On the face of it, this is a thoroughly stupid thing to do. How can you realistically compare synths that cost close on £1,000, £2,000 and £4,000? They're bound to be miles apart in facilities and sound, you say, through a mouthful of ham sandwich.
Spot on, Gaucho, but the intriguing part is that they're also widely separated in thought — to produce all the classic synth voices, the factory programmers have to juggle the controls in a different way for each of the three machines.
So if we back track from a handful of familiar voices, those that appear on every polysynth, then we should gain an insight into programming, and its effectiveness. Thus the theory goes.
The three polys chosen for this exercise are the Roland Juno 60, the Roland Jupiter 6, and the Memory Moog... selected mainly because in their running order they have one, two and three banks of oscillators. In fact the Memory Moog is rare in having three VCO banks. The majority have two, but when the latter was designed, Moog wanted to make a polyphonic version of the Mini Moog, and that, as you'll recollect, is a threesome type.
They were all fed through the same studio monitoring system and occasionally tweaked with a little echo or chorus but otherwise were on their own. MIDIs out, no conferring. Your starter for ten. Make me some strings...
The earliest string section imitations were made using a chorus unit. This took the plain notes from an organ and, essentially, wobbled them around. By continuously shifting the pitch ever so slightly up and down, you gained the impression there was more than one organ note, maybe two or three marginally out of tune with each other. Tuning is one of the ways in which the brain counts. One reason it knows that an orchestra's string section has got 12 violins instead of one, is because they are never at exactly the same pitch.
Like many single oscillator polys, the Juno 60 has a built-in chorus unit, but string imitation has come several feet in the last few years. Naturally you now have an envelope generator to shape the sound, producing a softer attack, but the secret of a deeply fabulous string setting is to make it full and constantly shifting, without being cyclic. Nothing gives away a string synth more than hearing the chorus unit whirring up and down like an old bicycle wheel.
Since you can't produce a random chorus unit it helps if you can at least introduce another regular pitch or tone changing effect at a different speed so the movements begin to even out.
So the Juno, and many of its similarly priced contemporaries, use pulse width modulation. This technique of broadening and narrowing the shape of a square wave using a slow running LFO, closely resembles the beating of two oscillators when they're detuned. The Juno has the choice of three waveforms — pulse width, ramp and triangle. It adds in the ramp to thicken and enrich the tone. Triangle is a smoother alternative with fewer harmonics and wouldn't be much help in this case.
The Jupiter has a similar approach with its choice of waveforms and pulse width mod, but has the advantage of an extra envelope generator and a second set of oscillators. The detuned VCO's create another inner cycle, and the twin envelope generators supply a shade more punch in the attack without losing the graceful intro.
But it doesn't have a chorus unit, and that does make a difference. Sure, if you turn off the Juno's built-in effect, then the remaining string sound is thin and ineffective. But since we're judging instruments on the noise that emanates from the output jack, then the Juno in fact sounds sweeter, smoother and richer. That division rapidly disappears when the Jupiter meets an outboard chorus pedal.
There is something distinctive about a Roland string sound — it's bright, clear, very up-front with plenty of presence. Both Roland keyboards exhibited that characteristic which must be a quality of the oscillators and the filter (resonance 'off', cutoff frequency about half way in both cases).
The Moog has a different way of thinking. It also uses pulse width mod, though not controlled from the LFO. On the Memory Moog, it's possible to tune down oscillator three so low that it reduces to a click, in fact it becomes a Low Frequency Oscillator. But since it's still connected to the keyboard, each note will be an LFO of a different frequency.
If you now run the pulse width mod of oscillators one and two from oscillator three, then each note you press will have a unique rate. Hold down a chord and the PWMs are all churning away at individual speeds — the cycles vanishes entirely, the strings become a soft, unfocused swell. It's almost pointless attaching a chorus unit to the Memory Moog because the voices are already so 'loose'.
The Jupiter 6 also has the facility to untie one of its oscillator banks, but that only leaves one other set of VCOs on which it can work. The Moog still has two, so keeps its detuned effects.
Much has been said, written and telexed about the Moog filter, but it remains an enigma. It's softer and can be toned down for mellow violins and cellos to imbue a gentle, soothing feel that the Roland's can't quite match. Though the JP6's filter can also be muted, it seems to remove harmonics deeper within the voice, which the Moog leaves intact.
A few weeks ago, the boss of a rival weekly publication, and meself were having a drunken verbal brawl (one of our more civilised exchanges) about what noises people expect from synths these days. He argued that pre-sets such as 'Fender Rhodes' and 'Hammond Organ' were largely pointless since very few people buying a Juno 60 would have come across either. Both belong to an older lexicon of instrumentation.
I replied with a left hook and the theory that though the keyboards themselves no longer appeared in band line-ups, the sounds they produced have survived because they've been transferred to polys. It is
the sounds that are important — Hammond and Rhodes are convenient handles and standards by which you can judge newcomers.
So, though we trip back a good 20 years, the Hammond organ persists as a worthwhile barometer because it shows up plenty of programming tricks.
For those who don't know, the Hammond tone is typified by a percussive click at the beginning of the note — a byproduct of noisy key contacts in the original machine. In poly's it's re-created in a couple of ways. On the Juno, which has only one ADSR section, your finger becomes the envelope generator. The note starts immediately the key is pressed down, finishing when you release it, just like a 'real' organ. The ADSR and filter are used to produce the click. The ADSR is completely zeroed so that only a brief croak can escape from the filter.
The Juno sets the frequency and resonance of the filter at around half way since it still needs a tone control for the waveforms. The chorus unit contributes the other recognisable chunk of the Hammond throat — the Leslie speaker. The Juno's chorus unit has two settings, and this time the weaker one is applied so the organ tone doesn't become too swimmy.
The Jupiter 6 has a few more subtle suggestions to make. The waveforms and filter settings stay the same, but now as there are two envelope generators, the second one is aligned to give an immediate attack and decay, and a full sustain.
Leslie speakers instil a vibrato like shift in pitch — a doppler effect as the speaker swings round to face you then whirls off again. But they also have a tremolo effect — a rise and fall in volume that's mimicked on the JP6 by linking the VCA (Voltage Controlled Amplifier) to an LFO.
The two detuned oscillator banks fatten the sound, but again, ironically, I preferred the Hammond à la Juno, partly because it had the benefit of a suboscillator. This is an electronic division of the main oscillator at an octave lower, and it certainly added growl and grumble to the Juno.
The Moog goes off on its own, as you'd expect, and the Hammond test was perhaps the most telling when trying to see into the minds of American and Japanese designers.
Oriental design philosophy, where synths are concerned, can probably be summed up in the word 'perfection'. It's rumoured, for example, that Yamaha withdrew the CS80 not because it was out of date, but because it had a reputation for packing up. Even though many synth players swore by it at the time, and others were clamouring to get hold of one, Yamaha couldn't abide the thought of there being a product on their books that was anything less than 100 per cent reliable.
Roland have a similar mentality. The Juno and Jupiter are as near perfect as possible — immaculately matched circuits all the way down the line, uniformity of control, locked tight oscillators that rarely drift, and if they do, retune in less than two seconds (the Moog takes 15 or more).
The Memory has moments of... er... irrationality. But when you're attempting to copy as accidental and irregular a sound as the Hammond, you've got a head start. Perhaps that's why Roland are doing better with this decade's bands. We've come to expect razor sharp high gloss sounds, and the JP6 spits them out with spikey detachment.
You see, the Memory Moog is not an exact science. For example, when you wind up the volume of each of the oscillators to maximum, there's a good chance you'll start overdriving the filter or the output stages. There's not a control
for distortion, but it happens if you know how to get it. That's why most of the Memory Moog's organ programs are thoroughly impolite. But so was the Hammond, because it used to distort its Leslie cabs.
What can this mean? All three keyboards are analogue, sure enough. Digital, in this particular paragraph, is another sound handle. Synths such as Yamaha's DX7, boast a brand of bell-like and metallic settings that are a direct result of their form of tone generation — in the example of the Yamaha, many complex waveforms are created by the frequency modulation of several sine wave oscillators. So can analogue synths impersonate them?
This is out on the Juno. With only one bank of oscillators there's nothing you can do. But on the Jupiter and the Moog there are ways into this field.
The Jupiter applies cross modulation where the oscillators 'interfere' with each other — an effect similar to ring modulation, harsh, discordant noises flourish as the harmonics go apeshit. What I liked about the Roland's method was the speed with which you could produce such voices — simply move the cross modulation slider from zero, and they begin.
Not so the Moog which has no cross modulation but still 'chimes'. As there are three oscillators you can tune each one to a different pitch, then with the filter in oscillation as well, that's one note plus three manufactured harmonics which can be adjusted in level on the Moog's mixer section. It's slower, but it can be more successful. Chords still hang together, for example, whereas on the Roland, the cross modulation rapidly breaks up into a jumble.
But the JP beats the Memory hands down for glass rattling special effects. It can sound mental, furious and explosive. Don't argue with it.
Another digital accomplishment comes with noses... can't really think of any other way to describe it. A lot of Casio voices, for example, have an extra "whiff' of sound behind them, a cross between a musty pipe organ and a clarinet with a cold.
The Memory Moog and the JP had the same solution waiting. When you talk about syncing oscillators it normally implies you're going to shift the pitch of one of them with an LFO, subjecting the other to a screaming, flanger-like effect. It doesn't have
to be that way. Sync the oscillators, tune one high to an odd interval and it will collect a nasal overtone (trés digitalesque) which can be mixed into the desired amount. Introduce a slight attack and the result will be close to that 'hhunf' of a thwarted chip.
To round off, it's worth devoting a few sentences on performance, because this is where the Jupiter scores extensively over the more expensive Moog.
It features six note polyphony, same as the Memory Moog, but the keyboard can be split and the patches stored in the memory. Though the Moog probably has more features, they are all called up via various code words or numbers on the calculator style pad on the left of the control panel. This can involve punching three or four buttons, and the Jupiter invariably does the job in one.
It's also lighter, much faster to tune up, and on balance, easier to understand though comments about 'the simplest synth' are generally redundant. As soon as you've spent two or three hours coming to grips with any new machine it makes a lot more sense. No-one spends three and a half grand on a new poly then gives up after five minutes.
What sort of ultimate decisions can you come to when the subject is as nebulous as sound? There's as much fashion in our ears as in our eyes. The point that all three keyboards make (particularly the Juno) is that it is possible to work to a standard and a price without cutting corners.
In the coming year, when manufacturers battle to slice costs even further, be on the look out for the wrong economies — one overall filter instead of one for each note (back to the brass machine days), organ tone generators instead of VCOs, and so on. This feature should at least tell you what you should expect to HEAR at each price level.
That's what counts. As they say, listen then decide.
1. The Juno 60 is a solid, workmanlike synth, and the inclusion of the chorus unit not only circumnavigates the problem of a single bank of oscillators, but can be better suited to certain voices than detuned VCOs would be. Simple sounds, but strong ones.
2. The Jupiter 6 is the action man's machine — fast changes, big sounds, split keyboards, programmable arpeggiator — a great, busy synth. If you could only take one synth on stage, there's not much this wouldn't do.
3. The Memory Moog — finally, the analogue sound against which all others have to be judged.
Jupiter 6: £2250
Juno 60: £1199