Login

Sample Rate Discussion (off of AITB Germanos EQ Thread)

If you need to write about anything else please do it here..

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Postby tumburu » Wed Feb 16, 2011 4:53 am

I always said that the major inconvenience with Nebula SRC is loading time. That's why I'd like to have a similar SRC in NAT for converting the 96k libraries to 44. Can't really understand why the two SRC algorithms are different.
User avatar
tumburu
Vip Member
Vip Member
 
Posts: 430
Joined: Tue Mar 30, 2010 11:00 pm

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Postby yr » Wed Feb 16, 2011 1:06 pm

I have to say I disagree with some of the recent remarks. Personally, I started testing the Nebula src long ago, after being dissatisfied with the results I was getting at 44.1kHz using the consoles. It has nothing to do with the recent "aliasing panic" on Gearslutz. I work and mix at various sample rates for different clients (and different goals). Obviously when I get an OMF file that needs to be edited/mixed for broadcast the next day I don't bother up-sampling before using Nebula...

The interesting question for me is whether I'm getting the best results I can from Nebula when using sensitive material, and whether I can hear the difference at different sample-rates. As an example, I enjoy a bit of coloration when mixing classical music, but using the console presets at 44.1kHz (I own libraries from different developers) almost always leads to inferior results when compared to upsampled material and even the source material.
Reuven | post-production & sound-design | scenography |
website | nebula presets
User avatar
yr
Beta Tester
Beta Tester
 
Posts: 438
Joined: Mon Mar 29, 2010 9:04 am
Location: Amsterdam

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Postby Stolle » Wed Feb 16, 2011 2:39 pm

I think Yr is raising some very interesting points and argues in a constructive manner.

We all know Nebula rocks. It's about how hard we can make it rock.

I must say that I'm really surprised by AlexB's audio examples. The difference ain't subtle when compared in the DAW. I much prefer the 96-96 -version.
Ohh wait! I wrote while listening. The 44-44 version is lower in volume than the 96-96 -version (as is the 96-44). After the tracks are level match I must say it's a close tie between the 44-44 and the 96-96. I belie I can actually hear the sound of the SRC in the other examples now. Strangely I hear that effect subtly in the 44-44 example to.
I can't understand how I can hear a difference between 96-96 and 44-44. Surely, at age 31, my hearing can't reach those upper frequencies. What gives?
Stolle
User Level VI
User Level VI
 
Posts: 68
Joined: Mon Mar 29, 2010 3:27 am

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Postby Mercado_Negro » Wed Feb 16, 2011 4:11 pm

SRC will always lead to inferior results, it doesn't matter if you're using the best algorithm in the universe, like AlexB said, you can't beat the laws of physics. I mix at 48kHz, does this mean developers should also release 48kHz versions? Seriously, I couldn't care less after tons of listening tests I've done here. When comparing the sound quality and results I get from Nebula with some other plug-ins I realize I wasted my time in such tests, time I should've invested in learning how programs react/sound at my preferred sample rate so I can decide when/how to use them.
i7 3770k :: Asus P8H77-V LE :: 16Gb DDR3 @1600MHz :: Geforce GT 520 :: OCZ-Vertex 128Gb :: WD Black Series 1Tb and Green Series 1Tb :: F*******e Liquid56 :: REAPER 64bit and StudioOne 64bit (both latest versions) :: Win 10 64bit
User avatar
Mercado_Negro
Beta Tester
Beta Tester
 
Posts: 1353
Joined: Sat Mar 27, 2010 9:30 am

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Postby yr » Wed Feb 16, 2011 4:44 pm

Actually if you are working at 48kHz you have a dilemma since there are no native 48kHz presets. From my experience, using the Nebula conversion to go 44.1kHz-48kHz actually sounds better then 96kHz-48kHz.

Personally I don't understand why you would disregard test results if you can hear the difference (as opposed to only seeing it). I know in my case, I completely stopped using the console presets for 44.1/48kHz projects, when my main concern is clarity and transient response. This kind of findings may or may not also influence which presets you buy next time if you decide to commit to a certain sample-rate.

As for the AlexB test- you can't draw any conclusions comparing the 44.1-44.1 vs 96-96 files. The reason is that the 44.1kHz file was down-sampled first prior to being processed by Nebula, which will effect the results...

A better test would be to use an identical sound-source which was recorded once at 96kHz and another time as 44.1kHz. I used a mechanical music box to do similar tests in the past.
Reuven | post-production & sound-design | scenography |
website | nebula presets
User avatar
yr
Beta Tester
Beta Tester
 
Posts: 438
Joined: Mon Mar 29, 2010 9:04 am
Location: Amsterdam

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Postby ngarjuna » Wed Feb 16, 2011 5:51 pm

I have yet to see (or read) a SRC comparison study that was anywhere near acceptable. Enough of the anecdotal "OMG you can totally hear the difference!!" and enough of the "Obviously the laws of physics support higher sample rate!!".

The thing I don't understand is that increasing sample rate/decreasing sample rate conversions is probably the tiniest, least significant thing you can do to improve your mixes. And yet instead of learning more about their craft, about electronics themselves, about waveform physics, people spend hours a day writing the next great American novel about sample rate.

Sample rate is a total red herring. It has never made a good mix bad or a bad mix good. Anyone who focuses on sample rate is probably producing inferior results (if you're so focused on that particular detail good luck seeing the trees through the forest).
User avatar
ngarjuna
Expert
Expert
 
Posts: 778
Joined: Tue Mar 30, 2010 5:04 pm
Location: Miami

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Postby yr » Wed Feb 16, 2011 6:10 pm

We are not talking about complicated src issues or a very complex quest. Any user can devote 1 hour to test this matters for themselves and come to their own conclusions.

All we are trying to figure out is whether we can/can-not hear the effects of the Nebula src and if that would effect our choices. It's obvious that people are coming to different conclusions based on their hearing, needs, cpu resources, tastes etc. What may seem insignificant to you while creating a rock mix using many Nebula instances to get a certain sound, may be a huge deal for someone else trying to get a different sound with completely different materials.

I'm not sure why some are trying to make the test procedure into a complex matter. It's actually a very simple thing to do: use one of your favorite Nebula chains (multiple presets) to process audio using Nebula src and compare it to the results you get when up/down-sampling the audio material prior to Nebula (to avoid the Nebula src). It takes about one hour and will give each user a solid answer rather then relaying on opinions, declarations etc.
Reuven | post-production & sound-design | scenography |
website | nebula presets
User avatar
yr
Beta Tester
Beta Tester
 
Posts: 438
Joined: Mon Mar 29, 2010 9:04 am
Location: Amsterdam

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Postby ngarjuna » Wed Feb 16, 2011 6:29 pm

yr wrote:We are not talking about complicated src issues or a very complex quest. Any user can devote 1 hour to test this matters for themselves and come to their own conclusions.

All we are trying to figure out is whether we can/can-not hear the effects of the Nebula src and if that would effect our choices. It's obvious that people are coming to different conclusions based on their hearing, needs, cpu resources, tastes etc. What may seem insignificant to you while creating a rock mix using many Nebula instances to get a certain sound, may be a huge deal for someone else trying to get a different sound with completely different materials.

I'm not sure why some are trying to make the test procedure into a complex matter. It's actually a very simple thing to do: use one of your favorite Nebula chains (multiple presets) to process audio using Nebula src and compare it to the results you get when up/down-sampling the audio material prior to Nebula (to avoid the Nebula src). It takes about one hour and will give each user a solid answer rather then relaying on opinions, declarations etc.


Nobody is trying to "make the test procedure into a complex matter". Testing these kinds of differences is a complex matter because you constantly have to account for observation biases which will creep into your test at any point that you allow it.

Unless you have two identical systems that you can actually (quickly!) switch back and forth to A-B properly (this is all assuming you even have the experience/skills/practice A-B-X'ing, which frankly most people don't to any degree that we should accept their results as even reasonably objective) you haven't actually done a comparison test, plain and simple. In the amount of time it takes you to reset the sample rate and load the project in a different format your ability to retain specific details from the previous listening round is completely eroded and you are now comparing your subjective recollections of that prior listening session to what you're hearing now. The fact that you think you can leave a huge time gap in between audio tests (and yes, many seconds is a huge gap to your weak, human-based perceptual memory) and still have a valid comparison is precisely my complaint about this thread and threads like it: you aren't testing anything near what you think you're testing.

That does not in any way satisfy even a pseudo-scientific survey upon which to base a hypothesis. Study some more about psychological biases, the inherent weakness in human perceptual memory and the importance of dotting every "i" and crossing every "t" when doing double blind audio testing and you might come to a similar conclusion that I did: the rigors of actually testing the SRC differences are so complex as to not be worth the time spent for most working engineers. After all, we have work to do!

Again, I stand by my original point (that I've made before in many other threads like this on many forums) that has never been refuted by anyone:

SRC and sample rate considerations are possibly the least important detail you'll be worrying about in your mix. SRC/sample rate changes have never made a good mix sound bad or a bad mix sound good.
User avatar
ngarjuna
Expert
Expert
 
Posts: 778
Joined: Tue Mar 30, 2010 5:04 pm
Location: Miami

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Postby yr » Wed Feb 16, 2011 6:36 pm

There is absolutely no bias or placebo-effects once both samples share the same final sample rate (44.1kHz in my example) and loaded into an ABX program. I majored in psychology, so I'll be the last one to ignore any mind-tricks. I stand by my last post-one hour work to test this very case specific issue. I hope you take the time to do that...
Last edited by yr on Wed Feb 16, 2011 8:39 pm, edited 1 time in total.
Reuven | post-production & sound-design | scenography |
website | nebula presets
User avatar
yr
Beta Tester
Beta Tester
 
Posts: 438
Joined: Mon Mar 29, 2010 9:04 am
Location: Amsterdam

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Postby ngarjuna » Wed Feb 16, 2011 6:43 pm

yr wrote:There is absolutely no bias or placebo-effects once both samples share the same final sample rate (44.1kHz in my example) and loaded into an ABX program. I majored in psychology, so I'll be the last one to ignore any mind-tricks. I stay by my last post-one hour work to test this very case specific issue. I hope you take the time to do that...


Which still does not address really any of the concerns that make this kind of testing tricky.

If you want to pretend that your one hour "testing" regiment actually proves something, by all means. But if you're going to proselytize on the forums based on incomplete anecdotal results then you should be prepared for people to disagree with your results and methods.
User avatar
ngarjuna
Expert
Expert
 
Posts: 778
Joined: Tue Mar 30, 2010 5:04 pm
Location: Miami

PreviousNext

Return to Other things

Who is online

Users browsing this forum: No registered users and 3 guests