Paul White offers a little practical advice on the application of signal processing during recording and mixing.
Over the years, SOS has covered the workings of just about every type of signal processor imaginable, and new ones are coming along all the time. But it's not always easy to decide when these devices should be used and when they're best left switched off. If you've just spent the best part of a grand on a new compressor designed to make everything sound warm and cuddly, there's a great temptation to use it on everything — but this isn't a good idea. Processors are basically tools designed to do a job; if the job doesn't need doing, there's no point in using the tool. The idea of this article is to look at the more common types of signal processor used in the studio and to see where and how they might best be employed.
Compressors
Compressors remain one of the least understood, most abused processors in the recordist's armoury. The main job of a compressor is to reduce the dynamic range of a signal — low‑level signals can be brought up and high‑level signals brought down — but under what circumstances is this a good idea, and what, if any, are the trade‑offs? One of the side effects of compression is that the sound being processed changes: at the very least, its envelope is modified as the compressor goes through its attack and release phases. Perhaps the most benign side effect is that of added 'warmth', but go too far and you can introduce audible pumping, which doesn't usually sound that great unless you need it as an effect.
Most vocalists need compression to keep their levels even; bringing down the peaks means that you can make the average signal level higher, and this results in a tighter, more confident sound that sits well in a mix. However, don't just rush in and apply maximum compression — listen to the singer run through the song, watch the record level meters, and try to decide for yourself whether the voice needs a lot of control or simply a light touch. In any event, it's better to under‑compress during recording, because you can always apply more compression when you come to mix, whereas over‑compression is virtually impossible to reverse.
It's when you're processing vocals that the subtle differences between compressors come to light. Some models, such as the Focusrite Green I tried recently, manage to even up the level without making the process in any way obvious — it just sounds as though you have a beautifully controlled vocalist — whereas Joemeek units, for example, introduce a deliberate 'character' to the sound, which makes the vocal sound slightly larger than life. I can't say which is best, because different jobs need a different approach: the ideal situation would be to have at least one fairly transparent model and one 'warm' compressor in your rack. Some units attempt to do both, but few are entirely successful.
As a rule, if you want to make a transparent compressor produce more of an obvious effect, you can increase the ratio and at the same time shorten the release time. The shorter the release, the more obvious the gain pumping effect of the compressor. As you make these adjustments, watch the gain reduction meter and adjust the threshold control to get the amount of gain reduction you need. When you're recording, 5 or 6dB of gain reduction may be enough; if you're compressing a track that's been recorded dry, you might want to use as much as 12dB of gain reduction, or even more if you're after a strong effect. A tip here — if you're not too sure about adjusting the compressor attack and release times, use a model with an Auto setting. If you don't have one, set the attack to its fastest, and the release time to between 300ms and half a second. Soft‑knee compression usually provides the smoothest compression, but hard‑knee models give tighter control and may be better for creating hard compression effects.
Acoustic guitars sound smoother and have better sustain when compressed.
Before moving onto other areas of application for compressors, I need to mention noise. Compressors are not inherently noisy devices, but because they reduce the dynamic range of the signal being treated, some make‑up gain is needed to bring the peak signal level back to where it originally was. In other words, although compressors really just turn down loud sounds, once you've adjusted the make‑up gain control, the loud sounds are back where they were and the quieter sounds are much louder. The quietest of quiet sounds is noise, so if you're compressing to achieve 10dB of noise reduction, any noise that happens to be part of the signal will also be increased by 10dB for all input levels that are below the threshold. Periods of silence between words or phrases are most vulnerable, as it's here that the compressor gain is highest. As well as starting off with the cleanest signal you can, it might be wise to gate the signal immediately before it enters the compressor. Some compressors have built‑in expander gates for this purpose: used properly, they really can make a difference.
Other sound sources you might want to compress include acoustic guitar, clean electric guitar, and bass guitar. Bass in particular tends to fluctuate in level by a considerable amount, so don't be afraid to hit it hard with the compressor. As well as keeping those peaks from clipping your recorder, compression will fatten up the tone and save weakly played notes from getting completely lost. Acoustic guitars sound smoother and have better sustain when compressed; with clean electric guitars, you can increase the compressor attack time slightly to get a nice 'ping' at the start of each note.
Compression can give mildly distorted guitars more sustain, but I've never seen much benefit from compressing heavily distorted guitars — the mechanism of an overdriven amp or pedal essentially produces a limited output level anyway. All that will happen is that, during any pauses, the background hum and hiss will rise even further than usual.
Acoustic instruments tend to reveal compressor artifacts more obviously than electronic instruments, so for 'invisible' gain control, use your most transparent compressor, set to soft‑knee if there's a choice, and a fairly low ratio — probably 4:1 or even less. Set the threshold to give you just as much gain reduction as you really need, and if there's any sign of pumping, increase the release time until it stops. A fast attack time will retain the natural attack of an instrument or voice; if you need a more defined attack for a percussive instrument, though, the compressor attack time can be slowed slightly.
When it comes to electronic sound sources, you can compress these in exactly the same way as you would their acoustic counterparts, but be aware that many of the sounds used in drum machines and synths are heavily compressed already. By all means experiment by adding more compression, but don't be surprised if you can't add much more punch to a sound.
Compressors should be switched to stereo link mode when they're treating stereo material, and I know some engineers and producers who like to compress all their finished stereo mixes, to push the average energy level even higher. Whether this is a good thing to do is almost entirely a matter of taste, but I strongly suggest that if you plan to compress your entire mix, you actually monitor the compressor output while mixing, rather than simply creating a stereo DAT tape then compressing it later. The reason for this is that compressors can change the way the balance of the mix sounds, and vocal levels in particular can seem to vary after global compression. If you compress as you mix, and monitor post‑compression, at least you'll know exactly what you're getting.
Gates
Gates are relatively simple devices, and they can't perform miracles, but they can be a great help in the fight against noise, particularly at the mixing stage. Gates and expanders can only silence noise in the absence of a wanted signal — they can't 'skim off' the noise from an existing sound. Think of them as a switch that turns off when the signal level falls below a threshold, and you have the right idea. Because a badly set‑up gate can fail to open in the presence of quiet sounds, or close too quickly, chopping off the tail end of a steadily decaying sound, gating while recording is risky. You might get it right — but on the other hand, you could ruin the only good performance of the day. Far better to gate when mixing, as you then have the opportunity to run through the track as often as you like when setting up the gate. Furthermore, any noise added to the track during the recording process will also be gated
It makes good sense to leave adding EQ to the mixing stage wherever possible, because if you add the wrong EQ it can be hard or even impossible to compensate for it later.
There are three main gate parameters, controlling attack, release and threshold. A fast attack means you won't miss any sharp transients, such as drums, but on a more slowly attacking sound you might hear a click as the gate opens. For this reason, try to match the gate attack time to the attack characteristics of the sound you're dealing with. Similarly, the gate decay sets how quickly the signal level fades away when the gate closes; if this is set too fast, a long decay, such as a reverb tail, might get cut short. On the other hand, if you use a long release time with a drum beat, you'll hear the noise that follows the drum beat fade gradually rather than quickly. Again, match signal release and gate release times for the best result. A useful tip is to always gate a signal before adding reverb or other time‑delay effects, if you can. Not only will this prevent the reverb decay being clipped by the gate, it'll also help disguise any gating artifacts that might affect the dry signal.
If you're mixing a number of separate audio tracks, and maybe a few virtual MIDI driven tracks as well, noise can soon build up. Even if the individual sounds are fairly quiet, the added noise from 16 tracks of tape plus a rack of synths can be significant. Ideally, any instrument or voice that isn't playing all the time should be gated, so that when it's resting, the track really is completely silent. Sadly, this means a lot of gates patched into your console insert points — you can't just use a single gate and connect it to your aux send. Fortunately, you can usually get by without gating every track, simply by subgrouping the sounds that tend to play together, then gating the subgroups. Stereo subgroups will require a pair of gates set to stereo link mode. You can also put individual gates on problem tracks, such as that noisy electric guitar or hissy synth module. Having a couple of quad gates to hand is very useful, even if they're fairly simple. It's nice to have key filters, external keying, attenuation controls and so on, but, for most routine jobs, you just need a simple gate.
Some people try to gate a whole stereo mix, then wonder why it doesn't get any quieter. As I said earlier, when a gate is passing signal, it can't do anything about the noise, and most mixes have something going on all the time, even if it's just the reverb at the end of a drum fill. In most cases, the gate will open at the start of the song and stay open until the end. In this respect, a gate can be useful for cleaning up a noisy song start, and for ensuring that the mix fades into true silence at the end, but that's about all it can do.
Equalisers
EQ is basically just a jargon term for tone control, which essentially allows you to turn the level of selected sections of the audio spectrum up and down. The subject of EQ is a controversial one, but I'd like to simplify it as much as possible. Firstly, natural sounds invariably sound most natural if they're well recorded and then subjected to as little EQ as possible. Change the mic, change the mic position, even change the player, but don't turn on the EQ until you really need it. Secondly, gentle EQ rolloffs at either end of the spectrum sound fairly benign, and you can also make fairly deep cuts over a narrow region without it sounding odd, but significant EQ boost over narrow regions normally sticks out like a sore thumb, especially with cheap equalisers. From this you can extract the general rule that boost should, where possible, be wider and less pronounced than cuts, which can be almost as tight and deep as you like. It also makes good sense to leave adding EQ to the mixing stage wherever possible, because if you add the wrong EQ it can be hard or even impossible to compensate for it later.
The other universal rule is that the people who could really use decent EQ don't have the access to it. You might think that those knobs on your mixer labelled EQ are up to the job, but in my experience even the better mid‑price console EQ falls a long way short of what you get from a quality outboard equaliser. If you use a really nice EQ, you'll find you can add more boost without messing the sound up, and, more importantly, the sound seems to stay natural, rather than becoming honky, boomy, nasal, or any of those other adjectives commonly used to describe what you hear when you've tried to do any major equalising using the EQ on a home recording mixer.
If you're planning on using a lot of EQ — and sometimes heavy‑handed EQ is the only way to get a result — treat yourself to at least one really good outboard EQ box. That way, you can use your console EQ for rolling off excessive low end, or for fine‑tuning sounds, and your good EQ for occasions when a sound needs to be completely reshaped.
Next month I'll be moving on to effects, to see where they are of benefit, and where they can cause more harm than good.