All-digital Programmable Polysynth
A British-built additive synthesis instrument with multiple keyboard splitting and overlaying, and up to 16 voices at its command. A dream come true? Simon Trask has the answer.
From the wilds of West Yorkshire comes a new computer-based instrument that could be Britain's answer to the PPG. We present an exclusive preview.
Last month we previewed an upmarket British instrument, the Anvil drum machine, which on spec looks set to give the Linn 9000 rather more than a run for its money. Well, there must be something in the air, because now comes news of another British instrument with upmarket pretensions, the AS32 digital synthesiser.
The people behind AS32 are a company by the name of Microskill, and the instrument has already taken them some three years to develop. The company's main line of business is, in fact, developing real-time computer systems for the gathering of aircraft flight data information - not exactly the sort of background you'd expect to produce a serious, contemporary musical instrument. But just think for a moment. Your average digital polyphonic synth makes a pretty hefty demand on both hardware and software designers, with the emphasis well and truly on real-time processing. So in a sense, the development of an instrument such as the AS32 isn't totally unrelated to the everyday encounters of computer folk like Microskill.
Company directors Peter Smith and Peter Sutcliffe (the latter being head of AS32 development) recently visited E&MM's offices, bringing with them one of only three pre-production samples of the AS32 for your curious reviewer to investigate. They also sampled the delights of the local curry house and lived to tell the tale.
First things first. The 'AS' bit stands for Additive Synthesis (the instrument's chosen method of sound creation - a feature it shares with the Anvil), whilst the '32' indicates the number of sound channels available. And yes, this does mean it's possible to play 32 notes at once (either through application of the forearm smash or - much gentler - the sustain pedal), but this entails using the machine's minimum voicing resources. The synth can in fact be 32-, 16- or eight-voice polyphonic depending on the number of channels allocated to each voice. All will be explained later.
For the technically-minded, Figure 1 provides an overview of the AS32's system architecture, and shows there to be no fewer than three processors running concurrently. At the heart of the system lies a 2MHz 68B09 with 64K of ROM and 96K of RAM, whilst a 1MHz 6803 handles keyboard scanning, and a 16-bit custom chip running at 10MHz handles all the sound-generation chores. The custom processor can perform several calculations in parallel ('pipelining seven-deep' is apparently the correct term here), resulting in a mighty impressive 70 million calculations per second. And I thought our Accounts Department was quick.
The AS32 has a six-octave (F-to-F) touch-sensitive plastic keyboard. Our sample had only attack velocity implemented, though I gather release velocity will also be supported on the production model. Apparently, the Italian keyboard Microskill are using isn't capable of registering aftertouch, which is certainly an unfortunate state of affairs for a modern upmarket synth. Some alteration of the keyboard's action might not go amiss, either; there's far too much give during key depression, and it's a bit worrying that fully depressed keys lie flush with the rim of the outer casing. All in all, I don't think this keyboard is likely to get a very favourable reaction from the pro players it's aimed at.
If you've cast your eyes over the accompanying AS32 mugshot, you may well have noticed a fatal absence, or to put it another way: wot, no wheels? Fortunately, these will be catered for on the production model, with the standard pair of wheels being positioned left of the keyboard. One will be assignable to pitch bend or stereo bend (the panning of voices in the stereo spectrum) while the other will be allocatable to modulation or 'space stereo' (the width of the stereo spectrum used). All very well, but in these days of multiple controller assignments, a bit more imagination on the part of the AS32's designers probably wouldn't go amiss.
It's time for that familiar guided tour of controls and their layout. This time, though, you shouldn't pay too much attention to the paintwork. The layout has been finalised, but the facia currently comprises no more than a bromide sheet overlaying touch-sensitive pads - such is the state of pre-production things.
Starting at the left, we have tuning, touch-response and volume selectors. Separate increment, decrement and display selection pads are available for each function, together with pads for toggling each function on and off. The tuning function is interesting: it stretches the tuning at each end of the keyboard to mimic the way a piano is tuned.
Next come four Effects pads that enable selection of different assignment options for the two performance wheels. To the right of these is a so-called Palette section. This displays voices currently on-palette and those currently selected, and enables voices to be called onto the palette and selected either singly or in certain combinations (see Figure 2). The Instrument section to the right of this enables selection of Preset, Organ, Synth and Palette voices.
There's a numeric keypad that works in conjunction with a nine-character LED display of the Microskill's Voicing Editor section, situated centre-panel. This section is clearly laid-out, with an array of selector pads (a maximum of two parameters per pad) surrounding the display. The relevant LED blinks at you whenever you select a parameter for editing - a nice touch.
A column of four pads next to the Editor section allows for the introduction of good ol' shifted functions, and it'll be interesting to see how the Voicing Editor layout resolves clarity when faced with the increased number of functions Microskill are planning.
Figure 2 provides an overview of how the AS32 structures its voices. As you can see, sound data is held in a Waveshape library, whilst all parameters governing treatment of that data (together with pointers to individual waveshapes) are held in a separate Voicing library. This means that any waveform can be assigned to any set of voicing parameters, though careful planning is needed as editing a waveform obviously affects all voices to which that waveform has been allocated.
But if you want to make use of a voice, you have to call it up to a location on the palette. Up to six voices can be held on the palette at any one time, and from there they can be placed on the keyboard either individually or in overlaid combinations 1+2, 3+4 and 5+6. It's also necessary to place a voice on the palette before it can be edited, so it's helpful to regard said palette as one big edit buffer. One useful consequence of this is that any alterations made to a voice can't be lost (even during power-down) until a new voice is placed in the same palette position.
At the time of writing, the AS32 has three voice types: Presets, Organs and Digisynths. Not an inspiring array of terms, to be sure, and soon to be changed. The organ voices, which have their own set of parameters, will probably be subsumed within one or both of the other voice types when the Microskill reaches the production stage. Whatever happens, you'll still be able to mix all types together on the palette, though presets are limited to Positions 1 and 3.
Each voice consists of two channels, each of which has its own waveshape and parameter assignments. If one of the channels has its volume zeroed (ie. is effectively turned off), then 32 notes are available simultaneously. If both channels are active, 16 notes are available. All pretty logical, so there are no prizes for guessing that if voice combinations are called onto the keyboard (in the 1+2, 3+4 or 5+6 combinations mentioned earlier), a total of eight voices become available.
Eventually, the AS32 will be capable of operating in three modes, though only the one outlined above was in existence on our pre-production model. Each mode will have its own waveshape and voicing libraries, but Mode 1 voices will be compatible with Modes 2 and 3, and Mode 2 voices with Mode 3. And if everything goes according to plan, you'll be able to store up to 40 entries in each library in Modes 1 and 2, and seven in Mode 3.
You're probably pretty confused by now, so I'll explain the future modes a little further. Mode 2 will allow each octave (C-to-C) of the keyboard to be assigned its own voice or voice combination. This won't, however, give true multiple-split points, as voicing values are scaled across the octave from one set of parameters to another. The object of this is to enable a finer degree of voice editing over the keyboard range than is normally possible (a bit like the principle of assigning multiple samples across the keyboard, á la Emulator II and Kurzweil), but the more mischievous among you will no doubt try assigning completely disparate sounds at each octave just to see what the system comes back with. We shall see.
Mode 3 will go almost all the way and back again, by allowing true splitpoints to be assigned every three notes. The mind boggles. And in Modes 2 and 3, each keyboard 'map' is conceived as one voice, which means that palette organisation is the same for all three modes, including selection of voice combinations. Access to voice parameters in these modes is accomplished by keying a note in the appropriate octave or three-note group. Simple keyboard splits can be accomplished by assigning the same parameters to any number of three-note splits (a copying facility would come in useful here), or by taking a Mode 2 voice and 'tidying up' a particular splitpoint in Mode 3. It has to be said - in case you hadn't already gathered - that both methods are rather long-winded methods of achieving a simple split.
The only notable omission I can think of, which really should be rectified post haste, is the provision of a palette library for each mode. Still, overall it's an impressive set of voice organisation facilities.
We already know that the AS32 uses additive synthesis as its method of sound creation. What this means is that timbres are created by building up harmonics (or partials), each with their own definable amplitude. The AS32 offers an impressive total of 64 partials, which are stored in RAM as tables of harmonic coefficients. When a voice is called onto the palette, a Fourier synthesis operation is performed on the appropriate waveshape library entries, and the result placed in the waveshape store of the palette RAM. The appropriate voicing parameters are then read into the voicing store of the palette RAM.
As already mentioned, each voice can have up to two channels assigned to it, but each of these channels must be assigned its own waveshape, and editing of a waveshape is accomplished via selection of a channel (or 'digitone', to use Microskill's unfortunate choice of term). Using the numeric keypad, values from zero to 99 may be entered for the amplitude of each partial, with quick access to any given partial being afforded by the keypad's increment and decrement pads. Any adjustment is instantly available for your aural delectation.
There are 16 voicing parameters currently available for each 'digitone', though apparently Microskill have more up their collective sleeve. These include attack, decay and release rates, sustain level and slope, modulation rate and depth, detune and overall volume level. Also available is keyboard-tracking of both ADSR and waveshape.
There are two modes of transposition: one is available during playing, and is specific to a particular palette position and voice (with a range of up or down two octaves), whilst the other is available during editing, and is specific to a particular voice tone (with a range of down three octaves and up two).
'Response Curve' is a useful feature which, when set to maximum, reverses normal touch response, so that, for instance, heavier playing results in a quieter sound. The amount is individually adjustable for each digitone, which allows subtle variations in balance to be achieved through variations of touch on the keyboard. The facility could usefully be extended to allow variations between digisynth voices when two are selected on the palette.
At some point, you'll want to store those wonderful voices that the AS32 would seem to be capable of producing. What facilities are Microskill offering?
First off, let's say that the concept of tape storage doesn't seem to be something that's entered the collective brain-cells of the AS32's developers. If you're used to developing aircraft control systems for a living, perhaps this isn't surprising.
What Microskill are planning to use instead is RAM cartridge storage. One cartridge will hold about a third of the internal waveshape and voicing library capacity for each of Modes 1 and 2, and a pair of waveshape and voicing entries for Mode 3. Not a tremendous amount, by any means, and the fact that only certain library entries can be stored could prove to be a nuisance, necessitating a certain amount of re-ordering via the palette.
Fortunately, Microskill do seem to be aware of the importance of developing a computer-based extension for the AS32, both for Fairlight-style display and manipulation of sounds and for saving of all libraries to disk via the computer. What isn't clear at the moment is whether this extension will take the form of a dedicated computer (in the manner of PPG's Waveterm) or whether software will be developed for a gaggle of popular home micros.
More unusually, the company are also planning to implement a high-speed parallel interface which could be used for memory access to a Winchester disk or for sampling. Sounds interesting.
A quick move of operations to the AS32's back panel. From left to right, we find left, mono and right audio outs, with a three-state adjustable level switch; a headphone socket (similarly-equipped); MIDI Out, Thru and In (currently awaiting implementation); two stereo jack foot-switch sockets (each capable of accepting two footswitch inputs); and three six-pin DIN sockets that are said to be connectors for external keyboards. Cartridge port, on/off switch and power socket are to the far right. The future status of the keyboard connector sockets is apparently uncertain, as they were conceived before the company was properly aware of MIDI, and may now be an unnecessary extra.
The left-hand footswitch socket is currently unallocated (likely candidates for implementation here are volume and modulation facilities), but its opposite number is currently assigned to sustain and selective sustain functions. Selective sustain is equivalent to the third pedal on a piano, ie. all notes sounding when the pedal is depressed are sustained, but any notes played after the pedal is depressed aren't affected - a very useful feature.
One thing is clear. Now that the software and hardware expertise have manifested themselves in something resembling a musical instrument, what that instrument now needs is the musical know-how to make it complete. There's no doubting Microskill have a powerful machine on their hands, but they'll need to listen to plenty of feedback from keyboard players before it'll become something musicians will actually want to use. To give the company credit, they are actively seeking such feedback, and the apparent ease with which they can incorporate resulting software changes is very encouraging. They've got a lot of things right, and I particularly liked the palette approach - though a palette library is a must, guys.
At the moment, the AS32 is lamentably short of anything like good sounds, but it's undoubtedly capable of producing them: what is there at the moment sounds extremely promising. A 'terminology gap' will also need to be closed if the AS32 is not to alienate potential customers.
More specifically, the advent of dedicated MIDI sequencers and MIDI software packages means the lack of any onboard sequencing facility is not in itself a great problem, but Microskill would be well advised to make the most of MIDI's multiple voice assignment possibilities when they get round to implementing the standard. A machine in the sort of price category currently being discussed - four to five grand - deserves nothing less.
Given their commercial background, it isn't surprising that Microskill are very aware of the importance of a computer extension to the AS32, and a good bet would surely be to apply their hardware and software knowledge to the production of a dedicated add-on similar to PPG's Waveterm. That would ensure both maximum flexibility and freedom from the home computer minefield.
Aside from all that, I'd say the AS32's biggest potential problem is that its sound-creation processes are rather time-consuming. Some effort expended on how to improve current editing facilities probably wouldn't go amiss. However, the other side of the coin is that additive synthesis is a very rewarding method of programming. Even in its simplest form, it offers the musician a means of coming to grips with the nature of sound in a far more concise way than subtractive or FM techniques will ever manage.
Still, AS techniques aren't exactly overfamiliar to the majority of synth programmers, so a lot of people (Microskill included) are going to have to do a fair bit of educational tutoring in an effort to get the system accepted.
It's a shame Microskill haven't implemented a filtering control of the sort found on both the PPG Waveterm system and the Anvil Percussion Synthesiser. There's no reason why more traditional facilities shouldn't be included, and the fact that filtering can be implemented in software should result in a system of some flexibility, if things are done properly.
Anyway, enough of my ramblings. Microskill will be exhibiting at the British Music Fair, by which time the AS32 should be in a more advanced state than it is at the time of writing. Try it out for yourself. For a non-music company's first-time offering, it's an impressive - if occasionally flawed-creation.
Further information from Microskill at Unit 3F, (Contact Details).
Review by Simon Trask
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!