Is dithering a must?

Have an urge to learn, or a calling to teach? Want to share some useful Youtube videos? Do it here!
Post Reply
RobC
Posts: 1833
Joined: 10 Mar 2018

25 Apr 2019

Some say, do it, so the lowest levels don't get distorted.
Others say there's no need for it, cause tops engineers can detect that distortion with special devices.

...and what's the point in case of the loudness war participants. There's barely such a thing like low level audio in those materials.

Recorded (analog) music has its own noise; so does even purely electro if it goes onto vinyl for example (cause there's vinyl noise anyway).

Maybe you do electro, but in a dynamic way - still, all made with synthesis, so it will be clean. Won't the noise add more dirt than attempting to bring out what the listener won't hear anyway?

I've heard so many different professional opinions on this matter, that now I'm unsure.

User avatar
Marco Raaphorst
Posts: 2504
Joined: 22 Jan 2015
Location: The Hague, The Netherlands
Contact:

25 Apr 2019

dithering on 16 bit is a safety net. it's very hard to hear differences between dithered material and without dithering but I personally won't take the risk and will always add dithering to 16 bit files. noise is harmless.

User avatar
Loque
Moderator
Posts: 11175
Joined: 28 Dec 2015

25 Apr 2019

I would lie if i say i ever heard a difference. There is probably a theoretically negative impact to the sound without it. And i guess you know, that yous should dither only once.
Reason12, Win10

User avatar
nooomy
Posts: 543
Joined: 16 Jan 2015

25 Apr 2019

RobC wrote:
25 Apr 2019
Some say, do it, so the lowest levels don't get distorted.
Others say there's no need for it, cause tops engineers can detect that distortion with special devices.

...and what's the point in case of the loudness war participants. There's barely such a thing like low level audio in those materials.

Recorded (analog) music has its own noise; so does even purely electro if it goes onto vinyl for example (cause there's vinyl noise anyway).

Maybe you do electro, but in a dynamic way - still, all made with synthesis, so it will be clean. Won't the noise add more dirt than attempting to bring out what the listener won't hear anyway?

I've heard so many different professional opinions on this matter, that now I'm unsure.
It is a must if you are going to upload the music as mp3 or any other compressed format, like spotify, soundcloud, itunes.

Dithering is used to randomize quantization error, and makes your music sound more correct when you convert it to a lossy compression format Like mp3

https://en.wikipedia.org/wiki/Dither

RobC
Posts: 1833
Joined: 10 Mar 2018

28 Apr 2019

Thank you all, for the info! While I had the feeling that only the final export should be dithered, I still did wonder that in case it could save a little something that gets lost when rendering a synthesizer sound, then what if even smaller parts of a song would need it, but from the looks of it, not necessary for high quality audio.

In case of lossy formats, I wonder if it saves the stereo damage that happens, especially with mp3. Or at least I once checked a youtube vinyl rip's 'side channel', and instead of a clean sound (you normally hear), it was a total mess. But that may be just the encoding work.

All in all, I guess the masters can optionally stay clean; or be dithered, while what gets encoded, can benefit from dithering, thus it should be applied for that.

User avatar
moneykube
Posts: 3447
Joined: 15 Jan 2015

28 Apr 2019

interesting info... I have never used it yet lol :puf_wink:
https://soundcloud.com/moneykube-qube/s ... d-playlist
Proud Member Of The Awesome League Of Perpetuals


User avatar
Marco Raaphorst
Posts: 2504
Joined: 22 Jan 2015
Location: The Hague, The Netherlands
Contact:

28 Apr 2019

but then again: break the rules :)

RobC
Posts: 1833
Joined: 10 Mar 2018

30 Apr 2019

I didn't use it before either, thinking it's just for some 8 Bit oddness, but I see there's more and more to it.

Read the article ~ never thought about how the distortion can build up - and considering how much I will re-sample during sound design, I'll definitely have to consider using it.

I'm not a by-the-book type of person, but I don't rebel against it either, just look for the most efficient possibilities. I need to be careful though where I dither, cause some odd effects, such as what I called 'inverted clipping' where anything below a set threshold gets distorted - there the dither noise will needlessly color the sound, achieving the opposite of what I want (or at least I think so ~ need to test that). So in some rare cases, the quantized "sterile" sound might be preferred.

User avatar
Marco Raaphorst
Posts: 2504
Joined: 22 Jan 2015
Location: The Hague, The Netherlands
Contact:

30 Apr 2019

good test is use a sine wave, set volume low and export it to 16 bits. without dithering it doesn't sound good. dithering can totally change that.

RobC
Posts: 1833
Joined: 10 Mar 2018

30 Apr 2019

I understand that, and I'm pretty sure I will use dithering whenever I'm completely certain that I won't be using any low level clipping effect, cause that would amplify the dither noise otherwise.

RobC
Posts: 1833
Joined: 10 Mar 2018

24 Sep 2019

There is a method to apply the dithering noise only, if it falls below the threshold where it would get quantized, but I guess that would have a slight gating effect to it.

User avatar
Chizmata
Posts: 918
Joined: 21 Dec 2015
Contact:

27 Sep 2019

careful people, dithering is useless if it is NOT the absolute last step in the production. dont dither when you export from reason and use the file in another DAW again.

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

27 Sep 2019

RobC wrote:
30 Apr 2019
I understand that, and I'm pretty sure I will use dithering whenever I'm completely certain that I won't be using any low level clipping effect, cause that would amplify the dither noise otherwise.
As others keep correctly repeating, dither is the LAST thing done to the audio signal. You cannot amplify the dither noise if you're adding it last.

If you're going to do ANYTHING to the file later in your workflow, don't add dither (and don't reduce bit depth either).

This is one rule I wound't suggest can be "creatively" broken, unless you want to creatively add noise (or additional low level harmonics from quantization errors). Otherwise, the choice to dither or not IS a creative option, but like always, you should be aware of the ramifications of each decision you make and then make (or break) them intentionally. :)
Selig Audio, LLC

RobC
Posts: 1833
Joined: 10 Mar 2018

27 Sep 2019

Chizmata wrote:
27 Sep 2019
careful people, dithering is useless if it is NOT the absolute last step in the production. dont dither when you export from reason and use the file in another DAW again.
Honestly, that super low level distortion doesn't matter to me, so I probably won't dither ~ except for mp3 and sites that compress in any way - be it streaming, youtube, sound cloud, and alike...
selig wrote:
27 Sep 2019
As others keep correctly repeating, dither is the LAST thing done to the audio signal. You cannot amplify the dither noise if you're adding it last.

If you're going to do ANYTHING to the file later in your workflow, don't add dither (and don't reduce bit depth either).

This is one rule I wound't suggest can be "creatively" broken, unless you want to creatively add noise (or additional low level harmonics from quantization errors). Otherwise, the choice to dither or not IS a creative option, but like always, you should be aware of the ramifications of each decision you make and then make (or break) them intentionally. :)
Alright, but what do I do with the fact that Reaktor only wants to record in 16 bit in stand alone mode?

User avatar
Kalm
Posts: 554
Joined: 03 Jun 2016
Location: Austin
Contact:

27 Sep 2019

It is anytime you're changing between bit depths due to the removal of your lowest bit space, effectively they removal of your quietest sounds. Hence why it won't make much of a difference when editing. You're constantly changing your bottom line (See what I did there). So adding it as the last step ensures that your lowest bit is properly heard when conversion happens and doesn't just karate chop off your last 1 or 0. many people don't do it from 32+ F -> 24 due to other jargon I can go in but that's a jist.
Courtesy of The Brew | Watch My Tutorials | Mac Mini Intel i7 Quad-Core | 16 GB RAM | Samsung 850 EVO 250 GB | Reason 11 Suite | Studio One 5 Professional | Presonus Quantum | Komplete Kontrol 49 MK2 | Event Opals | Follow me on Instagram

RobC
Posts: 1833
Joined: 10 Mar 2018

29 Sep 2019

Kalm wrote:
27 Sep 2019
It is anytime you're changing between bit depths due to the removal of your lowest bit space, effectively they removal of your quietest sounds. Hence why it won't make much of a difference when editing. You're constantly changing your bottom line (See what I did there). So adding it as the last step ensures that your lowest bit is properly heard when conversion happens and doesn't just karate chop off your last 1 or 0. many people don't do it from 32+ F -> 24 due to other jargon I can go in but that's a jist.
Now that you pointed the bottom line out, I needed to restart reading the sentence about the lowest bit. x D

I can understand the case of 24 bits and not dithering, mainly due to twice the possible levels compared to 16 bits. Then again, the dithering noise level would be so low in case of 24 bits, that it can't hurt to add.

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

29 Sep 2019

RobC wrote:
27 Sep 2019
Alright, but what do I do with the fact that Reaktor only wants to record in 16 bit in stand alone mode?
Don't use Reaktor to record important audio that will be processed further! Not sure what use case there would be for using Reaktor as a front end for sampling - how are you using it? Surely there's another way? These days I more often run Reaktor in Reason because of all the advantages (for me, at least) of having Reason and Reaktor working together. Would that work out for what you're doing?
Selig Audio, LLC

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

29 Sep 2019

RobC wrote:
29 Sep 2019
I can understand the case of 24 bits and not dithering, mainly due to twice the possible levels compared to 16 bits. Then again, the dithering noise level would be so low in case of 24 bits, that it can't hurt to add.
Two things…

#1: 24 bit audio does hot have "twice the possible levels" as 16. First, bit depth affects only the Dynamic Range of a digital audio signal.16 bit audio has 96 dB dynamic range, 24 bit has 144 dB dynamic range (48 dB more, not 96 dB more). So 24 bit has 1.5x the dynamic range as 16 bit.

#2: The dither noise would not be present on the 24 bit file, so it would not be "so low in the case of 24 bits". What actually happens when you add dither is you add it at the DESTINATION level, so when adding dither to 24 bit audio in preparation for a bit reduction to 16 bits, the dither signal is added at -90dBFS, or at the level of the last bit, which covers the range from 90-96dBFS (each bit is 6dB).

It is important for dither to remain at the level of the lowest bit for it to "do it's thing", which is why you should not further process audio after adding dither, since it would no longer be at the proper level to have the desired effect.
Selig Audio, LLC

RobC
Posts: 1833
Joined: 10 Mar 2018

29 Sep 2019

selig wrote:
29 Sep 2019
Don't use Reaktor to record important audio that will be processed further! Not sure what use case there would be for using Reaktor as a front end for sampling - how are you using it? Surely there's another way? These days I more often run Reaktor in Reason because of all the advantages (for me, at least) of having Reason and Reaktor working together. Would that work out for what you're doing?
I think I mentioned once, that I want to create a synthesizer that samples itself (I heard there were digital hardware synths that did this). It would be great if with the push of a button, a synthesizer would record, or render a couple of seconds of its own generated, static signal.
If I want to rework a sample multiple times, I think it needs oversampling, so it won't end up suffering too much damage (especially when it comes to pitch-based effects).
As far as I know, I can tops render a 24 bit @ 192000 Hz file in Reason.

So, that self-sampling would definitely speed things up, if there's any possibility.
...and I'm not sure how much processing it takes until even a seemingly highly detailed file like 24/192 turns into a lo-fi mess.
selig wrote:
29 Sep 2019

Two things…

#1: 24 bit audio does hot have "twice the possible levels" as 16. First, bit depth affects only the Dynamic Range of a digital audio signal.16 bit audio has 96 dB dynamic range, 24 bit has 144 dB dynamic range (48 dB more, not 96 dB more). So 24 bit has 1.5x the dynamic range as 16 bit.

#2: The dither noise would not be present on the 24 bit file, so it would not be "so low in the case of 24 bits". What actually happens when you add dither is you add it at the DESTINATION level, so when adding dither to 24 bit audio in preparation for a bit reduction to 16 bits, the dither signal is added at -90dBFS, or at the level of the last bit, which covers the range from 90-96dBFS (each bit is 6dB).

It is important for dither to remain at the level of the lowest bit for it to "do it's thing", which is why you should not further process audio after adding dither, since it would no longer be at the proper level to have the desired effect.
Oh... Welp. I did learn that first one, but forgot. Thanks for correction.

Now, that exact info about dithering (how it's done), I didn't know or read about at all, but makes total sense now!

So, going down from 32 bits would add noise at -138 dBFS I presume ~ which is what I meant, that it would definitely not be bothersome.

Post Reply
  • Information
  • Who is online

    Users browsing this forum: No registered users and 6 guests