The long DIVA thread

[GruvSyco]

According to twitter, he installed the prototype a couple days ago (and installed a new SSD in the wifes iMac). Really people, it's so easy to stalk someone these days... catch up would ya

[petedako]

Oooh! Ooooh! I can't wait! ...and I hope there's a "LUSH-cious" Filter on DIVA so I can crank up to 11

[Urs]

We're working hard on her!

The main issue was making the filters run in realtime. As you know, Diva kind of transformed from "emulating this or that" into a "case study for zero-delay-feedback filters". The idea was, filters would commonly run "normal" but optionally run in zero-delay mode.

Due to the tremendous amount of computation we expected Diva to play 1 single note per cpu core, which is why we thought we'd add multicore support directly within the plugin. The other option was to calculate 2 or 4 voices at once, using SSE (which does have disadvatnages).

Now to the good news: My 4+ year old Clovertown (pre-Nehalem/i7/Westmere/Whatnot) manages up to 6 voices per core, with dual filter mode. That's because we use all 4 SSE registers for a single voice, having multiple problem solving algorithms compete with each other. (We haven't yet implemented multicore support though)

Unfortunately we haven't got a version with perfectly tuned levels and frequencies. Thus if we did a demo combining a Juno oscillator with an MS-20 filter it wouldn't exactly sound like the real deal. It would have too much of this or that or too much of something else lacking.

So basically, we're almost about to set the feature set in stone, and then start tweaking the algorithms to "fit".

Then we'd do a proper gui and sound examples.

Urs

[Bronto Scorpio]

Great
I can't wait to hear the sound examples

Cheers
Dennis

[News From The Sky]

Due to the tremendous amount of computation we expected Diva to play 1 single note per cpu core

What's the advantage of zero-delay-feedback filters? Is the resonance more convincing or something? How noticeable is the difference?

[3ee]

Great
I can't wait to hear the sound examples

Me too!

[Urs]

Due to the tremendous amount of computation we expected Diva to play 1 single note per cpu core

What's the advantage of zero-delay-feedback filters? Is the resonance more convincing or something? How noticeable is the difference?

Well, for most parts it's simply an academic question. With some filter types there is no measurable difference at all if the resonance and/or cutoff is low.

However, as soon as high frequencies, high resonance or fast modulations enter the scene, the picture changes.

If it's worth it, remains to be seen and my vary from user to user. It's only about 2-3 times as much cpu as ACE or TyrellN6 and there'll be an option to switch it off. In latter case Diva will be as cpu hungry as ACE or TyrellN6.

Urs

[3ee]

Maybe in the near-future you might consider having a dedicated hardware unit to run CPU heavy u-he software!

[Breeze]

Maybe in the near-future you might consider having a dedicated hardware unit to run CPU heavy u-he software!

Noooooooooooo! We've come this far liberating ourselves from the shackles of dedicated hardware, I'd hate to see us go backwards.

I'd rather see accessory processing like CUDA or Firestream than dedicated hardware. There's also time: AFAIK Moore's law is still valid. 24 cores anyone?

IMO, the fact that we can generate and modify high quality sound outside the confines of traditional hardware is an huge step forward.

[djanthonyw]

Are there any audio demos for DIVA yet?

[Urs]

Are there any audio demos for DIVA yet?

Nope... too early. We haven't modelled the VCAs yet.

[IrionDaRonin]

And the GUI? Is there any GUI or sketch?

Have a good day and good luck with this synth, i just noticed today about it.
Cheers.

[risome]

Easy Easy... patience children.

[ford442]

[IrionDaRonin]

Have a good day