KVR Audio

bmanic · Post by **bmanic** » Mon Jan 31, 2005 2:15 pm

Sounds good wide boy! Yeah, I missed that thread back in november.. darn. Anyways, Christian took the first step towards something very nice indeed. The ultimate goal would be to be able to sample outboard gear dynamically at a user selectable amount of samples (length and dynamics). This would put sintefex in trouble tough but it's only a matter of time before this technology goes native plugin so why not now?

Just for everybodys information, those who have not heard a sintefex FXx000 unit can not understand how damn good this technology sounds. Take your favourite plugin compressor and add times 10000000000000000.. that's how much better IMHO it sounds. It really is the "missing" link to get truly professional sound (think big time USA/UK gloss). Same with the EQs and I think Tritonedigital might hit the big time with their upcoming plugin (wish I could get the beta demo to work!

).

Cheers!
bManic

bmanic · Post by **bmanic** » Mon Jan 31, 2005 2:27 pm

More on convolution patents.

http://psplab.csie.nctu.edu.tw/invboard ... /t979.html

olafmol · Post by **olafmol** » Mon Jan 31, 2005 4:58 pm

bmanic wrote:Sounds good wide boy! Yeah, I missed that thread back in november.. darn. Anyways, Christian took the first step towards something very nice indeed. The ultimate goal would be to be able to sample outboard gear dynamically at a user selectable amount of samples (length and dynamics). This would put sintefex in trouble tough but it's only a matter of time before this technology goes native plugin so why not now?

Just for everybodys information, those who have not heard a sintefex FXx000 unit can not understand how damn good this technology sounds. Take your favourite plugin compressor and add times 10000000000000000.. that's how much better IMHO it sounds. It really is the "missing" link to get truly professional sound (think big time USA/UK gloss). Same with the EQs and I think Tritonedigital might hit the big time with their upcoming plugin (wish I could get the beta demo to work! ).

Cheers!
bManic

afaik current CPU's still cannot mach the needed processing power for dynamic convolution (f.e. the sintefex uses 8 SHARC dsp's for each mono channel)

Olaf

jens · Post by **jens** » Mon Jan 31, 2005 5:05 pm

olafmol wrote:

afaik current CPU's still cannot mach the needed processing power for dynamic convolution (f.e. the sintefex uses 8 SHARC dsp's for each mono channel)

Olaf

this thread is about a plugin that already exists and nontheless you're doubting that it is possible?

Kishon would have had fun with you...

bmanic · Post by **bmanic** » Mon Jan 31, 2005 5:40 pm

Yeah, FYI Christians plugin already works and uses MINIMAL cpu, all in real time. However, the sintefex "samples" are much more complex though. They all use 1024 samples length at 44.1khz and I have no idea how many dynamic layers but probably quite many.

Also, the sharc chips in the sintefex units are quite old by todays standard. The sintefex unit itself was designed on an old Intel P3 computer and even back then they could run some simple algos in realtime (according to the sintefex technical paper and the history of the FX units).

Cheers!
bManic

olafmol · Post by **olafmol** » Tue Feb 01, 2005 9:56 am

jens wrote:
olafmol wrote:

afaik current CPU's still cannot mach the needed processing power for dynamic convolution (f.e. the sintefex uses 8 SHARC dsp's for each mono channel)

Olaf
this thread is about a plugin that already exists and nontheless you're doubting that it is possible?

Kishon would have had fun with you...

never believe anything you read on the internet

Christian Budde · Post by **Christian Budde** » Tue Feb 01, 2005 1:24 pm

Yeah, FYI Christians plugin already works and uses MINIMAL cpu, all in real time. However, the sintefex "samples" are much more complex though. They all use 1024 samples length at 44.1khz and I have no idea how many dynamic layers but probably quite many.

It only works with minimal CPU if you have small Impulse responses. For 1024 Samples it's damn slow!!!
The number of layers doesn't matter at all. It's just a question of RAM.

Maybe with SSE and 3DNow code, the plugin might work fast enough.
But someone can try avoiding the patents AND the speed bottleneck, by separating the dynamic impulse responses in a prefilter convolution, a dynamic part (which in this case is somehow limited in IR size) and postfilter convolution.

The pre- and postfilter are standard, linear convolutions. And the dynamic part is (let's say it in the words of unidyN) "capricious convolution".

Unfortunatly the sampling process is part of the patent as well, but i'm sure, that there are other ways to do that...

Regards,

Christian

Blue Days · Post by **Blue Days** » Tue Feb 01, 2005 1:41 pm

What exactly is it good for?

You need a good dynamics processor to begin with, to "sample",
and do you really want to put your music through that FFT process?

Why not just try to design a good digital dynamics processor, without the hassle of analog noise, distortion etc..

Btw, this is what I am working on, and you can listen to the dynamics processor, rather specifically, adaptive limiter, in this post

http://www.kvraudio.com/forum/viewtopic.php?p=948042

The latter track features a quite overloaded limiter indeed..

Would be interesting to hear your opinions on this.

jens · Post by **jens** » Tue Feb 01, 2005 2:22 pm

oh no - not you again

(he's the guy who once tried to charge a few hundred $ for a (bad) synth-edit made limiter...(without any custom modules)

)

Blue Days · Post by **Blue Days** » Tue Feb 01, 2005 2:26 pm

Ahh it Djens again.

You shouldn't believe anything people (or other entities) tell you Djens!

whyterabbyt · Post by **whyterabbyt** » Tue Feb 01, 2005 2:30 pm

Blue Days quoth What exactly is it good for?

Your own failure to comprehend the utility of a particular tool or process does not mean that tool or process has no value.

You need a good dynamics processor to begin with, to "sample",

Only if you want to sample good dynamics processors. Or use samples of good dynamics processors belonging to other people.

and do you really want to put your music through that FFT process?

Why would there be a specific issue with that compared to any other DSP processing?

Sicklecell666 · Post by **Sicklecell666** » Tue Feb 01, 2005 2:36 pm

jens wrote:oh no - not you again

(he's the guy who once tried to charge a few hundred $ for a (bad) synth-edit made limiter...(without any custom modules) )

Thanks, Jens..way to start the day off!

Blue Days · Post by **Blue Days** » Tue Feb 01, 2005 2:40 pm

whyterabbyt wrote:Blue Days quoth What exactly is it good for?

Your own failure to comprehend the utility of a particular tool or process does not mean that tool or process has no value.
...

and do you really want to put your music through that FFT process?

Why would there be a specific issue with that compared to any other DSP processing?

It's true I do not comprehend why people would want to put their full mixes through a FFT process. If you can't hear the fatiguing effects of low FFT lenghts, or sludging effects of high, I guess you are a happy camper with FFT.
Good for you

I, on the other hand, would like my signal to be as high fidelity as possible, and I don't see how it can be called high fidelity after running it through FFT.
Ofcourse for reverbs, you don't hear these sideeffects that much, but for a full mix, the situation is dramatically different.

whyterabbyt wrote: You need a good dynamics processor to begin with, to "sample",

Only if you want to sample good dynamics processors. Or use samples of good dynamics processors belonging to other people.

Well ofcourse you are fully entitled to sample BAD dynamics processes ofcourse.

Hey.. your choice dude.

whyterabbyt · Post by **whyterabbyt** » Tue Feb 01, 2005 2:50 pm

Blue Days quoth
It's true I do not comprehend why people would want to put their full mixes through a FFT process. If you can't hear the fatiguing effects of low FFT lenghts, or sludging effects of high, I guess you are a happy camper with FFT.

And you're saying this is the case no matter the window size, or windowing method?

Good for you
I, on the other hand, would like my signal to be as high fidelity as possible, and I don't see how it can be called high fidelity after running it through FFT.

I, on the other hand, would like to have as wide a range of signal processing tools as possible at my disposal. Dynamic convolution is yet another signal processing tool, and not one that is explicitly linked to 'sampling dynamics processors'. As I say, your failure to see value in the process is a limitation of your own imagination.

Ofcourse for reverbs, you don't hear these sideeffects that much, but for a full mix, the situation is dramatically different.

Really...

Of course it must be true because someone claims it is.

Well ofcourse you are fully entitled to sample BAD dynamics processes ofcourse. Hey.. your choice dude.

How is your blinkered obsession with dynamics processors relevant?

jens · Post by **jens** » Tue Feb 01, 2005 2:53 pm

it's pronounced 'yenns', not 'Djens'

Dynamic convolution plug