Repro-1 (out now)

[Urs]

hakey,Are you saying you followed the same procedure on RePro One as Urs' video?

Yes- the first test with increasing resonance. Filter 4 is closer to this behaviour than filter 3, which changes pitch noticeably.

Filter 3 may be the more expensive algo, but it doesn't accurately replicate every aspect of u-he's Pro One. Presumably the intention is to capture generic Pro One behaviour rather than that of a specific instance (though I seem to recall that the latter was hinted at somewhere along the way).

When I change the waveshaper in the resonance path to behave like our Pro-One (self osc goes up by half a semi tone or os), 3 sounds correct but 1 and 4 explode. I had to "defuse" some parts of the model in order to make the more approximating algorithms work... however, we have example audio from Pro-Ones that drop in frequency by as much as a fourth, hence I thought it was okay to do so.

[Urs]

Having said all that, I like Diva's filters much better. There is no contest.

Perhaps that's not surprising if you keep in mind Diva's filter code was also (if not primarily ??) the work of Clemens Heppner. Given Diva's reception, perhaps Urs feels under some pressure to replicate Diva's success in equivalent products. However a key ingredient for that success, Clemens, is no longer there.

Well, I showed Clemens our latest method just yesterday and he went wow. He didn't really work in isolation back then, he implemented and optimised the algorithms after the methods we designed together on the whiteboard.

[Urs]

Why vote

Just add in the final release an extra knob, above the mastertune, which you can select a filter type suitable for you project. Everyone is happy

This won't be possibly I'm afraid... It took a lot of tweaking to make 1, 2, 4 and 5 work *at all*. Unfortunately, once I had 3 working I thought it would be a no-brainer to implement the other algorithms. But it wasn't - I had to change a lot of things, i.e. not let the filter open fully, lowpass modulation a tiny tad, and as I posted before, use a slightly different internal distortion. Once I put them back in, 1, 2, 4 and 5 will only work at 32x oversampling such that 3 will be the CPU-saving algorithm that only needs 8x oversampling.

Things are a bit more complicated than they may seem

[pdxindy]

Why vote

Just add in the final release an extra knob, above the mastertune, which you can select a filter type suitable for you project. Everyone is happy

This won't be possibly I'm afraid... It took a lot of tweaking to make 1, 2, 4 and 5 work *at all*. Unfortunately, once I had 3 working I thought it would be a no-brainer to implement the other algorithms. But it wasn't - I had to change a lot of things, i.e. not let the filter open fully, lowpass modulation a tiny tad, and as I posted before, use a slightly different internal distortion. Once I put them back in, 1, 2, 4 and 5 will only work at 32x oversampling such that 3 will be the CPU-saving algorithm that only needs 8x oversampling.

Things are a bit more complicated than they may seem

Then I suggest you go with #3... the 'cheap' model

[db3]

Why vote

Just add in the final release an extra knob, above the mastertune, which you can select a filter type suitable for you project. Everyone is happy

This won't be possibly I'm afraid... It took a lot of tweaking to make 1, 2, 4 and 5 work *at all*. Unfortunately, once I had 3 working I thought it would be a no-brainer to implement the other algorithms. But it wasn't - I had to change a lot of things, i.e. not let the filter open fully, lowpass modulation a tiny tad, and as I posted before, use a slightly different internal distortion. Once I put them back in, 1, 2, 4 and 5 will only work at 32x oversampling such that 3 will be the CPU-saving algorithm that only needs 8x oversampling.

Things are a bit more complicated than they may seem

Then I suggest you go with #3... the 'cheap' model

I thought #3 read as the cheap one in Urs' previous post. Confused Lol

[urosh]

When I change the waveshaper in the resonance path to behave like our Pro-One (self osc goes up by half a semi tone or os)

Wow, I was certain that 3320 uses plain old OTA for reso control based on scarce datasheet info. And judging by saw shape under high-mix-level high-reso condition I would guess that RePro uses something tanh-ish. So, you had to "flatten" reso path transfer function or what, to behave like 3320 when using exact solver?
Have you tried with tanh + large offset? Based on abysmal reso CV feed-trough info in datasheet, if OTA is used for reso it should have lousy specs, up to 10-15 mV of DC offset.

[PAK]

The RePro and Pro One filter FM clips Urs posted sounded close enough to me.

Which is fine.. the audio's there for anyone to listen and see what they get out of it. As mentioned, the amount of stepping the on-screen controls do (without shift) also gets in the way of perception with that type of sound, and regular MIDI CC will fair worse. It's a pretty objective thing to hear when you notice it, and best removed for a better sense of the engine. I didn't get around to that..

As I also said previously, spaceship invasion noises maybe aren't the best measure of what matters most in the end anyway.. it's merely something which is there to compare for now. Another reason why I'd rather wait for a completed product and more "regular" sounds to judge too

[PAK]

Quite how people were supposed to give a sensible answer, if the choices were between varying degrees of muffled or "artifacts", only on resonance and resonance modulation, on an otherwise identical sounding synth, I'm not sure.

???

Unless someone knew to check resonance/modulation, they wouldn't have heard any difference, except in their own mind. Perhaps not the best way to generate meaningful responses.

Even if they figured that bit it out, the difference was often a case of choosing between differing modulation patterns. I like this one more because it bleeps three times instead of twice. Three times is more analogue.

When it finally it got down to stuff that isn't as subjective, it was a choice between high cut frequencies (2 and 5) or "artifacts", with three sat in the middle. So, which sounds more analogue to you? Muffled resonance, or artifacts in resonance? Err..

If you didn't hear it like that, then well done you. As I also said above with hakey, some of this is about how much controls shift values at those sorts of settings, and (at least some of) the perception created is down to this.

It'd be a good test for people to assign high res controls and see what difference it makes. In this case, especially with those extreme settings, it matters more than usual.

[PAK]

Well, I showed Clemens our latest method just yesterday and he went wow. He didn't really work in isolation back then, he implemented and optimised the algorithms after the methods we designed together on the whiteboard.

I didn't think he did Urs, and it's good to know you're both still able to do that!

[fmr]

hakey,Are you saying you followed the same procedure on RePro One as Urs' video?

Yes- the first test with increasing resonance. Filter 4 is closer to this behaviour than filter 3, which changes pitch noticeably.

Filter 3 may be the more expensive algo, but it doesn't accurately replicate every aspect of u-he's Pro One. Presumably the intention is to capture generic Pro One behaviour rather than that of a specific instance (though I seem to recall that the latter was hinted at somewhere along the way).

When I change the waveshaper in the resonance path to behave like our Pro-One (self osc goes up by half a semi tone or os), 3 sounds correct but 1 and 4 explode. I had to "defuse" some parts of the model in order to make the more approximating algorithms work... however, we have example audio from Pro-Ones that drop in frequency by as much as a fourth, hence I thought it was okay to do so.

Hi Urs. There must be something in the Pro-One circuit that is causing that anomay, but it's not the filter chip. The chip is the same that is used by the Synthex (and others), and, AFAIK, the Synthex didn't show that anomaly (besides impementing a lot more filter variety than the Pro-One). Care to elaborate what's different?

[moscom_electronics]

Quite how people were supposed to give a sensible answer, if the choices were between varying degrees of muffled or "artifacts", only on resonance and resonance modulation, on an otherwise identical sounding synth, I'm not sure.

???

Unless someone knew to check resonance/modulation, they wouldn't have heard any difference, except in their own mind. Perhaps not the best way to generate meaningful responses.

Even if they figured that bit it out, the difference was often a case of choosing between differing modulation patterns. I like this one more because it bleeps three times instead of twice. Three times is more analogue.

When it finally it got down to stuff that isn't as subjective, it was a choice between high cut frequencies (2 and 5) or "artifacts", with three sat in the middle. So, which sounds more analogue to you? Muffled resonance, or artifacts in resonance? Err..

If you didn't hear it like that, then well done you. As I also said above with hakey, some of this is about how much controls shift values at those sorts of settings, and (at least some of) the perception created is down to this.

It'd be a good test for people to assign high res controls and see what difference it makes. In this case, especially with those extreme settings, it matters more than usual.

I think it is Urs (and probably other people) who hinted in the initial stages of the alpha test that Repro should be tested at extreme settings because that is what the Pro-One is all about. On my own, having never owned a Pro-One, I would have never figured this out.

But guided by this hint, the differences between filter models became obvious, and 3 was clearly the most musical, and analogue if you like, with the least artifacts. Again, 2 and 5 were just not adding anything to the sound, and actually it must be because their structure is designed to avoid artifacts. 1, 4 and 3 really reacted in their own ways to modulation, and brought new colors to the sound. And 3 sounded the best to me because it produced the less artifacts when modulated at extremes, i.e. it was the best compromise. The difference between 3and the others was not about 3 bloops instead of 2. It was just night and day... And also, part of the idea behind this test may have been to see whether people would care for the amount of artifacts vs sound coloration. the poll clearly indicates that the latter wins, artifact-producing 1, 3 and 4 being now far ahead of muffling 2 and 5.

Finally it is an alpha test, and there might be some tuning to do to actually improve the filters (keytracking, overall response,..) etc... So, a muffled response could be tuned but it probably would sound muffled still, while I am sure that there are ways to reduce again the amount of artifacts we hear.

And let's be honest, these mods are really extreme. Urs stated that the modulations in the Pro_one have an amplitude of 12 octaves, 144 semi-tones if you prefer. The minimoog has an amplitude of 2 octaves, 24 semitones.On my side, given what I heard with Repro, I hope that there will be technology transfer towards all of U-He's synths

[urosh]

Unless someone knew to check resonance/modulation, they wouldn't have heard any difference, except in their own mind.

Let me see if I got this right: everybody who participated in this shootout imagined that there are audible differences between models?

[beely]

As has been stated many times on the thread already, at modest settings, you aren't going to notice a huge amount of differences between implementations, it's only when filter models are pushed to extremes that the differences start to become more clearly audible. This is in the nature of how these things work.

So if people weren't testing out the more extreme settings (useful or not you may consider them to be for sound design), then they simply didn't know what to look for, and so their voting is more based on other factors (many of which are more psychological in nature).

It comes back a bit to what I said before (and was called on a few times) - when it comes to audio, a lot of people simply don't know what they are hearing, and thus making objective decisions is *really* hard. (And yes, I count myself as being one of them, though obviously with more experience than a typical consumer.)

I was also fully expecting to be "wrong" in this test, because I don't think my ears are nearly as well-developed as the DSP folks doing this stuff every day. It turns out my vote of "3" was arrived at guided more by my ability to test well (I'm a developer, and good beta-tester, as well as having a fair amount of experience in sound design), and after that, some intuition, and I'm pleased I went in the right direction - as I said in the beginning, I found this an interesting process in general, but also what I learned about my own processes and abilities in this regard.

And even then, I wasn't going "Oh, it's clearly 3, it's night and day better than all the others". It was pretty difficult to arrive there, and there was still a certain element of doubt.

Still looking forward to Urs' write up, not just of the technicalities, but also his findings and impressions of the test.

I'm also looking forward to having this test repeated for the Prodigy's hard sync bends...

[Urs]

Hi Urs. There must be something in the Pro-One circuit that is causing that anomay, but it's not the filter chip. The chip is the same that is used by the Synthex (and others), and, AFAIK, the Synthex didn't show that anomaly (besides impementing a lot more filter variety than the Pro-One). Care to elaborate what's different?

Every Pro-One sounds different. Every voice of a Synthex sounds different.

http://urs.silvrback.com/specimen

I have 20+ CEM 3320 chips, and when I swap them in the Pro-One the Pro-One sounds different. It's the chip as much as the circuitry around it.

[PAK]

Let me see if I got this right: everybody who participated in this shootout imagined that there are audible differences between models?

No. Feel free to misread this reply too. Those testing for differences, who didn't use resonance in a sound, weren't hearing much of anything between 1 and 5. Even if they kept the res low, it would also limit the differences. That's a large range of "typical" sounds, where testing for differences wouldn't necessarily have shown much, if anything.