Fixing The Mix, Part 1: Salvage Techniques

Conventional wisdom states that you can't 'fix it in the mix', but that doesn't mean you can't try! Paul White discusses a few practical salvage techniques in the first part of this short series. This is the first article in a two‑part series.

In an ideal world, a producer decides what is to be recorded while the engineer is left to set levels, sort out routings, programme effects and so forth. In the typical project studio, however, the same person usually ends up having to do both jobs. If everything has been recorded perfectly, and if the musical arrangement is appropriate to the song, then it should be possible to set up an initial mix fairly quickly, using little or no EQ. But what do you do if the session didn't go so well and some of the recorded sounds are less than ideal? Worse still, you may be called upon to take over work on a poorly recorded job that was started at someone else's studio. That's when your skills at salvage work are really put to the test.

Over the past couple of months I've discussed ways of making good recordings and achieving good‑sounding mixes [see 'Match Before You Mix' SOS May/June 1996], hopefully by avoiding most of the problems that I'm going to touch upon during the course of this article. I've also gone on record as saying that there's only a limited amount you can do to fix a bad recording, but provided not everything is disastrously recorded, it's usually possible to make some improvement on any mix. My approach here will be to look at specific problems and then propose possible solutions or partial solutions using both basic studio equipment and, where appropriate, more advanced hardware.

The first option is invariably the most satisfactory, but for whatever reason, musicians often seem very reluctant to take it — if a track sounds wrong, either musically or technically, then simply re‑record it. While this option may not be open to professional producers working with busy international artistes, it's usually possible to get Fred from down the road to come back and redo his out‑of‑tune bass part.

What follows is a catalogue of increasingly desperate strategies that may be applied if re‑recording is not feasible. Recording problems break down into many types, some clearly defined, others less so. The most obvious technical problems relate to noise, distortion, EQ and spill, while the more artistic considerations include the wrong choice of sounds, poor timing, out‑of‑tune notes or just plain incorrect notes. As you might expect, the more serious the problem, the more difficult it is to do address it, but by using samplers and hard disk recorders, even out‑of‑time and out‑of‑tune playing can be improved if you have the patience.

Gating & Muting

Probably the most common technical problem is 'noise', which could include tape hiss, instrument amplifier hiss, amplifier hum, and general digital grot from budget synths and effects units. You can attack noise on several different levels, and if the contamination is serious, you may need to use two or more of the following processes in combination.

A noise gate is a very effective means of removing unwanted noise during the pauses between sounds, and you'd be surprised how much you can clean up a mix by gating any channels that contain parts that do not play all the way through the mix. For example, if the lead guitar solo only pops up in the middle of the song, you don't want any background noise on that track contributing to the mix the rest of the time, so bang a gate on it. It's usually worthwhile gating vocals also, because there are always pauses between words or phrases where nothing useful is happening. However, don't feel you have to gate out traces of breath noise, otherwise the vocal track may well end up sounding unnaturally sterile.

With any slowly decaying sounds, make sure you set an appropriately long gate release time. Even if some gated sounds appear to be slightly unnatural, there's a good chance they'll sound fine when you add reverb, especially when the rest of the mix is playing.

Closely related to gating is MIDI muting, often found on upmarket mixers. Although it isn't usually practical to use this to clean up tiny pauses between individual words or phrases, it is very useful nonetheless for temporarily shutting down tracks when they're not required in the mix. Obvious uses are to mute the guitar solo track until it's needed and to mute the vocal tracks during instrumental passages.

Noise Reduction

A more subtle alternative to gating and muting is the single‑ended noise reduction (SNR) unit. Gates and mutes can only remove noise when they are closed, and no signal is allowed to pass — they can't do anything about noise (such as hiss) that's audible over the top of the wanted signal. Single‑ended noise reduction units, however, employ variable‑frequency low‑pass filters that open and close according to the level and frequency content of the music being treated. Though by no means perfect, a decent SNR can significantly reduce audible noise, both during pauses and when material is playing, without inflicting too much damage on the wanted signal.

You'd be surprised how much you can clean up a mix by gating parts that do not play all the way through the mix.

As with any processor, if you ask too much of it, side‑effects will show up. In the case of SNRs, typical side‑effects can include audible dulling of the signal (if you try setting the threshold too high) or the over‑processing of reverb tails. For this reason, whenever possible it is best to route signals through an SNR prior to adding reverb, and if you have several channels that you think would benefit from the treatment, route them all to a stereo subgroup and patch the SNR unit into the group insert points.

Equalisation

An often forgotten ally in the war against noise is the equaliser, but the EQ in a typical mixing desk may not be flexible enough to do the job properly. A good graphic equaliser is more versatile, but for the ultimate in control, you need a good parametric EQ.

Changing the EQ of a signal is obviously going to have some effect on the overall sound of that signal, but quite often you'll find your mix includes sounds that occupy only a limited part of the audio spectrum. For example, an overdriven electric guitar contains no really deep bass and the top end rolls off very quickly above 4kHz or so, because of the limited response of a guitar amp's loudspeaker. This being the case, you can apply a sharp top cut above 4kHz and a low cut below 100Hz without changing the guitar sound too significantly. You'll need to experiment to discover the exact frequencies, but you'll find this technique very useful, not only with guitars (and bass guitars) but also with warm synth pads from cheap or vintage instruments. Careful use of EQ may also help to reduce the effect of finger noise on stringed instruments.

If you don't have a parametric equaliser, you can use the side‑chain filters in a Drawmer DS201 gate as equalisers, simply by leaving the gate in Key Listen mode. This particular gate has very steep shelving filters which can be used to 'bracket' the wanted sound. The Low filter will keep out bass rumble and hum while the High filter will reduce hiss and digital grot (my technical term for non‑random noise and/or whining produced by budget digital synths and sound cards!).

It is vitally important to keep in mind that noise is cumulative, and though individual tracks may appear reasonably clean when listened to in isolation, when you bring up all the mixer faders, the overall background noise can still be quite significant, especially if you have 16 or more tracks (including virtual MIDI tracks) on the go. Anything you can do to reduce the noise contribution will greatly improve the clarity of your finished mix.

Off‑Line De‑Noising

Some of the more serious hard disk recording systems offer digital de‑noising, which tends to provide a far more effective solution than the simple analogue processing discussed so far. Roland build a stand‑alone digital SNR unit which is well worth investigating.

De‑noising systems such as Digidesign's DINR software (or the more upmarket CEDAR and Sonic Solutions' NoNoise packages) may be used to reduce the level of background noise by several dBs without introducing audible side‑effects. The simpler systems typically allow you to reduce the level of noise by as much as 5 or 6dB without any side‑effects showing, while the more sophisticated systems can probably double that figure.

Many of these systems work by first splitting the signal into a very large number of frequency bands. A software algorithm is then used to decide how much noise is present in each band and a separate threshold is set for each band (normally an automatic procedure handled by the software) below which any residual signal is assumed to be noise. It's almost like having separate intelligent noise gates working on several hundred very narrow frequency bands. If you push such systems too far, however, whatever noise is remaining may be heard occasionally peeping through the various filters. This produces a sound almost like distant wind chimes. Again, the more sophisticated systems are less likely to suffer from this problem, but if you hear it, at least you'll now know what it is.

These digital solutions can be applied to individual tracks (or sections of tracks), by copying them from tape onto hard disk along with their timecode reference. Using timecode is a good idea, otherwise you may find it difficult to copy the processed audio back onto the multitrack in the right place after you've cleaned it. If you must work without timecode, then deal with short sections and 'fly' each one in by hand as though you were playing back from a sampler. Ideally you should put the processed audio onto a spare tape track, but if you don't have any, you may have to risk recording it back over the original.

Non‑Random Noise

Mains hum or buzz from fluorescent striplights can be a serious problem, because as well as the fundamental frequency of 50Hz (or 60Hz depending on which country you live in), there's often a whole string of related harmonics present that run right across the audio spectrum. High‑pass filtering may well take be used to remove the hum component altogether, but you'll still be left with an irritating buzz.

Roland's single‑ended noise reduction devices are particularly good at dealing with this problem, because they employ a whole series of digital filters which create narrow notches in the frequency spectrum corresponding to both the odd and even harmonics of the fundamental hum frequency. A similar technique is employed by Digidesign's DINR software, when used in hum removal mode, and a clever search algorithm enables the filters to lock onto the hum automatically. However, attempting to attenuate the hum by too much can cause audible side‑effects — the notch filters start to work on the wanted audio, creating a slight 'phasey' effect.

I don't know of any technique that is entirely successful for rescuing a distorted lead vocal...

A lower tech approach is to use a multi‑band parametric equaliser with the lowest band tuned to the fundamental hum frequency (eg. 50Hz) and the following filters tuned to double (100Hz), three times (150Hz), and four times (200Hz) that unwanted frequency. The filters should be set for the highest Q (ie. a very narrow band) and the cut function applied until the hum problem is reduced.

This method will only be partially successful, because the multi‑band parametric won't have enough filter bands to tackle all the hum harmonics, but you should still be able to make a noticeable improvement. In most situations, the fundamental plus the first few harmonics will be the loudest elements, with the higher harmonics being much quieter. Each filter will have to be tuned in by ear and it may be useful to bypass all the filters apart from the one you're tuning, so that you can hear exactly what effect each filter is giving you. Once you've optimised the individual filter settings, you can switch all the other filters back on.

As mentioned earlier, the Low filter on a Drawmer DS201 gate can also be useful in removing the low frequency components of hum. What's more, the hum may only be clearly audible during pauses in the signal, so gating could be applied here in the same way as to tackle broad‑band hiss.

Distortion

Distortion is a particularly difficult problem to deal with and I don't know of any technique that is entirely successful for rescuing a distorted lead vocal or overdriven acoustic instrument sound. In theory, there are elaborate digital processes that could be used to reconstruct distorted waveforms, but in the smaller studio the most useful tool is probably the equaliser.

Before you reach for the equaliser, there may be an even better way out. If the distortion occurs only in one or two places, then my first thought would be to look for identical passages (or even individual notes) elsewhere in the music. If these exist, you can sample them and then drop them in over the top of the original sections. As always, work on a spare track using a copy of the original part if at all possible, because it's very easy to make something worse if the drop‑in doesn't go your way. If you have access to a hard disk system, this is where you'll come to appreciate the joys of cut‑and‑paste editing.

To tackle distortion successfully, you need to understand a little about what it is. Distortion is caused when a circuit is driven into clipping or when an analogue tape is driven into saturation; the usual outcome is that a whole series of harmonics are generated, some of which can sound very unpleasant. Because the harmonics are higher in frequency than the basic pitch of the sound being distorted, it is sometimes possible to reduce their effect by applying high‑cut filtering, but in most instances this will also affect the wanted sound. Even so, a slightly dull acoustic guitar might sound better than a distorted one.

Just occasionally, you'll find you can hide undesirable distortion by applying even more distortion to the whole track, but this is more likely to work on an artistic level when applied to guitars and synths than it is to, say, vocals or piano. Having said that, distorted drums and even vocals can be useful in certain styles of dance music, though you're unlikely to get away with it if you're mixing a ballad.

Remember. . .

Whatever you do, try to get away from the old idea that you can hide a serious error by recording something over the top to distract the listener's attention. If it is a drumstick click , you might just be able to overdub a vibraslap or a tambourine, but it has to be artistically appropriate — and even then, it will only work if the sins being masked are very minor. Back in the hippy '70s, we used to joke that if we had a lousy mix, we could always flange the whole song!

Next month, I'll be looking at the evils of vocal popping, out‑of‑time playing, tuning problems and musically inappropriate sounds.