Although we've run many articles on recording vocals in the past, one of the most common questions we get asked still concerns getting a decent recorded vocal sound — or processing it so that it sits well in the track. In a lot of cases it's the computer studio users who are most concerned about vocals, and in addition to techniques they raise subjects such as the quality of plug-in processing, the importance of having the right microphone, whether a good mic preamp makes a difference, and so on. To address some of these issues, I'd like to take a quick run through the whole recording and mixing chain so that I can concentrate on those areas where quality can be maintained or lost.
A vocal recording starts at the microphone, but before even getting into the issues of mic choice and mic placement, there's the matter of the recording location to sort out — and it goes without saying that this should be isolated as much as is possible from the physical noise generated by the computer's fans and drives. A lot of people think they need to buy better gear to sort out a vocal issue, but when you get to the bottom of the problem, it's often down to the room and its influence on the sound.
Photos: Richard Ecclestone
You don't need to do anything too fancy to record vocals, but the mic should be well away from any walls, and the area directly behind the singer should be non-reflective. This could be an area of foam tiles or it could be a duvet, but one point to watch out for is that, in rooms where a lot of damping material has been applied, you'll often find that it only absorbs effectively down to around 250-300Hz. So what actually happens is that frequencies below 300Hz are allowed to predominate, making the sound seem congested or boxy.
This is a particularly common fault in small vocal booths, where introducing some reflective scattering surfaces helps balance out the low-frequency resonances. Worse still are rooms that have the walls carpeted, as carpet only absorbs higher frequencies, so you end up with a sound that's both boxy and dull. It's often best then to record in an ordinary domestic room, probably with a carpet or rug on the floor and a duvet hanging loosely behind the singer. If the room is still too lively, hanging further absorbers to the sides of the singer will help, but it's very rare that you need to go much further than that.
Set up the mic a couple of feet from the centre of the room and make a test recording using no processing at all to see if the basic tonality is OK. If it's not, the chances are that the problem is with the room or the mic placement, so try more hanging absorbers and move the mic around relative to the walls.
Now the choice of mic. I've had people call asking if they should swap their £1000 tube mic for a £2000 tube mic, but in the majority of cases their dissatisfaction with their sound has not been the fault of the microphone. Yes, a good mic does sound better than a cheap one, but a well-designed budget microphone can produce superb results when used correctly — even a stage dynamic mic such as an SM58 can sound great! Although you can get away with using a dynamic mic, the low cost of imported capacitor mics is now such that anyone can afford one, and, as they're a lot more sensitive than dynamic mics, preamp noise is less of a concern, because you don't need to use so much gain to get a recordable signal.
Capacitor mics will give you a more open-sounding top end, and in the majority of cases they're the best choice. Note that some back-electret models — one example being the popular AKG C1000S — are designed primarily for live use, so their sensitivity is more akin to that of a dynamic mic than a true capacitor model. They're fine for recording close vocals, but perhaps less well suited to quiet or distant sound sources. From a noise and tonality point of view, it's probably fair to say that choosing one of the many large-diaphragm cardioid models on the market offers the best balance between cost and quality. Once you've perfected your vocal recording technique, then is the time to start checking out better mics and preamps.
It's always best to use a shockmount with any vocal mic, and unless low-frequency noise is a real problem, record without the low-cut filter, as you can do your filtering in software if need be. Equally important is the use of a pop shield, without which plosive 'B' and 'P' sounds are likely to cause popping. These particular vocal utterances are accompanied by a directional gust of air which, in the absence of a pop shield, slams into the diaphragm of the mic like a hurricane into a garden shed and results in a huge low-frequency signal that's often so high in level that it clips the preamp. Trying to shift popping at the mixing stage using filtering is a bit like trying to level a mountain using a garden strimmer — it's unlikely to work! The pop shield should be around three inches from the mic and the singer between six and 12 inches from the pop shield, though better-trained singers may vary their distance to help maintain a more even level.
If there are tonal elements in the voice being recorded that you don't like, try moving the mic relative to the singer before thinking about EQ. Often moving the mic up or down by a few inches helps, and in some cases moving the singer further away works too. On the other hand, if you have a singer with a quiet voice and you need to add a bit more vocal weight, working closer to a cardioid mic can help, as the proximity effect adds some useful bass-end reinforcement.
Once you're at this stage, check the peak levels on your preamp to ensure you're not approaching clipping on any of the loud notes and also check the recording level on your computer or recorder's meters. If you have any tube drive setting or other similar processing novelties, try bypassing them to start with. Now it's time for a trial recording using enclosed headphones to feed the singer their cue mix — open phones spill quite badly into the mic. If you can flip the phase of the headphone mix, try both settings to see if the singer can hear a difference. They often can because of the way the headphone sound combines with the direct sound they hear via bone conduction. Usually the singer will find it easier to sing with one phase setting than the other. Most singers also like a little 'comfort' reverb in the phones, but discuss this with them and set up your system so this is not recorded.
After the test recording has been made, listen carefully to the soloed vocal sound to ensure that it's clear, clean and free from unwanted room resonances. Any natural peaks in the voice that sound too 'forward' can be tamed later using narrow bands of EQ cut, but for now aim to capture an honest representation of the voice and don't be afraid to directly compare the acoustic sound of the voice with your recording so that you can see where any differences occur. You should be able to get to this stage using just about any decent mic and preamp, though a good-quality mic preamp will help you retain the quality you're getting from the mic.
Look at the waveform levels in your sequencer and see how much variation there is. If there's a big level fluctuation, you may need to add some compression when recording, but never add more than you'll ultimately need, as you can't take it off once it's been added. I usually stick to ratios of up to 4:1 and then adjust the threshold to give a gain-reduction reading of no more than 6dB on the peaks. Use a fast attack and a release time of around a quarter of a second, or the automatic mode if your preamp's compressor has one. If you don't have a compressor in your preamp, then record with no processing and use a software compressor when mixing. Excessive or inappropriate compression at this stage can lead to a congested, lifeless sound that's almost impossible to fix later. It also pays to bear in mind that compression brings up the effects of the room ambience in quieter passages, so while you may not hear the room on an unprocessed recording, it may start to intrude once you start to add compression. This is why it's so important to fix the recording space before you start recording.
It may be that some sections of the song, often choruses, are sung louder than the verses, in which case you might opt to record them separately on two different tracks so that you can match up their levels more easily. Other options include recording everything onto one track, then dividing that track into its verse and chorus sections before moving the chorus sections to a new track. Doing this makes the levels easier to manage without having to use automation, and you also have the ability to use two different compressor settings (or even two completely different compressors) on the two tracks. Often the louder section will need a higher ratio to keep it even, especially if the section includes a lot of natural dynamics.
Although compressors can compensate for many level problems, I find that there are occasionally words or phrases that are much too loud or too quiet and that compression doesn't completely cure. In this case it's invariably best to use your sequencer's level automation capabilities to balance the levels.
Getting a well-recorded vocal part to sit in a track relies on it having an even level, the correct tonal and level balance, and the right reverberant ambience. I like to start out by getting the compressor and gain automation settings right, so that the vocal sounds even enough to sit in the track, with none of the words or phrases sounding too loud or too quiet. Where you really want to add density to a voice, try upping the compression ratio to around 8:1 and then adjusting the threshold so that the gain-reduction meter just registers a decibel or two on average-level parts. This means quieter sections will be left alone and anything louder will be jumped upon fairly hard to bring it closer to the average level. The gain reduction that shows on peaks will depend on how much level fluctuation there is in the track — values of around 10-15dB are not uncommon on peak-level phrases. If your compressor has a choice of Peak or RMS limiting, RMS usually works best for vocals.
Next comes EQ, but in most cases you'll need very little unless you are after a special effect, or if the voice has some defects that you need to fix. The general rule is that, while EQ cuts can be made over quite narrow bands, boosting should be as gentle and wide as possible. If you think the voice does have an over-emphasised component, then set one of your EQ's peaking filters to full gain, with a Q value of around 1.5, and then sweep it through the frequency range until you hear it pick out the frequency that's been giving you trouble. You can then apply cut at this frequency, ideally using as narrow a filter (high Q) as possible to get the job done. If more than one frequency band stands out, then engage another filter section and do the same, but always apply as little cut as you can get away with, as less EQ is always better.
For more gentle shaping, try gentle cutting or boosting in the 100-300Hz band to add warmth or reduce chestiness. Also check the 1kHz region, as this is where nasal vocal sounds hang out, and if you need to add presence then try a gentle boost between 4kHz and 6kHz. The magical sense of air and space requires a very wide boost centred between 14kHz and 16kHz, but check what you're doing by comparing your EQ settings with the EQ bypass position, just to make sure you haven't gone too far.
If you use a pitch-correction product such as Antares Auto-Tune, the most natural results are usually obtained at the slower pitch-correction settings. The default correction rate seems a little too fast to me, as I can hear it working. If you can split up the vocal track into two tracks where the phrases and words that need tuning are all on the same track, then you can use pitch correction only on those sections that need it, which should result in a more natural-sounding end product. Pitch-correction may be applied a little more heavily to individual backing vocals, where any artefacts are more likely to be disguised by the layering of two or more voices.
That leaves reverb, and in my experience this is trickier to get right than you might imagine. For a start, few plug-in, host-powered reverbs have the clarity and character of a decent hardware or DSP-card reverb. However, if a plug-in is all you have, pick the best algorithm on offer and set up the reverb in your virtual mixer's aux send/return loop (with the effects mix set to 100 percent wet) so that you don't have to duplicate it for each vocal part. This will minimise CPU drain. If you have a good hardware reverb, you can patch this into any spare soundcard inputs and outputs you have and then route the virtual mixer's aux sends though it — a trick I often do with my Lexicon MPX550. The output of the reverb feeds back into the soundcard and then into a spare stereo mixer input.
Traditionally, vocals are treated using plate or room reverbs, but on the cheaper hardware boxes and most plug-ins, the presets tend to muddy the sound before they produce the desired thickening effect. One strategy that I have found to work well is to use a reverb algorithm where the early reflections level can be adjusted independently of the reverb tail, then turn the level of the reverb tail down by around 6dB. The early reflections have the effect of thickening the vocal without making is sound washy, and by turning down the level of the reverb tail, you can still get away with a fairly long reverb time (typically around 1.8 seconds) without losing clarity. Up to 100ms of pre-delay also helps add depth without clouding the picture, and you may even be able to drop the reverb tail level further if you're aiming for a more subjectively dry sound. If you have a commercial recording in a similar style, keep this on hand as a reference when you're setting up the mix, as it helps to compare general tonality and reverb settings.
The final test is to set up what you think is a good mix, then listen to it from outside the studio door. The vocals should be loud and proud, but not so loud that they sound 'pasted on', and you normally only need enough reverb to make the vocals sound as though they were recorded in the same acoustic space as the rest of the backing track. I've mentioned this 'standing outside the door' trick before, because it really is the best way to judge balance. A problem that might be completely unnoticeable when you're sitting in front of the monitors will jump out at you once you leave the room. At this stage you can make a test mix, burn a CD and then play it in as many different systems as possible, as this might reveal something that your monitoring environment didn't. Sorted — hopefully!