Zebra 3.0 - out now

[knob_monster]

Subsample phase wrap sharing is indeed the only way to avoid sync jitter in a modular graph. Without rendering both phases in the same block, zero-crossing extraction is always a sample late, causing digital aliasing. A dual-VCO module is the only clean architectural solution for that hardware-style sync sound.

[rACatkvr]

Apropos of nothing, I was just playing around with Zebra 3 and I was really struck by just the raw sound of the oscillators.

Just two oscillators set to saw slightly detunes, panned slightly and using two separate pitch settings for each oscillator with uncalibrate at different levels on each.

Sounds absolutely amazing.

Synthesizers are for making sounds?! News to me

Enough of your radical ideas.

You don’t want to derail this by talking about the actual (wonderful) sounds of Z3? /sarcasm mode off

I find Z3 to be great and, unlike those who are already asking for more this or more that or more this AND that, think that it’s deep enough to drown in for at least 4-5 years before hitting limits.

[liquidsound]

it’s deep enough to drown in for at least 4-5 years before hitting limits.

Not including Sleep, eat, work etc.

[cvale]

I do think it's worth saying: there are different expectations between a niche synth that does a few things well but stays in its lane, and a workhorse synth that tries to cover as much ground as possible. In that regard, I think when someone bemoans that Zebra 3 doesn't have X or Y feature, in a sense it's a rather large compliment. Because Zebra 3 is definitely a workhorse (work zebra?) and a really good one, and a great sounding one. If it were subpar, no one would care that it lacked features X or Y. Zebra 3 is soooooo painfully close to being the only synth you'd ever need that the narrow range of things it can't do are more conspicuous than they would be on a synth that can't do a hundred of the things that Zebra 3 can do.
What I'm saying is that there are a couple sides to this: on the one hand it's unfair to expect that it's going to be perfect for everyone, because that's impossible. On the other hand, wishing that it had X and Y features is a product of the implicit acknowledgement of how good it already is and that yes, it's natural to wish that something which is 99% perfect would just be 100% perfect. There isn't really a definitive solution to that. The best that can be hoped for is that the developers do what they can to meet the expectations they are capable of meeting, and from the other side, people are willing to compromise the expectations that just can't be reasonably met.

[exmatproton]

I do think it's worth saying: there are different expectations between a niche synth that does a few things well but stays in its lane, and a workhorse synth that tries to cover as much ground as possible. In that regard, I think when someone bemoans that Zebra 3 doesn't have X or Y feature, in a sense it's a rather large compliment. Because Zebra 3 is definitely a workhorse (work zebra?) and a really good one, and a great sounding one. If it were subpar, no one would care that it lacked features X or Y. Zebra 3 is soooooo painfully close to being the only synth you'd ever need that the narrow range of things it can't do are more conspicuous than they would be on a synth that can't do a hundred of the things that Zebra 3 can do.
What I'm saying is that there are a couple sides to this: on the one hand it's unfair to expect that it's going to be perfect for everyone, because that's impossible. On the other hand, wishing that it had X and Y features is a product of the implicit acknowledgement of how good it already is and that yes, it's natural to wish that something which is 99% perfect would just be 100% perfect. There isn't really a definitive solution to that. The best that can be hoped for is that the developers do what they can to meet the expectations they are capable of meeting, and from the other side, people are willing to compromise the expectations that just can't be reasonably met.

Exactly this

[pdxindy]

Zebra 3 is soooooo painfully close to being the only synth you'd ever need that the narrow range of things it can't do are more conspicuous than they would be on a synth that can't do a hundred of the things that Zebra 3 can do.

First, Zebra 3 is not close to being the only synth people would need. Lots of people use sample based synthesis, including granular and Zebra 3 doesn't do that at all. I can also do various stuff in Bazille that Zebra will never do.

Second, lots of people are not interest in or wishing for one synth that does everything.

On the other hand, wishing that it had X and Y features is a product of the implicit acknowledgement of how good it already is and that yes, it's natural to wish that something which is 99% perfect would just be 100% perfect. There isn't really a definitive solution to that. The best that can be hoped for is that the developers do what they can to meet the expectations they are capable of meeting, and from the other side, people are willing to compromise the expectations that just can't be reasonably met.

I disagree with the idea that more features will make a synth better. I disagree with the very concept of perfection. And as for what I hope, I hope that the developer doesn't do what they can to meet the expectations of those who want to satisfy feature checklists.

IMO, Zebra 3 is already complex enough. It already has enough modules and synthesis possibilities. (I'm considering an Arp/Seq as already a done deal, just not available yet.)

I can see a few workflow tweaks. Some users want a few theme options. I can envision a couple OscFX additions and a few additional Math Module modes. Then there's sorting out the A-D controls. But besides some refinements like that, I hope Zebra 3 doesn't appreciably change over the next year or two.

EDIT: Also, just in case it's not clear, I'm not saying your thoughts are wrong and mine are right, just that such thinking is a point of view and subjective. We have different interests in what we want from a synth.

[cvale]

Zebra 3 is soooooo painfully close to being the only synth you'd ever need that the narrow range of things it can't do are more conspicuous than they would be on a synth that can't do a hundred of the things that Zebra 3 can do.

First, Zebra 3 is not close to being the only synth people would need. Lots of people use sample based synthesis, including granular and Zebra 3 doesn't do that at all. I can also do various stuff in Bazille that Zebra will never do.

Second, lots of people are not interest in or wishing for one synth that does everything.

On the other hand, wishing that it had X and Y features is a product of the implicit acknowledgement of how good it already is and that yes, it's natural to wish that something which is 99% perfect would just be 100% perfect. There isn't really a definitive solution to that. The best that can be hoped for is that the developers do what they can to meet the expectations they are capable of meeting, and from the other side, people are willing to compromise the expectations that just can't be reasonably met.

I disagree with the idea that more features will make a synth better. I disagree with the very concept of perfection. And as for what I hope, I hope that the developer doesn't do what they can to meet the expectations of those who want to satisfy feature checklists.

IMO, Zebra 3 is already complex enough. It already has enough modules and synthesis possibilities. (I'm considering an Arp/Seq as already a done deal, just not available yet.)

I can see a few workflow tweaks. Some users want a few theme options. I can envision a couple OscFX additions and a few additional Math Module modes. Then there's sorting out the A-D controls. But besides some refinements like that, I hope Zebra 3 doesn't appreciably change over the next year or two.

EDIT: Also, just in case it's not clear, I'm not saying your thoughts are wrong and mine are right, just that such thinking is a point of view and subjective. We have different interests in what we want from a synth.

I feel like you misread my entire post, but I'm not going to lose any sleep over it.

[loctune]

Happy to hear notch filter will be added!

The issue with the sync OSCFX is it works only in its own osc and thus it's own waveform. It works nicely when it works, but having a dedicated osc sync function between osc would allow for much more flexibility to do things we use and like a lot on hardware, say something like the FM'ed sync effect on ARP Odyssey. While it might be technically possible to create things like this within a single osc by playing with waveforms, interpolations, and a bunch of OSC FX, having a dedicated sync function between oscs would also make the workflow much easier.

I agree that Zebra 3 has the ability to do many spectral things that others cannot. I have been using them and like them a lot. However, there's a critical thing missing in Zebra 3's spectral editor: it seems that Zebra 3 does not have the ability to transfer a given curve from time domain to spectral, or vice versa. So I can't draw a square wave in time domain and move to spectral to add some simple 2nd partial. I'd have to stay with the initial domain which I draw my curve in: do 500 partials for a square wave so that I can modify the 2nd partial in spectral mode, or use my brain to calculate what the waveform would look like in time domain when I add the partial...

I also have some other minor gripes about the spectral editing mode, like the not-so-pleasant UX for drawing in partials and the bug in precision math mode when you get unwanted partials. But I'd say the curve transfer is what I find the most annoying one.

Here's the thing, if you want analogue sync with analogue FM in Zebra, we need to add third oscillator type, which again would need to be a dual Osc. Neither the FMO nor the Spline based Osc will be able to handle Sync and FM.

The problem with analogue Sync is that it only sounds good if the subsample position of the trigger signal is known. This is where modular digital implementations fail. One can't just take the audio output of an oscillator and hope to extract the proper zero crossing from it. It jitters, resulting in artefacts. Proper Sync emulation requires that both oscillators are processed together, and the internal phase wrap position of one oscillator is used to wrap the other.

Hence, in order to do proper analogue Sync, we'd need a dual VCO module in Zebra. This was discussed for Zebra 2 many times and we have discussed it for Zebra 3. And each time we discuss it, we pretty much agree that it doesn't add enough value to Zebra, because the overlap between this and the other oscillators would be huge.

And we also think that people would expect such a functionality in the normal oscillators, where for reasons given we don't really see a chance to make it sound good. Digital oscillators are simply not good at sync, unless processed together, which in Zebra, by design, they are not.

#----

Spectral manipulation

Yes, it is not possible to represent both the time domain waveform and the frequency domain spectrum with a humanely manageable set of points. Nothing we could possibly do would change that.

The spectrum of a square wave for example would require 1000+ points. Similarly, the waveform representation of a spectrum with just one (!) harmonic (the highest possible) would require 2000+ points to be represented.

Therefore, what you can't do in Zebra 3 is "hey, I have this waveform, but I want to lower the third and fifth partial a bit, so I just flip there, edit those two and then go back".

The advantage of spline based editing is the infinite geometrical resolution that makes them independent from sample size, and the wide variety of shapes that are possible with a very limited number of control points. One can do a lot with it, and a lot of it sounds really good, but one can not do everything.

Switching between time domain and spectral domain in particular is not possible, hence we added two separate ways to interpret the curve, geometry and spectrum.

The alternative to this is a sample based editor, as also found in Zebra 2 (or Serum fwiw). In Zebra 2 it would have been possible to implement a switch between spectrum and waveform (but switching to waveform would eradicate half the spectrum), but there are only 128 samples. This is never enough resolution to sound crisp enough, which requires 2048 samples for a sweet spot (that's what Hive and Serum use, and what Zebra 2 has used internally when rendering its own splines).

So when being able to switch between waveform and spectrum, you don't get much of a chance to edit everything as it needs to present you with 1000+ harmonics for 2000+ samples. The sheer mass of that makes loads of manual operations tedious, if not impossible.

So yes, sample based editing lets one switch between spectrum and waveform (Serum even presents both), but it poses a conundrum: Who's gonna edit all that unless for a handful of tables?

Sure, many sounds require just 20-30 partials. Totally manageable. But here comes my gist: With that little amount of partials, it's already pretty easy to do it in Zebra 3, and we do have ideas on improving that further.

We'll continue to work on the editor and fix bugs. But we can't change the reality that switching between time domain and frequency domain in the spline editor is impossible.

Thanks for your detailed explanation Urs!

While the criterion regarding manageable number of control points makes perfect sense to me, I do think that having a way to manipulate partials would enhance the workflow a lot. Given we already have something like the spectral focus osc fx which lets us manipulate partials in specific ways, is it possible to implement a tool in the curve editor, which lets us add/reduce some given amount of a specific partial?

I also have a further question regarding osc sync. I have a number of soft synths which does both digital wavetable and some kind of osc sync (i.e. Serum, done through osc wrap). While I have no intention implying Zebra should do the same thing (and I am fine that Zebra doesn’t have it), I am curious about your opinion: is there something specific in Zebra which prevents this kind of implementation technically, or it just doesn’t work good enough?

[Urs]

I can't say much about more options to control Spectral Focus. I do not have an idea right now what else we can do there, or whatever OscFX we're maybe going to add in future that may or may not allow for more detailed harmonic control.

is there something specific in Zebra which prevents this kind of implementation technically, or it just doesn’t work good enough?

I can answer this one quickly:

There are implementation details, i.e. "the way Zebra works", that would compromise the sound quality of oscillator sync across modules.

By design, in Zebra the modules in the grid are processed from left to right, top to bottom. This is quite obvious, as it is also reflected by the way modules connect in the Grid: In the same row you can connect left to right, but not right to left, across rows you can connect top to bottom but not bottom to top.

It also means that all modules are processed separately, one after the other. Therefore, no two or more modules are processed in any kind of fused algorithm.

As explained, a good sounding digital osc sync implementation requires both oscillators to share low level data (phase wrap), not just the final output. Or else, jitter and - as pointed out by knob_monster - latency will compromise the effect.

Hence, because Zebra's modules are processed separately, low-level data can't be shared, which follows that osc sync can't be done in sufficient quality with the existing modules.

The separate processing is a vast CPU optimisation. It is the ultimate CPU optimisation for a concept with 50+ available modules.

[loctune]

Thanks for the explanation! It’s great to know more about the technical boundaries of Zebra.

I’ll just go with the other synths for hard sync then. No issue, I own more synths than I can ever use…

[EvilDragon]

What Serum does for sync is basically precisely what you already have in Zebra with Sync Osc FX... Literally does the same waveform repeat thing on a single oscillator, there's no cross-osc sync in Serum either.

[kraster]

I've never considered the shape of the oscillator imposing the sync on the other oscillator to be particularly important in hard sync since it's primarily acting as a clock/timing thing than having an innate spectral contribution to the sound.

[EvilDragon]

Depends how sync is done... could be on rising edge, on falling edge, or if it's soft sync it could be any number of other things...

[Urs]

I've never considered the shape of the oscillator imposing the sync on the other oscillator to be particularly important in hard sync since it's primarily acting as a clock/timing thing than having an innate spectral contribution to the sound.

The phase relation between main oscillator and synced oscillator has a minimal but audible contribution to the sound. That is basically the main topic here, in Zebra 3 an oscillator can sync to itself, but simulating a main oscillator would require that oscillator to have a fixed phase relation to the synced one. So we could solve this by taking the reverse perspective and allow one oscillator to start with the same phase distribution as one other. I'm afraid though it's not an intuitive way to go about this, and again, it adds very little yet needs a lot of explanation.

[loctune]

What Serum does for sync is basically precisely what you already have in Zebra with Sync Osc FX... Literally does the same waveform repeat thing on a single oscillator, there's no cross-osc sync in Serum either.

I stand corrected. Been working for too long yesterday and brain couldn’t work properly.

I do own other synths which does both wavetable and sync though, like the Korg Multi/Poly. I guess this one falls into the “processed together” category?