Login

More Sample Rate Conversion

Tips & tricks, working results, technical support

More Sample Rate Conversion

Postby justinmorell » Tue Mar 29, 2011 4:56 am

I know that SRC has been addressed quite a lot on this forum, but I'm still unclear on a few things. I am a new Nebula 3 Pro user, and trying to get a handle on this:

I have some mixes already at 48k, which I will ultimately need to convert to 44.1k. I have Sample Manager, which I know does a very good job with SRC. I'm trying to figure out the best option for mastering with 96k Nebula programs out of the few choices I can imagine:

1) keep the tracks at 48k for mastering, let Nebula do the SRC for the processing (48 to 96 back to 48), and then convert the final masters to 44.1k, or

2) convert the tracks to 44.1k in Sample Manager BEFORE Nebula processing, let Nebula do the SRC from 44.1 to 96k and back (this involves the same number of SRC instances as above, but avoids a POST-mastering SRC back to 44.1k), or

3) convert the tracks from 48k to 96k so that Nebula avoids a SRC at all, and then do the conversion from 96 to 44.1 after the mastering.

Any ideas would be much appreciated, especially in so far as the SRC conversion inside the Nebula engine compares to something like Sample Manager (iZotope).

Thanks!
--Justin

Mac Mini 2.4 GHz Intel Core 2 Duo
OS X v. 10.6.4
4 GB 1067 MHz DDR3

Running Logic Pro 9
Nebula 3 Pro
justinmorell
User Level 0
User Level 0
 
Posts: 8
Joined: Fri Mar 25, 2011 5:14 am

Re: More Sample Rate Conversion

Postby mathias » Tue Mar 29, 2011 10:20 pm

if your computerresources allow it, i would go for option 3).
this way you will have the best possible sound with nebula.
be sure to test the different downsampling options of the izotope src in sample manager.
a lot will be lost in this stage, but you can try to preserve as much as you can, by hearing the different src results.

the nebula src is leaning towards effeciency, because it has to convert all the kernels in a reasonable time.
it does a good job, but in your case you may get better results with upsampling. you can make a test by yourself with a snippet of your files and compare the results. i would be interested to hear it, when you have time to post examples.

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: 2116
Joined: Wed Mar 31, 2010 12:25 am
Location: South-West Germany

Re: More Sample Rate Conversion

Postby Tim Petherick » Tue Mar 29, 2011 11:45 pm

justinmorell wrote:Any ideas would be much appreciated, especially in so far as the SRC conversion inside the Nebula engine compares to something like Sample Manager (iZotope).



Even if the src In nebula was really amazing like izotope's, you'd still get not so good quality if you had -

more than one instance per track up sampling and or downsampling on each instance.(build up of artifacts)

Than-

Prefered sample rate ,by using high quality src like izotope and then using only neb progs that match that sample rate.Then you could use as many instances as you like with no constant up sampling or down sampling.

But I guess when you do need to use certain programs at a different sample rate . A better src in nebula would be great .I think nebulas is pretty good . I don't know if its near izotope's

I've been using option 3 in your post for a long time now and its much much much better quality.

Tim
User avatar
Tim Petherick
Expert
Expert
 
Posts: 1352
Joined: Sat Apr 17, 2010 4:07 pm
Location: Bath , Uk

Re: More Sample Rate Conversion

Postby mathias » Wed Mar 30, 2011 12:10 am

Even if the src In nebula was really amazing like izotope's, you'd still get not so good quality if you had -

more than one instance per track up sampling and or downsampling on each instance.(build up of artifacts)


i made a test on that issue, it is in one of the src threads (the files should still be available).
i could not find a real issue with a few nebulas in chain on three tracks, when you compare the cd-format-outputs (16 bit, 44,1 khz) of working in 96 khz or working in 48 khz with nebula-src. i did the downsampling with samplitudes internal src, which is not the best available, but quite good.

i would still recommend working in 96 khz, when your resources allow it, but there is no hearable buildup of artefacts or distortion by using nebulas src in more instances.
when you stay in 96 khz, you can really hear the difference, but it melts down to such a degree after downsampling, that i could not find great differences in the downsampled files.

i would be interested in examples from your side, when you found differences that you would describe as clearly audible. i did just one experiment which does surely not cover all variables. :)

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: 2116
Joined: Wed Mar 31, 2010 12:25 am
Location: South-West Germany

Re: More Sample Rate Conversion

Postby Tim Petherick » Wed Mar 30, 2011 12:20 am

I guess all I'm saying is its best to stick to a sample rate rather than upsampling and downsampling Through more than one nebula's .That was just in the Any ideas part of the original post.

When i said that the quality was much better I was refering to no.3 option in the original post.Which is to convert the audio files using a real good src . then sticking to the same sample rate nebula programs.

I only record at 96k now and use 96k nebulas.Or convert original audio to 96k if not already.

Yes I guess I theory it makes sense .Maybe it has had a placebo effect on me .Your right maybe I need to look at more varibles too. :) Sorry for a Very opinionated answer. I should have something to back it up but I don't right now
So i'm going to re-evaluate what i've said now and do some tests again.Hopefully this week If I have time . But I'm solidly recording.

All's good! Its a learning curve for me :lol:


Tim
Last edited by Tim Petherick on Wed Mar 30, 2011 12:38 am, edited 1 time in total.
User avatar
Tim Petherick
Expert
Expert
 
Posts: 1352
Joined: Sat Apr 17, 2010 4:07 pm
Location: Bath , Uk

Re: More Sample Rate Conversion

Postby mathias » Wed Mar 30, 2011 12:38 am

i am with you, it is a good choice to work in 96 khz with 96 khz nebula programs.

on the other hand we should not be fearful to work with nebulas src, when we have no evidence that there are really bad issues. that is why i commented on your statement. i am absolutely open to accept it, when somebody really finds something (hearable and proved so by others), so we can all learn and avoid potential pitfalls, when using nebula.

all the best,
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: 2116
Joined: Wed Mar 31, 2010 12:25 am
Location: South-West Germany

Re: More Sample Rate Conversion

Postby Tim Petherick » Wed Mar 30, 2011 12:39 am

I love this forum .

So many great people!!


Horaahhhh for nebula!!
User avatar
Tim Petherick
Expert
Expert
 
Posts: 1352
Joined: Sat Apr 17, 2010 4:07 pm
Location: Bath , Uk

Re: More Sample Rate Conversion

Postby mathias » Wed Mar 30, 2011 12:41 am

:D :D :D
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: 2116
Joined: Wed Mar 31, 2010 12:25 am
Location: South-West Germany

Re: More Sample Rate Conversion

Postby justinmorell » Wed Mar 30, 2011 1:36 am

Thanks guys for the ideas/discussion. Very helpful.

Right now I am working with the files in 44.1, but I'm going to batch convert the 48k source files to 96k and plug them into the same Logic file, with the same settings, etc., to hear the difference. There are two tracks that I'm mastering with two Nebula instances, so I think this could be a big improvement.

I'll be happy to make some excerpts available when I have them--curious to know what you all think!
--Justin

Mac Mini 2.4 GHz Intel Core 2 Duo
OS X v. 10.6.4
4 GB 1067 MHz DDR3

Running Logic Pro 9
Nebula 3 Pro
justinmorell
User Level 0
User Level 0
 
Posts: 8
Joined: Fri Mar 25, 2011 5:14 am

Re: More Sample Rate Conversion

Postby justinmorell » Fri Apr 01, 2011 6:26 pm

Okay, so I have excerpts from a mix here: http://www.justinmorell.com/filesforwhatever/conversion_comparison.zip

These are both 44.1 files, 24 bit. The one labeled "no_96k" was processed as follows: the original 48k file was converted to 44.1 in Sample Manager with izotope, then imported to Logic. I have 2 instances of Nebula 3 on this mix, one with the Alexb Digitub program, followed by the Otari 30 ips 0db Ampex 499 R2R program. Both of these are 96k samples, and therefore each instance involved a SRC from 96k to 44.1 and back. I also have a couple of other plugins on the track, including a SonEQ, a Logic compressor, and a Logic Adaptive Limiter, though I don't think those are relevant for the comparison since none of them introduces any SRC.

The one labeled "converted_96k" was processed as follows: the original 48k mix was upsampled to 96k using Sample Manager. I then imported this 96k file into Logic and used the exact same channel strip plugins as listed above. The difference, of course, is that the two Nebula instances did not need to SRC the files. I then bounced the file, used Sample Manager to run a SRC to 44.1k.

After all of this, I re-imported the files into a new Logic session and lined them up to run a phase test. They DEFINITELY did not cancel out, as expected, though I was a bit surprised at how far from cancellation they actually were. It's pretty hard for me to judge the quality, since I think they both sound pretty good. However, best I can tell, the file that was processed without using Nebula's SRC (i.e. with Nebula programs running in their native sample rate) sound a bit smoother in the highs and upper mids.

Curious to hear others' thoughts. . .
--Justin

Mac Mini 2.4 GHz Intel Core 2 Duo
OS X v. 10.6.4
4 GB 1067 MHz DDR3

Running Logic Pro 9
Nebula 3 Pro
justinmorell
User Level 0
User Level 0
 
Posts: 8
Joined: Fri Mar 25, 2011 5:14 am

Next

Return to Working with Nebula

Who is online

Users browsing this forum: Bing [Bot], daniddd, Pol Poil and 4 guests