U-HE Zebra 3 Alpha Prototype Developments

[gentleclockdivider]

Yes , that get you pretty much in the same ballpark area
fwiw
viewtopic.php?t=420194

[e-theory]

I'll send you the patch so you can see I'm not "doing it wrong". You can see the evidence directly that it's using frame based crossfades and not smooth interpolation. That has absolutely nothing to do with what you are talking about. If the interpolation was moving the points, it would move them, but it's not.

[e-theory]

It's funny how people still obsess over Saw-PWM.

The (visible) artifacts are only audible on very fast modulation - where some perceptional artifacts ("scratchiness" for lack of a better word) come in that are also present in the original hardware anyway.

(for the fun of it, I tired it in Z3, same as Zebralette 3. At normal PWM-style LFO modulation, the difference between a resolution of 800Hz vs 2kHz is inaudible, hence the visual artifacts are inaudible. At very fast modulation there's a slight difference and thus artifacts become audible.)

Thanks for clarifying urs. So based on that it is a crossfaded/interpolated frame system, which makes sense. The reason I asked was to see if Zebra 3 would use the same crossfaded/interpolated frames or something different.

The main reason I wanted to try saw-pwm is that it sounds good and it's a good test for how good an interpolation algorithm is.

What you say about the original hardware isn't really true. It doesn't produce the same effect when modulated quickly, it sounds distinctly different but I know you already know this, since you implemented it for DIVA and it's obviously totally different which is clearly visible when comparing the waveforms.

Actually you got me thinking. Turning off the frame interpolation would actually fix this since you could then never get 50/50 frames. It might work better for my case. I'll try it.

I love Zebra so thanks for all you do. I'm looking forward to v3 a lot. Thanks.

[e-theory]

You're doing it wrong , I can see from the screenshot that you're animating the points across the Y-axis , noticeable inthe first slots
I specifically wrote the X-axis , it's fundamental to get smooth transitions

I've attached the preset so you can show me what you think I'm "doing wrong":

[e-theory]

Yes , that get you pretty much in the same ballpark area
fwiw
viewtopic.php?t=420194

This isn't possible in Zebra as it removes DC offset from the waveforms. If you wanted to use AM to do this, the PWM would need to be centred around the bottom of the waveform, not it's centre, else you end up multiplying the saw by a negative number and flipping it, whereas in the MKS it's chopped to ground. If zebra supported offsetting the AM it would work, but I'm not aware of it being able to do this.

[Urs]

Zebra 3 Oscs do not by default remove DC offset, exactly for the reason that processors such as ring modulation could be used to combine waveforms in whatever ways.

[Urs]

While I personally find Saw PWM way less interesting than it was in 1987, when I bought my MKS-50, I still think the snarling artefacts of fast modulation on both the hardware and its software implementations by far out-ugly whatever artefacts there may be by frame blending in Zebra 2 or Zebra 3, which at high resolution and moderate modulation rate are demonstrably inaudible.

[e-theory]

Zebra 3 Oscs do not by default remove DC offset, exactly for the reason that processors such as ring modulation could be used to combine waveforms in whatever ways.

Sweet, that wasn't my experience when I tried this before, but thanks for the tip, I'll try it again. That might be the best solution, thank you.

[e-theory]

While I personally find Saw PWM way less interesting than it was in 1987, when I bought my MKS-50, I still think the snarling artefacts of fast modulation on both the hardware and its software implementations by far out-ugly whatever artefacts there may be by frame blending in Zebra 2 or Zebra 3, which at high resolution and moderate modulation rate are demonstrably inaudible.

Again, that's totally fine, each to their own, I still love me some good gabba hardcore techno so I still love a good hoover, but I disagree about audibility. I can easily hear the differences between the linear interpolation artifacts of a wavetable, and a genuine interpolation of the waveform points, but each to their own. I know you have amazing ears because I love how your products sound, but to me, the MKS-50 VCO still sounds appreciably and demonstrably different to anything I've achieved with a wavetable synth (though I admit, that might have to do with things other than wavetable interpolation).

Quick question. Is there any way to completely disable interpolation to force Zebra 2 to just switch between wavetable entries rather than blend them? Thanks. I've tried all the settings in the oscillator and none seem to do that.

[Urs]

Switching sounds worse. You have to take into account that digital representations of waveforms, regardless if sampled or generated, need to be bandlimited according to their frequency and the sample rate. Such that these waveforms in all but the deepest bass registers *never* have perfectly steep slopes, perfectly even plateaus or anything else where a sudden switch would make sense. Any change in timbre of a waveform will result in a change of slope and/or level, and thus a switch will alias. Two common things can overcome this: Crossfading (CPU friendly) and BLEP/BLAMP Synthesis (not as CPU friendly).

In Zebra, we opt for crossfading, simply because oscillator unison would discourage BLEP/BLAMP. And also, at high resolutions settings, this brings artefacts from crossfading down to pretty much 60dB below the fundamental in this particular case. 60dB is 1/1000th, and usually considered the border from audible to inaudible, hence the use of "RT60" in envelopes, reverbs, delays as the common designation of "here it finishes" even if it actually finishes a little later, but usually below the level of perception.

Hence, congrats if you can still hear it, you must have exceptional hearing! I do not hear a difference between Zebra 3's "800Hz" and "2kHz" setting with the example discussed, so for me this is way, way below my threshold of perception.

[audiouser720]

Will you implement analogue envelopes in Zebra 3? AFAIK Zebra 2 didn’t have this feature, and Zebralette 3 doesn’t have it either. What I mean by this is the singing envelopes in mono mode, where the new trigger doesn’t return to zero but starts from the previous release stage, like in classic analogue monos

[RobGee]

Will you implement analogue envelopes in Zebra 3? AFAIK Zebra 2 didn’t have this feature, and Zebralette 3 doesn’t have it either. What I mean by this is the singing envelopes in mono mode, where the new trigger doesn’t return to zero but starts from the previous release stage, like in classic analogue monos

I asked this before in the Uhe forum and Urs confirmed Zebra 3 will have them.

[Urs]

Yes

[GeorgK]

I am not sure if this is compatible with the way Zebra3 handles WTs internally, but:
Disabling WT interpolation makes sense when a wavetable is used as a set (or library) of distinct waveforms (like a switchable oscillator) without any intention to dynamically sweep through the waveforms during sound generation. Stepping through the wavetable is a very convinient way to try out different waveforms in sound design; but it is practically impossible to dial in an exact waveform if the WT generator interpolates (except one stores the same waveform several times consecutively in the table).

[Urs]

It does not have wavetables in the common sense.

(it has splines based curves which use a single cycle wavetable format as an intermediate storage format for further processing in the spectral or time domain)