The No.1 Website for Pro Audio
To dither, or not to dither
Old 29th January 2020
  #91
Lives for gear
 
didier.brest's Avatar
Quote:
Originally Posted by David Rick View Post
Here's an illustration of his algorithm for the LCR case.
  • Generate 3 independent RPDF dithers, RA, RB, and RC.
  • For the left channel TPDF dither, use TL = RA + RB
  • For the right channel TPDF dither, use TR = RB + RC
  • For the center channel TPDF dither, use TC = RA + RC


Why not using RA, RB and RC instead of TL, TR, TC ?

Quote:
Originally Posted by David Rick View Post
We've thus made 3 uncorrelated TPDF dithers for cost of 1.5
Unlike RA, RB and RC, TL, TR and TC are correlated (0.5 mutual correlation).
Old 29th January 2020
  #92
Quote:
Originally Posted by didier.brest View Post


Why not using RA, RB and RC instead of TL, TR, TC ?

Unlike RA, RB and RC, TL, TR and TC are correlated (0.5 mutual correlation).
Rx are RPDF, not TPDF.

Possibly you're right about the correlation Didier, but I think Tx are at least first-moment decorrelated. I posted the above algorithm from memory and perhaps I've mixed it up after so many years. Let me go find the original paper and post an update.

David
Old 29th January 2020
  #93
Lives for gear
 
Plush's Avatar
Thank you, David Rick

I would always look forward to reading an elucidation of a topic from Michael Gerzon.

The point is, for me, that using two sophisticated DAWS, SADiE and Sequoia, I have not had any problems with the stereo image (stereo picture) attributable to how dither is applied or attributable to dither. The level of the applied (global) dither is at a very low level.
Old 29th January 2020
  #94
Lives for gear
 
didier.brest's Avatar
Quote:
Originally Posted by David Rick View Post
Rx are RPDF, not TPDF.


Quote:
Originally Posted by David Rick View Post
I posted the above algorithm from memory and perhaps I've mixed it up after so many years.
It might be

TL = RA + RB
TR = RA - RB
TC = RC + RD

Then 3 uncorrelated TPDF (Triangular Probability Density Function) dithers are generated from 4 independent RPDF (Rectangular Probability Density Function) dithers instead of 6 ones in the conventional method for generating 3 independent TPDF dithers:

TL = RA + RB
TR = RC + RD
TC = RE + RF

6/4 would be the 1.5 cost reduction factor that you have mentioned in your post ?
Old 4 weeks ago
  #95
Gear Head
Ok so here's my question regarding dither.
Im running a hybrid rig.
When working on a project, I individually bounce tracks through my analog rig.
I use plugins on the way to the rig and on the daw's input bus on the way back.
It's definitely crucial to use eq to shape signals both before and after the rig which consists of culture vulture, fatso, vitalizer, m-a-s...
Most projects are running at 32bit float/96k
My apogee symphony is outputting at 24bit

Once I've got a pretty good mix happening, I then route the whole mix through my analog chain again and tweak away.

At what points do I need to dither?
Does it make more sense to run audio @ 24bit internally?

Sorry if this is covered already
Thanks in advance for any info
Old 4 weeks ago
  #96
Motown legend
 
Bob Olhsson's Avatar
 

Dither right before the feed to all D. to A. converters. Write 32 float files after anything has been digitally processed.
Old 4 weeks ago
  #97
Gear Head
Quote:
Originally Posted by Bob Olhsson View Post
Dither right before the feed to all D. to A. converters. Write 32 float files after anything has been digitally processed.
One minor complication. I run into my gear sum/difference style instead of LR.
If the dither must be last on the way out, it's going on the mid/side version of the audio which then gets converted back to stereo on the way back in.

Should there be any issue? the dither is meant for stereo files. Does anything get screwed up with the mid/side laden dither when it gets converted back to stereo?
Old 4 weeks ago
  #98
Lives for gear
 

Quote:
Originally Posted by Bob Olhsson View Post
Dither right before the feed to all D. to A. converters.
in practical terms: do you know if anyone does when feeding from a daw to a large analog desk for mixing?

i might give it a try but so far, i could not hear much of a difference between sending from my daw (with or without dither) or from sending from my hd-recorders (without dither) to my d/a converters... - at least it did not bog me down.

also, dithering on the way out to the monitoring? how comes speaker processors do not apply dithering? -
i'm not much conviced...

(i have to admit that i haven't been setting up a comparison of any kind lately)
Old 4 weeks ago
  #99
Motown legend
 
Bob Olhsson's Avatar
 

M/S and stereo get processed the same way so dithering works the same way. Remember that it prevents distortion that is louder than properly implemented dither. Almost all A to D converter chips dither to 24 bits even if they don't offer dithering to lower depths.
Old 4 weeks ago
  #100
Lives for gear
 

Quote:
Originally Posted by Bob Olhsson View Post
Almost all A to D converter chips dither to 24 bits even if they don't offer dithering to lower depths.
d to a converters you mean?

do the chips alone give any hints to what kinda dithering gets used?

and why then dither when converters do it anyway?
Old 4 weeks ago
  #101
Motown legend
 
Bob Olhsson's Avatar
 

No, A to D converters. Dithering must be done prior to the conversion to a fixed point signal in order to prevent the distortion. And yes, it is stupid that we even have to think about fixing people's lazy coding.
Old 4 weeks ago
  #102
Motown legend
 
Bob Olhsson's Avatar
 

Conversion to 24 bit fixed point is built into the A to D chip! It ought to be built into the DAW!
Old 4 weeks ago
  #103
Lives for snowflakes
 
12ax7's Avatar
 

Quote:
Originally Posted by Bob Olhsson View Post
Conversion to 24 bit fixed point is built into the A to D chip! It ought to be built into the DAW!
Reaper will do this!
Old 4 weeks ago
  #104
Motown legend
 
Bob Olhsson's Avatar
 

Samplitude had built-in dither 25 years ago. Pro Tools TDM had it as an option 20 years ago.
Old 4 weeks ago
  #105
Gear Head
Quote:
Originally Posted by Bob Olhsson View Post
Samplitude had built-in dither 25 years ago. Pro Tools TDM had it as an option 20 years ago.
My main DAW is digital performer which has a dither option in one of the dropdown menus. It's just a checkmark beside the word dither, there's no explanation where the dithering happens or what bit rate, nor is there an option for noise shaping.

So anyways I just started using Quan Jr with noiseshaping... lo and behold I notice a difference in the high frequencies! Smoother
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump