Setting up your gear for low noise and minimum distortion needn't be a nightmare. Martin Walker guides you through the process, and shows you how to stand tall, even without much headroom.
Most people understand that if they want to get the best audio performance out of their studio equipment, input levels from sound sources must be high enough to ensure that noise levels remain very low by comparison, but not so high that they overload the equipment and cause distortion. However, for the best results, this optimisation process must be carried out at each stage of the amplifying chain, and this is where the concept of gain 'structure' comes from — the tweaks are carried out from the very first input, all the way along the signal path, right to the end of the chain, whether the signal is being recorded onto DAT, or emerging from a loudspeaker.
If you record with acoustic instruments, the first stage for you will be to ensure that the input gain of your mixer's microphone input (or one of the fashionable stand‑alone mic preamps) is set correctly for the levels coming out of the mic itself. Most mixers, even tiny ones, provide PFL (Pre‑Fade Listen), and this allows you to monitor the level of a particular mixer channel after any EQ, but before the main channel fader. This is extremely useful when you want to listen to any sound in isolation, and also for initially setting up the input gain controls. When you press the PFL button, the signal will normally also be routed to one of the mixer meters, so that you can see its level. Even without PFL facilities you can achieve the same thing by first pulling all the channel faders right down, setting the master faders to unity (0dB), and then raising each channel fader in turn to the 0dB position.
Those with golden ears say that solid‑state amplifiers start to sound 'edgy' in the final few dBs before clipping sets in.
Once you have some typical signal levels going through the mixer, you adjust a channel's gain control until the meter is hovering around the 0dB level (for a reasonably steady signal), or a bit higher (+6dB or so) if there are a lot of transients in the signal, since its average level will then be somewhat lower. If you're dealing with closely miked drums, some input levels may be so high that even with minimum mixer input gain you still have too much signal level; in this case you may have to switch in a pad, or use a less sensitive mic. Plugging the mic into a line input is not recommended, since the impedance values will be wrong.
Most stereo mixer inputs, the ones often used for electronic instruments such as synths and samplers, only provide a switch labelled +4/‑10, instead of a fully variable gain control, and the best position of this switch can be determined in exactly the same way, using the channel PFL button. In most cases, if you can turn up the output level of your synth or effects unit to maximum, so that the switch can be set to the less sensitive '+4' position, you're likely to get slightly lower noise overall. Once all your inputs have been set up in this way, the channel faders are then used, with starting positions somewhere near the 0dB mark, to mix everything so that the final combined levels again peak at about the +6/+9dB mark on your output meters.
So far, so good — I'm sure most of you know about the above techniques already (although it's surprising how often I spot peoples' mixer output meters with only a couple of LEDs twitching near the bottom, or flashing red at the top of the range). What's a little more confusing is where different manufacturers choose to place the first red LED in their meter displays, and why. Many mixers have green LEDs up to 0dB, amber from 0dB to +6dB, and red for +9 and +12dB (the highest indicator). The idea is that if you see very occasional flashes of the +9dB LED on peaks, you'll be OK, but if the second red LED flashes as well (+12dB) you're approaching the point of distortion. In fact, all mixer manufacturers will have designed in a bit more headroom than this. Headroom is exactly what its name implies — a bit more space over the metered limit before things overload — and is traditionally the difference between average level and clip point.
This is where we find the huge difference between analogue and digital circuitry. If your mixer meter does occasionally go a little 'over the top', the mixer itself is unlikely to sound distorted, but if your mixer is feeding a digital recorder you'll almost certainly have to do another take. In a mixer, typically there will be at least another 6dB of output level available above the top LED before amplifier clipping occurs. Most mixers standardise on an output level of +4dBu (1.23V RMS) when the meters read 0dB VU. So when the top LED is just lit at +12dB VU, the actual output level emerging from the sockets will be +16dBu (4.9 volts RMS). If you look at your mixer spec to see its output level, it will give a figure of something like +22dBu maximum (9.76 volts RMS). This is 6dB higher than the top LED on the meter, and the extra headroom should ensure that your signals always emerge cleanly.
I say 'should' assuming that an amplifier will sound perfect right up to the clipping point, but this isn't always the case. Those with golden ears say that solid‑state (transistor or FET) amplifiers start to sound 'edgy' in the final few dBs before clipping sets in. If you have the impression that some of your gear doesn't sound quite as good as it should when you drive it close to the clip point, you may be right. Fortunately most modern gear is quiet enough to be calibrated to run at slightly lower levels, to give a cleaner sound. Effects units, however, which are often the noisiest devices in the studio, are sometimes temperamental about overload, even for a few milliseconds, and so benefit from special treatment (see 'The Effects of Noise' box).
With digital recording it's absolutely vital to avoid any overload at all. To be honest, although the mixer line‑up procedure already discussed will optimise the gain structure of your analogue electronics, once a digital recorder is involved, most people will religiously watch the meter on that like a hawk, rather than relying on a mixer's meters. This is because most digital recorders have a Margin indicator as well, which shows the highest peak level recorded since recording began, and which holds this value until the Margin Reset button is pressed. In addition, since digital electronics are so sensitive to overload, these meters normally have a much faster response than the meters on a mixer, and may therefore show different readings as well. So now that the mixer is lined up so well, to complete the chain you have to calibrate your DAT machine to your mixer.
The digital meter on the DAT is calibrated in a rather different way, with the top of the scale reading 0dB, rather than 0VU appearing about two‑thirds of the way up the mixer meter. You will often see a special mark on a DAT meter on or about the ‑12dB position. If you set up a 1kHz line‑up oscillator on your mixer and adjust its level to exactly 0VU on the mixer output meters, you can go into Record Monitor mode on your DAT machine, and then slowly increase its input level control until this ‑12dB mark is reached. (Due to the large gaps between calibration marks on a DAT, looking at the Margin readout is a far more accurate way to do this, as it normally changes in much smaller increments, such as 0.5dB.) At this point, where 0dB VU as indicated by the mixer is equal to ‑12dB relative to Full Scale on the DAT meter (0VU = ‑12dBFS), you've calibrated your DAT machine so that mixer meters just touch the top red +12dB VU LED as the DAT machine reaches 0dB.
This ‑12dB reference level is fine for those recordings where every level is extremely well behaved, such as MIDI or sample playback, but live instruments recording is rarely so predictable. Since you still have headroom on your mixer beyond its Full Scale reading, you could reduce the DAT reference level to ‑18dBFS with 0dB VU on the mixer, to allow for unexpected transients. Now that more 20‑bit converters are appearing, even on budget equipment, noise levels are also dropping, and there is a school of thought which says that using a reference level such as ‑18dBFS doesn't compromise noise levels significantly, whatever the dynamic range of your music, and at least it lets you return to looking at your mixer meters, without having to worry so much about the odd extra dB ruining a digital recording.
One thing to watch out for here: you may come across digital recorders that try to mimic their analogue counterparts by setting their internal reference level to typical mixer output levels. When your mixer reads 0dB VU (normally emerging at +4dBu), an Alesis ADAT may still only be reading about ‑15dB on its own meter. Although this gives you plenty of headroom, to reach 0dB on the ADAT meter will need +19dBu output from the mixer, which is getting perilously close to the clipping point of many small mixers — and, as already mentioned, your mixer may not sound quite so clean in the final few dBs before clipping. If you find this is a problem, you might try using the digital recorder at its ‑10dBV input sensitivity, which will let you 'go all the way' without risking output clipping of your mixer.
For both live and studio work it's also common to patch in a compressor at the mixer buss insert point, and in some cases an additional 'brick‑wall' limiter set to a level just below the digital clip point, to ensure that nothing gets through to overload the digital side. High‑end processors such as the TC Electronic Finaliser even include a fine level adjuster for the limiter, calibrated in 0.01dB increments below 0dBFS, and since many digital recorders don't have proper 'Over' indicators (see The Digital 'Over'), this ensures that you never get erroneous readings, since the record signal will never actually reach 0dB.
Now that digital recording is so much a part of the modern studio, a completely mathematical approach can be adopted. Since the digital signals are simply a stream of '1's and '0's, gain can be adjusted by multiplying or dividing digital values, and this is equivalent to amplification or gain reduction in the analogue domain. However, digital processes have one big advantage — by looking ahead in the waveform, compression/limiting algorithms can anticipate transients, rather than having to react to them as quickly as possible after they happen, as in the case of the analogue compressor.
Normalisation is another gain adjustment, but this time it's carried out after recording, by bringing the maximum peak level to the maximum allowable digital level, to make sure that the signal is as 'hot' as possible. It does not increase the dynamic range of the programme material, since low‑level signals will be brought up by exactly the same amount as high‑level ones, and neither does it ensure that every track destined for an album will end up at the same perceived loudness, since this depends on average levels, and not peak ones. Many recordings will only have a few occasional peaks approaching 0dB, and the average level can nearly always be brought up by at least 3 or 4dB, simply by compressing these short transients, without having an obvious audible effect. You can do this with a limiter, or using the software approach of plug‑ins like Waves' L1 Ultramaximiser, and Steinberg's Loudness Maximiser.
Ultimately, being thorough in your approach to gain structure will ensure that you only hear distortion when it's part of your music, and that quiet passages remain free of unwanted background hisses and hums.
If you're trying to make your recordings sound as 'loud' as commercial releases, it's best to monitor all digital signals through one high‑quality digital/analogue converter — listen to your CDs, hard disk recordings, and so on, and then you can compare them directly.
Normalisation does ensure that on cheaper playback systems, where system background noise is more of a problem, your recordings will make use of the top end of the dynamic range. It should always be carried out as the final operation, after any other digital editing, since you're asking for trouble if you tweak the digital signal any more once it contains peaks at the maximum theoretical level. Also, since both of the above plug‑ins can take advantage of noise‑shaped dithering (for better low‑level resolution), this can also raise levels, particularly at higher frequencies. For this reason, many people play safe, and normalise to a figure just below 0dB — Waves recommend ‑0.3dB when using their L1 Ultramaximiser during mastering for CD.
Ultimately, being thorough in your approach to gain structure will ensure that you only hear distortion when it's part of your music, and that quiet passages remain free of unwanted background hisses and hums. Noise gates and muting can remove all the background grunge once it falls beneath a threshold level, but with attention to detail and careful wiring (preferably with balanced lines) your music will sound more transparent even at normal levels if the noise floor is as low as possible. Sadly, digital artefacts often sound more objectionable than analogue ones, because rather than being random in nature (a steady background hiss that the ear tends to ignore if low in level), they tend to be tied into the signal itself, and are therefore more noticeable. At low levels, where the converters run out of resolution, smooth waveforms begin to resemble a staircase, which gives a gritty sound, known as quantisation noise.
The last year has seen many more affordable digital converters appearing in mid‑price to budget equipment, and these have lower noise floors than equivalently priced equipment that is a couple of years old. Now, more than ever, it's worth giving your studio the once‑over to optimise everything in the audio chain.
Setting the right DAT recording level: SOS January 1995.
Noise and how to avoid it: SOS May 1995.
A Concise Guide to Compression & Limiting: SOS April 1996.
The Mysteries of Metering: SOS May 1996.
Minimising Mixer and Effects Noise: SOS July 1996.
When you send audio equipment a signal large enough to overload its circuitry, each device will respond in a different way. Many people do overload their equipment for creative reasons, because the signal emerges with a different sound, and much development work is going on to produce computer software plug‑ins which mimic the 'softer' overload characteristics of tube circuitry. Whereas a guitar amp normally benefits in a musical way from being overdriven, few people enjoy the sound of digital overload, and this is because of the nature of the distortion produced. The singing sound of guitar overdrive tends to be predominantly second harmonic, and the human ear finds this fairly pleasurable. For a start, it's only an octave away from the input signal and therefore easy for the ear to 'attach' to the overall sound. As digital circuitry becomes overloaded, it neatly clips off the top of the waveform, generating lots of third‑harmonic distortion (not as nice as second, but still passable), but also lots of higher harmonics as well, extending to very high frequencies. Although the human ear can only pick up second‑harmonic distortion when it reaches around 0.5%, eighth‑harmonic distortion at as low a percentage as 0.01% is audible to humans — which could be part of the reason why valve amps with high measured values of THD (Total Harmonic Distortion) often sound far better than transistor amps with THD values of 0.01%.
Probably the easiest way for most people to achieve quieter mixes is to optimise the gain structure of their effect sends and returns, since effects units do tend to be the noisiest items in many studios. Try to ensure that most aux sends end up at about the 7 to 8 position, since this is normally the optimum position as far as mixer noise is concerned. However, since it's the output noise from the effect that can prove troublesome you should try to drive it as hard as possible at the input end. Many effects units have a 'Clip' or 'Overload' LED that comes on 5 or 6dB below clip point, but different units tend to react differently — some sound horrendous even if this is exceeded for a few milliseconds, while others are more tolerant. If you can increase the input level control on the effects unit by a few dB, the Return fader on your mixer can be reduced by the same amount, leaving effects levels identical, but with correspondingly lower noise levels. Once you've performed these tweaks you'll probably notice a big improvement in the most obvious places — at the beginning and end of tracks.
There is still much confusion between signal‑to‑noise ratio and dynamic range, especially where digital signals are concerned. Signal‑to‑noise ratio is the RMS level of the noise with no signal applied, expressed in dB below maximum level. Dynamic range is defined as the ratio of the loudest (undistorted) signal to that of the quietest (discernible) signal in a unit or system, as expressed in decibels (dB). Dynamic range is often said to be a subjective judgement more than a measurement — you can compare the dynamic range of two systems empirically with identical listening tests, by applying a 1kHz tone, and see how low you can make it before it is undetectable.
The maximum signal‑to‑noise ratio of a 16‑bit recording is 96dB, since each bit contributes 6dB to the total. If you leave 6dB of headroom on your digital recording, to prevent any unexpected peaks from causing clipping, you immediately reduce this to 90dB. The background noise level will depend on the converters, as well as the design of the rest of the circuitry. Due to the confusions, even between manufacturers, on how these figures should be measured, Crystal Semiconductor (the well‑known designers of A/D and D/A converter chips, as used by many companies worldwide) have suggested a standard method for the following measurement procedure. They define Dynamic range (DR) as the ratio of the full signal level to the RMS noise floor, in the presence of signal, expressed in dB FS. The addition of a low‑level signal (a suggested 1kHz sinewave at a level of ‑60dB FS) ensures that any noise‑gate circuitry is bypassed, but of course this signal must be notched out before the actual measurement is taken. The final figure quoted is also likely to be A‑weighted, which takes account of the characteristics of the ear, which is more sensitive to frequencies between 2k(dot)z and 4kHz. 'A‑weighted' figures tend to make comparing figures between different hardware components easier.
When dealing with digital recorders, any signal that flashes the Over indicator on the input level meter of the digital recorder will sound dreadful on playback — there is no headroom with a digital meter. In fact, very few digital recorders actually have a proper digital 'over' indicator — if you think about it, once the signal gets to 0dB, it just can't get any higher. Many so‑called Over indicators are actually measuring analogue levels, so that they can indicate a level which exceeded the calibrated digital peak level. This is why you can overload the inputs of some digital machines and the tape you've just recorded never shows any overload indication on replay. Other machines which do flash an overload may well be reading 0dB and assuming that the signal might have overloaded.
The clever machines use a different method to determine whether or not a real overload has occurred. Just as the highest peak of a signal touches the 0dB mark, a single sample will be recorded with a value of 0dB. If the input level goes any higher, several samples in a row will be at this 0dB point, and this is likely to be because of overload. So some manufacturers count consecutive 0dB values — the Over indicator on a Sony 1630 machine will indicate an overload if three samples in a row are detected at 0dB. However, at 44.1kHz three samples lasts only a few tens of microseconds, which is generally regarded as inaudible — other manufacturers use four, five or six samples in a row. The beauty of the sample‑counting Over indicator is that it will also work with a digital input, and pick up recordings that have previously been overloaded.