Ideal for the home or semi-professional recording studio, a four-channel compander offering 30dB improvement of signal-to-noise ratio, simultaneous encoding/decoding and LED peak indicators.

It's on the cards that within the next five years there'll be a sixteen-track digital cassette recorder, complete with a touch control mixer, in a box the size of Teac's Portastudio. Even now, at the top end of the synthesizer spectrum, there's a new Crumar programmable polyphonic synthesiser with built-in digital recorder (but yet to arrive in this country) and the amazing Synclavier II complete with sixteen-track digital memory and every editing facility under the sun, a snip at a mere £16,000!

The great advantage of digital recording is the absence of noise, and, if one's working with digital instruments, as in the case of the above polysynths, then there's no A/D conversion before the sounds are committed to 'tape'. However, at what might be described as the tail end of the analogue era, most of us are stuck with trying to get the best possible sound out of our trusty Revoxes, Teacs, or whatever. The main problem with such machines is their annoying habit of burying your latest creation under a blanket of tape noise as soon as you depart from the first tape generation. And, with the new Teac 3440, the basic quality is so fine that some way of preventing the build-up of tape and machine noise seems a pretty logical step to take.

Noise reduction systems reduce the irritating noise of tape hiss and so on by an encode-decode process. Quiet sounds, especially those at the top end of the spectrum, are easily swamped by tape hiss, so an encoder is used to artificially boost these signals before they are recorded. During playback the reverse process decodes the recorded sound back to its original state and rids the music of tape-generated noise. Up until recently, noise reduction systems have fallen into three distinct types: Dolby B (domestic), Dolby A (professional) and DBX (professional). However, the near future is likely to see a confusing proliferation of other systems offering various degrees of noise suppression, including: Toshiba's Adres system, Telefunken's High-Com and Telcom, Sanyo's Super D, Dolby's C and HX systems and Tandberg's Dyneq. If there's any sense in this race to the pinnacle of perfect music reproduction, then hopefully there'll be some common standards of operation agreed upon! Table 1 gives the S/N ratios obtainable from various recording mediums with and without different types of noise reduction.

The various systems of noise reduction available at present basically work on the principle of complementary compression of the on-tape signal and expansion of the off-tape signal. Compression involves reducing the dynamic range of the material that is being recorded, so that, with a 2:1 compression ratio, if the input to the compressor increases by 12dB, then the output of the compressor (on-tape signal) will increase by only 6dB. Conversely, expansion involves increasing the dynamic range, so that an increase of 6dB in off-tape level will result in a 12dB increase fed to a subsequent mixer, thereby restoring the original dynamic level of the music. At the same time, the noise introduced in the recording chain, in particular tape hiss, will be rendered inaudible on expansion since this unwanted signal is not subject to the initial compression treatment and is therefore expanded downwards way below the lowest dynamics of the music signal. This process is illustrated in Figure 1.

Another feature of the compression/expansion process is that it allows the recording of signals with a dynamic range approaching the limits of audibility, i.e., 100 to 120 dB. However, since modern-day musical experiences tend either to be restricted to the bottom end of the dynamic range (muzak) or stuck at the top end (rock and heavy metal), this facility may be more theoretical than practical!

With careful shopping it should be possible to make the complete four-track unit for around £55.

Circuit

The circuit diagram for the compressor and expander is shown in Figure 2. The power supply circuit is given in Figure 3.

The compressor input is routed via SW1a, either directly to the output in the 'out' position, or to C1 in the 'in' position. IC1 and associated components form a second-order high pass filter with a 12 dB/octave roll-off below 30Hz. This removes sub-audible signals (infra-sonics) that might be generated from record warps or sub-octave tracking VCO's. The reason for this filtering is that once audio frequencies descend towards DC, the response of tape recorders drops-off dramatically, and on playback a signal compressed in response to high level low frequency signals will be expanded, resulting in phantom modulation by the missing low frequency component lost during recording. The output of the filter is AC-coupled to a simple RC network (C4, R4) which forms a high frequency pre-emphasis circuit providing a 12dB treble boost. Without this pre-emphasis, and corresponding de-emphasis in the expander, a low level signal may be swamped by high level bass frequencies and typically results in a 'heavy breathing' or pumping effect as the expander attempts to adjust the gain accordingly. The signal is then applied to one half of the NE570 (IC2a) configured as a compressor using an internal variable gain cell and full-wave rectifier as well as an external output op-amp (IC3a). The variable gain cell is similar to a standard operational transconductance amplifier (OTA), except that, unlike OTA's, it is 'linearized' and therefore insensitive to temperature changes as well as offering low noise and low distortion performance. The signal at the output of IC3a is rectified and the resultant control voltage used to adjust the variable gain cell. By placing the gain cell in a feedback loop with the op-amp, a variable current generated in proportion to the input signal is used to adjust the overall gain of the op-amp. A 6dB increase in output level produces a 6dB increase in the gain of the variable gain cell, and, since this is effectively an expander inserted in the feedback loop, results in a 12 dB increase in feedback current to the input of the op-amp. Consequently, an increase in input level of 12dB results in only a 6dB increase at the output of the op-amp, thereby yielding the desired 2:1 dynamic range compression.

The current from the full-wave rectifier is averaged by an external filter capacitor (C11) with the result that the gain control is made proportional to the average value of the input signal. The speed with which this gain adjustment is made determines the transient response of the compressor and is a product of the value of the filter capacitor and an internal 10k resistor. The value of 1uF for C11 yields good transient response at average signal levels. However, at low signal levels, the gain of the op-amp increases and any mistracking that occurs between the compressor and expander will be magnified by the high gain levels. To improve tracking at low dynamic levels it is necessary to provide a level-adaptive circuit that speeds-up the response time. This feature is derived from the circuit built around IC4a and b with series diodes (D1, D2) shunting the output of IC4a to ground.

The RCR network (R8, C8, R9) around the op-amp, IC3a, provides DC feedback to bias the output at DC. C7 is an external compensation capacitor to provide stable operation over the audio bandwidth. It may seem curious to use an external op-amp when the circuit diagrams indicate that the NE570 has its own. This is because the op-amps in this 1C are equivalent to 741-types with slew rate, noise, bandwidth, and output drive capability that aren't really adequate for demanding audio situations. With weak signals, the compressor circuit operates at high gain and the NE570 op-amp runs out of loop gain. Furthermore, a slew rate of 600mV/us means that high frequencies will suffer. By using a J-FET op-amp, such as the LF351 with a slew rate of 13V/us, these problems are eliminated. Additionally, the output swing can be larger since IC3a is powered by a dual supply rather than from the single-rail supply required by the NE570.

The non-inverting input of the NE570 op-amp is biased by an internal reference voltage of 1.8V. In the case of the external op-amp, IC3a, this is accomplished by tying it to pin 8 via an RC decoupling network (R7, C5) which filters out noise from the NE570 reference voltage. Pin 8 also serves another important function, that of providing the means for trimming distortion generated by IC2a. Even harmonic distortion is produced by voltage offsets in the variable gain cell, and RV1 enables adjustment of the offsets for minimum distortion.

The function of R10 is to isolate the output of IC3a from the potential capacitive load of a long length of screened cable connected to the compressor output which could lead to oscillation. SW1b selects the 'in' or 'out' mode of operation.

Comparators IC5a and b provide an indication of the signal level at the output of the compressor. The inverting inputs receive the half-wave rectified output signal which is compared with reference voltages derived from the potential divider network, R21, R22 and R23. C13 and R19 determine the fast attack/slow decay operation of the comparators. IC5a and b respond to signal levels of, respectively, -3dBm and 0dBm.

The expander configures the other half of the NE570, IC2b, with a different arrangement of the various blocks. Once the off-tape signal has been routed via SW1c to C14, the signal is applied to comparators, IC5c and d, to provide an indication of off-tape levels, and simultaneously to the full-wave rectifier and variable gain cell. The rectifier produces a control voltage that is used to adjust the gain cell, with a response time determined by the level-adaptive circuit of IC4c and d tied to the rectifier filter capacitor (C22). An RC network (R30, C20) is connected in parallel with the op-amp, IC3b, to provide a treble cut of 12dB, therefore de-emphasizing the pre-emphasized signal emerging from the compressor via the tape recorder. When the input signal increases by 6dB, the gain cell control current is raised by a factor of 2, resulting in an increase in gain of 6dB. Since the input of the external op-amp, IC3b, is derived from the gain cell, the output level increases by 12dB, giving the required 1:2 dynamic range expansion. RV2 enables adjustment of gain cell offsets for minimum distortion, as in the compressor. Finally, R32 isolates the output of IC3b from subsequent screened cable, and SW1d selects the mode of use.

Construction

The unit is designed on a modular basis so that each PCB provides simultaneous compression and expansion for one channel. Single sided PCB's have been used to keep the cost down, though double sided PCB's could easily be made from the layouts provided, eliminating 29 links from each one. The PCB's have been specifically designed to fit into a West Hyde TEKO ALBA case, order code TEK A22L. This particular case has the dual advantage of colour (lobster red) and price (£3.60 + VAT), both of which have a brightening effect in these dull and inflationary times!

If you see red at the idea of lobster-coloured electronics, alternative colours (black, TEK A22K and grey, TEK A22G) are available.

In order that decoding should be the exact inverse of coding, it is important that components are well-matched. This is obviously no problem with resistors, but the notorious variability of electrolytic capacitors necessitates the use of closer tolerance components such as the mini electro-lytics available from Maplin. These capacitors offer ±20% tolerance at half the price of conventional tantalum types.

PCB designs and component overlays for the main board and PSU are given, respectively, in Figures 4 and 5. The threaded phono sockets suggested for the unit have the dual advantage of small physical size, enabling them to be fitted as rows of four on the back panel, and compatability with the connectors normally encountered in using Teacs and Revoxes. These sockets are mounted on the rear panel and connections to the signal pins made by short lengths of unscreened wire from the relevant holes on the PCB's.

The PSU is utterly standard, though it's important to note that mains earth is connected directly only to the front panel and indirectly via a 1k0 resistor (R41) to the 0V line. This should prevent the build-up of any hum loop when using the noise reduction unit with earthed equipment. Power line busses can be connected from the PSU to all four main PCB's.

Setting-up and use

The unit requires very little setting-up apart from adjustment of RV1 and RV2. The output level of a mixer is adjusted so that the compressor 0dBm LED's fire with louder dynamics. The record level is set to match the optimum requirement of the tape being used. Playback levels are then adjusted so that the expander 0dBm LED's fire at the same level as the compressor 0dBm LED's. This level matching isn't critical since the level-adaptive response time circuits take care of possible mistracking, but it does ensure really accurate decoding of the encoded signal.

A couple of points to note: the unit will not reduce the noise present in a noisy signal presented to the compressor input (this is territory best served by dynamic noise limiters) and any difference in the signal between compressor output and expander input introduced by the recording process will be exaggerated by expansion, including such horrors as common-or-garden dropouts. Therefore to get the best out of the unit scrupulous attention should be paid of alignment and cleaning of tape heads!

To optimise the distortion levels of the unit, apply a 1V RMS (equivalent to +3dBm) 10 kHz sinewave to the input of the compressor and expander in turn and adjust RV1 and RV2, respectively, for minimum distortion of the waveform viewed on an oscilloscope. This adjustment can also be carried out without testing equipment, but does require a good ear to get the audibly cleanest waveform.

Using the unit should be simplicity itself and basically you should be able to plug it in and forget all about it. Judging by the comments of two recording studios using the unit, it appears to be gentle and uncomplaining by nature, just like the ideal wife (or husband)!

Modifications

Apart from noise reduction, these circuits can also be adapted for use as a general purpose compressor-expander. The first modification to be made is to add a pre-emphasis/de-emphasis 'defeat' switch, bypassing C4 and R4 in the compressor and taking C20 and R30 out of circuit in the expander, so that the frequency response of the compressor or expander remains substantially flat. With this modified circuit, 2:1 compression can be performed on excessively wide dynamic range signals (guitar, piano, etc.) to get a more punchy sound, and 1:2 expansion can be carried out on previously compressed material (much rock music) to increase the dynamic range. However, this degree of compression or expansion is excessive for all but special effects and some means of altering the gain adjustment ratio is necessary. This is easily accomplished by introducing some variable feedback into the circuits using a couple of dual-ganged 4k7 potentiometers and some 4k7 resistors (see Figure 6). It should be possible to make these adaptations without too much destruction, but it will be necessary to break two tracks on the PCB, ie. that joining the track from C9, C10 and R12 with R10, and that from C15 and C14 to SW1a. It is important to note that the ganged potentiometers are connected so that the two halves operate in opposite directions.

The resistors and pots required to adapt a single compander board appear in the separate Modification Parts List.

With these adaptations, the compressor and expander sections will offer, respectively, a compression ratio adjustable from 1:1 to 2:1 and an expansion ratio ranging from 1:1 to 1:2.

Such a unit is similar to that marketed by dbx in the USA, and can be very useful in adding a bit of guts to that reluctant electric piano, or whatever. The dbx expander is claimed to restore the original dynamic range to recordings compressed during transfer to vinyl. I'd make no such claims, but it can be effective on some internally compressed keyboards and also yields interesting results if used with some of the over-compressed heavy rock and heavy metal discs around today!