Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[dcollins]

Ronny wrote on Sun, 14 August 2005 08:37

Probably because you don't know where dither is applied when you dither a 24 to 16 bit word reduction and assume like many people that it's appled to the 16th bit after the reduction, that ain't the way it works.

What?  It's applied to the signal before truncation at the amplitude that corresponds to the reduced wordlength.

Whether from 24 to 16 or 16 to 8, the only difference is the level of the noise added.

Always, always always.

DC

[Barry Hufker]

Steve,

When I read your posting I knew you were making a joke.  And I read the link in full. I guess I lost sight of that as I was frustrated with not being able to upload the files I wanted.  Sorry!  I do have a good sense of humor when I'm more clear-headed.

Barry

[chrisj]

OK, here  

http://www.airwindows.com/dithering/DitherAnalysis.html

Been meaning to do that.

Bit of a summary-

The high energy noise shapers produce a color. I don't care who does them- Alexey does them best and still gets a sort of 'illuminated' quality that people actually seek out.

Sound Forge is doing something interesting, that I want to know more about. It is producing a tone character that would be more to the taste of Brad, or anyone who is having objections to the noise shaper sound. It's an area I've been specializing in lately. Sound Forge was also VERY popular in the 2496 shootout for a non-high-energy noise shaper.

If you're going to use UV22, you might as well use Ultra, because you gain nothing by half measures and using Normal.

Cool Edit Pro 44.1 actually sucks- Alexey only likes it because he likes high energy shapers, and his is much better. Use C1 or C3 for that.

Barry Hufker is dead-on about the Pow-R dithers- 1 is your best best. 3 has funny colorations but all's fair in love and EQ, and 2 is a blast of very obvious hiss. None of them are really 'all that'. If you want high energy noise shapers Alexey is your man- if not, TPDF or Sound Forge is. Or maybe UV22 if you don't mind its color.

Don't use the highest energy noise shaper in Wavelab under any circumstances, it's really obnoxious. The lower energy ones aren't that much better. TPDF is your friend. Dither noise shaped wrong is MORE annoying.

Waves noise shapers are all annoying, reduce soundstage depth and produce a coloration, but the worst of them is not the most extreme but the least. DOn't even use the Low setting- worse than plain dither. Really, it'd be better to just use dither. Or go out at 24 bit and dither elsewhere.

*awaits brickbats and general gleeful abuse*

[Barry Hufker]

Chris,

Thank goodness you showed up and posted!

It is comforting to have someone else say they can hear a difference with POW-R.  I don't have much experience with other dithers and noise shaping.  And despite the limits of Alexey's examples, I too think he's got something.

Barry

[chrisj]

"Thank goodness"? Dude!  

You'll ruin my 'audiophile crank' cred, no matter how I try to bolster it with inflammatory, impolite statements  

Seriously, you must have good monitoring and good ears to be hearing this stuff since it is NOT obvious. However, so what? People can hear that which isn't obvious, some or most of the time. And the Pow-R stuff is as you described. It's even more obvious in Robin Schmidt's shootout. Kudos to you for hearing it not under controlled circumstances.

Me, I don't care if I'm deaf as a post so long as I can make computers draw purty pic-choors  

Dither is like EQing the sidechain, or something. It's a tiny trace of some sort of quality, and you get to pick what sort. It does very little to obviously affect anything, but it's pervasive as hell at 16 bit. I don't buy that there are perceptible differences at 24 bit. For my 24 bit output in my software I made it do plain TPDF on purpose (from 64 bit float) because I thought noise shaping to 24 bit was insane and pure wankery

Not that there's anything wrong with that, but when you're talking about stuff that far beneath any realworld noise floor, I just want the noise floor to be free of correlated distortion, and correct TPDF will do that very nicely.

Can't really go wrong with TPDF....

[Ronny]

dcollins wrote on Sun, 14 August 2005 17:57

Ronny wrote on Sun, 14 August 2005 08:37 Probably because you don't know where dither is applied when you dither a 24 to 16 bit word reduction and assume like many people that it's appled to the 16th bit after the reduction, that ain't the way it works. What?  It's applied to the signal before truncation at the amplitude that corresponds to the reduced wordlength. Whether from 24 to 16 or 16 to 8, the only difference is the level of the noise added. Always, always always. DC

Ok, I stand corrected on that. If it's applied at the 16th bit, wouldn't the dither be +48dB instead of around +3dB, that's what's confusing me. I can see the dither at -93dB when going from 24 bit to 16 when no audio is playing, so I know that it's in the 16th bit after quantization, but it is applied before the quantizer according to some dither papers, so I'm wondering about the bits 24-17, is the dither in those bits when it's applied before the quantizer? I can't picture empty bits on 24-17 and than dither only at the 16th before quantization.

[bblackwood]

Ronny wrote on Sun, 14 August 2005 23:05

dcollins wrote on Sun, 14 August 2005 17:57 Ronny wrote on Sun, 14 August 2005 08:37 Probably because you don't know where dither is applied when you dither a 24 to 16 bit word reduction and assume like many people that it's appled to the 16th bit after the reduction, that ain't the way it works. What?  It's applied to the signal before truncation at the amplitude that corresponds to the reduced wordlength. Whether from 24 to 16 or 16 to 8, the only difference is the level of the noise added. Always, always always. Ok, I stand corrected on that. If it's applied at the 16th bit, wouldn't the dither be +48dB instead of around +3dB, that's what's confusing me. I can see the dither at -93dB when going from 24 bit to 16 when no audio is playing, so I know that it's in the 16th bit after quantization, but it is applied before the quantizer according to some dither papers, so I'm wondering about the bits 24-17, is the dither in those bits when it's applied before the quantizer? I can't picture empty bits on 24-17 and than dither only at the 16th before quantization.

If dither is applied to the 16th bit, then by definition, everything beneath it is dithered - the dither level is the maximum amplitude...

[Ronny]

bblackwood wrote on Mon, 15 August 2005 12:51

If dither is applied to the 16th bit, then by definition, everything beneath it is dithered - the dither level is the maximum amplitude...

Ok, I understand, thanks. Nika's paper says that you add 9 bits of noise, I can visualize that, but I've also heard people say that you add the dither to the LSB at the 24 bit, Glenn M. among others. There is a lot of conflicting information on the net and it's hard for me to separate the collards from the weeds. Mucho explanations of why you need dither and what it does, but very little on the exact implementation of it. Maybe that's why there is so much discussion over something that as long as it's applied, can be less significant than moving a mic 6 inches one way or the other when tracking.  

[Sam Lord]

chrisj wrote on Sun, 14 August 2005 23:01

"Thank goodness"? Dude!   ....

Hey Chris, I also want to thank you.  I've never read one of your posts that didn't leave me nodding my head saying "Dude, you are SO righteous!"  No kidding, your dither comparison sits at the top of my "Technical Papers" file above about 200 others.  You go!  Sam

[dcollins]

Sam Lord wrote on Tue, 16 August 2005 11:00

No kidding, your dither comparison sits at the top of my "Technical Papers" file above about 200 others.  You go!  Sam

Has Chris started used conventional FFT's to test the dithers, or some kind of charts  that no one else uses?

DC

[dcollins]

Ronny wrote on Mon, 15 August 2005 10:49

There is a lot of conflicting information on the net and it's hard for me to separate the collards from the weeds.

And you wouldn't just believe me, right off the bat?  I'm hurt.

You really need to get the Pohlman book and spend some time with it, I promise it will clear up many misconceptions.

DC

[jfrigo]

dcollins wrote on Tue, 16 August 2005 14:35

You really need to get the Pohlman book and spend some time with it, I promise it will clear up many misconceptions.

Or Watkinson. It's amazing how much traction all of these dither misconceptions have. This stuff is not some new and little understood technology. It's been well understood for decades. It's all in the texts for those who want to read it. I don't take part much in dither discussions anymore. They're kinda boring.

[bblackwood]

jfrigo wrote on Tue, 16 August 2005 18:28

This stuff is not some new and little understood technology. It's been well understood for decades. It's all in the texts for those who want to read it. I don't take part much in dither discussions anymore. They're kinda boring.

Indeed, often quite remedial, but still important to clear up the misconceptions...

[Ronny]

dcollins wrote on Tue, 16 August 2005 17:35

Ronny wrote on Mon, 15 August 2005 10:49  There is a lot of conflicting information on the net and it's hard for me to separate the collards from the weeds. And you wouldn't just believe me, right off the bat?  I'm hurt. You really need to get the Pohlman book and spend some time with it, I promise it will clear up many misconceptions. DC

Yes, I agree I should get the Pohlman book. It's not that I don't believe you right off the bat Dave, I've learned a lot from you, but you are only one person, you are human, you make mistakes like the rest of us and at my age, my memory isn't as good as it was when I was younger, so I can't retain information like I used to and that prones me to making mistakes and why sometimes you notice me covering the same question that I might have asked a year ago. Also, there is much stuff written that conflicts with what you are compiling and relating from Pohlman, so in that regard it's hard for me to separate the wheat from the chaf. There are other factors involved, I don't have a college education and lack the skills to read complex mathematical formulas, this makes it harder for me to understand a lot of the information. I was a late comer to getting a computer compared to most of the folks on these forums, was recording analog for 25 years before I got my first PC, so although I believe that I can bring non-comprehensive stuff to psw forums regarding frequency interactions and tried and true recording techniques that aren't covererd in any of the standard books, they didn't have consumer computers when I was in school and consequently everything that I have learned is from my own experience, what people tell me and what I glean off of these newsgroups or the internet articles about digital audio.

I realize that some of this is remedial, Brad, but remedial used to mean supplying a remedy, the forum title is Mastering Demystified, not mastering for People That Already Know Everything. Please understand from my above post that it may take me longer to understand than others as I don't have a degree in electronics and just aren't going to understand everything as quickly as formally educated folks, without simple clarification.

[bblackwood]

Ronny wrote on Tue, 16 August 2005 19:18

I realize that some of this is remedial, Brad, but remedial used to mean supplying a remedy, the forum title is Mastering Demystified, not mastering for People That Already Know Everything. Please understand from my above post that it may take me longer to understand than others as I don't have a degree in electronics and just aren't going to understand everything as quickly as formally educated folks, without simple clarification.

What many people seem to have lost nowadays is that ignorance isn't an insult but a fact of life for everyone. We all have gaping holes in our knowledge, so the fact that this subject is remedial isn't insulting but rather meant as a reference point for those who know 1% of what you do...

It's not meant as an insult - we all have much to learn!

I just hope that when any of us tread on subjects we're not 100% sure of, we do so carefully, as our posts may unintentionally mislead someone trying to get a grasp on what we're discussing...