Taking the hiss?
Peter Rodgers describes a neat add-on for eliminating unwanted noise from musical effects.
While most musical effects units produce excellent sounds, some add a fair amount of noise as well. The worst offenders are circuits which include analogue delay lines (flangers, chorus units, etc), though there are others with a similarly poor noise performance.
It is possible to obtain a substantial improvement in signal-to-noise ratio using a circuit known as a "compander" — some effects units have one of these built-in. Circuits of this type can be found in most noise reduction systems used in tape recording. Where an effects unit does not have a built-in compander, it is usually possible to add one externally. This article describes such a unit.
Really a compander is two separate and complementary circuits; a 'compressor', which processes the signal before it is fed to the input of the effects unit, and an 'expander', which processes the output. The role of the compressor is to boost low level signals; the lower the signal level, the greater the degree of boost. The graph in Fig 1 shows the input/output characteristic of a compressor which has a 2:1 compression rate (ratio). A large input signal (at the 0dB level) produces an output signal which is at the same level. However, if we take a low input level of say -80dB, this gives a greatly increased output level of -40dB — a substantial increase in signal level, since a boost by 40dB is an increase of one hundred times.
The reason behind using compression is that low level signals, which would otherwise fall below the noise level, are boosted well above this level. For example, if we assume that the effects unit has a poor signal-to-noise ratio (at the -40dB level), any signal much below this level would tend to be lost amongst the noise, if the 2:1 compressor is used, signals at levels down to -80dB are boosted to -40dB or more and therefore equal to, or greater than, the noise level. Signals which would otherwise have been lost in the noise are brought up to a level where they are clearly audible.
Compression alone can effect a form of noise reduction, but there are several major drawbacks. One is simply that the compressed signal sounds very different to the original, which is not really surprising when one considers the drastic changes to the dynamic levels. Another is that the background noise level is not reduced — noise reduction is achieved by maintaining the wanted signal at a fairly high level so that it tends to mask the noise. The noise level might actually be increased, since any noise on the input signal, like any other low level input signals, will be substantially boosted by the compressor! The practical result of this is: the signal tends to be quite loud all the time and the noise level, during pauses in the signal, may well be worse than ever.
All the above problems can be overcome by using a complementary expander at the output of the effects unit. A 2:1 expander has the characteristic shown in Fig 2. An expander exaggerates any changes in input level, so that an input at say -40dB produces an output at -80dB, but a 0dB input gives a 0dB output. In the previous example (a 2:1 compressor), a signal at -80dB was boosted to -40dB. It should be apparent that the expander restores this signal to its original level. In fact, any input signal is restored to its original level at the output by the expander and there is no overall distortion of the dynamic levels. The important point is that the noise produced by the effects unit is greatly reduced by the expander. If we assume a noise level at -40dB again, reference to Fig 2 will show that the expander reduces this to -80dB. In other words the signal-to-noise ratio has been boosted by some 40dB.
Thus, noise is reduced at low dynamic levels where it would otherwise be very noticeable and objectionable, but is unaffected at high dynamic levels where it is not usually noticed. As far as anyone using the equipment is concerned, there is a large and genuine improvement in the signal-to-noise ratio.
An important point about a compander is that it only counteracts noise in the effects unit, it does not have any effect on noise present on the original signal. This noise is treated just like any other input signal, and always emerges from the compander at its original level. A compander is therefore of no value where a noisy signal needs to be 'cleaned' up; it is only used where noise from an electronic circuit of some kind needs to be reduced.
One way of processing a signal to produce better noise performance is to use a circuit called a 'noise gate'. Strictly speaking this is a circuit which cuts off ('gates') the signal if it falls below a certain level. However, an expander can be used to give a sort of noise gate action and may well give better results. The noise reduction is provided in exactly the same way as when an expander is used in a compander but as the processed signal has not been previously compressed the dynamic levels of the output signal are altered and exaggerated. A 2:1 expander would, in most cases, exaggerate dynamic changes to an unacceptable level and for use as a pseudo noise gate a more gentle rate of expansion is required, such as the characteristic shown in Fig 2.
The unit featured here is a 2:1 compander based on the NE570 integrated circuit which is specifically designed for use in applications of this type. The NE570 is the only active device in the circuit, which can also be employed as a pseudo noise gate. Fig 3 shows the block diagram for the Compressor/Expander.
The compressor is based on an amplifier which has its gain varied by a voltage controlled attenuator. Negative feedback, opposite in polarity to the input signal, tends to have a cancelling effect on this signal. Thus, the greater the level of feedback, the lower the voltage gain of the amplifier.
The control voltage for the VCA is obtained from the input signal via a full wave rectifier and smoothing circuit. With only a low input level the control voltage is small, the voltage controlled attenuator provides a high level of attenuation, the feedback level is low and the amplifier exhibits a high level of voltage gain. As the input level is increased, the amount of feedback is raised and the gain of the amplifier falls. This falling gain with increasing input level gives the required compression.
The expander uses three circuit blocks which are identical to those used in the compressor — they are merely rearranged slightly. The signal passes through the voltage controlled attenuator and then through the amplifier (the latter is really just used as a buffer stage in this case). The control voltage is again provided by the input signal via a rectifier and smoothing circuit. With a low input level the voltage controlled attenuator provides a high level of attenuation, but as the input signal level is increased (and the control voltage rises) the degree of attenuation falls, giving the required expansion characteristic.
For use as a noise gate, a bias is fed to the control input of the voltage controlled attenuator. This merely reduces the maximum attenuation that can be provided so that a reduced expansion rate suitable for this application is obtained.
Use of a dedicated integrated circuit means that few extra components are needed in the unit (Fig 4). Most of the components are for coupling or biasing. R2 and R3 are used to bias the amplifier in the compressor, but only internal bias components are used for the amplifier stage of the expander. R5 and VR1 provide the bias which gives a reduced expansion rate when the unit is used as a noise gate. VR1 enables the expansion rate to be adjusted to best suit the application. The bias is switched out using S1 when the unit is used as a compander.
The NE570 has a distortion trim input for each half of the device, but it is not essential to use this feature as the typical harmonic distortion without trimming is only 0.3% even at high input levels. This is perfectly adequate for use in any normal electronic music set-up.
Also, it is difficult to properly carry out the distortion trimming without the aid of some advanced test gear.
An aluminium box having approximate dimensions of 152 x 102 x 51mm is suitable as the case for this project. It would probably be possible to use a somewhat smaller case than this, but only if sufficient panel space for the sockets and controls can be found.
All the components are mounted on the printed circuit board and this is detailed in Fig 5. This also shows the hard-wiring of the unit. Construction of the board is quite straightforward, but be careful to connect the electrolytic capacitors with the right polarity. Also be very careful to connect IC1 the right way round, as this is by no means one of the cheapest of integrated circuits — it's a good idea to fit it in a socket.
The completed board is mounted on the base panel of the case using 6BA or M3 fixings. The board can be used as a sort of template when marking out the positions of the three mounting holes on the base panel of the case. The board should be positioned with R6-C2-R5 towards the front of the unit, and leaving sufficient space for the battery at the rear of the board. Spacers about 6 to 12mm long are used over the mounting bolts, between the board and the case, so that the connections on the underside of the case are kept clear of the metal case and are not short circuited.
The completed project does not require any setting up or adjustment of any kind. The circuit will operate satisfactorily with a wide range of signal levels and will take an input of over one volt without producing serious distortion. The noise level of the circuit is very low, and it will operate well even with input levels of just a few tens of millivolts.
When used as a noise gate the expansion rate (and degree of noise reduction) is decreased as VR1 is advanced in a clockwise direction. In most cases this control will need to be set quite far in the clockwise direction in order to prevent unacceptable distortion of the processed signal's dynamic levels. In the majority of applications the adjustment range of VR1 will be adequate, but in some situations a very low level of expansion might be required. This can be achieved by reducing the value of R5 to around 22k.
Feature by Peter Rodgers
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!