for a master buss rumble cut. Try out everything you’ve got and see what works best.

Next time – making bass translate.

How Loud is Loud?

Last time we touched on a few of the things to think about before sending your tracks off for mastering. This time I wanted to take a look in a bit more detail at the loudness issue – something of a hot button topic in this day and age.

“I want it loud but without losing dynamics”

This is probably the number 1 sentence used when clients try and tell us how loud they want it, and is the main reason we ask for reference tracks.

Essentially, this sentence sums up arguably the main goal of mastering, but there is a fairly wide spectrum within which that objective is met. And whether that’s loud enough for you depends on what genre you’re working in and where you stand on the infamous ‘loudness wars’.

One person’s loud is another person’s soft and vice versa.

A brief history of loudness

We all know the story by now so I won’t delve into it in too much detail, but it’s worth thinking about when considering how loud you want your record.

Put simply, pop music used to be much much quieter. Partly as a consequence of technology (they didn’t have brickwall limiters in the 50s) and partly as a result of good engineering practices.

As the technology developed it became easier to make things louder and louder. This became an arms race as everyone wanted their tune to sound the hottest on the jukebox and/or radio. This practice reached its nadir in the ’90s with the introduction of digital technology like brickwall limiters and the practice of clipping high-end A/D converters.

Nowadays there is a move back to more dynamic masters, but it’s a difficult transition to make as people still don’t want their tracks to sound quieter than their favourite records from the 90s and onwards. Added to that, new technology has made it easier to get to those levels without completely mauling the sounds (I’m looking at you Izotope IRC III).

So where do you stand?

So when considering how loud you want to go, it really is a case of deciding where you want to stand on this.

One of the biggest developments in mastering moving back to more dynamic recordings in the introduction of Bob Katz’s K meters. If you want to get really involved with this, take a look at a few of your favourite recordings using a meter that will display using these scales – Voxengo’s SPAN is one that does it for free.

We consider ‘loud’ to be just touching the red on a K-12 scale in the loud sections of a song. Although Bob Katz would argue that’s still too much (it’s way over on a K-14) we think it’s a nice compromise and our masters certainly don’t sound squashed at this level.

In dance music there is still a tendency to push things way over, and there’s no way round that as it will make a difference in the clubs. The sad thing is it still means plenty of DJs are pushing things into digital distortion. That’s where we draw the line, but dance still makes us see plenty of red on a K-12 meter.

A Paul Weller story

To sum up, we once did a track for a Mod revival band who sent us two reference tracks by Paul Weller. One was a cut from the 80s and another from the mid-2000s.

The former was very quiet, not even peaking to full scale (peaks were well below 0dB), the latter was bright, brash, harsh sounding and very loud. The band not having expressed a preference, we pitched it for about halfway between. They were very happy with the result.

I think that sums up nicely just about where we should be on the loudness curve. We don’t need to make things as quiet as they were pre-1990s as we have the technology to get things louder, using up all the bits, without mauling recordings like they did in the ’90s. But with that comes a responsibility to not push things as far as we possibly can, because that doesn’t make for an enjoyable listen.