The previous post was an attempt to explain/introduce the OP-1's Phase synthesis engine; how successful of an attempt it was is debatable, but it's at least an initial stab in the dark and will be updated in the future with corrections or supplemental info -- if you have any thoughts or ideas, please hit up the forum thread linked to at the bottom of the post.
Now that we've established -- no matter how shakily -- what's happening inside Phase, and put forth some guesses at to how, it's time to look with a critical eye and ask why its current design was chosen, and whether there are any avenues for improvement in future iterations of the software.
As with the FM engine, it bears repeating that phase distortion is not the most simple or intuitive of synthesis techniques, being both abstractly mathematical and also relatively unfamiliar to most. Again, Teenage Engineering have put forth an excellent effort, and Phase is more beautiful and approachable because of their careful work.
Having said that, it's also important to realize that nothing is ever perfect, and if there are any easily-addressed issues or other UI avenues that could bear exploration, it might be useful to identify them.
Visual System
The "interactive diagram" metaphor of Phase's graphical interface is both aesthetically pleasing and quite functional -- the current value of each parameter is immediately identifiable.
My main gripe with this visual presentation is that an actual proper interactive diagram would be preferable to the current setup, which only plays at being authoritative and diagrammatic; it illustrates but does not accurately or usefully represent certain parameters, paying only lip-service to the mathematical purity underlying the synthesis.
There's certainly nothing wrong with the current layout -- it's just that it would be infinitely cooler were the graphics actually an interactive mathematical diagram.
The display of the White control is especially irksome, as it is so close to being accurate: currently, when White is rotated CW, both endpoints of the white line descend from top to bottom. If instead the left endpoint remained fixed at the top, while the right endpoint descended, this would be a perfectly accurate depiction of White's action: shaping the volume of each period of the wave into a decreasing ramp/sawtooth.
Possibly the CPU effort required to warp the orange sinusoid into a decreasing ramp was more than TE could afford; certainly the current scaling along the y axis is somewhat cheaper to calculate than a properly enveloped shape -- but not by much!
This is a relatively unimportant criticism; more important is the visualization of Green/Blue. These parameters are, as our wild flailings and inevitably inaccurate guesswork in the previous post attested to, quite complex and interact in strange ways.
The current graphical treatment does nothing aside from communicating their current abstract state, i.e two 1-dimensional values situated between minimum and maximum bounds. This is the bare minimum requirement for visual feedback in a synthesizer, and doesn't do justice to the beautiful AMOLED screen.
Better by far would be graphics which communicated not just a numerical value, but the semantic value of each parameter: some sort of diagram which told the user what each control "means" and tries to communicate how the sound changes as the parameters are moved.
Of course, the most obvious suggestions are to show the phase distortion function and/or the waveshape resulting from applying this function to the sine wave. The latter literal display isn't always useful -- in the case of Cluster or FM, seeing the complex waveform is unlikely to help users understand anything about what's happening; advocating a realtime waveform display also moves us closer to the slippery slope of "all synths look like Dr. Wave".
Nevertheless, some sort of evocative interactive movement would be greatly preferable to the current graphics: two sinusoidal sprites which shift left and right, failing to illuminate the user.
Parameters
Moving now from the visual interface to the function of the four encoders/parameters, we again come back to our old nemeses, Blue and Green -- or in other words, if I have correctly surmised, the phase distortion function.
As mentioned previously, these two parameters interact in complex and utterly unpredictable ways; because of the nature of PD synthesis, minute adjustments of the distortion function can result in large changes to the resulting waveshape, creating sudden discontinuities (and the sudden jump in high-frequency content these produce) and then just as suddenly removing them.
One good example of this chaotic unfathomable behaviour is the following: start by setting Blue, White, and Orange parameters to minimum (fully CCW) and Green to maximum (fully CW). Now, while holding down a key (so that you can hear the resulting tone), slowly turn Blue from minimum to maximum. As Blue traverses its entire range, it crosses two "peaks": areas where a noticeable amount of high-frequency energy is added, making the sound buzzy/thin.
Currently the audible result could be described as low->high->low->high, where the "high" regions are those high-frequency peaks and the "low" regions are periods of relatively darker, rounder sound. If instead the sound evolved in a more linear and consistent way as the parameter increased -- low->low->high->high -- the control might feel more natural and sensible to the user.
Looking at this particular example, we can discern a general rule or approach to user control of the phase distortion function which might be more intuitive and simpler to use: constant monotonic adjustment of the parameters should result in constant monotonic changes in the audible result.
This is pure conjecture, but I suspect that currently Blue and Green are altering the phase distortion function in a mathematically linear way: perhaps Blue moves one of the functions' line segments' vertices across the x axis or along some other linear path. The problem is that linear movements in "phase space" (or whatever abstract space the distortion function is described in) translate into non-linear, unpredictable and strange changes in audible spectrum.
What if, instead of effecting linear movements in phase space, each control point of the phase distortion function was swept along a complex non-linear curve in phase space? The curves would be chosen such that the audible result of a linear encoder movement would be a "linear" change to the resulting waveshape's spectrum -- a more constant, comprehensible and monotonic movement than the current "wandering all over the place" behaviour.
By warping the input to the phase distortion function in such a way that the output changes in a more predictable and consistant manner, the Blue/Green controls could be made more useful, friendly, and intuitive.
As for how to best implement such a mapping, or whether such a mapping even exists, I have no idea. But I think the general concept is sound: if linear changes to the internal variables cause non-linear results, why not warp things so that the resulting non-linear changes to the variables produce linear results?
Anyway, it's a thought.
One final criticism of Phase's parameters is that in practice, only two knobs ever tend to be useful at once: if Orange is at minimum then Blue and Green change the sound in drastic ways, but as Orange is increased their effect is diminished, eventually being completely drowned out by the warble. This means that you can choose to tweak Blue/Green, or White/Orange -- each pair is very effective and useful at different times, but never together.
This collision/conflict of parameters may be inherent to PD synthesis or to the sync-type process controlled by Orange. Certainly there are always such redundant dead-end spaces in any parameter set; for example, when a lowpass filter's cutoff is turned all the way down (rendering the output inaudible), the controls which alter e.g oscillator shape are completely moot!
Still, it would be nice if all the parameters were useful and Orange wasn't such a "beat you over the head with warble-y 2-op FM sounds" type of affair.
In Conclusion
Phase is a great little synth, whose nerdy clinical exterior belies a wild and uncontrollable range. While it would certainly be a great tragedy to diminish its chaotic sonic nature, taming Phase's chaotic useability issues might be an improvement.
I think that some carefully considered adjustments to the UI could preserve all of its considerable power while rendering it much more controllable, understandable and approachable. In other words, improving the reliability of the saddle/bridle/harness would allow us to better ride this wonderful beast!
This concludes our look at Phase. As usual, if you have any feedback of any sort -- complaints, corrections, additional information, etc. -- please visit this thread on the forums to discuss it: http://ohpeewon.com/discussion/234/op-101-phase
