avatar

dandan

Silverado

Posts: 74 Member Since: 02/02/2012

Lead

Dec 6 16 8:51 PM

Tags : :

Hi, haven't been around here for a while, nice to see some familar faces. The title may be a little misleading, or not, bear with me, I dither......
Quite often to always, when I hear the tracks in context, I want to revise a mix. I need to replace just the audio, while keeping the placement, Title, ISRC etc. 
So it's revising mixes, sometimes for sonics, sometimes just for level,  not really dithering.....
Then to the real dither. My  understanding is that a level change, say in a DAW,  will expand the bit depth to the full depth that the software is capable of. 
So how does Pro deal with getting a gain changed track back down again to 16 Bit for CD? 
Ta in advance, DD

Last Edited By: dandan Dec 6 16 8:54 PM. Edited 1 time

Quote    Reply   
avatar

zmix

Aqua Marine

Posts: 4,285 Member Since:20/01/2011

#1 [url]

Dec 7 16 12:22 AM

When the word length is changed from 16 bits to 24 (or 32 or 48 or 64, etc) bits each additional bit is actually an additional Least Significant Bit, or LSB,  and not a MSB, so it does not affect the "gain" of the signal at all.  In fact, the upper bits of a PCM audio word are identical regardless of the word length, the only difference between 8, 16, 24 (etc) bits is that the noise level is (theoretically) lower.

When the signal is converted back from a longer to a shorter word length, for example from 24 bits to 16 bits, the lower 8 bits are removed by truncation.

This does not affect the resulting signal level, it simply reduces the ability to represent very small values accurately, effectively raising the level of what we call quantization error.

Adding dither eliminates any distortion to the signal caused by the truncation, effectively removing the quantization error.

Since the dither is typically around 1/2 the level of the LSB it is largely inaudible.

I hope this clarifies things for you..

-CZ

Last Edited By: zmix Dec 7 16 8:12 AM. Edited 2 times.

Quote    Reply   
avatar

chrisj

Platinum Blonde

Posts: 1,008 Member Since:22/02/2011

#2 [url]

Dec 8 16 5:16 AM

zmix wrote:

Since the dither is typically around 1/2 the level of the LSB it is largely inaudible.


Eh? That seems odd.

Chris Johnson, airwindows.com

Quote    Reply   
avatar

zmix

Aqua Marine

Posts: 4,285 Member Since:20/01/2011

#3 [url]

Dec 8 16 8:45 AM

chrisj wrote:
zmix wrote:

Since the dither is typically around 1/2 the level of the LSB it is largely inaudible.


Eh? That seems odd.
It does, and it's a simplification for the purpose of explanation. 
Dither isn't a steady state signal,  it is random, so if, for example,  you're adding 8 bits of dither (completely hypothetically) to a 24bit word  using TPFD, etc,  the average level of the  remaining dither in the 16 bit output is roughly ~1/2 LSB

.

Quote    Reply   
avatar

drknob

Gold Finger

Posts: 788 Member Since:02/02/2011

#4 [url]

Dec 8 16 9:34 AM

Dither is applied before truncation. I believe TPDF and RPDF are typically LSB only while Gaussian dither is LSB and LSB+1. It's function is to decorrelate quantization error by transforming it to noise.
Lavry on dither.
Wikipedia.
And for the brave, the seminal Vanderkooy & Lipshitz paper: Dither in Digital Audio.
Lavry is really good at making digital audio understandable: Lavry White Papers.

Harold Kilianski

Music Industry Arts
Fanshawe College

Quote    Reply   
avatar

bob olhsson

Aqua Marine

Posts: 3,352 Member Since:25/01/2011

#7 [url]

Dec 8 16 4:00 PM

Almost all signal processing expands bit depth to as close to infinity as the hardware and software are capable of. The signal then needs to be dithered back to the desired file or stream bit depth. It's always a choice between noise and distortion. Previously truncated noise can not act as dither preventing distortion. Tape hiss or console noise also can not act as dither and prevent distortion.

Lots of nonsense gets posted about this.

www.audiomastery.com Bob's room 615 562-4346 georgetownmasters.com Georgetown Masters 615 254-3233 www.thewombforums.com

Quote    Reply   
avatar

zmix

Aqua Marine

Posts: 4,285 Member Since:20/01/2011

#8 [url]

Dec 8 16 6:11 PM

bob olhsson wrote:
Almost all signal processing expands bit depth to as close to infinity as the hardware and software are capable of. The signal then needs to be dithered back to the desired file or stream bit depth. It's always a choice between noise and distortion. Previously truncated noise can not act as dither preventing distortion. Tape hiss or console noise also can not act as dither and prevent distortion.

Lots of nonsense gets posted about this.

Have a look at Nigel Redmon's  video demonstration at about 6 minutes in...


?t=369



Quote    Reply   
avatar

jerry tubb

Gold Finger

Posts: 659 Member Since:06/02/2011

#11 [url]

Dec 9 16 9:26 AM

dandan wrote:
Hi, haven't been around here for a while, nice to see some familar faces. The title may be a little misleading, or not, bear with me, I dither......
Quite often to always, when I hear the tracks in context, I want to revise a mix. I need to replace just the audio, while keeping the placement, Title, ISRC etc. 
So it's revising mixes, sometimes for sonics, sometimes just for level,  not really dithering.....
Then to the real dither. My  understanding is that a level change, say in a DAW,  will expand the bit depth to the full depth that the software is capable of. 
So how does Pro deal with getting a gain changed track back down again to 16 Bit for CD? 
Ta in advance, DD

Make the gain change to the 24-bit file, Before it's been reduced to 16-bit.
i always keep the 24-bit session available for any revisions.
then after any re-adjustments are made, do your bit depth reduction & dither to 16-bit.
best, JT 

Quote    Reply   
avatar

dandan

Silverado

Posts: 74 Member Since:02/02/2012

#12 [url]

Dec 9 16 1:47 PM

Safe

Yes Jerry, that is exactly what I do, to be safe/sure. I am hoping Pieter will spot this thread and tell us how DDP Pro functions with regard to Dithering. It is an option. 
It would be a great great thing for me if I could do final volume control within DDP, AND truncate and apply best quality Dither when outputting 16 for CD. 
DD

Quote    Reply   
avatar

jerry tubb

Gold Finger

Posts: 659 Member Since:06/02/2011

#13 [url]

Dec 9 16 10:36 PM

dandan wrote:
Yes Jerry, that is exactly what I do, to be safe/sure. I am hoping Pieter will spot this thread and tell us how DDP Pro functions with regard to Dithering. It is an option. 
It would be a great great thing for me if I could do final volume control within DDP, AND truncate and apply best quality Dither when outputting 16 for CD. 
DD

Different workflow here. I use Sonic soundBlade for project assembly, final level adjustments, dithering & 16-bit DDP generation.
then I use Sonoris DDP creator for the final DDP checking, wave file exports, and disc burns.
best, JT

Quote    Reply   
avatar

dcollins

Platinum Blonde

Posts: 2,373 Member Since:27/01/2011

#14 [url]

Dec 10 16 1:31 AM

zmix wrote:
bob olhsson wrote:
Who is Nigel Redmon? I go by Jim Johnston's words. He's the real expert.

What did JJ say on the subject of self noise and dither?
Probably that it doesn't de-correlate both moments as TPDF of the proper level does.

Quote    Reply   
avatar

zmix

Aqua Marine

Posts: 4,285 Member Since:20/01/2011

#15 [url]

Dec 10 16 2:31 PM

dcollins wrote:
zmix wrote:
bob olhsson wrote:
Who is Nigel Redmon? I go by Jim Johnston's words. He's the real expert.

What did JJ say on the subject of self noise and dither?
Probably that it doesn't de-correlate both moments as TPDF of the proper level does.

Yes, I'm sure, however, in Nigel Redmon's second video (posted above) he audibly demonstrates the residual difference signal at various truncation levels,  and asserts that 24 bit audio really doesn't require any dither, (which most of us know already).

I've been experimenting with bog standard white noise at ~1/2 LSB and found that it's effective as dither, next I'll try "proper" TPDF...



Quote    Reply   
avatar

chrisj

Platinum Blonde

Posts: 1,008 Member Since:22/02/2011

#17 [url]

Dec 10 16 6:28 PM

zmix wrote:

bob olhsson wrote:Who is Nigel Redmon? I go by Jim Johnston's words. He's the real expert.

What did JJ say on the subject of self noise and dither?

Same thing I would say: no such thing. Noise in the source is part of the signal. Dither is very specific additional degrees of noise, typically TPDF (two distinct one-bit ranged random sources added or subtracted).

Back in the day I blind tested full-intensity noise versus low-bit full intensity noise. I could tell it 10 out of 10. I think it might have been Bruno Putzeys who told me to TPDF dither the noise (but it could have been JJ, or Bob O.)

I dithered the noise, and then I couldn't tell the difference at all.

Self dither is an oxymoron, it just doesn't exist: especially if the noise level is GREATER than you'd need for flat dither. And flat dither isn't really technically correct, TPDF is the one to look at for technically correct performance.

Chris Johnson, airwindows.com

Quote    Reply   
avatar

dcollins

Platinum Blonde

Posts: 2,373 Member Since:27/01/2011

#18 [url]

Dec 11 16 1:53 AM

zmix wrote:
dcollins wrote:
zmix wrote:
bob olhsson wrote:
Who is Nigel Redmon? I go by Jim Johnston's words. He's the real expert.

What did JJ say on the subject of self noise and dither?
Probably that it doesn't de-correlate both moments as TPDF of the proper level does.

Yes, I'm sure, however, in Nigel Redmon's second video (posted above) he audibly demonstrates the residual difference signal at various truncation levels,  and asserts that 24 bit audio really doesn't require any dither, (which most of us know already).

I've been experimenting with bog standard white noise at ~1/2 LSB and found that it's effective as dither, next I'll try "proper" TPDF...



In RPDF you'll still have some noise-modulation.

Quote    Reply   
avatar

zmix

Aqua Marine

Posts: 4,285 Member Since:20/01/2011

#20 [url]

Dec 11 16 9:17 AM

Certainly noise can "self dither" and yes pure white noise can also dither.

When we're talking about TPDF, yes, it works "better" in the sense that it can accomplish the same function as random white noise, but at a largely inaudible level.

However, (as cited in the demonstration) once you have a word width that has a lower noise floor than the source by several bits, dither isn't helping in any meaningful way.

Quote    Reply   
Add Reply

Quick Reply

bbcode help