The process of adding dither in the studio

Orb

New Member
Sep 8, 2010
3,010
2
0
Hi Bruce,
in some threads we have had interesting discussions relating to aspects of ADC-DAC and I posted some technical links relating to dither to show its importance with regards to resolving transfer function related (quantization "step-staircase") and distortion-noise.
However in many threads and debates one aspect never really discussed is how the dither is implemented within the studio, some articles I posted-linked anecdotally infer that while dither is critical its implementation is also as important, not just the aspect of TPDF but also at what point applied and ensuring limited use of dither (not repeated).
This is compounded that from what I can tell there is no defined standard to its use for studios, and also truncation of 24-bit masters to 16-bit remix-remastering releases.

Any chance you could provide some insight into this, such as what stage it is implemented, studio considerations, state of a universal standard-guideline,etc.
Cheers
Orb
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
First of all, there are no guidelines. I know the Grammy P&E wing makes recommendations.
Most of the workstations today operate natively at 32-bit floating point. There are others that work in 64-bit and I think even in 80+ bit. Most plugins operate natively at 24/96... I think Waves comes to mind. I think there are others that operate at much higher bit-depth and sampling rate such as Algorithmix and Flux. Whenever you use one or many of these plugins, the bit-depth gets changed constantly. Working at such high bit-depts and going back and forth, dither is not used. Whenever these tracks get bounced out of the computer, then dither must always be used. The choices of dither is numerous. You can use everything from Pow-r 1-3, IDIR, Aqua, TPDF, square, noise shaping, triangle, MBIT+ and to further confuse you, it depends on if further processing is going to be applied.
So first rule of thumb is to always use dither when bouncing out of a computer. So this means dither is applied after you record a song, after it is mixed and another after it is mastered. Now see how much noise is added!!
 

Orb

New Member
Sep 8, 2010
3,010
2
0
WOW Bruce,
I never realised it had to be that much, and thanks for clarifying it is not just TPDF used and this surprises me I must admit - I had a feeling there was still no universal rule-guidelines and so I do wonder if part of this is the reason for some unfortunate releases as well from some studios (not talking about you but generally) or compounds it.

BTW just curious do you ever listen to a song being recorded without dither and compared, if so I assume either 32-bit or 64-bit.
Some articles suggest multiple reapplication of dither affects sound (mostly anecdotal reports), and importantly also the type; even if comparing flat TPDF to noise shaped TPDF, let alone the others.

Just to be clear, do any of the ADC modules inject dither or is it always used within the workstation and at certain stages of the recording-mastering cycle?
Talking within the context of how studios would do this normally rather than it being technically possible.

Thanks again.
Orb
 
Last edited:

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
For years Pro Tools had 2 audio engines, a dithered and an undithered engine. You could change which you used deep in the menu. Not a lot of people knew about it. I think they got rid of the choice around PT 8. I don't know, it still may be there.

During recording, I really don't hear a difference in dithered and non-dithered 32-bit or 64-bit engines. But when you start stacking dozens of tracks together, the noise starts to rear it's ugly head, especially if you are using a lot of plugins.

I have in the past not dithering a bounce to 16-bit. Was used mostly for metal/ska music.

I don't believe any ADC's use dither in the sense of word. DSD uses noise shaping which can be considered dither. Also analog tape hiss can be used as dither, since it's random noise.

Most studios that I've known over the years could care less about dither, much less which one! They just use the default of what came with the program and wonder why their mixdowns sound the way they do.
 

Orb

New Member
Sep 8, 2010
3,010
2
0
Thanks Bruce,
real eye opener as I was wondering how much of a headache the "wonder cure" dither could be when comes to implementation, this is definitely something those who discuss or experience with digital signal processing tend to forget.

Thanks again.
Orb
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
Bruce,

You have stated that you usually do not use dither when using Audiogate to convert DSD to PCM. Why not?

DSD already has enough noise... why add more?

Plus... we've done both hundreds of times and prefer no dither!
 

Gumby

New Member
Dec 16, 2012
10
0
1
Seattle, WA
DSD already has enough noise... why add more?

Plus... we've done both hundreds of times and prefer no dither!

That was quick. I suppose I would add noise if I could not hear it and/or the dither provided some benefit that outweighed the effect of the noise. I have not tried it both ways thinking I would not be able to tell a difference, plus it is time consuming. I will accept the results of your experience I think.

Thanks.
 

DonH50

Member Sponsor & WBF Technical Expert
Jun 22, 2010
3,947
306
1,670
Monument, CO
Dither is used in many ADCs. Among other things... But, I do not know commercial audio implementations that well. In RF systems dither is often band-shaped and applied before the ADC, and after the DAC. Some systems add dither before the DAC (in the digital domain). Some add it inside the ADC and/or DAC deep in the circuitry. There are many, many implementations, many proprietary, and I should note it is used in both conventional and delta-sigma converters (ADCs and DACs).
 

opus111

Banned
Feb 10, 2012
1,286
3
0
Hangzhou, China
According to my understanding of dither theory (which is fairly limited) dither is only required when quantization is being performed. So at the input of an ADC and when ever the word length is truncated, dither is called for. DSD -> PCM is a word length expansion, so dither's not required.
 

Orb

New Member
Sep 8, 2010
3,010
2
0
Yeah with Opus; dither needed for word length truncation (PCM->PCM), and also integral to the transfer function/quantization/noise in ADC for PCM.

Just to add though,
also worth noting it also depends upon the bit depth relating to PCM; 24-bit still is not perfect without dither (appreciate you probably know this).
However it is interesting when one goes to 32-bit as in theory dither should not be needed, and this was Bruce's experience when I asked this in this thread; this may be one reason the optimum/transparent transcoding for DSD to PCM is 32-bits and 352.8khz/384khz as per recommendation out of Philips Labs (which Bruno mentions in his own general papers).

This thread came about due to discussion in a different one, where I provided some links showing technically why dither needed in a studio but importantly that it has issues (again wanted Bruce's experience on this as Puget is a great studio) both from being used too often and critically the dither type applied.

Look back at beginning I raised experience on dither and if he noticed if it was required at 32-bit (which DAWs can do and greater),It is pretty clear 24-bit still requires dither (seen from various tests-measurements showing it without/with), but wanted to know experience of using 32-bit and its Bruce's experience is that it is not needed.

Of course real world PCM audio releases are not 32-bit and not as efficient as say DSD when streaming/archiving for ADC-workstation-storage, but crucial is applying the right type of dither and as minimal as possible, as touched upon near beginning of this thread and some others where Bruce and myself discussed it.

Bruce, you had a DXD capable ADC in the studio once?
If so you remember how that handles dither and what options available as that is 24-bit from the ADC?

Also when you had the MSB ADC in, how/when did you apply dither?
Would add a lot to this discussion.
Thanks in advance.
Cheers
Orb
 
Last edited:

Gumby

New Member
Dec 16, 2012
10
0
1
Seattle, WA
According to my understanding of dither theory (which is fairly limited) dither is only required when quantization is being performed. So at the input of an ADC and when ever the word length is truncated, dither is called for. DSD -> PCM is a word length expansion, so dither's not required.

Based on my understanding of dither theory, which is also limited, your reply makes perfect sense.

Thanks.
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
Bruce, you had a DXD capable ADC in the studio once?
If so you remember how that handles dither and what options available as that is 24-bit from the ADC?
Orb

I still do... the DAD AX24. I don't recall it having it's own dither, but instead using the dither of the software Pyramix that it was attached to. I don't see an option for dither.

DAD.jpg
 

Orb

New Member
Sep 8, 2010
3,010
2
0
Excellent that is what I thought; this makes sense because I thought dither was moved sometime ago from being an external source input with/within ADC, to now as you say within the modern studio DAW itself after the ADC.

Cheers
Orb
 

o0o

Active Member
Dec 6, 2015
6
0
31
to dither or not to dither

For years Pro Tools had 2 audio engines, a dithered and an undithered engine. You could change which you used deep in the menu. Not a lot of people knew about it. I think they got rid of the choice around PT 8. I don't know, it still may be there.

During recording, I really don't hear a difference in dithered and non-dithered 32-bit or 64-bit engines. But when you start stacking dozens of tracks together, the noise starts to rear it's ugly head, especially if you are using a lot of plugins.

I have in the past not dithering a bounce to 16-bit. Was used mostly for metal/ska music.

I don't believe any ADC's use dither in the sense of word. DSD uses noise shaping which can be considered dither. Also analog tape hiss can be used as dither, since it's random noise.

Most studios that I've known over the years could care less about dither, much less which one! They just use the default of what came with the program and wonder why their mixdowns sound the way they do.

Hello Bruce!

I've been reading this forum and some others to understand whether I need to dither going from DSD to 24 bit PCM or not.

From what I understand in PCM domain whenever I reduce a bit depth I have to dither to avoid quantization errors. Going upwards doesn't require any dither at all.

My confusion arises when I have to deal with non-PCM streams, like lossy mp3, aac, vorbis, etc and DSD stream since most decoders provide a 32 bit floating point output and any resampling job will also boost the bit depth to 32 or even 64 float point, depending on software used.

From my "research" I figured out that Philips suggests to use 32 bit floating PCM output as default for DSD-to-PCM conversion.

So what's the right "scientific" way of converting DSD to 24 bit PCM dithered or not?

---

Btw did you ever tried the new handy foobar2000 SACD plugin?

I use it for easy DSD to AAC conversion (DSD 352.8 -> SSRC 48 -> AAC TVBR Q127) and it's good enough for very that purpose and no dithering required! :D

Regards,
Dmitry
 

o0o

Active Member
Dec 6, 2015
6
0
31
For years Pro Tools had 2 audio engines, a dithered and an undithered engine. You could change which you used deep in the menu. Not a lot of people knew about it. I think they got rid of the choice around PT 8. I don't know, it still may be there.

During recording, I really don't hear a difference in dithered and non-dithered 32-bit or 64-bit engines. But when you start stacking dozens of tracks together, the noise starts to rear it's ugly head, especially if you are using a lot of plugins.

I have in the past not dithering a bounce to 16-bit. Was used mostly for metal/ska music.

I don't believe any ADC's use dither in the sense of word. DSD uses noise shaping which can be considered dither. Also analog tape hiss can be used as dither, since it's random noise.

Most studios that I've known over the years could care less about dither, much less which one! They just use the default of what came with the program and wonder why their mixdowns sound the way they do.

Hello Bruce.

I'm trying to figure out if I should use dithering converting from DSD to PCM.

From my understanding any bit depth reduction requires dithering to avoid quantization errors. Taking in consideration the 32 bit float point output of most non-PCM decoders and 32/64 bit used in many re-samplers the output will always be > 24 bit hence dithering is a logical step, right?

On the other hand DSD stream by nature has enough low level noise + lots of ultrasonic shaped noise raising above 20-30kHz and it's not really good idea to add more, is it?

So I'm stuck with this dilemma and would love to have some simple straight to the point explanation if you please.

Regards,
Dmitry
 
Last edited:

o0o

Active Member
Dec 6, 2015
6
0
31
According to my understanding of dither theory (which is fairly limited) dither is only required when quantization is being performed. So at the input of an ADC and when ever the word length is truncated, dither is called for. DSD -> PCM is a word length expansion, so dither's not required.
That question was answered in post #10. In short: no. :)
Yes, Don, it was answered in terms of PCM steam but I wouldn't consider going from 1 bit DSD decoded as 32/64 bit float point PCM to 24 bit as an expansion. It's not like 1bit stream wrapped in 32/64 bit container, it's a genuine 32/64 bit floating PCM stream we have to deal with.

Dmitry
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
Yes, Don, it was answered in terms of PCM steam but I wouldn't consider going from 1 bit DSD decoded as 32/64 bit float point PCM to 24 bit as an expansion. It's not like 1bit stream wrapped in 32/64 bit container, it's a genuine 32/64 bit floating PCM stream we have to deal with.

Dmitry

PCM -> PCM, yes use dither

DSD -> PCM, no, use no dither. In layman's terms, dither is noise and DSD has enough noise already. You can try using dither, but in every trial I've done and given to my clients, no dither was always chosen. The jury is still out if you go all the way down to 16bit.
 

o0o

Active Member
Dec 6, 2015
6
0
31
Thank you, Bruce! I agree about noise, DSD has more than enough of it...

So let it be bold an clear converting DSD to PCM requires no dither!
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing