Q. Would you ever compress one channel of a stereo recording?

Published in SOS January 2013

I was called in half an hour before a concert to record a violin and piano concert at a school with the violinist Jennifer Pike (2002 Young Musician Of The Year). I opted for a stereo pair of Beyer MC930 mics in ORTF format on a tall stand, and recorded on a Tascam DR100 MkII portable digital recorder. However, I was very restricted as to where I could set up the mics, as the audience was already seated, and the duo ended up closer to the mics than I would have liked. As a result, the recording is heavily weighted in favour of the piano on the right-hand channel. The recording is primarily for a teacher who is very ill, but Jennifer Pike also wants a copy, so I’m very keen to make it as good as possible. I have tried to reduce the piano using equalisation, but that approach also compromised the violin, which is the star of the show. I also tried compressing the right channel on its own but I am in two minds about that approach. Does the theory say anything or is it down to the skill and judgement of the operator? I have no DAW and use a Yamaha AW2400 recorder for the editing.

Via SOS web site

SOS Technical Editor Hugh Robjohns replies: To answer your initial question, no, you wouldn’t normally compress just one side of a stereo signal, because, in the situation you describe, where a dominant instrument is well to the right-hand side, conventional unlinked stereo compression will cause the weaker source to wander about all over the place.

What happens is that when the piano gets busy, the compressor on the right-hand side will wind the level down, while the other channel’s compressor will be wondering what all the fuss is about and do nothing at all! Consequently, the violin, which was more or less central (and thus roughly equal in level in both channels), becomes much stronger in the left channel than the (now attenuated) right. That means that the violin’s position in the stereo image will scoot across to the left-hand side to get away from the scary loud piano!

When the pianist calms down and the right-channel compressor releases its gain reduction, the violin level will balance out again and its image will wander back towards the middle. If there’s a lot of room reverb or ambience, that will be similarly affected, and it will sound like someone is closing and opening heavy curtains down the left-hand side of the hall.

However, all is certainly not lost: there is a solution, and that is to process the signal in Mid/Side rather than left-right form. To do that, you’ll need to convert your ORTF recording to Mid/Side, then pass the Side channel through a mono compressor, and, finally, re-convert both Mid and Side signals back into normal left-right stereo. Note that the physical spacing of mics in an ORTF array can cause comb-filtering issues when the signals are combined to mono, so the Mid signal may suffer from the processing. Only experimentation will reveal whether this approach will deliver completely acceptable results, but it usually does.

In this setup, if your ORTF recording is converted to Mid/Side, the piano will be captured mainly by the side mic. The now louder piano will cause the compressor to attenuate, thus reducing the stereo width. The piano is now controlled in level, as desired, but does appear to move slightly inwards.

The way this approach works is that since the piano is heavily to one side of the left-right image, it will be very dominant in the Side signal and very weak in the Mid signal, while the reverse will be true for the violin. So in this way you can compress (and/or equalise) the piano in the Side channel without significantly affecting the violin’s dynamics (or tone), and thus control the piano’s maddest moments.

The other advantage is that, rather than suffering the blatant left-right image shifts that result from unlinked L-R stereo compression, this way of working in the M/S domain affects the stereo width instead. Consequently, the image will appear to ‘breathe’ in and out a little, getting narrower as the piano gets louder, and vice versa. Most people don’t notice this effect. If necessary, you can easily disguise the image width changes by overlaying some additional (ideally, carefully matched) artificial reverb to maintain a constant ambient width.

The M/S conversion can be done with some convoluted wiring around an analogue console (the technique has been described often in this magazine and on the web), but is probably easier to achieve in a DAW these days, using dedicated plug-ins. I note that you don’t have a DAW, but perhaps you have a friend who could oblige, or find someone on the SOS forums willing to help.

If the whole M/S processing idea is not viable for you, the other way of dealing with the problem is to think of your location recording as comprising two mono mics, rather than one stereo array. Just use whichever track provides the most acceptable mono balance. You can then pan that to the middle of a stereo track as a mono source and add some artificial reverb to reinstate a sense of stereo room space again. It’s a bodge, obviously, but will probably work quite acceptably, and it’s a technique that has saved me on more than one occasion in my career, when unforeseen disasters have happened and I’ve had to stage a rescue. You could even blend back in some of the other channel at a low-ish level, panned as required to give the piano a sense of width.