Bazille: Rate of looped EG-cum-OSC not tracking keyboard properly

[themagicalkamja]

Another title for this post could be: Can I use a looped EG as an oscillator in Bazille? If so, how?

In trying to use ENV2 as an oscillator, I (1) sent it to the VCA, (2) set it to loop, (3) dropped Att and Sus to 0 and Dec so short that the EG was looping at audio rate, and (4) lastly, adjusted Dec to tune this EG-cum-OSC to the desired pitch. So far so good.

From there, the last step should be to set Rate Mod to KeyFollow and increase the amount. If successful, this should cause Dec to open/close as I press higher/lower notes, effectively changing the frequency of the loop so that the EG-cum-OSC's pitch tracks the keyboard, just like a regular oscillator.

Or so I thought, because it doesn't seem to work. Instead, there are two problems. First, the pitch doesn't increase evenly across the keyboard, like it does when keytracking a self-oscillating filter. If I hit two keys a whole-tone apart, say, the resulting notes may be more or less than a whole-tone apart.

Second, the keytracking doesn't stay consistent even on single notes. So if I set the Rate Mod Depth to X and hit the same key twice, the first time it might sound fine, but the second time it will be slightly detuned from the first time, despite that the Rate Mod Depth did not change.

Why do I get these strange results? Am I doing something wrong? Or is it just not possible in Bazille to use looped EGs as perfectly keytracked oscillators?

[Urs]

The envelopes run at Controlrate (1/4 samplingrate), they're not meant to be used as audio sources (hence no input sockets). Unfortunately it would drive CPU up crazy to render them at full speed plus bandlimiting.

I would still be interested in the patch though. Haven't thought of this application yet, and still brooding over a Maths-style update, maybe, one day.

[Urs]

Also of course... the rate modulation is scaled linearly (afaik), not exponentially. So it isn't comparable to a volt/octave scaling that would be required for the Keyfollow to take the desired effect.

(whereas recently I'm doing a lot of 2^x scalings for a lot of parameters)

[themagicalkamja]

The envelopes run at Controlrate (1/4 samplingrate), they're not meant to be used as audio sources (hence no input sockets). Unfortunately it would drive CPU up crazy to render them at full speed plus bandlimiting.

I would still be interested in the patch though. Haven't thought of this application yet, and still brooding over a Maths-style update, maybe, one day.

The patch, as requested!

https://nofile.io/f/o5DX1OtXnnH/EG-as-OSC.h2p

I used the technique in my first post to turn ENV2 and ENV3 into oscillators and tuned them to uhe's middle E3 and then approximately a fifth up (the resolution of KeyF in this instance is insufficient for perfect tuning). Then I sent these signals to different VCAs and hard panned them to allow for easy comparison between their tunings.

Load this patch and you should see the two problems I was talking about. First, hit E3 and it SHOULD sound reasonably harmonic (still not great), approximately like ENV1=E3 and ENV2=B3. But if you go one semitone up to F3, only ENV2 tracks, and not that well. You have to go all the way to G3 before ENV3 starts to track at all. Second, hit E3 repeatedly and you should hear that the tracking is unstable: ENV3's tuning jumps around a bit, while ENV2's stays steady. Very weird.

A quick patching note: In order for the pitch to increase with higher notes, Env Rate Mod Depth must be negative in this case, because higher notes = shorter decay times. I had to think about this for a second as I was initially setting it up, since I'm used to positive keyfollow values begetting increases up the keyboard. But here, higher KeyF means "shorter decay", not "higher pitch" directly.

Anyway, it would be awesome if this could be straightened out in a future update. If the math can be ironed out, Bazille would acquire 3-4 more uniquely shapeable oscillators without even touching the GUI or adding a single module!

[themagicalkamja]

I would still be interested in the patch though. Haven't thought of this application yet, and still brooding over a Maths-style update, maybe, one day.

The patch, as requested!

https://nofile.io/f/o5DX1OtXnnH/EG-as-OSC.h2p

Attaching the same patch here as a .zip for posterity, since nofile.io URLs seem to expire after a given period.

Urs, if you get the chance to inspect this patch, please let me know if you can figure out the answers to my questions above.

Also interesting to note and probably related, if you tune the looped EG as described in my first post and then set the LFO to lightly modulate the EG's rate, the "oscillator" tuning fluctuates vibrato-style (as you'd expect), but in a graded fashion with very noticeable steps rather than smooth oscillations of the pitch like you would get if you LFO-modded the frequency of a regular VCO. Why?

[ahanysz]

To shed some light on this, try holding down a note and moving your pitch bend wheel. You'll see that the envelope rate is quantised: it moves in steps, not smoothly. (For comparison, put noise through a LP24 filter, turn up the resonance and key-track the cutoff. When you move the pitch wheel, the pitch changes smoothly.) You'll notice the interval between steps is wider (in terms of musical intervals, e.g. semitones or steps) for higher numbers.

Running at 44.1 Khz, and putting Bazille through a tuner, I find that envelope 2 is generating pitches in Hertz of around 153,162,172,184,196,212,229,250,275,306,344,393,459,551,... Divide 44100 by each of these numbers and plot a graph of the results to see a nice straight line of slope -16 -> my conclusion is that the rate of an envelope is always a multiple of 16 samples. Not something you'd notice if you use the envelopes "normally" (-: For a low note, the period is long, so 16 samples is a fraction of a semitone. For a high note, 16 samples is much wider as a musical interval.