10 thoughts on “Phase Distortion Synthesis: Libration Dub”

1. Nice. Really nice work. One thing tho is that the CZ-1000 is not an analog synthesizer…

2. @Ryan Bite! I thought that might raise a few eyebrows. The CZ-1000 has DCOs (digitally controlled oscillators), like the Juno-106 does for example. However, if I’m not mistaken, the oscillators themselves are technically analog. In other words, they’re not PCMs.

3. Do you have a reference for that? I can’t see how one could implement phase distortion without using some sort of digital wave table lookup. Pretty sure phase distortion on the CZ is just changing the rate of lookup in a single cycle wave form table … I guess it’s technically possible to do it analog but given the polyphony etc on the CZ I doubt it’s the case… ?

4. I remember reading it somewhere, but it’s not coming to me. I’ll investigate further. Perhaps they are just 8-bit PCMs… Not to diminish the instrument of course.

5. Ryan, my apologies. I stand corrected! According to Wikipedia the CZ series has digital oscillators. Thanks for calling me out on that one. I am still going to look at it on the scope, but wikipedia has this to say:

   The CZ-101 and CZ-1000 had only eight digital oscillators. For patches using one oscillator per voice, this allowed 8-note polyphony, but if two oscillators per voice were used, this restricted polyphony to four voices. The CZ-3000, CZ-5000, and CZ-1 had sixteen digital oscillators, making them sixteen- or eight-voice synthesizers. Each of the oscillators in a two-oscillator patch could be independently programmed.

   http://en.wikipedia.org/wiki/Casio_CZ_synthesizers

6. I think there’s some confusion here. They are not really DCOs, contrary to what has been posted in other places. I’m certainly not an expert, but I understand Phase Distortion synthesis to be a purely digital process, very similar to what Yamaha calls Frequency Modulation (FM) as found in the DX7 and all its cousins, both on the Yamaha side and elsewhere. PD is somewhat different and better able to emulate analog tones because of the types of waves generated. But a DCO is a “digitally CONTROLLED oscillator’, i.e. an “analog” oscillator receiving digital clock information rather than voltage pulses (as in a VCO). It’s still “analog”; Roland Juno synths and the Korg Poly 800 and even monsters like the SCI Prophet T8 (and everything Dave Smith is putting out these days) all have DCOs, not VCOs, but they are not digital instruments. Their oscillator tuning is stable because it’s digital, not voltage controlled. However, CZ series synths have Digital Oscillators (everything is in the digital, computer world), not digitally CONTROLLED oscillators. (Sorry for all the CAPS, there’s no italics). But these are far from PCMs! No samples here. To put a fine point on things…if you whack the CZ 1000, it won’t make the tone waver. If you whack a Poly 800, it will give a wobble, because it’s subject to things happening in the physical world. But a digital oscillator is far from a sample, and different from “pulse code modulation” (PCM) used on Casio’s home keyboards. Capische? Walter out.

7. I’ll add further that it is confusing, because, even tonally, I’ve noticed that even the tones on the CZ 1000 sound similar to my Korg Poly 800 (which is actually an analog synth w/ analog lowpass filter), and they came out around the same year. The DCO synths (tend to) sound more cold and “digital”. I liken this to architectural design–sometimes the media or the tool determines the outcome; if you physically draw a design, it has a more human quality, if you use AutoCad, it may be easily recognized as something that originated in the digital world, since you were using digital tools. This is at least true in the world of theatre set design, and may remain true for a while. Just a thought…

8. Loud and clear, @Walter. I appreciate the detailed description. One question: if PD does not use PCMs, does it amount to a primitive sort of analog modeling? In other words, how are the waveforms being generated? Are they stored as data somewhere, or modelled by the internal software?

9. I believe PD was originally based on a table lookup algorithm. An oscillator is basically three things here. First, a table that’s just a sample of a sine wave. This table isn’t mutable, it never changes. Second, a function that generates ‘phase’ values to look up on the table. Third, a function that actually looks up table results for phase. The output of the oscillator is the results of the last function.

   For a sine wave, the ‘phase’ values just steadily increase at note frequency, let’s say it rises from 0 to 1 over an interval of note frequency. For 440hz, there’s just a line from 0 to 1 every 440th of a second.

   One can add in another step, though. PD introduces another function that takes the 0-to-1 phase value and returns another 0-to-1 value, but not a steady line; more of a segmented, jagged distortion of a steady line. This is done with very little computation – the math and the shapes are not very complicated, more or less just algebraic.

   Passing this distorted phase to the lookup table gets a wave that’s based on a sine cycle, but is sort of sped up or slowed down in segments of that cycle. By cleverly playing with the extra bit of math operating on ‘phase’, the various different CZ waveforms can be generated.

   I’ve omitted a few details but the basic diagram of functions here is the whole story of PD.

10. Thanks for the fascinating insight, Andrew!