Quantcast
Flat Dither Question Mbit+ - Page 2 - Gearspace.com
The No.1 Website for Pro Audio
Flat Dither Question Mbit+
Old 1st October 2012 | Show parent
  #31
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
Quote:
Originally Posted by Rossd25 ➡️
I am using Ozone 4, so is this the same as Ozone 5 (the most recent version)?
If your Noise Shaping is set to None, then it's the same. If you are using any non-zero MBIT+ noise shaping, then it has improved in Ozone 5, thanks to iZotope's collaboration with Bob Katz.


Quote:
Originally Posted by Rossd25 ➡️
So, if I use megabit + with high dither amount I get TPDF.
Yes, you get standard TPDF.


Quote:
Originally Posted by Rossd25 ➡️
I always use Type 2 with dither amount set to 1. Is that enough dither for it to be TPDF, too?
Not sure about Ozone 4, but in Ozone 5 this is a high-pass TPDF dither: not the same as standard TPDF.


Quote:
Originally Posted by Rossd25 ➡️
Waves IDR Type 1 with shaping to none is flat?
Yes, this is standard TPDF.


Quote:
Originally Posted by Rossd25 ➡️
Should normal dither amount Mbit + flat work well enough to eliminate quantization distortion?
Yes, it was designed for that. It is still a TPDF dither, but with a slightly lower amplitude than standard TPDF.


Quote:
Originally Posted by Rossd25 ➡️
I ask because I have auditioned the dither at 8 bits on and off and normal seems to eliminate quantization distortion. Low doesn't seem to be sufficient or enough to eliminate it, but I don't hear it with normal. High just seems to add a lot of hiss which is really distracting flat.
Dither amount is a trade-off between audibility of dither and elimination of distortion. I think that Normal is a sweet spot.
Old 4th October 2012
  #32
Gear Addict
 
🎧 10 years
Quote:
Originally Posted by Alexey Lukin ➡️
If your Noise Shaping is set to None, then it's the same. If you are using any non-zero MBIT+ noise shaping, then it has improved in Ozone 5, thanks to iZotope's collaboration with Bob Katz.



Yes, you get standard TPDF.



Not sure about Ozone 4, but in Ozone 5 this is a high-pass TPDF dither: not the same as standard TPDF.



Yes, this is standard TPDF.



Yes, it was designed for that. It is still a TPDF dither, but with a slightly lower amplitude than standard TPDF.



Dither amount is a trade-off between audibility of dither and elimination of distortion. I think that Normal is a sweet spot.
Alright, thanks. Explains everything I wanted to know in one single post.

I didn't realize that about IDR. I will just set that to flat from now on when I use the L2.
Old 12th October 2012 | Show parent
  #33
Gear Addict
 
🎧 10 years
I have yet another question:

What would be the difference between Normal dither amount and High? Would it be an advantage to use High rather than normal? Will High retain more of the detail of the track, or does it not matter?
Old 12th October 2012
  #34
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
I'll comment on the technical side: High is the standard, "textbook", 2 LSB peak-to-peak amount of dither for full prevention of distortion and noise modulation. Normal allows for some small noise modulation.
Old 12th October 2012 | Show parent
  #35
Lives for gear
 
dcollins's Avatar
 
Verified Member
🎧 15 years
Quote:
Originally Posted by Alexey Lukin ➡️
I'll comment on the technical side: High is the standard, "textbook", 2 LSB peak-to-peak amount of dither for full prevention of distortion and noise modulation. Normal allows for some small noise modulation.
Why would the Normal mode ever be used, as the whole point is to remove any modulation artifacts?


DC
Old 12th October 2012
  #36
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
The point is that slight modulation of the noise power is not really audible. It does not have any nonlinear distortion, just some variation of the noise level. So, Normal is a good trade-off between low noise level and absence of modulation.
Old 12th October 2012
  #37
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
I believed in low level and noise shaped dither until I started working on projects I'd recorded myself using 24 bits that had never been allowed to be truncated without textbook dither.

On these projects the textbook dither retained lots more subtle detail and depth. It was a real ear-opening lesson about always using top quality converters and textbook dither. It's a method that simply solves most people's complaints about digital audio.
Old 12th October 2012
  #38
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
Apparently everyone has his own preferences. I've heard from Bob Katz that he prefers shaped dither on many projects. He specifically requested finer adjustment of noise shaping amount for MBIT+ in the light to medium range.
Old 12th October 2012 | Show parent
  #39
Lives for gear
 
dcollins's Avatar
 
Verified Member
🎧 15 years
Quote:
Originally Posted by Alexey Lukin ➡️
The point is that slight modulation of the noise power is not really audible. It does not have any nonlinear distortion, just some variation of the noise level. So, Normal is a good trade-off between low noise level and absence of modulation.
Any d*ther with noise-modulation is broken.

Unless you are specifically using it as an effect.


DC
Old 13th October 2012 | Show parent
  #40
Gear Addict
 
JLaPointe's Avatar
 
🎧 10 years
Quote:
Originally Posted by dcollins ➡️
Any d*ther with noise-modulation is broken.
I don't know if its been fixed in recent versions, but there was a time when the default TPDF dither amount in Samplitude resulted in noise modulation.

Dither option anxiety!
Old 13th October 2012
  #41
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Fortunately you can set Samplitude's dither to 2 bits. At least it defaults to some rather than none!
Old 13th October 2012 | Show parent
  #42
Lives for gear
 
dcollins's Avatar
 
Verified Member
🎧 15 years
Quote:
Originally Posted by JLaPointe ➡️
Dither option anxiety!
It's just so unnecessary. Having an option that results in the system not working 100% correctly is confusing.

This is a separate issue from having different curves or weighting available, of course.

Not that I would use the fancy shaped types, but if they make people feel like they have really taken the reigns on signals below -90, so be it.


DC
Old 13th October 2012 | Show parent
  #43
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
Quote:
Originally Posted by dcollins ➡️
Any d*ther with noise-modulation is broken.
Then Waves offers a lot of "broken" dithers too.
Old 13th October 2012
  #44
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Broken dithers could sound better with early digital recording technology. That doesn't mean they aren't very broken when applied to the A to D and D to A technology of twenty years later.

An awful lot of these misguided concepts, especially the idea of dither being optional, only using noise shaping and "never dithering twice" were based on the distorted sound of 8 bit sampling keyboards. It's mind-boggling how many contemporary magazine articles and software manuals parrot uninformed information from decades ago.
Old 14th October 2012 | Show parent
  #45
Lives for gear
 
David Rick's Avatar
 
🎧 15 years
Quote:
Originally Posted by JLaPointe ➡️
I don't know if its been fixed in recent versions, but there was a time when the default TPDF dither amount in Samplitude resulted in noise modulation.
In Samplitude, the dither type (Rectangular, Triangular, or none) and dither amplitude are independently settable. Amplitude default is 0.5 lsb peak (not p-p), which is correct for the default Rectangular pdf dither. If you change to Triangular, you should change the amplitude value to 1.0.

I've never liked that the Samplitude dither dialog works that way; I think when you change the dither type, it should enter the correct default amplitude. But it's been that way for about 15 years, so maybe nobody else cares.

I think a lot of Samplitude users leave the dither on default settings until they're ready to burn a CD. At that point they choose one of the three Pow'R noise-shaping options, which auto-adjust their amplitude for the bit depth.

David L. Rick
Seventh String Recording

P.S: Thanks to Alexey Lukin for explaining the MBit+ settings. I'm sure some marketing 'droid thought thought the High/Medium/Low labels were "friendlier", but I think the opposite: they prevent skilled users from applying their prior knowledge.
Old 14th October 2012
  #46
Gear Addict
 
🎧 10 years
High is the standard? Huh? The Ozone guide specifically said that normal was recommended (and I assume the standard). I just thought high would be some kind of option for... I don't know... adding more dither noise. Anyway, lately I've been using the L2 more and using IDR with the noise shaping turned off.

It still has never been explained why I would want to apply noise shaping. I mean, I know what it does. It's just, should it only be applied to material where flat dither is really intrusive? Either way someone is going to have to have really good ears to hear it at 16 bits. I can't hear the dither hiss... only at 8 bits.

I feel really wary about using any kind of extreme noise shaping. I know high is going to shove the noise way up into the higher frequencies, but something tells me that this sounds more unnatural. After all, people have been listening to analog hiss for years and are quite used to low level noise. Besides that the L2 guide said something about Ultra not translating well onto cheaper CD players. It just seems ridiculous to shape noise that no one is going to hear anyway.
Old 14th October 2012
  #47
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Adding noise to preserve low level detail is counterintuitive to those of us with an analog recording background. The term "noise" is also misleading when it comes to the digital world. It means that low level information is missing as opposed to being obscured by hiss.

The flaws in early converters combined with our prejudice against hiss led us to believe that minimizing the audibility of dither by reducing its level or noise shaping would lead to more transparent results. In hindsight I think most of what it did was to hide low level artifacts that are not present in modern A to D converters.

Where I'm coming from and I assume Dave is coming from is that we really want people to produce better sounding recordings. The misguided recommendations about dither from two decades ago really need to be called out.
Old 14th October 2012 | Show parent
  #48
Lives for gear
 
dcollins's Avatar
 
Verified Member
🎧 15 years
Quote:
Originally Posted by Bob Olhsson ➡️
Where I'm coming from and I assume Dave is coming from is that we really want people to produce better sounding recordings. The misguided recommendations about dither from two decades ago really need to be called out.
Oh yes, absolutely agreed.

D*ther is incredibly mis-understood even today. It's a subtle thing, of course, but it all adds up. Also, as Bob says, it's one of those things that makes people thing digital is fatally flawed as it was either not used at all or used with converters that were so inaccurate at low signal levels that it couldn't work properly - ergo digital sucks, eats reverb for breakfast, is mostly mono etc.

I'm just confused how Alexey, who is a major genius and math expert, could get marketing advice that leads to a setting that is not 100% free of signal modulation?

That setting should come with a warning label or disclaimer.


DC
Old 14th October 2012
  #49
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
The goal of MBIT+ is to provide most options to the user, including some unique ones. Just like there is no warning signs on Waves Type1 and Type2 IDR dithers (one of them does not prevent modulation), there's no warning signs on different dither amounts in MBIT+. Defaulting to "Normal" dither amount is not a marketing trick, it is indeed a recommendation that we support. Let's keep it clear: this setting does not result in any signal modulation; the only modulation that happens is a slight modulation of power of quantization noise. When noise shaping is turned off, the amount of this modulation never exceeds 1 dB. When noise shaping is on, it is even less. We consider this to be a useful mode taking its niche between Type1 and Type2 dithers (i.e. between 1 LSB p-t-p RPDF and 2 LSB p-t-p TPDF). We have only heard positive opinions about this mode, so I'm frustrated to hear that you can't admit it does a proper job.

I do appreciate your criticism about lack of technical details in the docs on this setting. I'll write this section myself with our next update...
Old 19th October 2012
  #50
Lives for gear
 
David Rick's Avatar
 
🎧 15 years
Rethinking triangular dither

We use rectangular (RPDF) dither to prevent quantizer-induced distortion. We can use triangular dither (TPDF) to prevent signal-correlated modulation of the noise floor. Sometimes noise modulation is a problem, and sometimes not. It is clearly a big issue in the case of 8 or 12-bit PCM, but we don't use those for music. I also think it was an issue for 16-bit material before audio engineers started peak-normalizing everything. Even today, engineers who adhere to K-system standards when mastering CD's will end up releasing some material in what's effectively a 14-bit PCM format.

Historical aside: Thomas Stockham Jr. heard and solved noise modulation in 1980 when Fleetwood Mac were recording their Tusk album using his Soundsteam digital process. (This was a 14-bit digital recording format.) The solution was to use TPDF dither instead of RPDF; Stockham kept this as a trade secret for a number of years. TPDF dither was already known to theoreticians (Sripad & Snyder, 1977), but not to other audio engineers. Most audio folks learned dither theory from the work of Lip****z, Vanderkoy, and Wannamaker in the mid 1980's.

It's possible today (though not very popular) to release music at 24-bit resolution. Even in this format, I've often used TPDF dither simply out of habit. On reflection, I see no reason for using it with 24-bit word lengths and I'm going to stop. Who cares about noise modulation if the quantization noise is inaudible? Rectangular dither should be sufficient.

Likewise, noise shaping algorithms are only valuable when the quantization noise would otherwise be audible. I hear a clear benefit to the Pow'R algorithms when mastering carefully-recorded acoustic material for 16-bit releases. But for some other material, shaping the noise floor sometimes unmasks artifacts from earlier in the production process that are better kept covered up!

I'm still thinking about the implications of Alexey's 1.5 LSB TPDF dither, although I've sometimes liked it without knowing what it was! If you use the wrong-sized rectangular dither, the quantization distortion is not zero, but it's still lower than if you'd done nothing. I expect that a similar thing happens with triangular dither and noise modulation: there remains some correlation between the audio and the noise floor, but that correlation might be small enough to go unnoticed. I'd like to see curves for this similar to the ones Wagdy and Goff published for distortion.

David L. Rick
Seventh String Recording
Old 19th October 2012
  #51
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Maybe this is true with synthesizers but the improvement in low level detail and depth from 24 bit TPDF is a no-brainer with microphone-sourced audio.
Old 19th October 2012 | Show parent
  #52
Lives for gear
 
David Rick's Avatar
 
🎧 15 years
Quote:
Originally Posted by Bob Olhsson ➡️
Maybe this is true with synthesizers but the improvement in low level detail and depth from 24 bit TPDF is a no-brainer with microphone-sourced audio.
Compared to what? No doubt it's better than not having dither. Are you saying you prefer 2 LSB TPDF to 1 LSB RPDF when mixing to 24-bit fixed point? That would be interesting. It might persuade me to run some controlled experiments.

For the record, I do not record synthesizers. Most everything I do involves strings and/or Steinway.
Old 19th October 2012 | Show parent
  #53
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
Quote:
Originally Posted by David Rick ➡️
... It might persuade me to run some controlled experiments...
I've run a bunch of them. A lot of assumptions about dither from the 1990s turn out to apply mostly to 1990s 18 and 20 bit converters and not more recent 24 bit ones. A I wrote above, with modern converters the theoretical flat TPDF turns out to sound consistently better provided the signal is clean. The whole idea of adjustable dither seems questionable to me at this point.
Old 19th October 2012 | Show parent
  #54
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
Quote:
Originally Posted by David Rick ➡️
I'm still thinking about the implications of Alexey's 1.5 LSB TPDF dither, although I've sometimes liked it without knowing what it was! If you use the wrong-sized rectangular dither, the quantization distortion is not zero, but it's still lower than if you'd done nothing. I expect that a similar thing happens with triangular dither and noise modulation: there remains some correlation between the audio and the noise floor, but that correlation might be small enough to go unnoticed. I'd like to see curves for this similar to the ones Wagdy and Goff published for distortion.
We do have some figures for the Normal dither amount in MBIT+. Here is the table for 44.1 kHz:


It shows that setting dither amount to Normal instead of High reduces the noise power by approx. 0.5 to 1 dB (see first two rows). The amount of noise modulation is ±0.5 dB without noise shaping and even less with noise shaping (see the third row). It reduces to virtually nonexistent with stronger noise shaping, while still offering 1 dB lower noise floor than High dither.
Old 19th October 2012
  #55
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
The main thing 2 LSB d*ther does for me is to better preserve extremely low level information. That's more than worth an extra half dB. higher noise floor 96 dB. down. The difference is information audibly going away. I hear this happening in a 24 bit mix where the additional noise is at an absurdly low level. I'm not sure this has anything to do with noise modulation other than maybe it acts like a noise gate down there.

When you look at such numbers, there is a tendency to assume that program material should mask anything that low but that simply isn't how our hearing works.
Old 20th October 2012 | Show parent
  #56
Lives for gear
 
Cellotron's Avatar
 
Verified Member
3 Reviews written
🎧 15 years
fwiw -
Amount of projects I've worked on in 20 years of mastering where I was called on to do a revision and where changing the dither type was the key to getting reference approval from the client:

ZERO.

Hope that helps.

Best regards,
Steve Berson
Old 21st October 2012 | Show parent
  #57
Lives for gear
 
David Rick's Avatar
 
🎧 15 years
An aha! moment on dither

Quote:
Originally Posted by Bob Olhsson ➡️
A lot of assumptions about dither from the 1990s turn out to apply mostly to 1990s 18 and 20 bit converters and not more recent 24 bit ones. A I wrote above, with modern converters the theoretical flat TPDF turns out to sound consistently better provided the signal is clean.
I was thinking about this while riding my bike today, Bob. (Some people walk and chew gum at the same time. I pedal and do mathematics at the same time. It's a geek thing.) I have a mathematical theory for why TPDF might be more useful for high-resolution files today than it was in 1990.

Recall how optimal dither eliminates distortion: The error function for a uniform quantizer looks like a sawtooth wave. Its Fourier transform is an equally-spaced set of impulses whose spacing is inversely proportional to the converter's LSB size. Dither is a random variable whose amplitude described by a probability density function. The Fourier transform (i.e. characteristic function) of rectangular PDF dither has a "sinc" or sin(f)/f shape with periodic nulls. When you work out the expected value of the quantization error, it turns out to be independent of the input signal because the nulls in that sinc function cancel the impulses from the transformed error curve. (All but one, anyway.) For this cancellation to be perfect, the rectangular dither has to have exactly the right peak-to-peak amplitude (1 LSB), and the quantizer has to be perfectly uniform: all LSB's the same size.

What we had in the 1990's were single-bit sigma-delta converters. A single-bit quantizer yields perfect linearity, so all LSB's are the same size. But single-bit converters had other problems like overloads and limit cycles which led converter designers to switch over to the multi-bit architectures that predominate today. Multi-bit converters don't have perfect linearity. Not only can the LSB steps be different from one another, they even change size from instant to instant. Consequently what were once zero-width impulses in the error characteristic get spread out and leak out of the dither nulls.

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.

It used to be that we liked triangular dither because its sinc-squared characteristic function decorrelated the second moment (power) of the quantization error from the signal. (That is, it prevented noise modulation.) But in a 24-bit system, the quantization noise is way too low to hear, so noise modulation shouldn't matter any more. Distortion still does, though. I think what matters today is that triangular dither is better than rectangular dither at coping with the non-uniform quantization steps that come with multi-bit delta-sigma converters.

Hopefully Alexey is still watching this thread, because I'd like to have his thoughts on this.

David L. Rick

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.
Old 21st October 2012
  #58
Motown legend
 
Bob Olhsson's Avatar
 
Verified Member
🎧 15 years
I also have a few clients who can spot dither changes or broken dither in a heart beat. I asked Jim Johnston about this and he told me it was to be expected if the people had hearing damage.

A few years ago I posted these examples of the same -110dB. tone dithered vs. truncated to 16 bits. While it is 16 bits and noise shaped dither, it makes the point of why dither is important.

https://gearspace.com/board/4685247-post30.html

The word "distortion" is a bit misleading. What happens is that low level information is lost.
Old 22nd October 2012
  #59
Gear Addict
 
🎧 10 years
Sorry to resurrect this thread. I do notice that megabit+ with a high dither amount is about the same amount of dither as wave arts final plug dither and type 2 flat. What I don't get is that if megabit+ high is normal TPDF, then why does it sound so different from type 2 which is static sounding? And also, why is it that the harmonic suppression option comes up with type 2 used flat with number bits set to 1? This bothers me because this option only comes up in megabit when it is set to none and low, so is type 2 flat not doing its job properly?

Overall, I am just going to use megabit+ flat with dither amount set to high.
Old 22nd October 2012 | Show parent
  #60
Lives for gear
 
stinkyfingers's Avatar
 
🎧 10 years
Voxengo TPDF...FTW!!!!


edit: just kidding, F*TPDF, i use the Gauss...
📝 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