To me there seems to be a different approach to the way that Protools is designed. You guys talk about the fallacies of an uneducated programmer, well maybe, just maybe, that is all we're talking about here. That the designers of the Protools system seem to have programmed thier platform the right way paying close attention to detail, and perhaps, just perhaps the others are still learning(or maybe not).
After all Cubase has always been touted as a Midi Production enviroment, audio fidelity may not have been their selling point, hence the VST Standard. On the other hand, Protools has always had the selling point to be about Recording(the replacement of the analog tape machine), so the fidelity part is quite important. However, if this is the case, then the debate over Float Vs Integer is indeed a moot point.
The real issue at hand would be the Steinberg(or native) programmers -Vs- The Digidesign Programmers, if it is indeed as you all have stated. So, to shed some light on the two programming units, lets see if one company has programmed better than the other. Ill be putting my results up shortly.
PS. I didnt intend for this thread to become an all out war of the programmers, but it has shed some light on programming techniques and disciplines. I commend u all for your programming efforts, because if it wernt for the likes of you all, we would not be on this forum at all.
Okay, I know this has been done before, on another board, but id like to conduct a summing test of my own. The point would be to kill this subject once and for all. However the source tracks will be a little different from the other boards test.
All source files come from vstis and samples.
We'll conduct this test in the same manner as the others did. I will provide the sound files and also my results and anybody whos willing to participate can doe the same.
These files will not need to be touched in any way except for importing into your platform.
The 2 platforms that i wish to compare is:
Cubase SX or Nuendo 1,2, or 3 and the Protools HD system.
Cubase is 32bit float based and Protools HD is 48 bit integer based.
Since all of the daw platforms, with exception of Protools HD, is 32bit float based, Cubase SX or Nuendo will suffice.
If anyone is interested then please say so here or PM me.
Alright, I need someone with the protools HD system to get this thing started,
-------------------------------------------------------
Alright folks, once again here are the rules
1. the test is for 32 bit float(CUBASE) VS 48bit integer(PROTOOLS HD).(u can test other software.Include what software u used)
2. Do not alter the files in anyway. ie. normalize, eq,etc
3. Put a link to the mixdown in this thread along with what format u used.
4. the pan law will be -3db
heres the link. http://rapidshare.de/files/7634800/stef.rar.html
NATIVE(CUBASE) PROGRAMMING VS PROTOOLS HD PROGRAMING
-
- KVRist
- Topic Starter
- 227 posts since 14 May, 2005 from Atlanta
Last edited by YISH313 on Sat Nov 19, 2005 4:49 pm, edited 3 times in total.
---------------------------
http://www.thaproducerz.com
http://www.thaproducerz.com
-
- Banned
- 82 posts since 28 Oct, 2005
hi bud.
Sounds Good
Why dont you include Sonar 5 64 bit double precsion.
Also I know for a fact Samplitude has a great rep WRT their mix engine.
Anyhow i hope this gets going, im very interested in this shit.
Sounds Good
Why dont you include Sonar 5 64 bit double precsion.
Also I know for a fact Samplitude has a great rep WRT their mix engine.
Anyhow i hope this gets going, im very interested in this shit.
-
- KVRist
- 92 posts since 6 Feb, 2005
Which word processor has the best spacebar? I think the spaces on word 2000 are "warmer" than on word xp. Anyone have other word processors so we can test?
-
- KVRAF
- 12555 posts since 7 Dec, 2004
i do think for audio alone, the extra accuracy isnt that big of a deal. please find monitors and an output device, and an insultated cable rated with a noise floor lower than 120db. i dont think you will. but, if you are using 24 bits per sample, it means using float you loose bits as soon as you mix anything. float can only fit 24 bits (it uses an extra for sign, so yes, technically it is 25 bits ) so every time you double the number of channels added you lose one bit. mixing 8 channels with vst and a float buffer, you get only 20 bits? - no, this is wrong. the function for finding eqiv. bits is more complicated, but i'm just trying to outline the problem here not discribe it exactly. ah ok, i'll add this and then i'm stopping. i think that every power of 10 means you spread the accuracy across that greater range. so the accuracy from 1.0 to 10.0 is the same as the accuracy from 0.0 to 1.0. the accuracy from 10.0 to 100.0 is the same again. so each time your numbers get 10^n, you have 1/10th the accuracy.
anyway, it is more about implementation than about which datatype is used, as in all cases. "whoever's" integer mixer might just be a better implementation with quality dither and so on, and to actually lose the bits as i discribed would require the worst-case implementation. so i really can not see any difference in using a mix buffer with many more bits as long as you use a quality implementation. when you move out of the range -1.0 to +1.0, by adding an offset or using a range larger than that, the accuracy drops.
anyway, it isnt _THAT_ big of a deal, you'd have to be a major asshole like me to notice the difference. i've said all this before..
) so every time you double the number of channels added you lose one bit. mixing 8 channels with vst and a float buffer, you get only 20 bits? - no, this is wrong. the function for finding eqiv. bits is more complicated, but i'm just trying to outline the problem here not discribe it exactly. ah ok, i'll add this and then i'm stopping. i think that every power of 10 means you spread the accuracy across that greater range. so the accuracy from 1.0 to 10.0 is the same as the accuracy from 0.0 to 1.0. the accuracy from 10.0 to 100.0 is the same again. so each time your numbers get 10^n, you have 1/10th the accuracy.anyway, it is more about implementation than about which datatype is used, as in all cases. "whoever's" integer mixer might just be a better implementation with quality dither and so on, and to actually lose the bits as i discribed would require the worst-case implementation. so i really can not see any difference in using a mix buffer with many more bits as long as you use a quality implementation. when you move out of the range -1.0 to +1.0, by adding an offset or using a range larger than that, the accuracy drops.
anyway, it isnt _THAT_ big of a deal, you'd have to be a major asshole like me to notice the difference. i've said all this before..
Last edited by aciddose on Mon Nov 14, 2005 8:12 am, edited 1 time in total.
-
- KVRAF
- 6519 posts since 13 Mar, 2002 from UK
nuffco Audiophile salutes you.
nuffco Audiophile - where maths is purely advisory.
nuffco Audiophile - where maths is purely advisory.
Now with improved MIDI jitter!
-
- KVRAF
- 6478 posts since 16 Dec, 2002
Call me stupid, but listen up,
why not simplify the test a LOT?
Doesn't traktion2 have a choice between 32bit and 64bit (float) mixing? Why not simply test a complex mix inside T2, export it properly and do a blind test? Couldn't this also be done in Sonar5?
that would eliminate arguments about host mixer implementations and different "sounds" ie Logic sounds better than Cubase bleha blah blah.
And seriously. f**k protools.
why not simplify the test a LOT?
Doesn't traktion2 have a choice between 32bit and 64bit (float) mixing? Why not simply test a complex mix inside T2, export it properly and do a blind test? Couldn't this also be done in Sonar5?
that would eliminate arguments about host mixer implementations and different "sounds" ie Logic sounds better than Cubase bleha blah blah.
And seriously. f**k protools.
-
- KVRAF
- 2583 posts since 26 Aug, 2002 from here
yep that the way to do it - and when no-one can tell the diference can we finally lay this to rest
I believe every thread should devolve into character attacks and witch-burning. It really helps the discussion.
-
- Banned
- 82 posts since 28 Oct, 2005
Kingston wrote:Call me stupid, but listen up,
why not simplify the test a LOT?
Doesn't traktion2 have a choice between 32bit and 64bit (float) mixing? Why not simply test a complex mix inside T2, export it properly and do a blind test? Couldn't this also be done in Sonar5?
that would eliminate arguments about host mixer implementations and different "sounds" ie Logic sounds better than Cubase bleha blah blah.
And seriously. fcuk protools.
No this will not be sufficient.
Each hosts calculations are not identicle.
They will have to be tested individualy.
Might i add that the mix should contain lots of panning automation, volume automation, and as high a plugin count as possible.
-
- KVRAF
- 6478 posts since 16 Dec, 2002
Oh well. I tried.
You people might as well go ahead and do your masochistic tests with X^2 number of hosts. I just hope you do realise none of this has any real world value whatsoever.
You people might as well go ahead and do your masochistic tests with X^2 number of hosts. I just hope you do realise none of this has any real world value whatsoever.
Last edited by Kingston on Mon Nov 14, 2005 10:34 am, edited 1 time in total.
-
- KVRAF
- 6478 posts since 16 Dec, 2002
Doug,Doug1 wrote:not sufficient, each host sounds different...bleha blah blah.
What did I *just* say?
"that would eliminate arguments about host mixer implementations and different "sounds" ie Logic sounds better than Cubase bleha blah blah."
