R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 ... 6 7 [8]   Go Down

Author Topic: Pro Tools: I was wrong. Yep. Wrong.  (Read 28208 times)

Fergal T

  • Newbie
  • *
  • Offline Offline
  • Posts: 42
Re: Pro Tools: I was wrong. Yep. Wrong.
« Reply #105 on: May 22, 2005, 04:12:31 AM »

Nika Aldrich wrote on Sat, 21 May 2005 22:16

Then it will be determined to have been indicative of the answer for that mix but not necessarily representative of the formats as a whole.

Nika


Well, how about we do a worst case scenario type mix, where the first four inserts on a channel are 6dB high shelving eqs and then the last's a compressor... - If we don't notice any difference there, then we can safely assume it really isn't the issue loco makes it out to be.

-Fergal
Logged

Fergal T

  • Newbie
  • *
  • Offline Offline
  • Posts: 42
Re: Pro Tools: I was wrong. Yep. Wrong.
« Reply #106 on: May 22, 2005, 04:18:41 AM »

pipelineaudio wrote on Sat, 21 May 2005 23:55

where in PT is it 48 bit? The track itself is 24 bit as are the plugin inserts right?


Almost all calculations are calculated to 48bit precision - the mixer for instance, and in almost all plugins. However, all data transfer to and from plugins on the TDM bus is at 24bits.

-Fergal
Logged

Patrik T

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 833
Re: Pro Tools: I was wrong. Yep. Wrong.
« Reply #107 on: May 28, 2005, 08:46:44 AM »

Hello everyone.

I do my recordings and my mixing in Sonar. It is using 32 floating bits. If plugins are used, some are having 64 bits of internal processing, others less. I am most often just saying to myself: "Ok, well done - 64 bits of precision, the plug actually sound quite good compared to that old thing". But it does not interfere with my intentions music-wise. It does not restrict me to use only "this" or "that".

I find the mixers internal mathematics to be similar to a continuously expanding formula. I know it happens. I am aware.

BUT.

It is not: "Oh god, I just automated a fader, the calculations have expanded, I'm lost in truncation."

Instead it is: "Oh well, I just automated a fader, the calculations have expanded, it had to be done, the music needed it."

I've mixed down without any dithering before. Records have been released.

Now I have used flat dither, triangular, 24 bits, no noise shape on my mixbus, when mixing down and out-putting everything to a fixed 24 bit file. This has been the only 24 bit dither used in a project, since everything else float around in 32 bits. I tell myself there is slightly more depth in the result. And there is slightly less grain. Especially when it's later converted to 16 bits, using another, maybe noiseshaped dither.

So for me, it's all about evaluation. Not until I hear my new, mastered CD-project, I can tell if the 24 bit dither took my compositions further to where they were when swimming in the 32-pool. But I tell myself the effort will probably be worth it. And if it is, I'll continue to dither my huge 32 floating bits to 24 fixed ones.

Until then, I hear only music, not much noise. And the power, variations, message and body of the mix is far more important than truncation at levels belo -110 dB.

Evalution.












Logged

Bob Olhsson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3968
Re: Pro Tools: I was wrong. Yep. Wrong.
« Reply #108 on: May 28, 2005, 12:00:54 PM »

According to some of my nerd friends most native floating point processing is actually executed with 80 bit precision so, provided the developer has half a brain, the numbers game is pretty meaningless except for how the data is sent and returned from the host application. I've seen references to this by the Samplitude folks too.

Far more important is watching out for routines that can't handle input above digital zero. Not as dense a sonic mine field as a fixed-point system but you still can't assume floating point means infinite head room.

Giovanni Speranza

  • Full Member
  • ***
  • Offline Offline
  • Posts: 240
Re: Pro Tools: I was wrong. Yep. Wrong.
« Reply #109 on: May 29, 2005, 07:16:23 AM »

Floating point has a huge advantage on fixed point. The resolution of the signal is always the same, independently of it's gain. So as during a mix there are tons of gain changes, floating point wins hands on over fixed point.

henchman

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 549
Re: Pro Tools: I was wrong. Yep. Wrong.
« Reply #110 on: May 29, 2005, 12:10:38 PM »

I don't care how much theory is onvolved.
All I know, is that I find mixing in floating point much easier and find myself not having to worry so much about clipping any of the signal path.

KHStudio

  • Newbie
  • *
  • Offline Offline
  • Posts: 9
Re: Pro Tools: I was wrong. Yep. Wrong.
« Reply #111 on: May 30, 2005, 11:11:02 PM »

I've been following this thread and have learned a lot but I still need input. I've had an enlightening experience with dithering and have started a new thread here:

http://recforums.prosoundweb.com/index.php/t/5487/9619/?SQ=a 8509bc55f577c8f71691238e7c24b19

Please check it out
Logged

yhomas

  • Newbie
  • *
  • Offline Offline
  • Posts: 18
Re: Pro Tools: I was wrong. Yep. Wrong.
« Reply #112 on: November 08, 2005, 09:21:12 PM »

Nika Aldrich wrote on Sat, 21 May 2005 16:22

Fergal Toohey wrote on Sat, 21 May 2005 05:29

Erm... I've been following this thread for the past half a month, since it started, and I don't recall Nika ever saying that having 24bitfixed busses is better than having 32bitfloat busses...  


This one is probably project-dependent.  There are probably cases where each would shine.  24-bit fixed point and properly dithered busses would have an inherent benefit over 32-bit floating point and inadequately dithered busses except for in situations wherein the distortion buildup could be kept lower than the dither noise of the 24 bit system.  And the requirements of that type of project would be too specific and complicated to elucidate upon.

Nika


---

Nika, or others, I have a question on the 24-bit fixed TDM vs the 32-bit Native issue:

Hypothetical situation:
Take a 24 bit (fixed) input file with peak of X dB below full scal (0dB).  Import the file into a track and insert a gain reducing plugin such that the output is Y dB lower than the input.  Using the mixer, raise the gain so that the peak value is 0 dB.  Store the output back to a 24-bit file.

In the above example, would a TDM (fixed point) system or a Native system achieve better results for various values of X and Y.

(We should assume that the TDM plugin and Native plugin as well as the TDM mixer and Native mixer use 48 bit precision.  The only difference is in the 24-bit fixed TDM bus vs the 32-bit float Native bus.  Also, we should assume that when possible, the proper dithering is always done.)

---

The TDM processing will result in a higher noise floor but no trunctation distortion.  The the Native processing will result in a lower noise floor but more truncation distortion.

I assume that for small enough values of X and Y, the TDM processing will generally be superior; so my question is this:  At what point (values of X and Y) will the Native system become superior?

It seems to me that with values of around 6 dB for X and Y, the Native system would already be superior, but I am not really sure.

Any comments?
Logged

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: Pro Tools: I was wrong. Yep. Wrong.
« Reply #113 on: November 09, 2005, 07:47:41 AM »

yhomas wrote on Tue, 08 November 2005 21:21



Nika, or others, I have a question on the 24-bit fixed TDM vs the 32-bit Native issue:

Hypothetical situation:
Take a 24 bit (fixed) input file with peak of X dB below full scal (0dB).  Import the file into a track and insert a gain reducing plugin such that the output is Y dB lower than the input.  Using the mixer, raise the gain so that the peak value is 0 dB.  Store the output back to a 24-bit file.

In the above example, would a TDM (fixed point) system or a Native system achieve better results for various values of X and Y.




Well, there is no such thing as a "24 bit fixed" buss in the fixed point systems such as Pro Tools HD or TDM. EXCEPT WHEN YOU ARE GOING TO AND FROM THE PLUGIN ENVIRONMENT, where the buss width has been reduced to 24 bis. So if you are just changing gain and using the 48 bit mixer, I'd say that for normal signal ranges and levels, the sonic results of the native (32 float) mixer and the fixed point mixer would be very similar. For very abnormal signal levels and gains, the floating point mixer (Native) might be superior sonically, as long as you created a normal level signal at the end of the chain before converting to fixed point 24 bit.

But getting back to your question about the gain reducing plugin. In my opinion, the native system would be superior as you maintain the full range of the 32 bit float numbers and do not have to convert. But if you used a well-designed dithering double precision plugin to reduce gain in the TDM system and did not do this too many generations, you would probably not notice any degradation.

Years ago, I noticed a "coldness" in the sound when boosting gain by 10 dB in one fixed-point DAW system. I did some very careful comparisons and listening comparisons and found that using the Waves Ren EQ as a booster sounded better than using the DAW itself. This situation has improved and I'm talking about 10 year old digital technology.

BK
Logged
There are two kinds of fools,
One says-this is old and therefore good.
The other says-this is new and therefore better."

No trees were killed in the sending of this message. However a large number of
electrons were terribly inconvenienced.

yhomas

  • Newbie
  • *
  • Offline Offline
  • Posts: 18
Re: Pro Tools: I was wrong. Yep. Wrong.
« Reply #114 on: November 09, 2005, 11:38:18 AM »

It is an honor to speak with you Bob.

I totally agree about the mixer, which is irrelevant.  The only reason that I mentioned it is that at the end, I want the output audio to use all 24 bits.

My post is attempting to quantify the tradeoff of using a 24-bit fixed bus vs a 32-bit float bus between plugins.

Nika's dither paper establishes that it is not possible (or at least impractical) to perfectly dither to floating point.  I suspect that most respectable Native plugins will internally use a higher precision than 32-bit float, but when the audio exits the plugin, it will be truncated back to 32-bit float (with either no dither or an ineffective dither).  As I understand, this truncation will cause a loss of information (distortion).

Similarly, the fact that the output of TDM plugins is always 24-bits fixed results in a loss of information if the level of the audio is sufficiently low.  For example, a TDM plugin that reduces the gain by 48dB results in 8 bits of audio being moved from above the noise floor to below the noise floor.

In my opinion, it is pretty clear that if using values of 24 dB for "X" and 24 dB for "Y", the Native 32-bit float processing would prove superior.  However, I don't have a rough idea of at what point the 24-bit fixed takes the upper hand.

The last 7 pages of discussion are very interesting, but in my opinion, the end result (of 32-bit vs. 24-bit being analogous to dictionary vs. bycicle) is not very satisfying or useful.

I am trying to craft a scenario where we can quantify (in a rough, primitive way) the relative strengths/weaknesses of the 24-bt fixed TDM plugin output vs. the 32-bit float Native plugin output--thus giving a more concrete, satisfying (and possibly useful) conclusion to this discussion.

Logged
Pages: 1 ... 6 7 [8]   Go Up
 

Site Hosted By Ashdown Technologies, Inc.

Page created in 0.054 seconds with 19 queries.