A New Kind Of Synthesizer
SOS reader Colin Muir tells why he wants a new kind of synthesizer for Christmas.
In recent years we have been blessed with a proliferation of new keyboards, mostly splendid sounding machines capable of producing a seemingly endless variety of sounds — so what's the problem? Why do I want a new kind of synthesizer?
Where a lot of today's synthesizers seem to fall down is that as more and more sophisticated digital techniques and systems are introduced, we move further and further away from intuitive programming methods. FM is a classic example — a very powerful and original method of synthesis, quite rightly a runaway success in the mid-Eighties, but as far as FM programming goes, very few could claim to know what they are doing in an intuitive sense when creating their own sounds. Ultimately, you do get used to adjusting modulators and carriers if you spend enough time at it, but the end results are seemingly still the product of some hidden sonic alchemy. Indeed, the best FM sounds I have ever programmed were invariably a result of serendipity rather than planning.
The same comments could be applied to other methods of synthesis, such as Casio's Phase Distortion (PD). And as for iPD synthesis (as used in the Casio VZ1) — I reckon that anyone who cracks that deserves a medal!
So what about Additive synthesis? Theoretically wonderful, but in practice it requires a marathon effort to programme just one sound. In fact, some have even resorted to using computed Fast Fourier Transformation conversions from samples to harmonic series just to coax decent sounds from Kawai's K5 additive synthesizer. But I believe that if it requires that much hassle to get a good sound out of a box, then something is fundamentally wrong with the concept.
Perhaps the biggest concession we have had to programmability in recent years has come along in the various guises of 'sample+synth' instruments like the D50, M1, K1, and so on. Here, the hard work of imitative synthesis — generating the crucial attack portion of a sound — is neatly sidestepped by providing short samples for each instrument required. However, there is a down side to this. I have never found these beasts very satisfying to programme, because the sample half of the equation is ultimately limited in both the number of alternative samples on offer and the timbral variation that may be introduced. Yes, you can improve things with velocity crossfading and, if you're lucky, your system might be able to apply a velocity sensitive low-pass filter (the D50 can't, the K4 can), but the final result is never anything like as flexible as I would wish.
So where does this leave us? When it comes to intuitive programming, I don't know of any system which is better than the good old low-pass subtractive filter envelopes, as found on oldie-but-goodie analogue synths. They are easy to use and quite effective. We can all relate to filters to some degree — we have them on our hi-fi and mixer, and their effect is easily understood. But if Subtractive synthesis is a nice friendly system to use, why has it been largely overtaken?
I think the answer must be due to the complexity of sound. Wavetable sequencing, AFM, PD, etc, are not intuitive synthesis systems, I would argue, but they are certainly powerful, offering vast dynamic change within one sound, temporally as well as through the influence of various modifiers (velocity, modulation, etc). Indeed, most of today's synths do sound great, but programming them is largely a matter of prodding, tweaking, and hoping!
So if that's my problem, what's my solution? (Cue fanfare of synthesized trumpets) Subtractive Dynamic Interpolate Filtration, or SDIF [© Colin Muir 1990], Obvious, isn't it?
Just in case you haven't figured out all the fine points from the name alone, let me explain what this SDIF synth could do. It may sound a bit complicated at first, but the end result is intended to be something far more intuitively programmable than today's synthesizers yet, at the same time, sonically outstanding. To achieve all this we will employ little more than a humble subtractive filter. Let's also give our synth a hi-res colour LCD display for visual editing and a tracker ball to drive it. Firstly, we simply draw a 'snapshot' frequency response for the filter at a given instant in time. Much the same as setting up a graphic equaliser but with hundreds of bands and a zoom in/out facility to make broad or specific changes. OK so far?
We've now specified the filter response at one instant in time. That time might be the start of the note, for example. Let's define another response graph at a later point — say, a quarter of a second later. As a starting point, we might take a copy of the first plot. Now we change this plot as we please, incrementally or drastically. We then define as many more of these snapshot filter responses as we wish, at any time in the envelope of the sound. (I haven't mentioned what sort of sound source we will feed into this filter yet, but I'll come to that a little later.) Still with me?
Now the clever bit (and you don't have to do a thing). This is where the 'interpolation' comes in. A big word but it just means that a computer is working out what the filter response should look like at all the time points in between each of your snapshot filters. Thus the filter is continuously mutated throughout this time, from the original to the new response. All this number-crunching and filtering will be going on with our blissful ignorance within a custom DSP (Digital Signal Processing) chip. Now, if two successive filter responses look completely different, then there will be a profound dynamic change applied to the source by the filter, over the time interval between them. Beginning to sound interesting?
At the risk of confusing you by over complication, let me also suggest that our system has Waveform Interpolation Sources (WIS). These work by specifying a waveform source at a specific time (either from a preset library within the machine or perhaps by drawing an original one) and a different wave at another time, and having the system interpolate between the two (or three or four...). Now this is not fundamental to the concept of SDIF, and it isn't particularly intuitive, but it does give the prodders and tweakers scope for some counter-intuitive fun!
I have plenty of suggestions for other enhancements to SDIF. I haven't room to go into them all here, but interpolative velocity sensitivity for the filter is an absolute must; non-linear (exponential) interpolation curves would be nice; looping between two filters would be interesting, as would the ability to frequency shift up or down all or part of a frequency response between successive time points. Stacking of filters would be handy too...
Enough! I hope this article has whetted some appetites. I believe a Subtractive Dynamic Interpolative Filtration synthesizer could offer visually intuitive sound programming in a way that none of today's machines can match, giving more dynamic sound movement for far less effort. We have the technology (I believe) and we have the market (I'm sure). If anyone wants to design one of these wondersynths, I would love to be consulted and would be more than happy to programme some of the factory sounds!
As I said, I want a new kind of synthesizer. Anybody want to build one?
Opinion by Colin Muir
Previous article in this issue:
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!