Roland's old SH101 monosynth is still a popular synth some ten years after its launch, but there were other SH synths before it - like the SH7.

Back in the days when polyphony meant Hammonds, Clavinets and Solina String Ensembles, a new Japanese company called Roland started making a range of self-contained monosynths with the prefix SH. First came the SH5 and SH3A, and two preset monophonics, the SH2000 and SH1000. You can tell these four apart from later SH models by the oval-headed slider buttons similar to those on Roland's System 100 modular system (not to be confused with the slightly later and much dinkier 100M).

Then came a small brother and big sister (with slightly more up-to-date slider tops) in the shape of the SH1 and SH7, and these were followed by the SH2, the SH09, and finally, as a last fling, the SH101, which brings us up into almost living memory. (There are rumours of an SH10 somewhere, but I've never seen any evidence.) Some of the numbering makes sense - the SH2, for instance, is a two-oscillator offshoot from the SH1; and there's a definite increase in complexity, facilities and price from 3A to 5 to 7. But some doesn't follow that sort of logic at all - the SH09 is a cross between the SH1 and SH2, and the SH101, resplendent in its choice of colours and sling-on-ability, had nothing at all to do with the modular 101 keyboard.

The SH7 was definitely designed to be the flagship of the range. Not only did it have a duophonic capability, it also had no less than 40 sliders, 14 rotary knobs, and 22 switches. And it has a place in my heart as the first synthesiser I ever owned.

The duophony wasn't exactly original, as ARP had introduced it with the Odyssey, and the Octave Electronics Cat was around at the time with the same function. But it was useful, in the days when the only way to go for any genuine synthesiser polyphony was by splashing out well over £3500 (equivalent today to around £11,000) on an Oberheim 4-Voice. The trouble was that the keyboard contacts weren't particularly reliable. Roland always seemed so good at producing solid, well-made instruments, even in those early days, that I suspect the need for a slightly more complicated switching system may well have caused the problems. When it was on form, though, it was a pleasure to be able to play non-consecutive thirds, fourths and fifths without needing two keyboards to do it.

You could, of course, switch to monophonic playing, in which case one key triggered both oscillators - and when you used the Sample and Hold facility or another synth or sequencer to trigger the notes, you had no choice in the matter - one (two VCO) note or nothing. Each oscillator could be switched independently to triangle, sawtooth, or square wave, and to 2', 4', 8', 16' or 32'. The pulse width on the square wave could be altered on either or both VCOs by a slider, or by modulation from the LFO or Envelope 1. That wasn't quite the end of it, either. Oscillator 1 had a second section of five square waves from 2' to 32', which you could bring in on sliders, giving a little bit of the feel of drawbar manipulation to organ-type sounds.

Each oscillator's pitch could then be modulated by LFO, sample and hold, or Autobend. Every good synth has at least one thing that sets it apart from the competition, and with the SH7, Autobend was one of those things. All it meant was that you could program in a little pitch variation at the beginning of each note - either a swoop up, or an overshoot before settling down to the correct pitch. A global control knob affected the autobend time, and you could switch between swoop or overshoot. The amount by which each oscillator was affected by this was set by two sliders, and you were away, with some good human voice and whistling effects, but also with some other useful sounds - these were admittedly the days when any lifelike imitation of real instruments was out of the question, but you could still get some flute, trumpet and fretless bass -like sounds more easily with autobend than by manipulation of the mod wheel. One of the few useful modulation features that Roland chose not to implement on the SH7 was envelope control of pitch, but the autobend feature goes a long way to making up for that.

You can also alter the pitch of VCO 2 relative to VCO 1, for ready-made musical intervals, or to thicken the sound. Plus there's a sync switch, which I must admit on my SH7 is a bit of a disappointment. I love classic sync sounds, where (in a typical case) one VCO is set an octave higher than the other, and then sync'd to it. The result should, with some strategic knob-twiddling, be that amazing sweeping phasey sound with different harmonics being picked out in turn. On my SH7, the damn switch seems to take its sync job too seriously, and ends up pulling the oscillators so tightly together that not half as much happens as, for instance, on a Moog or a Prophet. Whatever the technical reasons, it just doesn't sing like an American synth - or, by all accounts, like Roland's own SH5.

The two independent oscillators, though, did provide another of the SH7's unique selling points. Using them independently, you could do amazing things with the built-in ring modulator. Because it produces new frequencies, this rarely harmonises with conventional melodies or chord sequences, but it's great for adding colour and effects. You can also combine one of the VCOs in the ring mod with whatever's plugged into the external input - to usually disastrous but occasionally brilliant effect.

"What gives analogue synths their place in today's world is the fact that every control is continuously variable in real time."

Along with the two VCOs come two full ADSR envelope generators - a cut above all its rivals like the Odyssey and Cat, which have one ADSR but one AR envelope. When I bought the machine, I felt absolutely sure that you'd be able to set one ADSR to control one of the oscillator outputs, and one to control the other, so that you could (in monophonic mode) have a second sound coming in over the top of the first. It's not so and it's a big disappointment. (I had to wait another six years for a Poly 800 to get what I wanted.) There was a way round it, using the CV and gate out sockets on the SH7 to control another synth; but in those days even tiny monophonics cost several hundred pounds.

Other sound-making facilities onboard the SH7 include pink and white noise, and the ability to put the VCF into self-oscillation for a pure sine-wave. There's also the external audio input, for mucking about with electric guitars, Wurlitzers, string machines and the like.

But right now, filters. The SH7 was reasonably well-endowed in this department, with a high-pass filter as well as low-pass. The filters don't sound anywhere near as powerful as Moog filters - or Oberheim or Prophet, or Roland's own SH5, come to that - and the suspicion is that they are two-pole 12dB/octave rather than meatier four-pole 24dB/octave jobs. But they do a reasonable job, and the low-pass filter comes with a magnificent assortment of modulation possibilities. Here we go: you can modulate the filter with: envelope 1, positive or inverted: the LFO or S/H; the keyboard voltage or a pedal; VCO2 or noise; and an envelope follower positive or inverted. This envelope follower takes the sound coming into the external signal input, and sends a proportional voltage to the filter. Good for auto-wah, and also some other interesting effects using speech.

Back to the filter: you couldn't sensibly ask for a more comprehensive set of modulation possibilities. What Roland did was provide practically all the useful routings from a modular synth in a simple and reliable form - the amount of each of these modulations can be set by just five sliders. It's just a shame the filter wasn't powerful enough to take full advantage of 'em.

Tucked in the corner next to the VCF section come the VCA controls: a couple of sliders to control how much the LFO modulates the VCA for tremolo effects, and how much of a constant level there is on the VCA (as on the ARP Odyssey and System 100M). This is useful not just for drones and when you're processing an external sound source, but also to boost general level if you've got the low-pass filter way down. There's also a switch for selecting which of the envelope controls works the VCA. With Env 1 selected, the same envelope controls the VCA and VCF, while with Env 2, the VCF is still controlled by Env 1, but the VCA is controlled by Env 2 - useful, for example, for trumpet-like sounds, where the filter envelope voltages should be different from the amplifier envelope voltages.

"The SH7 was the flagship of the range - not only did it have a duophonic capability, it also had 40 sliders, 14 rotary knobs and 22 switches."

Moving from the bottom right-hand corner of the panel to the top left, the sample-and-hold section can sample voltages from a sawtooth, triangle or random source (the noise generator, in fact), and you can smooth off the voltage changes with an output lag slider. Next to this you can set the LFO to sawtooth, square, or sine-wave, and the sine-wave has a dedicated slider for controlling delay. This is basically for delayed vibrato, which at the time was a pretty good thing to have. Thinking about it, there are a lot of otherwise good synths that came along in the 15 years following the release of the SH7 which either didn't have the facility for delayed vibrato at all, or are a bit of a pain to set up for it.

What else does the SH7 have to justify its original list price of around £1170 (present-day equivalent over £4000)? Not much, really. A decent array of input/output sockets; a somewhat subdued colour scheme involving grey and a sort of tasteful khaki (if such a thing is possible); a transposition switch - one octave up or down from normal; a total volume control sensibly placed for routine and/or emergency use; and just two other unique selling points.

First, a portamento control that works on either upward or downward intervals, as well as on both - very impressive at the time. And second, a well-specified "bender" section. Bender is one of those unfortunate terms which Roland have persevered with over the years; what it meant in 1977 was "all-purpose modulation control". The way this is achieved is a model of brilliant design - one modulation lever, click-stopped in the middle and spring-loaded, along with three rotary controls for the amount of the modulation effect on the VCO, VCF, and VCA, and three-way switches for each of these, to affect either CV or LFO, or neither. This means that one lever has a lot of possible uses - or, as the excellent manual puts it, "Bender effects on the VCO, VCF and VCA may be used completely independent of each other, or simultaneously in any combination". As with filter modulation, Roland spared no expense to produce an easily-worked version of modular synth capabilities. (It's a great shame that the SH2000's touch-sensitivity didn't have the same sort of flexibility of control.)

Talking of the manual reminds me to say thanks to Roland UK. Nine years ago, my manual lost itself in the deepest recesses of Exeter University. Frantic calls to Roland UK at the time met with no response. When I came to write this article, I thought I'd give them another ring, just in case. I don't know if the air down in Wales is more invigorating than Brentford exhaust fumes, but I ended up with a pristine copy of the owner's manual of this rare old synth by first class post. To be truthful, there isn't a great deal that anyone who was brought up on subtractive analogue synthesis would need the manual to explain, because the SH7 is logically laid out. But as an introduction to all that old-fashioned stuff, it is exemplary.

The synth itself is really recommendable as a learning tool as well. It has practically all the modulation possibilities it's worth experimenting with, in an easy-to-use form; it's extremely solidly made, with components of high quality. And if the keyboard itself is a bit suspect, you can always run it with a MIDI/CV box, like the Roland MPU101. I'm sure Kenton Electronics could fit up an internal MIDI interface if you require - they have certainly retrofitted the SH5.

The SH7 is also a good buy as a source of interesting and usable sounds, and if you want polyphony it's easy to set up a good sound on the SH7 and then sample it. Things it does well are bright modulations (as on the Korg MS20 but with subtle variations much more easily achievable); elaborate sample-and-hold doodles; classic early '70s sequencer sounds (more mid-range than chunky bass, but usable nevertheless); and reasonable brass sounds which get more than reasonable and positively good when you get to the French horn and tuba end of the spectrum. All the wind sounds are usable, especially when doubled with an FM, LA, sampled or AWM sound - maybe just taking one line of a chordal brass pad, or doubling a trumpet part.

Like all early synths, the lack of memories is both a pain and a blessing. You have to fiddle for a long time with analogue sliders to duplicate the precise mix of control voltages that produced last week's killer sound but you're a lot more likely to go for the exact sound you want for a particular piece of music, instead of going through presets or memories and sticking with the one which does a decent job - even if you've used that sound any number of times before. What gives non-memory-equipped analogue synths their special place in today's world is the fact that every control is continuously variable, that it's intuitively easy to vary them, and that you can do it in real time while the sound is actually coming out of the speakers.

It's hard to place the SH7 in a league table of monophonic synths. It's a long way from being the most characterful or smooth-sounding instrument ever, but it is solidly made, intelligently designed, and presents an Aladdin's cave of analogue possibilities. All I can do is make a comparison with cars from an earlier era.

If the Minimoog was like an XK Jag, and the ARP Odyssey like a '60s Porsche, then the SH7 was like a two-tone Ford Zodiac with radio, sunroof and every other conceivable extra thrown in as well.