The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Good dither practices, what are yours? Dynamics Plugins
Old 7th January 2017
  #1
Gear Guru
 
Karloff70's Avatar
 

Good dither practices, what are yours?

Hey gents,

I have been following some chat by Bob Olhsson and Chris from Airwindows in another thread about dither and it brings to my attention the fact I am a blithering dither moron of truncating proportions. I would like that to change, and I think this here High End section is a good place to ask who dithers how in their workflow. Should be very interesting.

I'll go first. I have basically spent an awful long time in audio and to be embarrassingly honest never given a ****e about dither.

I am curious, do you guys put dither plugs at the end of a chain of plugs when bouncing that in place to audio? Do you put dither on the end of a chain involving hardware insert plugs to bounce to audio?

Who dithers where, and why? Go......
Old 7th January 2017
  #2
Gear Addict
 
PVJesper's Avatar
In the same way as you, the threads following airwindows dither releases have been a wake up call for me, regarding dither. I have always only used it for when bouncing 24 bit projects as 16 bit. But having had it explained to me by chris and bob, it seems blindingly obvious why it's neccessary even at the same bit depth when applying processing. Im not that unfamiliar with the inner workings of DSP myself, so i feel i should have been able to figure it out even on my own a long time ago.

I don't have so much to contribute in the way of good dither practices except that i'm about to do follow them. Well, its never too late to improve.
Old 7th January 2017
  #3
Quote:
Originally Posted by Karloff70 View Post
Hey gents,

I have been following some chat by Bob Olhsson and Chris from Airwindows in another thread about dither and it brings to my attention the fact I am a blithering dither moron of truncating proportions. I would like that to change, and I think this here High End section is a good place to ask who dithers how in their workflow. Should be very interesting.

I'll go first. I have basically spent an awful long time in audio and to be embarrassingly honest never given a ****e about dither.

I am curious, do you guys put dither plugs at the end of a chain of plugs when bouncing that in place to audio? Do you put dither on the end of a chain involving hardware insert plugs to bounce to audio?

Who dithers where, and why? Go......
Which thread? I'd like to read it (never use dither here - usually leave that for the ME)
Old 7th January 2017
  #4
Gear Guru
 
Karloff70's Avatar
 

Quote:
Originally Posted by [email protected] View Post
Which thread? I'd like to read it (never use dither here - usually leave that for the ME)
This one:

https://www.gearslutz.com/board/new-...-pc-vst-2.html

For me it was Bob's post no41 that threw the live cat spinning into the flock of pigeons.
Old 7th January 2017
  #5
t_d
Lives for gear
 
t_d's Avatar
this is an eye-opening presentation on digital audio, especially the part on dither (which begins around 12:00.. but do spend 25 minutes and watch the whole thing)


Old 7th January 2017
  #6
Lives for gear
 
apartment dog's Avatar
 

Quote:
Originally Posted by Karloff70 View Post
This one:

https://www.gearslutz.com/board/new-...-pc-vst-2.html

For me it was Bob's post no41 that threw the live cat spinning into the flock of pigeons.
Good idea to start this thread, Karloff
I was also amazed to read this thread;

Kazrog releases masterDither

I resumed the info for myself;

Use Dither;

-When rendering a mix to 24 or 16 bit
-When sending audio out to a D/A converter to use analog equipment
-On the mastertrack when monitoring
-When recording audio while inserting software plugins in the recording
-When rendering tracks (like vsti's) to 24 bit

The last one I found here, from Ian Shepherd;

http://productionadvice.co.uk/when-to-dither/

'And if you’re bouncing tracks internally, to remove plugins and reduce CPU overhead – you should definitely dither ! (Unless you’re saving at 32-bit floating point)'


I think not many people know about this.

I will read the thread about Airwindows NaturalizeDither and follow this one.
Old 7th January 2017
  #7
Gear Guru
 
Karloff70's Avatar
 

Quote:
Originally Posted by apartment dog View Post
Good idea to start this thread, Karloff
I was also amazed to read this thread;

Kazrog releases masterDither

I resumed the info for myself;

Use Dither;

-When rendering a mix to 24 or 16 bit
-When sending audio out to a D/A converter to use analog equipment
-On the mastertrack when monitoring
-When recording audio while inserting software plugins in the recording
-When rendering tracks (like vsti's) to 24 bit

The last one I found here, from Ian Shepherd;

When should you use dither ? - Production Advice

'And if you’re bouncing tracks internally, to remove plugins and reduce CPU overhead – you should definitely dither ! (Unless you’re saving at 32-bit floating point)'


I think not many people know about this.

I will read the thread about Airwindows NaturalizeDither and follow this one.
lol. Dither why? Because too much certainty. Love it!

Thank you for the great contribution! This is getting interesting.
Old 7th January 2017
  #8
Lives for gear
 
toledo3's Avatar
 

The only time you ever need to think about noise shaping dither is when *you* are reducing word length. End of story, thread's done, heh

When mixers internally jump up to a higher bit depth, they also use a proper non noise shaping dither to output the bit depth you're working in. You don't have to do anything, and *should not* do anything extra to out think the DAW.
Old 7th January 2017
  #9
Gear Maniac
 

Dither before any change to fixed bit depth, to DAC output or a reduced file (always the case if you use any ITB processing, including gain changes).

So for most people that's before hardware inserts and on master for monitoring/bounce.

As Bob likes to say, as soon as you introduce any ITB processing, you no longer have a fixed bit depth. Calling a project 24 bit for example is misleading.

apartment dog covers other cases above.
Old 7th January 2017
  #10
Lives for gear
 
toledo3's Avatar
 

Quote:
Originally Posted by js230 View Post
Dither before any change to fixed bit depth, to DAC output or a reduced file (always the case if you use any ITB processing, including gain changes).

So for most people that's before hardware inserts and on master for monitoring/bounce.

As Bob likes to say, as soon as you introduce any ITB processing, you no longer have a fixed bit depth. Calling a project 24 bit for example is misleading.
Not really - because even if the DSP bumps word length up to 32, 56, 64, the mixer itself will output 24, if you're working in 24. You will still be listening to 24 bits (less really by the time your D/A handles it). At least with any major DAW I've used.

The only reason it bumps up is to ensure that your 24 bit dynamic range will be untouched. It's not so now you can have some mega wide dynamic range, or anything like that. It's just a coding trick, to give you equivalent output dynamic range as you had at the start.
Old 7th January 2017
  #11
Lives for gear
 
JulenJVM's Avatar
I´ve also learnt to dither everytime there´s truncation, and that includes working at 24 bit in a 32 bit float/64 bit DAW. The threads on the Airwindows / Kazrog dithers have been very revealing. I was amazed about the effect of dither on the mixing stage, in fact, so amazed that I did a little comparison of the Airwindows dithers in this thread. There´s a link to the wav files in case the mp3 don´t offer enough detail for the effect of the different dithers to be noticed.
Old 7th January 2017
  #12
Lives for gear
 
toledo3's Avatar
 

Quote:
Originally Posted by JulenJVM View Post
I´ve also learnt to dither everytime there´s truncation
What does this functionally mean to you?

When do you perceive that truncation is typically occurring in your DAW, where the DAW doesn't add a non noise shaped dither, aside from bouncing your 24 bit session down to a 16 bit stereo mix?
Old 7th January 2017
  #13
Lives for gear
 
JulenJVM's Avatar
Quote:
Originally Posted by toledo3 View Post
What does this functionally mean to you?

When do you perceive that truncation is typically occurring in your DAW, where the DAW doesn't add a non noise shaped dither, aside from bouncing your 24 bit session down to a 16 bit stereo mix?
I´m not trained enough to discern when there is truncation distortion going on, but I´ll get there eventually. What happend is that I was working on a 24 bit mix when I came accross the threads that I mentioned, and found them very informative. Up to now, I had only dithered when exporting for CD format, so I gave a try to applying 24 bit dither to my mix (working in Cubase, so the audio is calculated at 32 bit float point).

I felt the change instantly, the guitars´position in space seemed a bit more defined all of a sudden, the lows felt warmer and it felt as if everything had a bit more detail to it. I´m not sure if this helps, but I thought the mix sounded a bit more analogue. And that was working at 24 bit, with 24 bit dither applied.
Old 7th January 2017
  #14
Dithering is really just one half of technically correct word-length truncation. IMHO it makes absolutely no sense to separate it from the truncation process, it just leads to (harmless) confusion.

It's at the DAW's responsibility to dither when needed. This is the case when it writes the ASIO buffers (being then sent to the fixed point audio output), or any other fixed point export. To my knowledge, all modern DAWs do. This process is well defined technically, the optimal dither is a "triangular probability density function" (TPDF) with an amplitude of exactly 2 bits. No secret, fully proven! This is the only method that guarantees, over the full audible band, to eliminate any nonlinearity originating from the truncation. BTW, TPDF is just a ridiculously snobbish name for "roll two dice and take the average".

Any deviation from the flat TPDF curve (which really is no more than an EQed TPDF) will either increase noise at certain freqs, or reduce noise - at the cost of slight nonlinearities (added partials originating from truncation) over these regions. Nothing's free.

My personal opinion about the matter: While dithering is a vital part of truncation, making it look like rocket science or a mystery is comparable to silver audio cables, directional cables, astrology and other snake oil. If there's any benefit, it can be demonstrated easily, in a reproducible, ABX-able manner. Take note how few really do this, if they don't, you can safely expect that they can't.

Last edited by FabienTDR; 8th January 2017 at 01:47 AM..
Old 8th January 2017
  #15
Gear Guru
 
Karloff70's Avatar
 

Thanks for all the contributions guys, keep it coming, all angles on this are interesting to me.

One thought of mine is, sure, the DAW should likely do this in situations where necessary, but are we trusting that? Or is there a benefit in slipping something like any of the Airwindows dithers into workflows still? I expect the answer would totally vary depending on DAW myself, but that is guessing based on how much faith I have in different DAWs and the impression of priorities they give me.....
Old 8th January 2017
  #16
Quote:
Originally Posted by Karloff70 View Post
One thought of mine is, sure, the DAW should likely do this in situations where necessary, but are we trusting that?
A good question. And certainly worth further investigation.

We can simply do a test similar to the one shown in Monty's video, i.e. take a simple sine wave, truncate and analyze the results via spectrum analyzer.

To me, this would be much like a DAW's SRC forgetting to bandlimit the incoming signal before decimation (and relying on all sorts of third party Nyquist filter plugins instead). IMHO A no-go. But who knows.
Old 8th January 2017
  #17
Dither in the high end.. love it..

Here's a decent vid:

Old 8th January 2017
  #18
Lives for gear
 
grrrayson's Avatar
 

I'm highly skeptical and I think the issue is overblown.

I don't think Mr. Olhsson is suggesting you insert a dither plug-in after every single processing plug-in. (But hey, try it and see if you like the sound better.)

I am under the impression that DAWs (and A/D and D/A converters) dither when they need to, as dither is an integral part of digital audio being able to work.

I've done tests where I could clearly hear a tonal difference between dithers. It's subtle yet noticeable; however, in my experience it's minimal and of lowest priority.

Sure, if you use dither that shifts energy into the high range, it's basically like adding a little EQ down at -80dB. If I wanted that I could just nudge an actual EQ further up the chain. I could spend an hour auditioning every possible noise shaping curve on every song to see which fits the music best but I don't.

These days I just select flat TPDF when it's an option (e.g. the "Save" dialog in Izotope RX) and move on.

Quote:
Originally Posted by Karloff70 View Post
sure, the DAW should likely do this in situations where necessary, but are we trusting that?
With respect (I generally enjoy your posts here a lot, by the way!), this is ridiculous—do you trust that your 16-bit saved file is really 16 bits and you aren't being jipped a bit or two? Or what if you select 320kbps for a client email ref mp3—do you worry that Pro Tools might generate a 240kbps mp3 instead? Do you worry whether or not your 24-bit session is really a 24-bit session? Of course not. Programmers overlook things occasionally as all humans do but in general the software we use is rather established in concept.

And moreover to my last point of whether it's worth worrying about—LISTEN. Compare. Do you hear a difference? If so, great—modify your workflow accordingly. If not, then you need not worry about it.
Old 8th January 2017
  #19
Lives for gear
 

FabienTDR answered the general question.

As for when I use dither -- well, when truncating bit depth between plugins, mix bus/mixer, etc. the system takes care of itself (if it doesn't, it's time to go to something else.) I leave everything in 32bit FP on output of the final file out of ProTools or Cubase. If I need to send it on truncated, I do the work in RX Advanced because I prefer the noise-shaping and truncation there (plus I like to see the spectrograph of the post truncation file).

I leave everything untruncated as long as possible until delivery requirements require otherwise at which point dither and truncation happen.
Old 8th January 2017
  #20
Gear Guru
 
Karloff70's Avatar
 

Quote:
Originally Posted by grrrayson View Post

With respect (I generally enjoy your posts here a lot, by the way!), this is ridiculous—do you trust that your 16-bit saved file is really 16 bits and you aren't being jipped a bit or two? Or what if you select 320kbps for a client email ref mp3—do you worry that Pro Tools might generate a 240kbps mp3 instead? Do you worry whether or not your 24-bit session is really a 24-bit session? Of course not. Programmers overlook things occasionally as all humans do but in general the software we use is rather established in concept.

And moreover to my last point of whether it's worth worrying about—LISTEN. Compare. Do you hear a difference? If so, great—modify your workflow accordingly. If not, then you need not worry about it.
Thank you, and with equal respect back, of course not. But I also do not believe that all DAWs deal with things like this in the identical and identically effective manner. For example Samplitude have been bragging forever about doing certain things right, keeping the audio in a better state than others. Enough about that as I don't want this to go to the dark DAWs sounding different place, but .......I think some are better than others at 'keeping happy'.

Couldn't agree more about just listen....lol. My preferred method. I will indeed do some experiments of my own, this thread isn't instead of that, it is curiosity about everyone's habits in regard to dither is all.
Old 8th January 2017
  #21
Lives for gear
 

plugs like sonnox at least introduces dither with the processing, it helps a little, not that i ever use plugins mind you.
Old 8th January 2017
  #22
Lives for gear
 
Silvertone's Avatar
We had this discussion over at PRW. While we rarely talk about audio on that forum we spent 11 pages talking about something no one can hear. I found that quite funny.

Daniel Weiss told me many, many years ago that it is okay to dither a source many times. After that I did not worry about dither so much. When someone of Daniels caliber tell you it's okay, then it's okay.
Old 8th January 2017
  #23
Gear Guru
 
Karloff70's Avatar
 

Quote:
Originally Posted by Silvertone View Post
We had this discussion over at PRW. While we rarely talk about audio on that forum we spent 11 pages talking about something no one can hear. I found that quite funny.

Daniel Weiss told me many, many years ago that it is okay to dither a source many times. After that I did not worry about dither so much. When someone of Daniels caliber tell you it's okay, then it's okay.


So your position is, if in doubt do one more dither, won't do any harm?
Old 8th January 2017
  #24
Lives for gear
 
Silvertone's Avatar
Quote:
Originally Posted by Karloff70 View Post


So your position is, if in doubt do one more dither, won't do any harm?
Ha ha, very funny.

I've been in this for way too long to have any "doubts" left... about anything.

I dither when appropriate.

www.silvertonemastering.com
Old 8th January 2017
  #25
Lives for gear
 
Plush's Avatar
What most engineers do not do is dither a 24 bit signal when it is being saved as a 24 bit signal after processing.

They should be using dither then, however. As cited in several posts this should be a TPDF at minimum. Also recommended are the POW-R # 3 and the Prism SNS #2 .

There is no penalty for multiple dithers.

Gain changes, bounces, renderings, using plug ins, eq's, ---all require dither to 24 bits again.

Output from your stereo buss to a d/a converter does NOT require dither. The reason is that your signal would have already been dithered when you followed good practice as recommended above.

There is no signal degradation or added jitter in a digital transfer such as sending to a d/a converter.
Old 8th January 2017
  #26
Lives for gear
 
toledo3's Avatar
 

The mixer itself will be adding a non noise shaped TPDF dither when reducing from 32 bit or greater for processing, when working in a 24 bit session.

A proper DAW program (pretty much all of them at this point in time) does not require you to keep adding dither just because you inserted plugins.

You should definitely NOT add noise shaped dither. The results of adding a noise shaped dither many times, IS actually audible, and not in a good way. POW-3, in particular, will add a bright edge to program material, especially if you mistakenly keep adding it successive times.
Old 8th January 2017
  #27
Gear Maniac
 

Quote:
Originally Posted by toledo3 View Post
Not really - because even if the DSP bumps word length up to 32, 56, 64, the mixer itself will output 24, if you're working in 24. You will still be listening to 24 bits (less really by the time your D/A handles it). At least with any major DAW I've used.

The only reason it bumps up is to ensure that your 24 bit dynamic range will be untouched. It's not so now you can have some mega wide dynamic range, or anything like that. It's just a coding trick, to give you equivalent output dynamic range as you had at the start.
We're not talking about after every plugin.

Point that Bob continually makes is that you have truncation when you reduce from the higher word length in the examples described above, even if you are working and listening and bouncing in 24 bit. The audible penalty from too much dither is insignificant compared to distortion from truncation, so unless you know for certain a path through converters or capture to fixed bit does this for you, best to dither.
Old 8th January 2017
  #28
Lives for gear
 
JulenJVM's Avatar
Quote:
Originally Posted by toledo3 View Post
The mixer itself will be adding a non noise shaped TPDF dither when reducing from 32 bit or greater for processing, when working in a 24 bit session.

A proper DAW program (pretty much all of them at this point in time) does not require you to keep adding dither just because you inserted plugins.

You should definitely NOT add noise shaped dither. The results of adding a noise shaped dither many times, IS actually audible, and not in a good way. POW-3, in particular, will add a bright edge to program material, especially if you mistakenly keep adding it successive times.
The thing is that I´ve been checking Cubase 8.5 Pro Manual, and it doesn´t say anything about the DAW applying dither automatically to renders below 32 bit float (see attached capture below)

Quote:
Originally Posted by Cubase Manual
Dither effects allow you to control the noise produced by quantization errors that can occur when you mix down to a lower resolution.
Dithering adds a special kind of noise at an extremely low level to minimize the effect of quantization errors. This is hardly noticeable and much preferred to the distortion that otherwise occurs.
NOTE
As Cubase internally uses 32-bit float resolution, all integer resolutions are lower. However, the negative effects are most noticeable when mixing down to 16-bit format or lower.
It then goes on to explaining how to use the UV22 plugin to properly add dither. So in theory, there´s quantization noise being added when working at 24 bit in Cubase, unless the user actively adds dither. It might not be very noticeable, but if one renders at lot of tracks, this truncation distortion can pile up and take a lot of life out of the track.

If you want a non obtrusive dither, have a look at Paul Dither or Naturalize from Airwindows.
Attached Thumbnails
Good dither practices, what are yours?-cubase-dither.png  
Old 8th January 2017
  #29
Lives for gear
 
toledo3's Avatar
 

I have to admit, I haven't used Cubase in years. Always best to read the manual. However, if you're working at a 24 bit resolution in the project settings, there is a big likeliness that the mixer itself adds TPDF to bring it to 24 bits, after the DSP steps and the gain adjustment at the fader.

TPDF is the textbook, industry agreed upon non obtrusive dither, with all due respect to Airwindows.
Old 8th January 2017
  #30
Lives for gear
 
JulenJVM's Avatar
Quote:
Originally Posted by toledo3 View Post
I have to admit, I haven't used Cubase in years. Always best to read the manual. However, if you're working at a 24 bit resolution in the project settings, there is a big likeliness that the mixer itself adds TPDF to bring it to 24 bits, after the DSP steps and the gain adjustment at the fader.

TPDF is the textbook, industry agreed upon non obtrusive dither, with all due respect to Airwindows.
I´m trying to confirm about the dither on Cubase, but what I´m reading so far indicates otherwise.

About the Airwindows plugins, you mentioned that dither becomes audible when applied accross many tracks, that´s when those plugins come in.
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