Quantcast
Flat Dither Question Mbit+ - Page 3 - Gearspace.com
The No.1 Website for Pro Audio
Flat Dither Question Mbit+
Old 23rd October 2012 | Show parent
  #61
Lives for gear
 
Cellotron's Avatar
 
Verified Member
3 Reviews written
🎧 15 years
Quote:
Originally Posted by David Rick ➑️
P.S. to Steve Berson: I'm fortunate to have one client who can distinguish between different dithers and noise shapers (better than I can), but I recognize that he's highly unusual.
It's incredibly easy to distinguish between different dither types if you want to - you just need to crank tails and fades up to levels that would make your ears bleed if you played standard scale material at them!

I in fact spent a great deal of hours about 7 years or so ago while working at Europadisk doing exactly that while beta testing the various Sonoris Dither types developed by Pieter Stenekes back then, and doing comparisons with the various competing dither algorithms that were available or being marketed as superior back then. Spending these hours (along with null tests, listening at normal volumes, abx's, etc.) made me understand via actually hearing what dither does to sound - and what it doesn't. And LORDY - there's a heckuva lot of claiming in various forum threads - often by those who should really know better - that dither does things it really doesn't. Does it change the sound of the least significant 1 to 2 (or maybe even up to 4 bits if you apply noise shaping) - certainly. Does it change the sound of the bits above that if the algorithm is not in fact broken?? No, it doesn't.

I'll say this: those who focus on dither options for any more than say 30 seconds per job are in fact showing gross negligence in prioritizing their mastering time. ANYTHING you do to the most significant bits (i.e. eq, comperssion, limiting, etc.) will in fact be of utmost more importance in terms of how people appreciate the end result than what you choose to do to the least significant bits. There's a reason they call them "least significant" after all!

Tons and tons of experimenting has shown that more satisfactory results are gained by adding dither noise while requantizing form 24 to 16 bits. So - my advice to folks is to just take the 10 seconds to enable doing that - and focus your rest of the time on the more meaningful decisions. Of course people are welcome to spend the hours that I have done in the past discovering exactly what dither choice can or can not do for you if they wish. Just shut off the chat rooms - crank up the fades for some of the time - and then be done with it. And in the end you're back to the 10 seconds that enabling dither takes in todays workstations.

Sorry for the rant folks. But I figure there are some that are glad I went ahead and said it for them while they bit their tongue on yet another thread on this old and overly picked through topic.

Best regards,
Steve Berson
Old 23rd October 2012 | Show parent
  #62
Lives for gear
 
Tarekith's Avatar
 
Verified Member
1 Review written
🎧 10 years
Well said.
Old 23rd October 2012 | Show parent
  #63
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
Quote:
Originally Posted by David Rick ➑️
Now consider optimal triangular PDF dither. It's made by adding two optimal rectangular dither signals together, so their PDF's get convolved (yielding a triangle) and their characteristic functions get multiplied together yielding a sinc-squared function. The nulls are still in the same places, but they are wider than in the rectangular case. This means that triangular dither will do a better job at cancelling signal correlated errors from non-ideal quantizers.
David, I understand your reasoning: somewhat wider notches of a triangular PDF's characteristic function at LSB levels could be good for properly dithering a slightly nonlinear system. Although they would not "fix" the nonlinearity, of course. I have to admit I know too little about sources of nonlinearity in multibit converters. The THD figures of good modern converters are around -105 dB, which is much worse than the theoretical limit of a 24-bit or even a 20-bit quantization. Same with noise. Some of it certainly comes from op amps. I doubt that 24-bit dither shape mentioned by Bob would have any audible effect in presence of such noise and distortion. But if we speak of 16 bits, then yes, probably.
Old 23rd October 2012 | Show parent
  #64
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Quote:
Originally Posted by Alexey Lukin ➑️
... I doubt that 24-bit dither shape mentioned by Bob would have any audible effect in presence of such noise and distortion. But if we speak of 16 bits, then yes, probably.
I've put numerous people's jaws on the floor demonstrating the effect of 24 bit TPDF dither on the imaging and depth of a mix. We can hear right through most noise and distortion to much lower level information than would ever seem logical. Our hearing is non-linear while our logic assumes linearity.

Using 24 bit dither also prevents the crunchy sounding results that people hate about digital audio processing. I've been able to tell people didn't use the Pro Tools dithered mixer countless times and never been wrong when I believed that to be the case.
Old 23rd October 2012
  #65
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
I've read similar thoughts in the book of Bob Katz. While I don't understand them, I appreciate you sharing your experience with us.
Old 23rd October 2012 | Show parent
  #66
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Jim Johnston has told me about some IEEE papers about research into what we can hear and masking effects. This research was an integral part of the development of modern lossy digital audio encoding.
Old 24th October 2012
  #67
Deleted 691ca21
Guest
So, both Bob's seem to be saying that dithering from 32bit float to 24 bit fixed can result in an audible and positive difference, as compared to just truncating from 32 bit float to 24 bit fixed, am I right? If that's the case, then it's a shame that more dithers don't offer a 24 bit output.
Old 13th June 2013 | Show parent
  #68
Lives for gear
 
🎧 15 years
I'm using lots of analog outboard as inserts (analog compressors, eqs, ..). So if I understand this right, it would be advisable to dither every signal that gets converted to analog.

Is there a small good dither plug-in available for this? I have Ozone5, but inserting the whole lot seems a bit overkill and pumps up the latency.

thanks!
Old 13th June 2013 | Show parent
  #69
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Quote:
Originally Posted by Babaluma ➑️
So, both Bob's seem to be saying that dithering from 32bit float to 24 bit fixed can result in an audible and positive difference, as compared to just truncating from 32 bit float to 24 bit fixed, am I right? If that's the case, then it's a shame that more dithers don't offer a 24 bit output.
Although I knew this to be theoretically correct I was only able to verify it when I started doing 24 bit multi-track music recording after only doing analog multi-track along with digital post production and mastering from 16 bit sources for a decade.

What I found with my own carefully recorded projects was that I could clearly hear a difference in mixes using my mastering monitor system but only provided I used the flat full level TPDF dither that theory calls for.

Check out Stochastic resonance - Wikipedia, the free encyclopedia This is why even an analog magnetic recording with its dramatically worse signal to noise ratio can preserve more spatial information than truncated digital audio. If the audio has ever been truncated at some point, the effect of the dither becomes far more subtle because you can't bring back the low level information that was lost in the truncation. Working with truncated audio was what I think confused us with noise shaping, reduced dither levels, etc.
Old 13th June 2013 | Show parent
  #70
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Quote:
Originally Posted by cowudders ➑️
I'm using lots of analog outboard as inserts (analog compressors, eqs, ..). So if I understand this right, it would be advisable to dither every signal that gets converted to analog.

Is there a small good dither plug-in available for this? ...
Don't we all wish! The closest thing I've found is the 20 bit Pow-R1 plug-in that comes with Pro Tools. It sounds better than nothing to me but not as good as Xeon, Ozone or the Sonnox dynamics. Their regular flat dither which was the obvious choice was disappointing.

Now you don't need to dither any tracks that have no gain changes or other digital signal processing being done to them.
Old 13th June 2013 | Show parent
  #71
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
please delete!

Last edited by Bob Olhsson; 13th June 2013 at 06:34 PM.. Reason: wrong button, please delete!
Old 13th June 2013 | Show parent
  #72
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by Bob Olhsson ➑️
Don't we all wish! The closest thing I've found is the 20 bit Pow-R1 plug-in that comes with Pro Tools. It sounds better than nothing to me but not as good as Xeon, Ozone or the Sonnox dynamics. Their regular flat dither which was the obvious choice was disappointing.
Using Ozone 5 in IRC I (takes the least processing power) and MBIT+ without Shaping and dither=High seems to be the best option then.

Quote:
Originally Posted by Bob Olhsson ➑️
Now you don't need to dither any tracks that have no gain changes or other digital signal processing being done to them.
Which would be really rare, as I'm using most analog stuff on busses, via Logic's I/O plug-in.

Thanks Bob, very appreciated.
Old 14th June 2013 | Show parent
  #73
T***o Dawn Labs
 
FabienTDR's Avatar
 
Verified Member
🎧 5 years
Quote:
Originally Posted by Babaluma ➑️
So, both Bob's seem to be saying that dithering from 32bit float to 24 bit fixed can result in an audible and positive difference, as compared to just truncating from 32 bit float to 24 bit fixed, am I right? If that's the case, then it's a shame that more dithers don't offer a 24 bit output.
Converting a 32bit floating point number into a 24bit integer does not result in less precision in general (in the sense of a 8*6 dB higher noise floor). Fact is, a properly normalized 32bit floating point file can be converted to an (nearly) equally precise 24bit integer format.

The question if these minimal conversion losses truly require additional 24bit dithering is still wildly discussed.

However, in the case of a non normalized floating point signal, several bits will be lost during the conversion process. Without proper 24bit dithering, this truncation will generate the distinct un-harmonic distortion patterns know from "normal" format truncation processes (simply because the source signal doesn't cover the target format's full scale).

A 32bit floating point number is stored in such a way:

[Sign bit] 1 bit
[Exponent] 8 bit
[Significant Precision] 23 bit

A 24bit integer number is stored like this:

[Sign bit] 1 bit
[Significant Precision] 23 bit

You probably see the similarities, the exponent part of the floating point number can be imagined as the scale of the core value (i.e. position of the decimal point). Reality is more complicated and makes the conversion between both number representations a complicated and lossy process. But as mentioned above, it does not introduce the obvious and easily measurable quantization distortion un-dithered truncation introduces (e.g. 24bit to 16bit). It is unclear if dithering improves the situation at all in the case of properly scaled (i.e. normalized) signals.

As for the subjective effects mentioned in this thread, keep in mind that noise can also reduce certain kinds of distortion. Noise has the effect to make un-harmonic, yet regular distortions less regular (and substantially lower!). Noise alone can clearly add all the holographic whatever, completely without truncation.

Or in other words, does this:

32bit fp -> 24bit dither -> Truncation to 24bit

really sound better than:

32bit fp -> Truncation to 24bit -> 24bit dither

I can't hear the difference and can't measure any improvements. And it indicates that this specific conversion process doesn't benefit from dithering. The material itself does benefit from added noise, but the effect itself seems to be completely unrelated to the actual conversion process.
Old 14th June 2013 | Show parent
  #74
Deleted 691ca21
Guest
Quote:
Originally Posted by cowudders ➑️
Is there a small good dither plug-in available for this?
Yes, I use Sonoris Dither. It's a tiny little non-expensive VST plugin, highly recommended. It won't dither to 24 bit, but I've asked Pieter about this and he has said he will include a 24 bit option in the next update.

http://www.sonorissoftware.com/catalog/dither-p-32.html

Last edited by Deleted 691ca21; 14th June 2013 at 08:25 AM.. Reason: Added link
Old 14th June 2013 | Show parent
  #75
Deleted 691ca21
Guest
Quote:
Originally Posted by Bob Olhsson ➑️
Although I knew this to be theoretically correct I was only able to verify it when I started doing 24 bit multi-track music recording after only doing analog multi-track along with digital post production and mastering from 16 bit sources for a decade.

What I found with my own carefully recorded projects was that I could clearly hear a difference in mixes using my mastering monitor system but only provided I used the flat full level TPDF dither that theory calls for.

Check out Stochastic resonance - Wikipedia, the free encyclopedia This is why even an analog magnetic recording with its dramatically worse signal to noise ratio can preserve more spatial information than truncated digital audio. If the audio has ever been truncated at some point, the effect of the dither becomes far more subtle because you can't bring back the low level information that was lost in the truncation. Working with truncated audio was what I think confused us with noise shaping, reduced dither levels, etc.
Thanks for the detailed reply Bob. Seems that never truncating, and always using TPDF should keep the most audible information.
Old 14th June 2013
  #76
Deleted 691ca21
Guest
And to FabienTDR too, thanks for the detailed exposition!
Old 14th June 2013 | Show parent
  #77
Lives for gear
 
🎧 15 years
How to only use the Dither in Ozone 5

Just got some good info from Joel (iZotope support):

'Lastly, it is important to note that the active or bypassed state of the maximizer applies only to the Maximizer processor. Dither is independent of the Maximizer. When the Dither type is set to "None", there is no dither or word length reduction applied. To turn dither and word length reduction on, click the "None" button to select MBIT+, Type 1 or Type 2 as the dither type.'
Old 9th July 2013 | Show parent
  #78
Lives for gear
 
Filthrill's Avatar
 
🎧 10 years
Hey guys. Nothing too heavy. I've been using Sonnox Limiter's dither for at least the last 6 months. My setting is Type I dither, depth=40%.

Curious if anyone on here uses the Sonnox's dither on a "regular basis" also.

Wondering if I'm a loner on the Sonnox dither island.
Old 9th July 2013 | Show parent
  #79
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Quote:
Originally Posted by FabienTDR ➑️
I can't hear the difference and can't measure any improvements. And it indicates that this specific conversion process doesn't benefit from dithering. The material itself does benefit from added noise, but the effect itself seems to be completely unrelated to the actual conversion process.
Where I hear the biggest difference is after additional processing although on some microphone-sourced mixes I can hear an obvious loss of depth when I turn the dither off and my clients always ask "what happened?" I also have mastering clients who can spot different dithers or broken dither immediately due to the damage done to their ears on stage. When a band is missing from somebody's hearing, there's nothing to mask the distortion! We all love linear concepts but hearing is totally non-linear!

If 24 bit dithering is properly done, it has always been inaudible so to me it makes no sense to treat dither as optional because I can't assume others won't hear the difference. The facts of life are that we are all flying deaf to a certain degree which makes proper engineering very important even when we can't hear any difference from doing it wrong.
Old 9th July 2013 | Show parent
  #80
T***o Dawn Labs
 
FabienTDR's Avatar
 
Verified Member
🎧 5 years
Quote:
Originally Posted by Bob Olhsson ➑️
[...] because I can't assume others won't hear the difference. The facts of life are that we are all flying deaf to a certain degree which makes proper engineering very important even when we can't hear any difference from doing it wrong.
I agree with it in general.

The specific case I mentioned (conversion from 32bit fp to 24bit fixed) is an extreme example where engineers can't properly explain or measure any significant benefit (at least to my knowledge, feel free to correct me). I don't question the use of dithering before truncation. I question if dithering is really needed in the case of a typical 32bit fp to 24bit conversion scenario. Simply because there is no precision difference between both files (both have 24bit precision) in a standard mastering scenario. This process is NOT comparable to simple truncation. Again, both 32bit fp and 24bit fixed have an effective precision of 24bits!
Old 10th July 2013
  #81
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Jim Johnston of Bell Labs told me to "not even think about not dithering because the math simply won't be correct." He also told me it could require as much as 10 dB. of additional noise to mask truncation distortion than simply using TPDF dither to prevent it in the first place.

I'll take JJ's DSP engineering credentials over virtually anybody else in the world. I'm thankful that we had him and other world renowned DSP experts posting to internet newsgroups during the late 1980s and early '90s. I'll certainly grant that I probably only learned enough to be dangerous.
Old 10th July 2013 | Show parent
  #82
Gear Head
 
🎧 10 years
And i really agree with that. It's basically very hard to hear Dither but once you miss it (being it not is there) the result is just not so smooth to human ears. We hear a lot more then we are aware off and it clouds our judgement.

Dithering is a form of saturation (not exactly though it's more in the form off a really low noise) for the human ears to make sence about the loss off information that our ears expect. And it works IMO.
Old 10th July 2013
  #83
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
It is not a form of saturation. It prevents the program material from modulating the bottom bit like a chattering noise gate which I'd argue sounds more like saturation. It also preserves low level material that would be completely chopped off by truncation. A good example is to record a sin wave at -110 and compare truncation to 16 bits to dithering to 16 bits. You will hear the tone in the hiss.
Old 10th July 2013 | Show parent
  #84
Gear Head
 
🎧 10 years
For the ear it is. Technically it's not saturation. In dutch we have a great term for it. Dithering or noise on a low level saturates the brain. No listener is going to listen to dither.
Old 11th July 2013 | Show parent
  #85
Lives for gear
 
Adam Dempsey's Avatar
 
Verified Member
🎧 10 years
Quote:
Originally Posted by Misja ➑️
No listener is going to listen to dither.
And yet they are very likely to hear (via use of dither) the absence of cumulative degradation/distortion.
Old 1st August 2013 | Show parent
  #86
Lives for gear
 
Filthrill's Avatar
 
🎧 10 years
Hey slutz. Sorry to be a "dither" bug but a producer friend & myself have been discussing this for a few days now & are reading conflicting theories everywhere so figured I'd ask here. I posted this on another thread but I'm looking to get some immediate feedback please:

who here would say that it's a bad idea to bus & print your mix internally in Pro Tools HD (version 8) thru a master fader & an aux track w/ a dither plug-in applied last & set to 16 bit, onto a new stereo audio track? Then exporting (not bouncing) that mix from the new audio track as a 44.1k 16-bit file for burning to CD for a client.

Let's say the PT session is at 44.1k 24-bit.

1) Is this double dithering?
2) Is it "harmful" or something you could live with doing yourself?

Or, who would do this & who would not do this & why?

Thanks in advance.
Old 1st August 2013
  #87
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
"Double dithering" means adding dither to a totally unprocessed file.

If you change the gain even .1 dB., you need to dither in order to eliminate the distortion. There are only two choices available, dither or edgy-sounding distortion that is more audible and accumulates faster than the noise from dither does.
Old 1st August 2013 | Show parent
  #88
Lives for gear
 
Filthrill's Avatar
 
🎧 10 years
Quote:
Originally Posted by Bob Olhsson ➑️
"Double dithering" means adding dither to a totally unprocessed file.

If you change the gain even .1 dB., you need to dither in order to eliminate the distortion. There are only two choices available, dither or edgy-sounding distortion that is more audible and accumulates faster than the noise from dither does.
Exactly! So it's okay that Pro Tools does a dither when you export (not bounce) an audio region to a 16 bit file, assuming you're working in a 24 bit session. I guess the automatic dither on export (that u can't turn off) is nothing to be afraid of then, when the dither gets printed internally to an audio track & then exported to a 16 bit file to burn to CD.

Just to be clear = that's 2 instances of dither. One from printing it w/ a dither plug-in & the other from PT's automatic dithering when "exporting region as file." Bob, you're still saying that this is desirable? Would u be okay w/ this fact if this was your product that you were delivering to a client?
Old 2nd August 2013 | Show parent
  #89
Lives for gear
 
1 Review written
🎧 5 years
Quote:
Originally Posted by Alexey Lukin ➑️
When noise shaping is none, the applied dither is flat (non-shaped) TPDF.
That's why I like the sound of noise shaping = none, thanks!
πŸ“ Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 55 views: 32139
Avatar for IM WHO YOU THINK
IM WHO YOU THINK 13th October 2020
replies: 4 views: 4094
Avatar for Maraxvx
Maraxvx 1st February 2011
replies: 1296 views: 181486
Avatar for heraldo_jones
heraldo_jones 1st February 2016
replies: 1591 views: 126899
Avatar for Rafat Kamal
Rafat Kamal 8th February 2021
Post Reply

Welcome to the Gearspace 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…

Forum Jump
Forum Jump