Repro-1 (out now)

[Urs]

Update Dec 8th 2016: Repro-1 has been released!

https://www.youtube.com/watch?v=vhQST401uYM

Update Nov 2nd 2016: A public beta version is available:

---> viewtopic.php?p=6603621 <---





Update May 2nd 2016: Our solution & conclusion available in a geeky little PDF:

http://www.u-he.com/downloads/UrsBlog/R ... veiled.pdf

Hi all,

As predicted during Superbooth and elsewhere, today we're unleashing a new Researchware plug-in, RePro-Alpha.



It is an excessively CPU hungry anti-optimized monophonic synthesizer, in essence a stripped down version of the Sequential Circuits Pro-One we're currently modeling.

The research part is, we have implemented the same pretty extreme model of the CEM3320 Curtis filter chip in our vintage Pro-One using 5 different numerical methods, each costing a very different amount of CPU. We wish for you to spot the "most analogue" sounding method. By that we wish to see if it is worth spending a lot of CPU or if we could get away with something cheaper. Furthermore we wish to discuss the following questions:

• what differences do you spot between the models?
• when do these differences become audible, i.e. which settings promote these differences?

Please download the plug-in, check it out, it's free forever (but won't be updated other than to the final commercial version), try some stuff, build an opinion, take your time, discuss and vote here in this thread.

Repro_001_4365_Mac.zip (AU/VST/VST3)

Repro_001_4365_Win.zip (VST2/VST3)

Also, we're happy to discuss further questions about RePro-1 in our company forum, we would like to concentrate on just the filter differences in this thread. I.e. if your Pro-One (or Synthex whatever uses CEM3320) sounds "utterly different", we'll be happy to request audio examples from you, but this isn't the focus of this thread.

Note: We are sorry if some of you can not run this plug-in due to CPU consumption. This isn't an indicator to the final version being CPU hungry as well, it is just a necessity for the trial - several filter algorithms are always run in parallel so that one can not spot the most accurate one by CPU hunger.

Enjoy,

- Urs

[bpblog]

Woohoooo!



Here's a quick demo:

https://www.youtube.com/watch?v=WfoCs_zaX90

[Elektronisch]

Its so hard to decide between 1 and 3.

[core]

There goes the afternoon

[Boardwalk]

Uhmm.. well.. minimum 16GB?

[tungee78]

I love number 1 the most. Then 3 is nice also. But 1 is more wild.

But CPU usage is extreme

[Totolitoto]

So... no expiration date ???

[Urs]

So... no expiration date ???

Nope.

[3ee]

people, don't rush with your votes, get to know the filters! try different modulations at different resonances.

for me... 4 is great but 5 takes my vote! If you ask me, I'd also like to have 4 as an alternative "asymmetric" character.

I particularly like 5 because the resonance gets softer with high freqs/cutoff values... which makes it ideal for heavy modulation.

4 is a bit more thick/chunky "fat" with simple/basic settings which is also great to have...

IMhO, 1 is the worst mode of them all without a doubt! ... followed by 2. which leaves 3,4 and 5 as the better ones.

[do_androids_dream]

1 and 2 are the most 'convincing'. They both have a chunky, weighty quality that 3,4,5 have less of. 1 and 2 are quite different from each other at different resonance settings and respond differently as the resonant peak kicks in as the filter is closed. I like them both equally - they are the most pleasing to me and have more analog 'appeal' than 3,4 or 5 imo.

[penguinfromdeep]

So far I find 4 the best, I think the first two ones (especially the first one) sound really harsh in the high frequencies and are more static sounding also

[EvilDragon]

As mentioned in the u-he beta area, for me 3 and 5 take the cake. 1 has too much noise and too nasty artifacts I don't think are happening on the real analog (my guess is that this is the non-ZDF filter).

[beely]

This is quite interesting... !

(Not just the point of the research, but what it tells you about your own biases and decision making processes - it's create easy to convince yourself that number X is the cheapest sounding filter, and then reinforcing that bias because that's what you expect to hear, until you keep investigating and try and remove those biases and make a better judgement.)

It's worth spending some time and effort before casting your votes, i think. Sometimes quick judgements based on instinct and gut feel on how you think it sounds is valuable, other times some investigative effort make a judgement more considered.

Like I say, interesting...

[Urs]

(my guess is that this is the non-ZDF filter).

There might be more than one of those... in a stricter sense, all but one are non-ZDF to some degree.

On a side note, this filter is 8x oversampled at 44.1/48kHz which is often quoted as the borderline from which ZDF and non-ZDF doesn't make an audible difference anymore. The fact that these filters sound different shows that - under certain circumstances - this is not so.

[Quadron]

With these settings there is an extreme difference between the filter models