Login

Neb 3 Pro and Neb 3 Pro x64 sound different

Tips & tricks, working results, technical support

Re: Neb 3 Pro and Neb 3 Pro x64 sound different

Postby ngarjuna » Thu Mar 31, 2011 8:35 pm

Mercado_Negro wrote:
fuseburn wrote:What I don't get either is: Why don't 2 instances of nebula with identical programs - meaning 2x x64 or 2x x86 with R2R or a console) - null out ? The difference signal is always down at -40 till -70 and it sounds like the program rate intervals are not synced (Overtones/distortion). That means you can't even verify if your plugin processing sounds just like the freeze you did 3 days ago, let alone the things I'm talking about in here (comparing x86 vs x64). Why is that ?


Because 2 hardware units don't sound the same either, ever. You'd use your ears in this case to verify if your unit is still in good shape. Null tests with Nebula is pointless and a major waste of time man, there are several hundreds of samples per programs and they are not the same on each pass; you can actually verify this by listening the same passage over and over, it just won't sound the same each time. Been there, done that. That's one of the things I like the most about Nebula, even in that it behaves like hardware :)

I have to run some tests to confirm what you say about differences between 32-bits and 64-bits versions but to be honest, I consider it highly unlikely since compiling is a pretty straightforward step after coding and having all the environment set up. In any case, I'll give it a go for fun :)


What you're describing sounds like random behavior which, while indeed potentially present with analog equipment (the behavior in that case isn't actually random but it's affected by factors which are quite invisible without a lot of test equipment), I don't understand how Nebula could accomplish that. My understanding was that Nebula used program input to determine which of the sample sets a particular programatic input would use. How would a common signal apply different Nebula samples (using a particular program) from one pass to the next? Is there some kind of pseudo-randomizer? Not really looking for code-specific details, but is this really what Nebula is doing? On one hand that would be pretty cool but from another perspective it's a bit frightening to think I could dial up my "dream settings" only to get a different sound on my render pass (yes, I do realize that is precisely what happens often with fussy hardware). I thought only time based effects would have this feature.

I have certainly experienced what you describe (listening to a pass and thinking to myself "Was that different? The transient on that third snare hit sounded a little different to me...") but I always attributed that to the poor auditory memory and extreme susceptibility to cognitive biases that humans are endowed with; the difference has never been so extreme that it couldn't be explained as psychological.

I'm curious, though, perhaps my understanding is incorrect.
User avatar
ngarjuna
Expert
Expert
 
Posts: 778
Joined: Tue Mar 30, 2010 5:04 pm
Location: Miami

Re: Neb 3 Pro and Neb 3 Pro x64 sound different

Postby fuseburn » Thu Mar 31, 2011 8:49 pm

giancarlo wrote:Anyway 2 nebula tracks sound different even if you use the same plugin.


Alright - but why is that ?

Another strange thing I thought I heard: Does the session BPM affect Nebula processing in any way ? Try setting your BPM to 30 BPM and playback a WAV through nebula, then turn it up to 390 BPM (without changing playback speed obviously ;-) ) and listen again. Could there be a difference ? But maybe I'm just hearing ghosts :mrgreen:
User avatar
fuseburn
User Level X
User Level X
 
Posts: 133
Joined: Mon Apr 05, 2010 6:19 pm

Re: Neb 3 Pro and Neb 3 Pro x64 sound different

Postby ngarjuna » Thu Mar 31, 2011 9:09 pm

fuseburn wrote:Another strange thing I thought I heard: Does the session BPM affect Nebula processing in any way ? Try setting your BPM to 30 BPM and playback a WAV through nebula, then turn it up to 390 BPM (without changing playback speed obviously ;-) ) and listen again. Could there be a difference ? But maybe I'm just hearing ghosts :mrgreen:


Out of curiosity, I tested this and it's true, it only nulls down to about -40dB at its highest.

If what Mercado_Negro was saying (about the pseudo-randomness) is what's happening, then that makes sense without much more explanation needed. But if my understanding of how Nebula chooses its samples is correct, then I don't get why that's happening.

But that issue (the pseudo-randomness) is definitely key to a bunch of questions floating around about Nebula right now.
User avatar
ngarjuna
Expert
Expert
 
Posts: 778
Joined: Tue Mar 30, 2010 5:04 pm
Location: Miami

Re: Neb 3 Pro and Neb 3 Pro x64 sound different

Postby hopsing » Thu Mar 31, 2011 9:53 pm

That sounds all pretty interesting. Fwiw, I once tried to do the sample rate conversion of the emt 140 programms to 44,1Khz in NAT and to verify the results I nulled two Instances of Nebula. Up to 2,5 seconds of reverb tail I could null the NAT converted and the Nebula converted programms completly (In Logic audio). Longer reverb tails were cut by NAT, but that seems to be a known bug of NAT.
So I was under the impression, that Nebula calculates reliably if all the variables are the same.
Tobias
hopsing
User Level I
User Level I
 
Posts: 15
Joined: Sat Apr 03, 2010 1:36 pm

Re: Neb 3 Pro and Neb 3 Pro x64 sound different

Postby Mercado_Negro » Thu Mar 31, 2011 10:24 pm

ngarjuna wrote:What you're describing sounds like random behavior which, while indeed potentially present with analog equipment (the behavior in that case isn't actually random but it's affected by factors which are quite invisible without a lot of test equipment), I don't understand how Nebula could accomplish that. My understanding was that Nebula used program input to determine which of the sample sets a particular programatic input would use. How would a common signal apply different Nebula samples (using a particular program) from one pass to the next? Is there some kind of pseudo-randomizer? Not really looking for code-specific details, but is this really what Nebula is doing? On one hand that would be pretty cool but from another perspective it's a bit frightening to think I could dial up my "dream settings" only to get a different sound on my render pass (yes, I do realize that is precisely what happens often with fussy hardware). I thought only time based effects would have this feature.

I have certainly experienced what you describe (listening to a pass and thinking to myself "Was that different? The transient on that third snare hit sounded a little different to me...") but I always attributed that to the poor auditory memory and extreme susceptibility to cognitive biases that humans are endowed with; the difference has never been so extreme that it couldn't be explained as psychological.

I'm curious, though, perhaps my understanding is incorrect.


That's how I understand it and apparently, according to what I hear, it doesn't depend on the number of samples in a program only but the amount of kernels as well. The more kernels you have, the greatest 'randomness', so to speak (I'm not sure if 'random' is a good word in this case, though). A couple of years ago, when I did all kind of null tests and stuff, I could almost null programs with a low amount of samples and just a couple of kernels. If I recall correctly, none of the 'Input' programs from AlexB's consoles could be nulled against themselves (e.g., IL1 vs IL1).

Hopefully Giancarlo or Enrique will appear soon and clear this out for us.

[EDIT] I wouldn't loose any sleep over loosing my "dream settings" since the differences are super mega subtle, just like it happens with units in good conditions or recently serviced :)
Last edited by Mercado_Negro on Thu Mar 31, 2011 10:37 pm, edited 1 time in total.
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
Global Moderator
Global Moderator
 
Posts: 1353
Joined: Sat Mar 27, 2010 9:30 am

Re: Neb 3 Pro and Neb 3 Pro x64 sound different

Postby giancarlo » Thu Mar 31, 2011 10:33 pm

it's how nebula works and it's not a bug. It happens also in analog world. It's the key for analogueness, and it's not random. Simply there are many variables, one of them is time, so if you process a thing twice those files will be slighly different. Now you understand why it's cpu intensive...
User avatar
giancarlo
Founder
Founder
 
Posts: 9153
Joined: Mon Sep 21, 2009 10:40 pm
Location: Italy

Re: Neb 3 Pro and Neb 3 Pro x64 sound different

Postby Mercado_Negro » Thu Mar 31, 2011 10:37 pm

giancarlo wrote:it's how nebula works and it's not a bug. It happens also in analog world. It's the key for analogueness, and it's not random. Simply there are many variables, one of them is time, so if you process a thing twice those files will be slighly different. Now you understand why it's cpu intensive...


There we go :) Thanks Giancarlo.
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
Global Moderator
Global Moderator
 
Posts: 1353
Joined: Sat Mar 27, 2010 9:30 am

Re: Neb 3 Pro and Neb 3 Pro x64 sound different

Postby fuseburn » Thu Mar 31, 2011 10:52 pm

Mercado_Negro wrote:Because 2 hardware units don't sound the same either, ever. You'd use your ears in this case to verify if your unit is still in good shape. Null tests with Nebula is pointless and a major waste of time man, there are several hundreds of samples per programs and they are not the same on each pass; you can actually verify this by listening the same passage over and over, it just won't sound the same each time. Been there, done that. That's one of the things I like the most about Nebula, even in that it behaves like hardware :)


Sure, but the differences nebula provides aren't derived from the original hardware. Otherwise all the samples from digital units wouldn't show these differences, but they do. So the effect seems to be artificial and therefore unneeded - at least imo ;-) Best solution would be to have that particular behaviour optional.

Right now it's mostly an obstacle to identify erratic behaviour. You can't prove that it sounds the same as yesterday. I just noticed that for example hitting the RELOAD button changes the sound significantly in some cases (can't really say precisely which cases yet), and the BPM-changes-sound-affair - I'd really love to be able to do scientific tests on that, you know :D , but I can't.

giancarlo wrote:it's how nebula works and it's not a bug. It happens also in analog world. It's the key for analogueness, and it's not random. Simply there are many variables, one of them is time, so if you process a thing twice those files will be slighly different. Now you understand why it's cpu intensive...


Same here - why do the S*L Duende programs sound slightly different aswell then ? No analog, no mojo, just digital algorithms ;-)

What exactly causes these differences, if they're - like you said - not random ?
User avatar
fuseburn
User Level X
User Level X
 
Posts: 133
Joined: Mon Apr 05, 2010 6:19 pm

Re: Neb 3 Pro and Neb 3 Pro x64 sound different

Postby Mercado_Negro » Thu Mar 31, 2011 11:01 pm

fuseburn wrote:Sure, but the differences nebula provides aren't derived from the original hardware. Otherwise all the samples from digital units wouldn't show these differences, but they do. So the effect seems to be artificial and therefore unneeded - at least imo ;-) Best solution would be to have that particular behaviour optional.

Right now it's mostly an obstacle to identify erratic behaviour. You can't prove that it sounds the same as yesterday. I just noticed that for example hitting the RELOAD button changes the sound significantly in some cases (can't really say precisely which cases yet), and the BPM-changes-sound-affair - I'd really love to be able to do scientific tests on that, you know :D , but I can't.


Significant differences? Wow man, when I read your posts I really doubt I can hear well hehe To me those differences are less than a problem when mixing. I've experienced the same thing when mixing OTB so I'm just probably used to it :) Just think you're using hardware and make your life easier, forget about all these tests and have fun. You know, after long periods of mixing sessions I start hearing monsters and ghosts as well. After some years I've learned that nothing sounds the same the next day :lol:
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
Global Moderator
Global Moderator
 
Posts: 1353
Joined: Sat Mar 27, 2010 9:30 am

Re: Neb 3 Pro and Neb 3 Pro x64 sound different

Postby ngarjuna » Thu Mar 31, 2011 11:36 pm

fuseburn wrote:...

What exactly causes these differences, if they're - like you said - not random ?


Different pseudo-RNGs work differently. One that is pretty common takes the time clock and gets its "random" numbers from there (injecting them from the ten thousandths decimal of a second measurement, for example, so that it happens at a speed that couldn't be contrived or easily predicted). As long as you have a stream of digits it's a fairly easy method to let the user's call (a moment in time) determine the numbers that will seed the PRNG. While it's not truly random it happens at a level which is far enough from the user's observation that it seems random-ish.

I'm guessing Nebula's engine isn't just chaotically, purely pseudo-random, it probably follows rules about which samples of a program to use for a particular sample of audio and then uses its seeding method to choose between acceptable values or something like that. So you should never have to worry about getting really erratic results (which would suck) but you still get the benefit of very subtle, slight variations (which can indeed seem analog-ish/hardware-ish as intended).
User avatar
ngarjuna
Expert
Expert
 
Posts: 778
Joined: Tue Mar 30, 2010 5:04 pm
Location: Miami

PreviousNext

Return to Working with Nebula

Who is online

Users browsing this forum: Bing [Bot] and 11 guests