Old 11-26-2010, 05:01 PM   #1
Fabian
Human being with feelings
 
Fabian's Avatar
 
Join Date: Sep 2008
Location: Sweden
Posts: 5,028
Default Dither vs Noise shaping

I do not really understand this, so I'm asking...

When rendering to audio CD, I have the choice of dithering as well as noise shaping. What's the difference?

Dither, I though I understood (stochastic noise below audible threshold to simulate higher bit depth, or something like that), but "noise shaping"?
__________________
// MVHMF
Vote for the Starter Marker FR here http://forum.cockos.com/project.php?issueid=4371
Let's finally get rid of those ugly odd time measures.
Fabian is offline   Reply With Quote
Old 11-26-2010, 05:06 PM   #2
Analogy
Human being with feelings
 
Join Date: Sep 2009
Posts: 870
Default

Noise shaping filters the spectrum of the noise used for dithering to approximate the equal loudness curve of your ear. Basically, it's intended to make the dithering noise less audible.
Analogy is offline   Reply With Quote
Old 11-26-2010, 05:09 PM   #3
Fabian
Human being with feelings
 
Fabian's Avatar
 
Join Date: Sep 2008
Location: Sweden
Posts: 5,028
Default

Quote:
Originally Posted by Analogy View Post
Noise shaping filters the spectrum of the noise used for dithering to approximate the equal loudness curve of your ear. Basically, it's intended to make the dithering noise less audible.
OK, thanks.
So why would I not want to use noise shaping?
__________________
// MVHMF
Vote for the Starter Marker FR here http://forum.cockos.com/project.php?issueid=4371
Let's finally get rid of those ugly odd time measures.
Fabian is offline   Reply With Quote
Old 11-26-2010, 05:12 PM   #4
timlloyd
Human being with feelings
 
Join Date: Mar 2010
Posts: 4,713
Default

You only want to use noise shaping when dithering to your final output bit-depth. If you're dithering stems or converting individual tracks to a lower bit-depth, you should use standard TPDF dither.

This is just what I've gleaned from research about various things in the past. I could be wrong, but as I understand it, noise-shaping does not de-correlate truncation distortion harmonics completely, while TPDF dither does.

edit-

I'll just add to this............I honestly do not understand why any DAW still gives the user the option to not dither when truncating to a lower bit-depth....................it's pointless. We need dither. All the time in digital audio. It suffers from being surrounded by some amount of confusion. It should not be an option. When it needs to happen in order to retain the mathematical audio fidelity of a signal, it should happen.........no options, it should just happen.

anyway..........that's enough internet for me for the evening!

Last edited by timlloyd; 11-26-2010 at 05:32 PM.
timlloyd is offline   Reply With Quote
Old 11-26-2010, 09:30 PM   #5
DBMusic
Human being with feelings
 
DBMusic's Avatar
 
Join Date: Sep 2009
Location: Illinois
Posts: 1,181
Default

Basically, noise shaping algorithms allow you to shift dithering noise to less obtrusive (i.e. higher frequency) areas of the audio spectrum.

In standard pop/rock/county/etc. recordings, this is rarely an issue as any dithering noise, if audible at all, would be masked. However, in something like a very dynamic classical recording, noise shaping may be a preferable option (it really depends on what best compliments the music).

Regards,

DB
__________________
My Stuff
DBMusic is offline   Reply With Quote
Old 11-26-2010, 10:16 PM   #6
Analogy
Human being with feelings
 
Join Date: Sep 2009
Posts: 870
Default

Quote:
Originally Posted by timlloyd View Post
I honestly do not understand why any DAW still gives the user the option to not dither when truncating to a lower bit-depth
Because you might be using a plugin that is adding its own dither, in which case you wouldn't want the DAW to be adding its own dither.

In any case, the importance of dither is VASTLY overstated. Even at 16 bits, there's really very very little benefit to using dither. I dare you to actually manage to ABX the difference between typical pop/rock material with and without dither.
Analogy is offline   Reply With Quote
Old 11-26-2010, 10:50 PM   #7
Xenakios
Human being with feelings
 
Xenakios's Avatar
 
Join Date: Feb 2007
Location: Oulu, Finland
Posts: 6,518
Default

Quote:
Originally Posted by timlloyd View Post
I honestly do not understand why any DAW still gives the user the option to not dither when truncating to a lower bit-depth
Because dither is useless (what it does is inaudible) in any real-life scenario, so better give the user an option to save some CPU cycles.
__________________
For info on SWS Reaper extension plugin (including Xenakios' previous extension/actions) :
http://www.sws-extension.org/
https://github.com/Jeff0S/sws
--
Xenakios blog (about HourGlass and λ) :
http://xenakios.wordpress.com/
Xenakios is online now   Reply With Quote
Old 11-27-2010, 05:10 AM   #8
timlloyd
Human being with feelings
 
Join Date: Mar 2010
Posts: 4,713
Default

Quote:
Originally Posted by Xenakios View Post
Because dither is useless (what it does is inaudible)
Perhaps, I think all professional mastering engineers would disagree with that statement though. It's not useless, it's an integral part of any digital system.

Particularly if you were to truncate a whole mix's worth of audio tracks to 16-bit (for some reason)...........if you did not dither, once you had EQ-ed and compressed them individually and in busses; I think if you then replicated the process exactly with properly dithered audio files, you would be able to hear the difference between the two mixes.

Analogy's point about using a plug-in is a fair one! However, if dither really is as pointless as you're both suggesting, then why bother with a plug-in? Wouldn't it still be better for it to just happen in the background automatically? What's the point in saving a negligible amount of cpu use when rendering?
timlloyd is offline   Reply With Quote
Old 11-27-2010, 05:55 AM   #9
Mich
Human being with feelings
 
Join Date: May 2009
Posts: 1,265
Default

Quote:
Originally Posted by timlloyd View Post
Perhaps, I think all professional mastering engineers would disagree with that statement though. It's not useless, it's an integral part of any digital system.

Particularly if you were to truncate a whole mix's worth of audio tracks to 16-bit (for some reason)...........if you did not dither, once you had EQ-ed and compressed them individually and in busses; I think if you then replicated the process exactly with properly dithered audio files, you would be able to hear the difference between the two mixes.
I think what he meant to say is that for 16bit (or in this case 24bit) you won't hear it. If you can then please take the challenge here http://www.ethanwiner.com/dither.html

However if Xenakios would render his favorite songs to 8bit once dithered once without dither he would have to admit that it is not useless and I think the fact that the same principles that apply to 8bit also apply to 16bit or more bits make dither universally valid and not useless.
Whether or not one can hear the outcome is a different matter. One best described with the analogy of whether to paint your whole room or leave the spots where your furniture is unpainted because no one will be able to see behind it anyway.

Quote:
Originally Posted by timlloyd View Post
Wouldn't it still be better for it to just happen in the background automatically?
REAPERs dither isn't the best in the world, so wouldn't it be better to have the user decide which dither to use?
Further more what if I would want to use REAPER to convert file e.g. from wav to aiff without any processing should REAPER then dither again or automatically recognize that no processing took place and disable it? Or wouldn't it be great that I as the end user had full control over everything?
Mich is offline   Reply With Quote
Old 11-27-2010, 06:11 AM   #10
timlloyd
Human being with feelings
 
Join Date: Mar 2010
Posts: 4,713
Default

Quote:
Originally Posted by Mich View Post
what if I would want to use REAPER to convert file e.g. from wav to aiff without any processing should REAPER then dither again or automatically recognize that no processing took place and disable it?
If no bit-depth truncation takes place, neither should dither. If it does, then it should. They're both pcm files, if you don't change the bit-depth they will be bit-identical on playback.

Quote:
REAPERs dither isn't the best in the world, so wouldn't it be better to have the user decide which dither to use?
Interesting. So it's not possible to hear 16-bit dither anyway, but the dither in REAPER isn't as good as some other inaudible dither, so you would prefer to have the option to choose?

If REAPER's dither is not as good as other algorithms out there, then ultimately it needs to be improved imho. Well integrated Red-Book PQ encoding and POW-R dither/ns are things I would love to see in REAPER
_________________________

Believe me, I know what you're saying about the perceptual difference between 16-bit dither and 16-bit truncation distortion. However, if I have the option to click once and fundamentally improve the fidelity of my audio, I'm going to do it. Even if you can't hear it under certain circumstances...........it is better dithered than not.
timlloyd is offline   Reply With Quote
Old 11-27-2010, 06:41 AM   #11
Mich
Human being with feelings
 
Join Date: May 2009
Posts: 1,265
Default

Quote:
Originally Posted by timlloyd View Post
Interesting. So it's not possible to hear 16-bit dither anyway, but the dither in REAPER isn't as good as some other inaudible dither, so you would prefer to have the option to choose?
Yes, I would prefer to have the option to choose what dither I don't hear. Remember there are differences between different dithers, thus the best dither varies with the material it is applied to.
EDIT: I may be using the term dither loose here because in fact dither is just plain white noise and is identical (though even there can be differences based on the source of the randomness of the noise) but what makes a real difference is the noise shaping: http://audio.rightmark.org/lukin/dither/dither.htm (whether that can be heard or not shouldn't matter because it is undeniable different).

Quote:
Originally Posted by timlloyd View Post
Even if you can't hear it under certain circumstances...........it is better dithered than not.
No arguing with that, but if someone (for whatever reasons) doesn't want to dither why force it on him?
Mich is offline   Reply With Quote
Old 11-27-2010, 06:43 AM   #12
timlloyd
Human being with feelings
 
Join Date: Mar 2010
Posts: 4,713
Default

Well..............if that is the case, then they either don't know what dither is, or what they may actually want is a bit-crusher.
timlloyd is offline   Reply With Quote
Old 11-27-2010, 07:15 AM   #13
Dstruct
Human being with feelings
 
Dstruct's Avatar
 
Join Date: Jul 2006
Location: Dresden, Germany
Posts: 11,996
Default

Quote:
Originally Posted by Fabian View Post
I do not really understand this, so I'm asking...

When rendering to audio CD, I have the choice of dithering as well as noise shaping. What's the difference?

Dither, I though I understood (stochastic noise below audible threshold to simulate higher bit depth, or something like that), but "noise shaping"?
http://forum.cockos.com/showthread.php?t=23324
Dstruct is offline   Reply With Quote
Old 11-27-2010, 08:12 AM   #14
schwa
Administrator
 
schwa's Avatar
 
Join Date: Mar 2007
Location: NY
Posts: 9,156
Default

Independent of the endless issues around when or whether to dither, REAPER's dither and noise shaping algorithms are pretty good. Not to put words in his mouth but I don't think Mich was saying that REAPER's dither is bad, I think he was saying that it's good for the user to be able to choose to use the dithering algorithm in some other mastering plugin (which might be the "best in the world", or just better for the specific application) if they want to.

One comment I'd make is that both dither and noise shaping add energy to the audio signal: dither adds white noise-like energy to mask artifacts, and noise shaping adds energy to shift the frequency spectrum of the artifacts. All else being equal, you should only apply this energy if you need to. Exactly when you "need to" is the subject of a lot of debate, but in my opinion the user should actively decide to apply dither or noise shaping, rather than having it applied automatically.

Last edited by schwa; 11-27-2010 at 10:20 AM.
schwa is offline   Reply With Quote
Old 11-27-2010, 08:22 AM   #15
timlloyd
Human being with feelings
 
Join Date: Mar 2010
Posts: 4,713
Default

Ok fair enough! It's good to hear your opinion.
timlloyd is offline   Reply With Quote
Old 11-27-2010, 10:08 AM   #16
probez
Human being with feelings
 
probez's Avatar
 
Join Date: Oct 2010
Location: Paris, FR
Posts: 13
Default

For those interested in this matter i remind this link :

http://www.ethanwiner.com/dither.html

It also contains a link to a discussion about it in 3daudio forums and it's also worth reading imo.

I'll also give this advice : do not apply dither to a track that might be sent to a mastering studio, they will certainly use their own (even on 16bits files... sigh...) and the combination of both might result in a loss in the higher frequencies because of phase issues between the two algos and your tune.
__________________
REAPER | NI Komplete 7 | MOTU Traveler MK3 | MACKIE MR8 | i7 950 @ 4.1Ghz | 6gB Ram
probez is offline   Reply With Quote
Old 11-27-2010, 12:01 PM   #17
Analogy
Human being with feelings
 
Join Date: Sep 2009
Posts: 870
Default

People spend way too much time worrying about dither for the benefit it actually delivers. We're talking about a signal 96 dB below full scale here. Even at relatively high monitoring gain (which would blow your ears off if you actually listened to any of the music you would be using dither on), your computer fan is probably a good 20-30 dB louder than the quantization noise or dither. -96 dB is *really fucking quiet*.

Oh, and by the way, unless you have a plugin that's adding dither, you're not hearing any sort of dithering while you mix. Reaper has to convert to integer to get the audio into ASIO, and your audio driver/hardware is truncating to whatever the bit depth of its DACs is. You've been monitoring with quantization noise, run for the fucking hills.
Analogy is offline   Reply With Quote
Old 11-27-2010, 12:44 PM   #18
timlloyd
Human being with feelings
 
Join Date: Mar 2010
Posts: 4,713
Default

Quote:
Originally Posted by Analogy View Post
Oh, and by the way, unless you have a plugin that's adding dither, you're not hearing any sort of dithering while you mix......(snip)........You've been monitoring with quantization noise, run for the fucking hills.
Well yes.........however............I monitor at 24-bit, and always record and mix with 24-bit files. So in my case it's not an issue if there is no playback dither, because the truncation is only to 24-bit...............that's perhaps a little different to monitoring at 16-bit with no real-time dither. I'm running Apple CoreAudio at 24-bit on my laptop, so for monitoring there will be no even vaguely-audible real-time truncation going on unless I do it on purpose. Perhaps it's different with Asio, I don't know.

Regardless, it doesn't take time to worry about dither Me, pedantic? Yes. Does it help in this industry to be pedantic? Sometimes, yes.

Fabian, I apologise for starting an argument in your thread. I hope you've had your question answered satisfactorily.

Last edited by timlloyd; 11-27-2010 at 12:55 PM.
timlloyd is offline   Reply With Quote
Old 11-27-2010, 02:19 PM   #19
Fabian
Human being with feelings
 
Fabian's Avatar
 
Join Date: Sep 2008
Location: Sweden
Posts: 5,028
Default

Quote:
Originally Posted by Dstruct View Post
Ah! The oracle... Somehow the search missed that one... Thanks.

Quote:
Originally Posted by timlloyd View Post
Fabian, I apologise for starting an argument in your thread. I hope you've had your question answered satisfactorily.
No worries... interesting read.
__________________
// MVHMF
Vote for the Starter Marker FR here http://forum.cockos.com/project.php?issueid=4371
Let's finally get rid of those ugly odd time measures.
Fabian is offline   Reply With Quote
Old 12-04-2010, 04:05 PM   #20
Simple Simon
Human being with feelings
 
Simple Simon's Avatar
 
Join Date: Nov 2009
Location: New Zealand
Posts: 186
Default

Quote:
Originally Posted by Analogy View Post
People spend way too much time worrying about dither for the benefit it actually delivers. We're talking about a signal 96 dB below full scale here.
This comes close to my understanding of dither as being most useful in music with wide dynamics and/or a lot of "space". I did an experiment a few years back in which I recorded a single acoustic guitar chord as it decayed. I rendered it from 24 bit 16 bits twice - once with dither and once without. What I found what that the tail end of the delay on the dithered version was noticeably smoother on the dithered version (which sounded a bit grainy), although I had to have them both playing at quite a high level before it was apparent.

My theory, although I have never tested it, is that this smoothing of low-level audio information would probably impact on sound-field imaging in recordings such as some classical, jazz and acoustic, but probably matters less with more electric/electronic and heavily produced sounds. I'd be interested in the thoughts of others on this.

Last edited by Simple Simon; 12-05-2010 at 12:21 AM.
Simple Simon is offline   Reply With Quote
Old 12-04-2010, 08:05 PM   #21
nfpotter
Human being with feelings
 
Join Date: Feb 2009
Location: Portland, OR, USA
Posts: 994
Default

Quote:
Originally Posted by Simple Simon View Post
This comes close to my understanding of dither as being most useful in music with wide dynamics and/or a lot of "space". I did an experiment a few years back in which I recorded a single acoustic guitar chord as it decayed. I rendered it from 24 to bit bits twice - once with dither and once without. What I found what that the tail end of the delay on the dithered version was noticeably smoother on the dithered version (which sounded a bit grainy), although I had to have them both playing at quite a high level before it was apparent.

My theory, although I have never tested it, is that this smoothing of low-level audio information would probably impact on sound-field imaging in recordings such as some classical, jazz and acoustic, but probably matters less with more electric/electronic and heavily produced sounds. I'd be interested in the thoughts of others on this.
Pretty much my understanding as well.

That said, I always dither when going from 24 to 16 bit (I record everything at 24/44.1K). Even very "thick" music. It's just habit, and I've never noticed dither doing anything "bad" to my mixes.
__________________
"A fly was very close to being called a land, 'cause that's what they do half the time."

-Mitch Hedberg
nfpotter is offline   Reply With Quote
Old 12-28-2013, 09:42 AM   #22
DuX
Human being with feelings
 
DuX's Avatar
 
Join Date: Jan 2007
Location: Underworld
Posts: 1,081
Default

I have a little question about dithering that I couldn't find an answer to by googling it because all I get is dithering and noise shaping threads, which is understandable. I googled "dither noise and hardware noise difference".

I understand dithering and noise shaping very well and I have no problems with that. I usually use TPDF dithering and a bit of noise shaping when needed [quieter material] on the master when truncating to either 24 or 16 bit audio.

However, what often makes me wonder is: if all the channels in the mix are already noisy enough since they've been recorded through a preamp, many with a mike, why would you need to dither when there is already enough noise in the mix? When I gate the channels I usually gate them just down to ~-80dBFS, not to infinity.

In other words, is there any difference between normal noise produced by hardware and dithering noise? To me, it seems logical that since you have more than enough "natural" noise, even at ~-70dBFS on some channels and the audio is more than sufficiently randomised, why would you need to add some more noise to it? Does in that case dithering makes any sense at all?

Cheers!
__________________
Goodnight, thank you, and may your God go with you.

Last edited by DuX; 12-28-2013 at 09:55 AM.
DuX is offline   Reply With Quote
Old 12-28-2013, 10:23 AM   #23
richie43
Human being with feelings
 
Join Date: Dec 2009
Location: Minnesota
Posts: 8,330
Default

Quote:
Originally Posted by DuX View Post
I have a little question about dithering that I couldn't find an answer to by googling it because all I get is dithering and noise shaping threads, which is understandable. I googled "dither noise and hardware noise difference".

I understand dithering and noise shaping very well and I have no problems with that. I usually use TPDF dithering and a bit of noise shaping when needed [quieter material] on the master when truncating to either 24 or 16 bit audio.

However, what often makes me wonder is: if all the channels in the mix are already noisy enough since they've been recorded through a preamp, many with a mike, why would you need to dither when there is already enough noise in the mix? When I gate the channels I usually gate them just down to ~-80dBFS, not to infinity.

In other words, is there any difference between normal noise produced by hardware and dithering noise? To me, it seems logical that since you have more than enough "natural" noise, even at ~-70dBFS on some channels and the audio is more than sufficiently randomised, why would you need to add some more noise to it? Does in that case dithering makes any sense at all?

Cheers!
Awesome question, and I personally believe that there is not a "right" or "wrong" answer. Your assertions make total sense to me, and I have A/B tested dithered vs non-dithered music until it made me insane, and honestly can usually not tell the difference. Render a few of both and listen, or maybe get extra geeky and look at the 2 files on Voxengo Span..... In the end, I say do what makes most sense to you.
__________________
Quote:
Originally Posted by Coachz
Rather than try to carpet the world, put on some shoes.
http://soundyaudio.com/ The Sounds of the Hear and Now
richie43 is offline   Reply With Quote
Old 12-28-2013, 10:27 AM   #24
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 18,518
Default

Quote:
Originally Posted by DuX View Post
In other words, is there any difference between normal noise produced by hardware and dithering noise? To me, it seems logical that since you have more than enough "natural" noise, even at ~-70dBFS on some channels and the audio is more than sufficiently randomised, why would you need to add some more noise to it? Does in that case dithering makes any sense at all?

Cheers!
I would think analog noise would be natural dither provided it never dips below the dither floor. There are likely conditions that sway that a little. I'd be curious in the situation that creates if you count on that analog noise then end up snipping it out in silent spots for example.
__________________
Your whole life people will tell you what you can't do. Getting past them is the first step to actually getting things done.
karbomusic is online now   Reply With Quote
Old 12-28-2013, 10:52 AM   #25
brainwreck
Human being with feelings
 
brainwreck's Avatar
 
Join Date: Jul 2006
Location: The 90's
Posts: 18,009
Default

I wonder if introducing some noise (the right kind) to the signal would help with amp sims...how overdrive just fizzles out when transistioning to clean.
__________________
The media are misleading the public about...pretty much everything.
brainwreck is offline   Reply With Quote
Old 12-28-2013, 10:55 AM   #26
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 18,518
Default

Quote:
Originally Posted by brainwreck View Post
I wonder if introducing some noise (the right kind) to the signal would help with amp sims...how overdrive just fizzles out when transistioning to clean.
I think it would require too much noise to mask if it is already bothering you when the fade is at normal listening levels. If it is literally down where the analog or dither floor is I suppose you could.
__________________
Your whole life people will tell you what you can't do. Getting past them is the first step to actually getting things done.
karbomusic is online now   Reply With Quote
Old 12-28-2013, 11:12 AM   #27
DuX
Human being with feelings
 
DuX's Avatar
 
Join Date: Jan 2007
Location: Underworld
Posts: 1,081
Default

Quote:
Originally Posted by karbomusic View Post
I would think analog noise would be natural dither provided it never dips below the dither floor. There are likely conditions that sway that a little. I'd be curious in the situation that creates if you count on that analog noise then end up snipping it out in silent spots for example.
My thoughts exactly. I'm currently toying with some tests, trying to see [hear! ] the difference with a noisy track [noise @ -80dB RMS] truncated to 16 bit with and without the dithering on the master. Yeah, very important point is to keep the noise always in, above the dither floor. That's why I never gate the channels digitally all the way. In the old times even the gates had noise. LOL

Regarding the digital amps and even reverb, if the audio is digital [no noise] I always like to put some noisy plugin in the chain before going into the FX. Like de la Mancha "imperfection", Klanghelm IVGI, Sonimus Satson, or Voxengo "Insert", for instance. I think it sounds better that way, but I might be just too old school and thinking too much in terms of an analog studio workflow, and I've learned to not care too much about the noise that is down @ ~-70dBFS RMS on the master out. However, I do prefer it a bit lower these days, at between -80 to -90dBs RMS.

In those days, I never thought I would be putting the noise back in some day. LOL But when you think about it, noise is natural. It is all around us all the time. Total, complete silence to -300dB or even -144dB is not natural. Even if put in a sensory deprivation chamber you would hear some natural noise, your blood flowing and heart beating at least.

Cheers!
__________________
Goodnight, thank you, and may your God go with you.

Last edited by DuX; 12-28-2013 at 11:41 AM.
DuX is offline   Reply With Quote
Old 12-29-2013, 02:24 AM   #28
planetnine
Human being with feelings
 
planetnine's Avatar
 
Join Date: Oct 2007
Location: Lincoln, UK
Posts: 7,522
Default

This video by Chris Montgomery of xiph.org is one of the most acccessible that covers dither (and other aspects) in digital audio.

https://www.youtube.com/watch?v=cIQ9IXSUzuM

It explores and amplifies quantisation noise, dither and noise-shaping.

I would add that I do partially disagree with his stance on 24-bit downloads, if only to look to the future with much lower-level audio masters due to loudness normalisation, but in general he very clearly (and humerously) speaks much sense and truth about digital audio.


It dispels more than one myth about the subject and is worth linking to in DAW forums from time to time



>
planetnine is offline   Reply With Quote
Old 12-29-2013, 03:12 AM   #29
gavriloP
Human being with feelings
 
Join Date: May 2008
Posts: 676
Default

Remember that if you are using some tape or other "vintage equipment" emulation plugin you are probably already dithering with it: if there is any hiss in the emulation (like in TB ReelBus or VoS's NastyDLA, for example) it is doing just that among other things.

Dithering may be practically inaudible but if it increases precision, why not use it? Isn't that what's digital is all about? REAPER's dither is fine for me BTW (when I'm not using ReelBus etc.) IF you have to boost that signal later it can become very audible. If you have ever worked with quiet samples that have that digital gargle of no dither, you know what I mean.
gavriloP is offline   Reply With Quote
Old 12-29-2013, 03:21 AM   #30
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 18,518
Default

Quote:
Originally Posted by planetnine View Post
This video by Chris Montgomery of xiph.org is one of the most acccessible that covers dither (and other aspects) in digital audio.


>
Seen it before but don't remember and don't have time to watch it all for a single question. Does a noisy analog line, that is louder than the dither would be, mask truncation or not?
__________________
Your whole life people will tell you what you can't do. Getting past them is the first step to actually getting things done.
karbomusic is online now   Reply With Quote
Old 12-29-2013, 04:53 AM   #31
planetnine
Human being with feelings
 
planetnine's Avatar
 
Join Date: Oct 2007
Location: Lincoln, UK
Posts: 7,522
Default

Quote:
Originally Posted by karbomusic View Post
Seen it before but don't remember and don't have time to watch it all for a single question. Does a noisy analog line, that is louder than the dither would be, mask truncation or not?

I think that depends on the level and the audio content of the noise.

Dither is added digitally because it is in a digital system, but also because it is added at the bit-level required to add "randomness" to the truncation point. If analogue noise is random at the 16-17 bit level, then I guess the answer is yes.

Remember that what we are trying to mask is harmonically related quantisation noise created by truncating tonal patterns, ie the noise components are related to the signal content. If the noise is random at that truncation point then I see no reason to add dither, other than it can't hurt



>
planetnine is offline   Reply With Quote
Old 12-29-2013, 08:51 AM   #32
timlloyd
Human being with feelings
 
Join Date: Mar 2010
Posts: 4,713
Default :)

Correctly applied dither does not work by 'masking' quantization error.

If done correctly it randomises quantization error, de-correlating it from the signal. The error still exists in the 'new' signal, but it is random noise rather than inharmonic distortion.

This is why, with correct dither, is is possible to preserve undistorted signal components that peak below the LSB ... the audibility of such signal components can be improved with noise shaping.

In audio, 'masking' is an entirely different thing. Masking can be relevant when discussing the audibility of signal vs dither noise, but it is not the mechanism by which dither works.
timlloyd is offline   Reply With Quote
Old 12-29-2013, 08:59 AM   #33
richie43
Human being with feelings
 
Join Date: Dec 2009
Location: Minnesota
Posts: 8,330
Default

Ironically, I got just got an email from PSP in regards to a new dithering plugin that they are just now releasing....... I have always used Tonebooster's Barricade since it is usually the final plugin on my mastering chain and have "heard" that it has good dithering/noise shaping algorithms. I have never figured out the need for a dithering-only plugin, especially one with as many options (that I don't understand yet) as this PSP plugin. But maybe these are the sort of options that timlloyd is referring to.
__________________
Quote:
Originally Posted by Coachz
Rather than try to carpet the world, put on some shoes.
http://soundyaudio.com/ The Sounds of the Hear and Now
richie43 is offline   Reply With Quote
Old 12-29-2013, 09:00 AM   #34
planetnine
Human being with feelings
 
planetnine's Avatar
 
Join Date: Oct 2007
Location: Lincoln, UK
Posts: 7,522
Default

It's added to the signal and it replaces it. Ok, masking has specific meaning in audio, I used incorrect terminology.

I was using general language, masking is hiding it, covering it up removing it. De-correlating it is a good enough gobo for me


>
planetnine is offline   Reply With Quote
Old 12-29-2013, 09:04 AM   #35
timlloyd
Human being with feelings
 
Join Date: Mar 2010
Posts: 4,713
Default

Quote:
Originally Posted by planetnine View Post
I was using general language
I know, (sorry for the pedantry) but even schwa typed 'masking' -- that could confuse people.
timlloyd is offline   Reply With Quote
Old 12-29-2013, 09:22 AM   #36
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 18,518
Default

Quote:
Originally Posted by timlloyd View Post
Correctly applied dither does not work by 'masking' quantization error.
Thanks for popping in Tim, however that isn't the question I care about right now. The question is, if the analog hiss from a noisy tube mic for example is reasonably louder than both the digital noise floor and truncation (or lets just call them any digital artifacts that meet said criteria), will it mask the need for dither? The question also is not "should I dither in this case" it is an as-is question and let's not get tripped up on the word mask either.

That can only be yes, no, or depends, which is it provided we know the analog noise is always louder and present? Any mistakes made in terminology while trying to say "yes, no, depends" isn't of our concern.
__________________
Your whole life people will tell you what you can't do. Getting past them is the first step to actually getting things done.

Last edited by karbomusic; 12-29-2013 at 10:07 AM.
karbomusic is online now   Reply With Quote
Old 12-29-2013, 10:30 AM   #37
timlloyd
Human being with feelings
 
Join Date: Mar 2010
Posts: 4,713
Default

Quote:
Originally Posted by karbomusic View Post
Thanks for popping in Tim, however that isn't the question. The question is, if the analog hiss from a noisy tube mic for example is reasonably louder than both the digital noise floor and truncation (or lets just call them any digital artifacts that meet said criteria), will it mask the need for dither?
Depends

Yes -- If the noise has sufficient energy (higher than the LSB of the target precision) at all frequencies that are present in the input signal, there is no need to dither.

If not, the signal will be distorted to some extent.

I don't have any references atm to point you towards to back this up, but as far as I am aware, it is highly unlikely that 'self-dithering' will occur in practice.

It's possible for noise and Dither to sound very similar or even the same, but for the noise to not actually be capable of functioning as Dither. There are many kinds of 'noise' and it takes a specific kind to 100% de-correlate quantization error: TPDF.

Last edited by timlloyd; 12-29-2013 at 10:43 AM.
timlloyd is offline   Reply With Quote
Old 12-29-2013, 10:41 AM   #38
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 18,518
Default

Quote:
Originally Posted by timlloyd View Post
Depends

Yes -- If the noise always has sufficient energy (higher than the LSB of the target precision) at all necessary frequencies (i.e. all frequencies that are present in the input signal), there is no need to dither.

If not, the signal will be distorted to some extent.

I don't have any references atm to point you towards to back this up, but as far as I am aware, it is highly unlikely that 'self-dithering' will occur in practice. It's possible for noise and Dither to sound very similar or even the same, but for the noise to not actually be capable of functioning as Dither.
I'm onboard with depends. It think the only thing above that peaks my interest is:

Quote:
(i.e. all frequencies that are present in the input signal),
I would think thermal and other analog noise in an analog circuit would pretty much cover the all frequencies requirement; I'm talking a tube mic here where the hiss is audibly high enough to be far above dither levels? It is the input signal when the instrument isn't playing. I'll chalk that up to meeting the sufficient energy requirement as well. However, I'm not too worried about it since it's more Dux's overall question than it is mine where for me it's more curiosity.

This is NOT a query to escape the need for dither in general practice, it is strictly academic assuming specific conditions.

Thanks Tim.
__________________
Your whole life people will tell you what you can't do. Getting past them is the first step to actually getting things done.
karbomusic is online now   Reply With Quote
Old 12-29-2013, 10:46 AM   #39
Billoon
Human being with feelings
 
Join Date: Sep 2006
Location: Arse end of the earth.
Posts: 2,980
Default

Here ya go, i did a little test.

My tube pres noise is at -88dB which is above the level of TPDF dither.

I recorded a single track at 24bit then rendered at 16 bit with and without dither (took the opportunity to try out the new PSP dither Richie linked above which is very nice, used preset included).

I've attached difference files, these were obtained by nulling the renders against the original recording - then gaining them both up by 68dB to make them audible. Only included a few seconds because the whole track sounds the same.

Clearly there is no audible inharmonic distortion in the non-dithered file, but the dithered file also benefits from noise shaping which makes it preferable to me.

So, yes you don't need to dither - but there are benefits.

EDIT : FTR, comparing it to flat TPDF dither, the undithered file is the winner - sounds basically identical(mono instead of stereo) but nearly 10dB quieter.
Attached Files
File Type: zip dither again.zip (245.2 KB, 37 views)
__________________
Fortune favours the prepared...

Last edited by Billoon; 12-29-2013 at 07:03 PM.
Billoon is offline   Reply With Quote
Old 12-29-2013, 10:52 AM   #40
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 18,518
Default

Quote:
Originally Posted by Billoon View Post
So, yes you don't need to dither - but there are benefits.
Thanks, I'll check out later.

I've never really concerned myself too much with the to dither or not to dither question. Typically, I would dither because that makes the most sense. In other words, it takes a huge amount of time to examine and confirm that every second of every track I record does or doesn't meet conditions where dither is the most or least helpful.

I've found that even if I took the time to do all of that, I'd eventually miss something, thus the entire exercise is a bit moot unless the material meets rare exceptions on a regular basis. Otherwise, just use dither as intended.
__________________
Your whole life people will tell you what you can't do. Getting past them is the first step to actually getting things done.
karbomusic is online now   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -7. The time now is 09:38 PM.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2017, vBulletin Solutions Inc.