wow, bouncing down Diva in 192khz

[olikana]

ofc "insane goodness" should also be a live mode not only a rendering mode. u need to be able to hear how it sounds in the mix before rendering. noone should render on... deaf faith.
with my 3 year old phenomII i bet i can run even a couple of "insane goodness" live and ppl with i5s and i7 should be able to run it very comfortably.
a better argument is...does it make any difference?... from the example i made yesterday for most sounds it seems NOT . diva allready takes care of pretty much all digital artefacts with it's current "divine mode" (which i guess includes allready a good amount of oversampling) to not warrant further oversampling..

the only sound in my test changing notably at different resolutions was the noise high hat and would be intereresting to know why.
here it is again
http://dl.dropbox.com/u/50331508/simplenoisehat.h2p


anyway tonight i'll think up of some patches which might unearth some digital artefacts....but i allready know with Diva that is indeed some task.

[cron]

I'm reminded of my very first post at KVR:

Hey all.. kvr newbie here. First post.

Wouldn't it be possible to build a super powerful render mode into a synth (similar to the draft function on zeta/NQ + HQ modes on pentagon) that you can just switch up when it comes to the mixdown? It might be a good idea for the super synth Dave was thinking about so you could preview/compose in real time, but switch to ultra high quality and get those oscs a swingin' in the final mixdown.

http://www.kvraudio.com/forum/viewtopic.php?t=12689

Seemed important 10 years ago, but with the quality of soft-synths today, I'm not so sure.

[ouroboros]

a brief cronology of render modes at KVR....

[cron]

[olikana]

in my previous test i saw that something wasn't right with a "white noise" based sound at different resolutions. so i looked further into it

WHITE NOISE TEST:
here is just white noise with open filter at 44divine - 96divine -192divine
http://dl.dropbox.com/u/50331508/44justnoise.wav
http://dl.dropbox.com/u/50331508/96justnoise.wav
http://dl.dropbox.com/u/50331508/192justnoise.wav
verdict : obviously something ain't right. (aliasing at 44 or just a bug when scaling at higher resolutions?)the difference is perceived as a lowering in volume as resolutions go up(but something tells me it ain't just volume)


EXTREME TESTS ( 44 divine vs 192 divine)
i did some tests to try and unearth as many differences as possible between running diva at 44k and 192k.
these are extreme tests and not musical..so watch your ears.

TEST 1: fm and filter modulation
http://dl.dropbox.com/u/50331508/44test1.wav
http://dl.dropbox.com/u/50331508/192test1.wav
verdict :hardly any difference (apart from osc being at different phases but that doesn't count)

TEST 2 : same as test 1 but with added shape modulation
http://dl.dropbox.com/u/50331508/44test2.wav
http://dl.dropbox.com/u/50331508/192test2.wav
verdict : aliasing (ghost artifacted tone-and even out of tune) at 44k !

TEST 3 : same as test 1 but with added white noise
http://dl.dropbox.com/u/50331508/44test3.wav
http://dl.dropbox.com/u/50331508/192test3.wav
verdict : as expected this was gonna sound different considering the results of the noise test.

all test patches are here :
http://dl.dropbox.com/u/50331508/DivaTestsOlikana.zip

conclusions:
-white noise is obviously flawed and needs looking into
-shape mod only aliases with a few shapes.
-FM and filter modulation -and pitch modulation too(not in these test files but i did test it!)- perform already extremely well at 44 divine.

"testing is the future" ;P

[Urs]

I would have to think about it, but with digital white noise, 50% of the energy is in the upper most octave. If you move the upper octave out of the hearing spectrum by doubling the sample rate, white noise will sound about half as loud despite having the same amplitude, because 50% of the energy is outside of the hearing spectrum.

I haven't given that much thought before, but we might have to render noise at 44/48kHz and upsample it (with no energy outside the hearing spectrum) in order to preserve the relation between amplitude and perceived volume.

Regarding shape modulation I'm not quite sure what I'm hearing. Maybe we can discuss this next week when Clemens is back from vacation.

[xh3rv]

Some posts from William Orbit on FaceBook discussing problems with the MDNA sessions ...

William Orbit: But alas, the time wasn't there. Great swathes of it taken up by the engineer and his assistant bouncing reverb tracks for hour after hour, night after night. Not to real tape or anything, where you could posit that there would be an advantageous sonic dividend (real tape can be magic) but all in the digital domain. A purely procedural thing. Although not a procedure I or any of my own colleagues in this game would want to squander time with.

William Orbit: I was just describing to a friend of mine, and whom I frequently work with Serban Ghenea (just look him up if you want to be awed) how it was done, and he thought I was kidding him!

This is just anecdote ... I thought it sort of pinged on a point about purely offline rendering inherently generating a sense of error to the user. Suddenly one can't trust real-time previewing, which is a source of error, and can't trust finely tuned decisions rendered offline since that can't be operated on in real-time. One could say the solution is bouncing a lot of options, but is that really a sane workflow?

Dunno, I could be making some bad assumptions as well

Also posted just to gossip about Madonna and William Orbit

[cosmicdawn]

I read this yesterday and I was wondering what he was talking about. It doesn't look like he wanted to spend time bouncing down reverbs, and that this was something he was forced to do? But what is the point, who would ask him to do that? I know record companies want stems of their tracks so they easily can hire remixers, make edits, archiving etc. But do you think they also want to have all reverb tracks on their own? I'm still wondering what he meant...

[Vectorman]

Interesting to hear William venting about some factors that contributed to MDNA being so disappointing. As far as his comment about M being distracted with fragrance lines and such, it does sound like the new album was just one more thing on a harried to-do list that she dashed off as fast as possible. How else to explain some of those vapid lyrics? I guess the magic she and William made on Ray of Light was a one-time thing that won't come again.

[xh3rv]

this was something he was forced to do? But what is the point, who would ask him to do that? I know record companies want stems of their tracks so they easily can hire remixers, make edits, archiving etc. But do you think they also want to have all reverb tracks on their own? I'm still wondering what he meant...

There was another bit where he complained about just being 'co-' producer, writer, etc., I read it as saying he couldn't assert himself. Which, isn't that really sort of why one would go to William Orbit in the first place?

[xh3rv]

I guess the magic she and William made on Ray of Light was a one-time thing that won't come again.

Not the same, but just to plug something that kills me every time I listen to it, William Orbit produced Katie Melua's The House.

example:

[Unaspected]

very keen to hear developer's detailed feedback, this thread has exploded

No, it just turned into nonsense as usually
192khz gimme a break Day and night differences, come on probably equipment choking, that's your difference, and all that 192khz for what?

More HD space taken up...

[Vectorman]

I guess the magic she and William made on Ray of Light was a one-time thing that won't come again.

Not the same, but just to plug something that kills me every time I listen to it, William Orbit produced Katie Melua's The House.

example:

That was lovely. Very nice!

[David Else]

that's all very good, but we have gone totally off topic!!

is this amazing mode being considered for the next release?

i have another question. in my experience downsampling is a major loss of quality. i assume all the audio in Diva is upsampled and then downsampled?

there must be a lot of compromise in the quality of the downsampling to make Diva run in real time. how do you think it would sound to use the recognized best quality free and open source SoX downsampling in the render modes (at max quality of course):

http://src.infinitewave.ca/
http://sox.sourceforge.net/

there is a good foobar plugin that uses this (for quick easy testing):

http://www.hydrogenaudio.org/forums/ind ... opic=67373

you might make it an option for those who really want the live mode to be identical to the render.

[xh3rv]

there must be a lot of compromise in the quality of the downsampling to make Diva run in real time.

The SoX utility does sample rate conversions between arbitrary sample rates, so interpolation routines are more costly and less streamlined than is typically done with plugs that up/down sample to fixed rates (like 48k * 2^n) internally. [e: Oh, on re-reading I think I might have mixed up what you're getting at? Downsampling post-Diva rather than internally. Seems like a DAW-centric thing though, rather than anything particular to Diva.]

Sorry for going off-topic, to restate a point without Madonna - I think there's a loss of pragmatic coherency with any render mode that's not plausible in real-time.

Additionally from the discussion, Urs's comments and olikana's samples I'm concluding that rendering at increased sample rates may actually get off-target (relative to tangible targets - sourced analog hardware). So, I guess I'm sort of unsure where 'insane goodness' would actually improve anything, what would justify a trade-off from what-you-hear-is-what-you-get.