Graphic Oscillator
2000-12-06 by Tkacs, Ken
The problem with a graphic oscillator is that the visual shape of a waveform doesn't have much real bearing on what the sound will be. And changes to that shape aren't the least bit intuitive, and are generally pretty useless. The concept sounds great at first---like you will be able to draw any waveform imaginable and have it at your fingertips. But the reality is that most sounds you get from it will be terrible. When I was in college I had proposed a similar Idea and an engineer friend mocked something up for me as a demo. It sounded like a great idea---draw ANY waveform!! When we played around with the prototype, our reactions were like... "oh." That approach is great for producing sounds akin to distorted 60-cycle hum (and not even in a good way...I like noise, but these just sounded like electronic 'problems'). We dropped the whole thing. That's why I get wary whenever someone comes out with a Wavetable-based oscillator boasting "99 banks of new waveforms," and then you look at their manual, web page, whatever, and they show scope traces of the "new" waveforms. "Here's the one that looks like a row of houses, here's the one that looks like an upside-down row of houses" (as if that would sound different), here's the one where every other cycle is a square or sawtooth..." And you *know* that they designed these new waveforms by their look, not with any regard to the useful portions of the spectrum that they offer. If they had really come up with some neat complimentary waveforms, I would expect to see an FFT of them, not just a scope trace. Ensoniq attempted to create some interesting waveforms for their ESQ machines. On their own, they are of limited use, but they were digitally designed to compliment the basic waveforms. These were then saved into the wavetables. They didn't have names like "row of houses" because they didn't look like anything simple. (Simple waveforms, like tri, sine, square, etc. are 'easily' made with analog anyway). Instead they were just called something like "Digital #2 with emphasized 9th and 10th partials." I'm not trying to be a wet rag on the idea; I'm just saying that the idea in practice isn't anywhere near as cool as it sounds. You think you will have the ultimate VCO but really it just grinds out nastiness. Additive is different---there, as you, let's say, move a slider up, you can hear that partial coming into play, and it sounds good and smooth and continuous as you move it, just like a lowpass filter sounds as it opens up (not "just" like---I'm trying to explain the 'continuity of it). Just changing a thin slice of a graphically-divided waveform wouldn't work anything like that. When you go out looking for new waveforms, you have to ask yourself why. Certainly you want something different that the Big 4 to compliment them. But simple bold waveforms are "bottom heavy" by nature, with very strong fundamentals. That's why their shapes are so defined, if you know what I mean (that and phase). With subtractive synthesis you end up "brightening" a tone by either adding another oscillator an interval higher (which is cool, but it imposes a 'spaced-out,' newly fundamental-heavy series higher up, and the ear can identify that), or else you highpass filter, which is still trying to pull-up the closely-packed, upper harmonics from a waveform that had very little energy up there in the first place. What would be cool would be to have waveforms *designed* to be up there, with harmonic series' to match where they land when being applied to the base waveform (I don't know how to describe this stuff without paper).