The Sound Man: Steve McNally On OASYS And Trinity
Steve McNally is one of the guys that creates the sounds you hear on Korg's new instruments. As Steve puts it, "sounds programmed by musicians for musicians. That's what I do."
In what way is the Prophecy derived from OASYS, and in what ways does it differ?
"OASYS is basically a computer, rather than a hard‑wired keyboard synthesizer. If you want to change how it works, you can load a completely different synthesis system from a hard drive. It's multitimbral as well as polyphonic, and can also be multitimbral in the sense that different types of synthesis can be positioned under different areas of the keyboard. For example, you can have a single patch which, when you play softly at the top of the keyboard, gives you an FM sound layered with an analogue sound, but gives you a physical model of a saxophone when you play a little bit harder, and a PCM sample of a pipe organ when you bring in the ribbon controller... or something. It's all completely controllable. On the other hand, the Prophecy is limited to five fixed algorithms derived from the OASYS. These are held in ROM, and can't be overwritten with programs from a hard drive. It's also monophonic."
So, despite its 'analogue' sounds, the Prophecy is entirely digital?
"Yes; a DSP generates a pseudo‑analogue signal with all the imperfections that analogue produces. As an example, take a sine wave... we all know what it looks like, but some people may be surprised to know that it's actually the result of a mathematical equation. The analogue model within OASYS can generate that equation every time that you play a key, although it could be generating the mathematical equation of a sawtooth, or whatever. The sound then passes through a 'wave shaper' which allows you to alter or distort the waveform. This alteration can be subtle or dramatic, and can help to 'dirty up' the signal to emulate the imperfections of analogue sound. It will also, of course, give you some of the imperfections of real instruments, such as saxophones.
How were the models developed?
"Have you ever heard of Synth‑Kit? Well, here's a bit of background. Physical modelling was first developed in universities by engineers, who proposed that the sound of every instrument could be broken down into a mathematical equation. To develop their ideas, they had to sit down at a computer, and type in the whole equation. The computer could take all night to crunch this information, only to produce a single note at the end which wasn't what the engineers had been after at all, and they'd have to reprogramme the thing all over again.
"But then someone within Korg developed Synth‑Kit, a system that runs on an Apple Macintosh and does all of that number‑crunching in real‑time. You can sit in front of the Mac and say, 'here's the bit of the equation that relates to density of a string, so I'm going to put a little slider on the screen, move the mouse up and down, and listen as the density changes'. OASYS and the Prophecy were designed using Synth‑Kit, and their models were derived from it. Our developers said, 'here's the thing, this is how it's done', and Synth‑Kit generated the waveforms. Analogue synthesis can be modelled like any other sound.
The Prophecy's Log is an innovative control mechanism, but — particularly surprisingly for a synth with brass and reed modelling — you've left out the breath controller...
"We've seen that competent use of the breath controller can be imperative on another manufacturer's physical modelling instrument. But many players can't get the hang of them, and it's a difficult thing to demonstrate in a store. So we consciously tried to design a keyboard that offered all the nuances of modelling using more familiar types of controllers... the log, modulation wheels and so on. Anyway, if somebody wants to use a breath controller, they can get an Anatek wind controller — it's not very expensive. Breath control is available for those who want it.
What about real‑time control?
"In addition to the pitch wheel, the mod wheel, and the log — a total of seven different 'directions' — there are the five knobs that we call the Performance Editors. Each of these can be assigned four sound parameters, and each parameter can have its own range and response curve. So you can have the filter opening up while the resonance is coming down, the panning is going from left to right, and the delay times increasing... just by turning one knob. And it can all be recorded and sequenced over MIDI.
"The controllers even have their own ranges — if all the way from 0 to 127 is too great a range for you, you just define the range over which a controller will operate, say from 50 to 60."
The Trinity workstation was previewed at the same time as the Prophecy, and I gather that it'll have a Prophecy motherboard built‑in?
"The Trinity will have an expansion board that will hold 64 of the Prophecy's sounds. The only difference will be that the effects will be the Trinity's rather than the Prophecy's. You'll even be able to use the internal Prophecy sounds in combination with the Trinity's PCM‑based sounds. It's going to be pretty cool..." [Stay tuned — Ed].
Modelling Competition
Yamaha launched physical modelling little more than a year ago. The VL1 caused a mighty stir in the press, excited a small handful of players with spare wads of cash and enough time to learn to play it, and then quietly went away again. But Yamaha is the master of designing hugely expensive instruments that prove to be test‑beds for technology that eventually becomes an industry standard. In the '70s, the GX1 begat the CS80, and soon after, the GS1 was the forerunner of the DX7. Nevertheless, this time Korg has beaten them to the punch. Notwithstanding the VL1's little brother, the VL7, the Prophecy is without doubt the next generation of physical modelling synths, both in its approach and its price. Yamaha's Self‑Oscillating Virtual Acoustic Synthesis (S/VA) demands breath control, and to some extent the VL1 and VL7 stand or fall by the ability of the player to blow into a mouthpiece, manipulate mod wheels, and play. The Prophecy is cheaper, more immediate, and more flexible. Whether Yamaha will now launch a killer synth in direct competition with Korg remains to be seen.
In the meantime, players have just four modelling synths from which to choose: the VL1, the VL7, the Clavia Nordlead, and the Prophecy. Which is more suitable depends upon the eventual use. The Nordlead costs nearly £1500 and only offers analogue‑style synthesis, but it does so with 4‑voice polyphony and 4‑part multitimbrality. The bitimbral VL1 and monophonic VL7 concentrate far more on orchestral imitations and perhaps have the edge in terms of imitative realism, but, at £3995 and £2195 respectively, they remain more than twice as expensive (per voice) as the Korg.
Despite its obvious similarities to the Yamaha VL7, the Prophecy will almost certainly be most often compared to the Clavia Nordlead, arguably the first physical modelling 'analogue' synthesizer. There's a comparison table elsewhere in this article which compares these two synths and (since it's such a yardstick) the Minimoog.
The Engineer — Phill Macdonald
Phill Macdonald is well known as one of Korg's programmers and technical support staff. But when he's not creating and manipulating PCMs, he leads a secret existence as an Electronics Engineer. He seemed the ideal target for some tough questions...
Spill the beans, Phill... what is 'physical modelling'?
"It's a difficult concept to get across, made harder because there are so many ways to model a signal. But all the Korg algorithms developed using Synth‑Kit treat a sound as a combination of a resonator, plus a driver that 'excites' it.
"In the plucked model, for example, the response of a string is simulated by a resonator constructed from a wave‑guide with two delays, plus feedback paths back into the middle. The amount of delay fed back into the resonator controls the decay characteristic of the string, and the amount of high‑frequency damping defines the loss within the note. The pulse that you fire into the 'joint' between the two delays is the driver, and the different pulse shapes control how the model is 'picked'.
So any sound can be derived by taking a resonator and applying an impulse to start it 'vibrating'?
"In the string model, the driver is a pick, and the output from the model will differ according to the shape of the pick, the aggression with which you hit the 'string', and the nature of the pick itself. After all, a harpsichord is quite different to a picked bass, which is quite different to a thrash metal punk guitarist using a 10p piece. The Prophecy offers control over all those parameters."
And the reed and brass models?
"They're entirely different — a lot more complicated. The brass resonator is an expanding cylinder modelled as a series of tubes of increasing size that are connected to each other to simulate a bell. The driver is also very complex — it's a mini‑model of the human lips and palate. This has its own set of complex parameters which simulate the way the instrument is blown, and hence control the tone.
"A physical model is just a concoction of various elements each designed to simulate a specific task, so the reed model is similar to the horn model, but the driver is different — it's a reed rather than the lips and palate, and it has a different set of parameters appropriate to adjusting the character of a reed. The tube part of the model is similar to the brass resonator, but reed instruments can be wooden as well as metallic, and some are cylinders rather than cones, so the reed model has a number of further differences. Essentially, to get the woodiness and tone of an oboe, the Prophecy uses the saxophone model — but then passes the sound through a filter block set up to resonate like an oboe. You can even remove the reed from the model, leaving you with just a tube. This gives you flutes and blown bottles. Some of the solo stringed instruments in OASYS have loads of filters acting as very complex resonators for the body of, say, a violin.
"It's called a physical model because you're building a computed replica of what actually generates the sound — the Synth‑Kit layout for a Hammond Organ model looks pretty much like a simplified Hammond circuit diagram! The model is just a representation of the real thing, and while the nature of some acoustic instruments is a bit hard to explain, especially if you try to account for everything that's going on, you can take all the physical characteristics of the instrument and try to duplicate them in the model.
"Unfortunately, people have got the idea that modelling is a hugely complex synthesis system developed by rocket scientists for other rocket scientists. But the parameters within the Prophecy are really quite straightforward, and certainly not impossible to understand. Whilst the algorithms may be complex, we've defined a set of points at which meaningful values can be inserted.
Could you have defined other parameters that would allow the player to tweak the sound in different ways?
"That depends upon the complexities of the models themselves. Up to 90% of the combinations of all possible parameters and values would create meaningless sounds. And within some models, there's a tendency for parameters to interact with each other. So we grouped those that perform usable functions together, and gave them a specific name. That way, when you manipulate a value, you could be changing three or four factors within the model. We've simply brought forward the best parameters and the most useful ranges.
How about the analogue model? Does its structure relate to any particular monosynth?
"I'm particularly pleased with the analogue side of things. It sounds good and, in the UK, it's probably the most important aspect of the Prophecy. Our demonstrations have concentrated on the analogue emulation, because it's perceived as the instrument's biggest market. But as for relating the Prophecy to a classic synth... not as such. Think about it. All three‑oscillator 24dB/octave analogue synths should sound the same, but they don't. The reason is this: a lot of these instruments have some unusual quirk that makes them individually desirable. That quirkiness, and therefore the synth's character, can be programmed into the Prophecy via the modulation matrix. The matrix also means that you can't relate the Prophecy to a simple synth such as a Minimoog or Odyssey. Look at its facilities: cross‑modulation, ring modulation, oscillator sync, pulse width modulation, three oscillators, six envelope generators, the serial and parallel digital filters [the filters can be run in series as a single 24dB/octave filter, or as two separate 12dB/octave filters, each offering high pass, low pass, band pass, and band reject modes, all with resonance — Ed] — short of a dreaded Moog Modular there's no analogue monosynth like it. But, on the other hand, if you want to make the Prophecy sound tinny and weedy like a TB303, and have all the knobs interact as they do on a 303, that's quite possible. So maybe the Prophecy is the monosynth to replace all your other monosynths — provided that you only need one at a time! I certainly aimed for that when I was creating some of the patches.
What happens to a signal once it's passed through the models?
"The models' outputs run through what is, essentially, the front end of a synthesizer like the O1/W. You still run the waveform through a wave shaper, mixer, and filters — you just create that initial waveform with a model rather than a PCM or an oscillator. Not that the electronics of the Prophecy are the same as an O1/W, I'm just using that as an example of the principle."
Why does the Prophecy need three DSPs?
"The sound source itself is generated in one DSP. Things like the filtering, wave‑shaping and mixing are dealt with by another one. And then the effects are handled by the third. But the processors aren't dedicated; the programming can change the tasks each is performing."