More Front-End

In the last instalment we looked at interfacing mics with the desk using transformers. By far the most valuable asset of a transformer-coupled front-end is its ability to 'match' impedances for minimum noise. This isn't the only way to achieve a balanced input, however.

The circuit in Figure 1 shows an op-amp wired in the differential mode: that is, the circuit has balanced inputs. The reason this circuit is not in universal use however is that it has a host of serious set-backs which makes it unworkable in practice. First of all, the common mode impedance between either input and ground is asymmetrical. In other words, the circuit unbalances the line, thus defeating its own object. This effect is particularly significant with floating balanced lines, the cable configuration used by the majority of microphone manufacturers. Secondly, the circuit requires extremely accurate (less than 0.1 %) resistor matching to achieve even moderate interference rejection; and thirdly, inevitable imbalances in the stray reactance between the inputs destroy totally any remaining rejection properties at high frequencies. This is ironic in that RFI (radio frequency interference) is the most irksome - and damaging - of all spurious signals.

Modifications

Modified two- and three-op-amp versions of the basic circuit (Figure 2 shows one of several variants) overcome many of these difficulties. In particular, the three-op-amp 'Instrumentation Amplifier' (Figure 2) makes direct and active control of the front-end gain feasible. However, the arrangement still demands stable, close tolerance components and careful design if it's to perform well at high frequencies. And even assuming this to be the case, achieving a well-balanced input isn't enough; there's also the question of noise performance. Without the matching benefits of a transformer, hiss is bound to be at a higher level than usual, though this is only a serious problem with mic inputs.

For the [line] level inputs, the transformerless balanced input can give excellent noise performance, but this opportunity is often missed, since it's both easier and cheaper for the manufacturer to derive a line level input by 'padding down' the mic input, rather than switching in another stage optimised for handling line level signals.

With many of the more dynamic instruments (eg. bass guitar), the extra hiss on each channel simply won't be audible, and even with, say, a quiet acoustic guitar solo, noise or buzzes which would be considered unacceptable coming from a studio desk will often be drowned by audience noise, for instance. It's important to remember that live mixing is very different from studio work.

However, at the same time the problem of noise in live situations should not be underestimated. It is possible to improve the S/N ratio of a desk with 'electronic balancing' (a misnomer, because balancing is always electronic, regardless of whether it's accomplished with a transformer or an op-amp) by using exotic input devices (eg. LM394/OP27/Jensen JE990). These exhibit a very low optimum source impedance, so that the circuitry will give lowest noise when it sees an impedance similar to that presented by a low-impedance mic. The greatest obstacle here is cost, however. The noise performance may be equal to that of a transformer-coupled input and the transient response and balancing exemplary, but total expenditure will far exceed the cost of a top-notch input transformer.

In any case, the capabilities of any balanced input stage will always be let down by the mic cables and multicore, which typically limit rejection to a fairly poor 40-50dB at high frequencies. This important parameter is rarely specified, but it's worth trying to seek out data on common-mode balance before parting with cash for good mic leads or multicore. Generally speaking, a cable with a tight, small-radius twist offers the best performance.

To sum up, a desk with transformerless balancing using only a single op-amp to handle the differential input simply isn't worth paying the extra for, as such balancing is of dubious value. There may also be undue hiss to contend with, since at this price level it's unlikely that the manufacturer will have adopted even a commonplace low-noise chip such as the NE5534 (itself relatively noisy in comparison to the esoteric super-chips needed for first-class results).

However, if the 'electronically balanced' input costs no more than the unbalanced alternative it's worth serious consideration, especially if you feel that any undue hiss will be obscured in practice during a live performance. Even the poorest balancing is better than none at all, provided that it's stable and doesn't have any untoward effects on the sound.

Additions

Additionally, don't forget that it's relatively easy to add transformers to budget desks and therefore upgrade their performance. Figure 3 shows a typical procedure for desks with unbalanced inputs. Single-IC electronically-balanced inputs can also be adapted in this fashion. Providing some care is taken to tune the square wave response with long cables in place - thereby quelling any tendency to peaking and instability - good transformers will give excellent results in most circumstances, though they aren't perfect.

If you move up to the next price-bracket, the two- or three-op-amp instrumentation-type balanced inputs should be seriously investigated. Try to judge whether the tradeoff in noise is compensated for by the undoubted improvement in percussion sounds, for instance. Preconceived notions aren't helpful here - in the end, the best desk is the one that sounds best, regardless of the technology it uses at its input. To give each front-end a fair hearing, leave the EQ flat, and keep the levels well below overload.

Filtering

In the context of a mixing-desk, filters come in two versions.

One type has a directly audible effect and comes under the EQ section, while the other - known as supersonic filtering - is essential for catching the massive signals arising from close-miking. The filtering required here should be placed as close to the input as possible, preferably at the input. The existence of fixed filtering will obviously show up in specifications as a drop in response above 20kHz. Ideally, the slope should be quite steep - say, -12dB/ octave - to ensure a useful degree of attenuation in the 20 to 35kHz region.

Adding filtering to an existing desk can be a relatively simple modification so long as a -6dB/octave slope is acceptable. This involves adding either a pair of empirically chosen RC filters across the inputs or a capacitor across the first op-amp's feedback resistor. Moving on to low- and high-pass filters for EQ purposes, these are used to reject either the entire upper or lower segment of the audible spectrum in order to limit overspill; ie. the spurious pickup of sounds from adjacent instruments. This facility is perhaps most important on cymbal and hi-hat mics, whose signal would otherwise be drowned by the remainder of the kit, or at best suffer nasty intermodulation effects. For this reason, the high-pass filter is the more useful of the two.

On budget desks, the filtering is often activated by a toggle switch, offering filtering below a single, fixed frequency, usually 100Hz. Of course, 'all singing, all dancing' consoles may offer a tunable turnover frequency, over, say, 30 to 350Hz. More important than the range of frequencies however is the slope of the filter, which should be a minimum of -12dB/octave.

Tidying up

To show how the abstractions we've looked into turn out in practice, Figure 4 shows the full front-end circuit of an AMEK M1000 PA console. The circuit complements what we've discussed so far and also raises one or two points that need tidying up.

The line input is obtained here by 'padding down' the sensitive mic input. The three resistors (R1-3) form a balanced attenuator which maintains the balancing of the line whilst ensuring that the input stage sees its usual 200 ohms. At the same time, the equipment at the other end of the line (which might not be capable of driving a low-impedance load) sees a 10k load, made up from R1 (or R2) in series with the intrinsic input impedance, which is much lower, of course.

Phase changing is useful as a means of modifying a sound creatively or of reducing overspill when using two mics on or close to one instrument. It can also be handy for correcting the sound of a 'zombie' mic, ie. one wired out-of-phase relative to the others. Phase inversion can be made to occur by switching around a unity gain - but inverting - op-amp stage. Using a switch on its own, however (as in Figure 4) is more useful. The 20dB pad was discussed earlier, but note the four-resistor network, with which it is easier to juggle with the attenuation whilst ensuring that the mic and input stage 'see' their usual (if only nominal) impedances, viz. 2000 ohms and 200 ohms respectively when the pad switch is thrown in. This is important if you are to avoid inadvertent changes in sound.

Going on to the op-amp, the active gain pot adjusts the amount of feedback to give gains from unity up to some 55dB. The theoretical hassles of having no actual attenuation available are avoided by the optional pad and, more significantly, by the adoption of a carefully designed gain structure in the succeeding stages of the console.

Lastly, and returning to the microphone/input socket interface, remember that the monitor desk, if simply wired across the stagebox feeds to out front, will reduce the impedance seen by the mics. This can impart some clearly audible changes to the sound of instruments, particularly when some of the more 'fussy' microphones are used. The remedy here is to use a splitter box, which allows everything to 'see' the correct impedance - a topic to be looked at later in this series when we get on to the subject of stage monitoring.