KVR Audio

fmr · Post by **fmr** » Tue Aug 01, 2017 2:28 pm

blacktomcat666 wrote:Meanwhile it sounds even better - but I'm thinking about cancelling the project at all.

WHY?

Please don't

Future1313 · Post by **Future1313** » Tue Aug 01, 2017 8:27 pm

18 Soundbanks (*.FXB) for the TerraTec - Komplexer VSTi

Converted Soundbanks from the Waldorf - MicroQ Synthesizer

INFO & DOWNLOAD

http://patcharena.com/downloads/comment ... 3e213ea2bb

fmr · Post by **fmr** » Wed Aug 02, 2017 10:04 am

Future1313 wrote:18 Soundbanks (*.FXB) for the TerraTec - Komplexer VSTi

Converted Soundbanks from the Waldorf - MicroQ Synthesizer

INFO & DOWNLOAD

http://patcharena.com/downloads/comment ... 3e213ea2bb

Thanks for these

blacktomcat666 · Post by **blacktomcat666** » Wed Aug 02, 2017 1:04 pm

fmr wrote:
blacktomcat666 wrote:Meanwhile it sounds even better - but I'm thinking about cancelling the project at all.
WHY? Please don't

It's too expensive for the CPU. At least for those of my computers.

fmr · Post by **fmr** » Thu Aug 03, 2017 12:34 pm

blacktomcat666 wrote:
fmr wrote:
blacktomcat666 wrote:Meanwhile it sounds even better - but I'm thinking about cancelling the project at all.
WHY? Please don't
It's too expensive for the CPU. At least for those of my computers.

Do you mind if I test it? My COU is not one of the latest, but it runs more or less comfortably ALL the current plug-ins, including those that tax more the CPU. I could give it a try and let you know how it behaved here.

But if you give up of that more ambitious project, please keep updating the current version, then...

kokotte · Post by **kokotte** » Thu Aug 03, 2017 6:45 pm

Blacktomcat666,
what is you computer cpu?
perhaps people who use your audioterm could donate and support your project ... i don't use a lot audio term but it have some superb options not avalaible anywhere for wt creation( filtering, formant over the time...) i' m ready to donate for a project like this. sometimes, the non commercial project are the best( i refer to,eg, tranzistow and audioterm).
best
& keep it up !

fmr · Post by **fmr** » Fri Aug 04, 2017 10:53 am

kokotte wrote:Blacktomcat666,
what is you computer cpu?
perhaps people who use your audioterm could donate and support your project ... i don't use a lot audio term but it have some superb options not avalaible anywhere for wt creation( filtering, formant over the time...) i' m ready to donate for a project like this. sometimes, the non commercial project are the best( i refer to,eg, tranzistow and audioterm).
best
& keep it up !

I would donate too, if you are comitting to come up with a project devoted to the creation of wavetables (even if not that Synclavier idea with had for AudioTerm 2). Of course, I would prefer a proper Windows GUI, but I'm OK with your interface now.

blacktomcat666 · Post by **blacktomcat666** » Sat Aug 05, 2017 4:11 pm

In fact the problem is memory access, not the CPU itself. All is fine as long the modulation is not too fast / in a narrow range - but beyond a certain range it rapidly jumps from 40% total load to 97% (on my Core2 E5500 and similar on my Athlon II x2 250 @ 3.6 GHz) because of the cache misses.

ThomasHelzle · Post by **ThomasHelzle** » Sat Aug 05, 2017 4:18 pm

blacktomcat666 wrote:In fact the problem is memory access, not the CPU itself. All is fine as long the modulation is not too fast / in a narrow range - but beyond a certain range it rapidly jumps from 40% total load to 97% (on my Core2 E5500 and similar on my Athlon II x2 250 @ 3.6 GHz) because of the cache misses.

Well, that sounds like rather ancient hardware - did you try it on a recent i7 with fast memory?
Should make a major difference!

Cheers,

Tom

fmr · Post by **fmr** » Sat Aug 05, 2017 5:01 pm

ThomasHelzle wrote:
blacktomcat666 wrote:In fact the problem is memory access, not the CPU itself. All is fine as long the modulation is not too fast / in a narrow range - but beyond a certain range it rapidly jumps from 40% total load to 97% (on my Core2 E5500 and similar on my Athlon II x2 250 @ 3.6 GHz) because of the cache misses.
Well, that sounds like rather ancient hardware - did you try it on a recent i7 with fast memory?
Should make a major difference!

Cheers,

Tom

I concur. You are really taking really ancient CPUs as a reference. A not so modern CPU like mine (i7 3770 @ 3.4 GHz) is capable of much more. And the current "state of the art" CPUs blow this one away.

Just to have a term of comparison, you can for example download DIVA and try to run it in divine mode - something the modern CPUs can achieve with more than oe instance) and see if your system is up to the task.

blacktomcat666 · Post by **blacktomcat666** » Sat Aug 05, 2017 6:10 pm

I cannot afford buying a newer computer. Even an i7 cpu alone costs more than I have for living per month. If I had some money left I wouldn't bother with coding.

Donation is also not possible because of my personal situation (health ... but I don't want to go into details here).

The frustrating thing is: Beside this problem everything else is nearly solved. I've added some (tweakable) noise to the FM without introducing inharmonic aliasing and managed to remove many unwanted artifacts (like the attack clicks when shifting the modulator phase):

https://dl.dropbox.com/s/pvd4aviqpodwmr ... zation.mp3

https://dl.dropbox.com/s/lkkpkoqfb9gqzyr/crescendo.mp3

This was a pitch lfo test, which also shows that there are no audible inharmonic artifacts.

https://dl.dropbox.com/s/f34tjm5cjztljw ... foTest.mp3

ThomasHelzle · Post by **ThomasHelzle** » Sat Aug 05, 2017 8:51 pm

Sad to hear that tomcat.
I hope you can continue on the project even so, it sounds intriguing.

All the best wishes!

Tom

blacktomcat666 · Post by **blacktomcat666** » Sat Aug 05, 2017 9:43 pm

Thamks. I don't want to start a lament on personal things, but I want to make clear that the circumstances of this project are a bit difficult. I know that trying such things on a that old computer may fail, but I have no choice at the moment.

One solution (but not the most elegant) would be to make a polyphony management similar to the old hardware. That means using chorus voices and wavescanning also reduce polyphony (at the moment it doesn't - one would need a synclavier with 256 FM voices). This would also mean loosing the "phasing free chorus" - you can detune the chorus voices and set them back without cancellations, clicks or volume changes.

But I cannot reduce the lookup table size and realtime computing of 256 x 64 sines / cosines @ 8x oversampling is much more expensive. The algorithm needs direct (integer) phase access per sample to apply the phase modulation, to reproduce wanted artifacts and to suppress unwanted aliasing. This is why I cannot use "best practice" sine generators or iFFT.

fmr · Post by **fmr** » Sat Aug 05, 2017 9:54 pm

You obviously have something working already. Why don't you let me and Thomas do the tests and mail you the results in our CPUs?

Since you managed to have these examples working, it's just a matter of being sure that it's workabel in a more modern computer, and we could do those tests for you.

blacktomcat666 · Post by **blacktomcat666** » Sat Aug 05, 2017 10:18 pm

It's only the plain test engine - no gui and still a lot of open end in code (but nothing that heavy like the osc core). My goal was something that (also) works on my machines - I do this primarely for myself, but I will share the results. But if it doesn't, it's pretty useless for me.

I also hate to give up that fast ... but the complexity of the code is high (using a lot of hq approximations, SIMD math, clock counters and interpolators and everything quite fine tuned). Larger changes will require a rewrite of most things and take some time.

And now the "health factor" kicks in: on a good day I can work 12 hours on this, on a bad day... well, a bad day can also last a month or so...

It's unpredictable how long even some minor changes will take - I don't want to put another 3 years in solving new problems only to find another bottleneck hidden on the home straight.

A small "Waveterm" for Terratec Komplexer (and other products)