Login

nebula sample rate conversion algorithm

Nebula feature suggestions

nebula sample rate conversion algorithm

Postby futur2 » Wed Jan 26, 2011 3:23 am

ok, in this thread http://www.acustica-audio.com/forum/ind ... c&start=20 niklas from analoginthebox claims that "the Nebula SRC algorithm ... is pretty unacceptable".

alexb on the the other hand claims that 96khz libraries sound better in 44.1 converted through the nebula SRC than if they were sampled in 44.1. niklas does not think so ;)

however if it is true that the nebula SRC algo is subpar - and i have no reason to doubt niklas words - it would be necessary to use a better SRC algo for nebula, right? because at least two of the big library developers (alexb and cdsoundmaster) make their libraries in 96khz.

please discuss!
futur2
Member
Member
 
Posts: 303
Joined: Sun Mar 28, 2010 8:16 pm

Re: nebula sample rate conversion algorithm

Postby analoginthebox.com » Wed Jan 26, 2011 4:03 am

Let's take a simple example.
Generate a test-tone and devonvolve it without any processing. We'll have only a look at the frequency + phase response.

Results are as expected when when loading a 44.1KHz program in a 44.1KHz Session or 96KHz -> 96KHz:

Image
Image

Image
Image

But:
:!: 96KHz downsampled to 44.1KHz :!:
Image
Image


Maybe I'll post some more examples later this week (including hardware, not just a dry-test). There are a lot of side-effects (not only caused by sample rate conversion)!
User avatar
analoginthebox.com
User Level XI
User Level XI
 
Posts: 155
Joined: Mon Apr 12, 2010 2:34 am

Re: nebula sample rate conversion algorithm

Postby futur2 » Wed Jan 26, 2011 7:01 pm

analoginthebox.com wrote:Maybe I'll post some more examples later this week (including hardware, not just a dry-test). There are a lot of side-effects (not only caused by sample rate conversion)!


thanks! that looks really bad.

i'm no expert in SCR algos but the listening tests i made with voxengo and sox sounded pretty good to me. so just to get it straight (you speaking of side-effects), would a better algorithm significantly improve nebula conversion from 96khz to 44.1/48 or no?
futur2
Member
Member
 
Posts: 303
Joined: Sun Mar 28, 2010 8:16 pm

Re: nebula sample rate conversion algorithm

Postby david1103 » Wed Jan 26, 2011 7:39 pm

this, and the fact that NAT does not reliably downsample libraries to 44.1 (at any quality) is a *massive* issue. The NAT bug is a bug that has been there for years as far as i can tell... YEARS!


everyone using 44.1 has to sit there waiting for the 96k samples to load every time when they could just resample in NAT if it worked.


would acustica audio consider please making a 'known issues' sticky section on the forum. its not fair on users to have to read the forum everyday to discover what is broken in their software and is widely known about.


when people know bugs they can work around them, or at least not spend hours working out whats going on themselves. there must be a list to copy and paste from somewhere :)
User avatar
david1103
Beta Tester
Beta Tester
 
Posts: 516
Joined: Wed Mar 31, 2010 12:26 am

Re: nebula sample rate conversion algorithm

Postby enriquesilveti » Wed Jan 26, 2011 8:56 pm

You are right. I running in a low priority level and NAT SRC "seek and destroy" bug hunting. I already done all CPCL SRC from 44 to 48, but I could find time for debug it.

I'll post a bug hunting into beta/3rd section and release a open for customer a bug hunting threat with the instructions how to do it and report it.

Sorry about this, we are very slow in everything...
Enrique Silveti.
Acustica Audio customer and technical support.

MBP 11.5 (i7-4870 | 16 GB | 512 SDD)
SP4 (i5-6300 | 8 GB | 256 SDD)
RME UFX | PS Lyra2 | SD USBPre2
VM U15 | VM W10 | VM OSX 10.12
N4/NAT4 | SPX3 | RX5 | LN2C | Smaart8 | R5 | PT12 | PX10
User avatar
enriquesilveti
Expert
Expert
 
Posts: 2662
Joined: Sun Mar 28, 2010 9:00 pm
Location: Lodi | Madrid | Buenos Aires

Re: nebula sample rate conversion algorithm

Postby giancarlo » Mon Jan 31, 2011 6:16 pm

Downsampling could be worse for particular samples than others. We are interested in average behaviour and loading time. The algorithm is very good and it will NOT be improved for all 2011 and 2012, unless you want other developments delayed forever. And I would not be so sure about results using other algorithms. That's all, take your conclusions.

My suggestion: use native libraries if they are available, otherwise accept it "as is".
User avatar
giancarlo
Founder
Founder
 
Posts: 9167
Joined: Mon Sep 21, 2009 10:40 pm
Location: Italy

Re: nebula sample rate conversion algorithm

Postby giancarlo » Mon Jan 31, 2011 6:17 pm

enriquesilveti wrote:You are right. I running in a low priority level and NAT SRC "seek and destroy" bug hunting. I already done all CPCL SRC from 44 to 48, but I could find time for debug it.

I'll post a bug hunting into beta/3rd section and release a open for customer a bug hunting threat with the instructions how to do it and report it.

Sorry about this, we are very slow in everything...



don't waste time. I will not debug that piece of code again. They are not bugs. It's the algorithm.
User avatar
giancarlo
Founder
Founder
 
Posts: 9167
Joined: Mon Sep 21, 2009 10:40 pm
Location: Italy

Re: nebula sample rate conversion algorithm

Postby mathias » Mon Jan 31, 2011 7:32 pm

thank you giancarlo!
that's a clear statement and i can understand your decision. so i will adapt my workflow to circumvent src, where i can not live with the soundchanges, that it causes.
we have enough options i think.
so let us not waste time by picking on the src-algorithm!

mathias
system 1: windows 8 32 bit - samplitude prox, tracktion6, reaper
system 2: mac osx yosemite - reaper(32+64bit), tracktion6(32+64bit)

both systems on: macbook pro (late 2009), core 2 duo 3,06 ghz, 4 gb ram, graphic: nvidia geforce 9600M GT 512 MB
mathias
Expert
Expert
 
Posts: 2113
Joined: Wed Mar 31, 2010 12:25 am

Re: nebula sample rate conversion algorithm

Postby scooter » Sat Feb 05, 2011 9:12 pm

Thought I'd double post this question over here, since it seems more appropriate.

What about using a 44.1khz library at 48khz?
That's what I find myself using the AITB stuff at.
Am I at a disadvantage by using AITB libraries this way?

Thanks!
-Scott
User avatar
scooter
Member
Member
 
Posts: 214
Joined: Fri Apr 02, 2010 8:24 pm
Location: Salt Lake City, Utah

Re: nebula sample rate conversion algorithm

Postby mathias » Sun Feb 06, 2011 2:47 am

you get a steep cutfilter at 22 khz, which is where 44.1 khz samplerate can not represent frequencies above.
so if you are not worried about the range between 22 and 24 khz, there should be no problem. maybe you get some artefacts of this filter when you have enough energy in this frequencyrange (from cymbals for example). you can use your ears, to check for it.

also the phase gets weird between 22 and 24 khz.
but in all my use of aitb-programs in 48 khz projects, i did not get unpleasant results, when it suited the material.

when downsampling 96 khz to either 48 or 44,1 khz,
vst-analyzer shows a little ripple in the frequencyresponse to the end of the spectrum, sometimes in the bassrange. soundwise this is not that much of a problem for me.
there is a more severe change in the phase-response.
where the phase is flat before, it rises or falls between 10 and 20 khz by a reasonable amount. this does not mean it sounds bad then, but it changes the sound of the program slightly.
so if you want absolut accuracy, you should work at the samplerate, the programs are sampled.

in other cases use your ears and maybe vst-analyzer, to get some visual-reference on what happens to the programs in up- or downsampling.

have fun :D :lol:
mathias
system 1: windows 8 32 bit - samplitude prox, tracktion6, reaper
system 2: mac osx yosemite - reaper(32+64bit), tracktion6(32+64bit)

both systems on: macbook pro (late 2009), core 2 duo 3,06 ghz, 4 gb ram, graphic: nvidia geforce 9600M GT 512 MB
mathias
Expert
Expert
 
Posts: 2113
Joined: Wed Mar 31, 2010 12:25 am

Next

Return to Requests

Who is online

Users browsing this forum: No registered users and 1 guest