Print

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

[Barry Hufker]

Dither *is* applied to the 24 bit signal, but I believe the only chance to hear it is after it has been applied and you are now listening to the 16 bit signal.

It is true, UV22 is akin to the bias signal of a tape recorder rather than the noise (shaped or not) called dither.  But in direct comparisons I've tried, its effects are also audible.

"Noise shaping" does not push the noise into frequencies we can't hear.  Take a look at the attached paper.  You can see there is an increase in dither noise that is in the clearly audible region.  This is true for POW-R2, L1-Ultramaximizer, L2.  By comparison, POW-R3 is in a category of its own!

So, not only is noise shaped dither in the audible band, but the extra energy is some reasons exaggerates frequencies.  As I said before, POW-R2 caused extreme sibilance in the recording I just completed.

After much study, I believe I am with Brad.  I am arriving at the conclusion that plain old TPDF is best.

Trivia: I read up on dither.  It comes from the British.  Bombsights, being built during WWII, were found to be more accurate in the air than they were on the ground.  The shaking of the bomber(dither)evened out innacuracies and made the sight better!

[Ronny]

TotalSonic wrote on Sun, 14 August 2005 11:42

Barry Hufker wrote on Sun, 14 August 2005 14:00 Here are a couple of interesting pages: http://audio.rightmark.org/lukin/dither/#samples and http://www.24-96.net/dither/ These particular examples are a fairly useless way to judge the effectiveness of the dithers because: 1) in the first case because they are 16 to 12bit conversions 2) in the second case because they are the dither noise alone just gain staged substantially up. Both of these examples do not represent a real world test for judging: 1) "effectiveness" at smoothing apparent distortion from quantization errors at the 16bit level 2) "transparency" of the dither the second test only allows you to judge: 3) "pleasantness" of the audible noise the dither imparts Best regards, Steve Berson

Good observations Steve. What we need is a blind dither test, performed by folks such as on this forum, where dither is applied at the real world level that we use it at. The test would have no influence from manufacturers that are testing other dithers and comparing it with their own. Apogee's dither examples are extremely misleading as I've analysed their examples and the output gains on the music examples are not the same, they apply dither to the 9th bit, so that you can hear the noise mixed in with the audio, that again tells us nothing, because we aren't using dither to increase noise in the signal, we are using it to eliminate audible artifacts that are a consequence of quantizing errors.

I'm still waiting to hear some physical reason as to how adding noise at the half bit of the 24th LSB, changes tones and dynamics on the final bit depth playback. I can only hear the differences between no dither and dither and only when I turn up a fade out or verb tail on the last note of a song, I can't hear truncation artifacts when the audio is above -40dB or so. I can't hear the difference between POW-R, UV22, IDR or even Cool Edit dithers when I apply them myself to material that I'm familiar with and compare the examples. I'm not saying that dither manufacturers are all trying to dupe us, but I'm quite certain that most of the dither types these days do an effective job and do not change tones anymore than one cd-r brand versus another does. Let's leave out opinions based on euphonics or the human ear for now, I just want to know how dither can alter frequencies. If I could hear the different dithers or at least the effects of the different brands, I'd still want to know why this is so and how it works.

[bblackwood]

Ronny wrote on Sun, 14 August 2005 10:22

bblackwood wrote on Sun, 14 August 2005 08:19 Ronny wrote on Sun, 14 August 2005 00:20 I hate to bring this up again, but the people that say that one TPDF dither sounds better than another TPDF dither, what exactly does it sound like? Is POW-R TPDF? The Weiss POW-R unit offers four dither types POW-R1, R2, R3 and flat TPDF.

Yes, but where are people saying "one TPDF dither sounds better than another TPDF dither"?

Quote:

Quote: Quote: And when the dither is noise shaped into the inaudible frequencies, please explain the physical reasons why you hear it at all, much less a difference. Curious, you don't hear any diff between UV-22 and POW-R? Which type is that Brad, POW-R1 has a curve that is very simliar to UV22, which BTW, Apogee doesn't call it dither noise. Also the reason why they call it 22 is because it's supposed to be noise shaped above the human hearing range.

OK, so you claim to not hear the difference between UV-22 and POW-R1?

Quote:

Quote: Quote: None of my dither blind tests support claims that people can hear dither when it's applied at the 24th least significant bit. Err, who said anything about dithering to 24? I don't know about where you apply dither, but when I apply it to reduce word from 24 to 16 bit it is applied at the half bit of the 24 bit signal, dither won't do a damn bit of good if you apply it to 16 bit "after the truncation" because quantization distortion from the reduction where it is needed remains.

I'm fairly certain you apply the dither at a level to toggle the LSB of the resulting wordlength, in this case the 16th bit...

[Barry Hufker]

Call me crazy, but if the human ear can pick up extremely small amounts of jitter at extremely low levels, why can't we hear the effects of the dither?  That's not impossible.  And to me it is the same as adjusting an EQ.  I had no bias when I listened to the three types of POW-R.

Ronny, use POW-R on something you know well.  Don't listen for the noise, listen for any timbral changes.  Tell me what you hear!

Barry

And oh by the way, The Town Halo, apparently this thread is no longer about you.  It is about *us*!  But I am glad you've found an answer on your own to your question!!

[Ged Leitch]

Ronny wrote on Sun, 14 August 2005 17:12.

I'm not saying that dither manufacturers are all trying to dupe us, but I'm quite certain that most of the dither types these days do an effective job and do not change tones anymore than one cd-r brand versus another does. Let's leave out opinions based on euphonics or the human ear for now, I just want to know how dither can alter frequencies. If I could hear the different dithers or at least the effects of the different brands, I'd still want to know why this is so and how it works.

The fact that certain users CAN hear the difference means something!!!
I myself when going from 24 bit to 16 bit with POW-R i cannot hear any difference.But thats what i WANT to hear, i.e least amount of signal change possible!!!
Sometimes when I've done a few blind tests I think i can hear differences, then sometimes I can't.It's one of those things that will go on and on.
As for scientific evidence well even if there were none to explain the differences it doe'snt change the fact that SOME people notice the difference.
Until a mass study is conducted as to why dithers can affect audio and others don't all we can do is speculate.

[Ronny]

bblackwood wrote on Sun, 14 August 2005 12:29

Yes, but where are people saying "one TPDF dither sounds better than another TPDF dither"?

Let me rephrase that to say one dither brand versus another, although both brands may be using TPDF.

Quote:

OK, so you claim to not hear the difference between UV-22 and POW-R1?

It's been awhile since I've used either and my I have to admit my testing wasn't complete just some quick observations with me and some friends, but no I don't remember hearing any differences. That's also why I'd like to see an evaluation test done by people that aren't selling their own types of dither. Adminstrated by someone who isn't going to be involved in profitting or losing profit from the results and used real world as Steve touched base on.

Quote:

I'm fairly certain you apply the dither at a level to toggle the LSB of the resulting wordlength, in this case the 16th bit...

As I mentioned dither is applied before the quantizer, if you applied it at the 16 bit, wouldn't the noise cover the dynamic range between -144 and -96dB? Although I've seen 2 LSB dither in the olden days, TPDF for example has a SNR loss of about 4.8dB. When I run dither with no audio playing at 24 bit, I can see the noise decibel level is at -141.4dB on Digichecks RMS +3dB meter. After the quantization the noise is still there but it's at the LSB of 16 bit and below -90dB. I can't hear -90dB at my normal monitoring levels. I'd still like to know how dither tonally or physically effects any frequency that is above -90dB on the final 16 bit file. I don't doubt that people hear it and I might also, if I took a more time to set up some tests that elimnate all factors but the dither applied to the music as we normally use it. Anyhoo, I haven't noticed where the difference between say POW-R3 and IDR are significant to the material that I work on.    

[Ronny]

Barry Hufker wrote on Sun, 14 August 2005 12:44

Call me crazy, but if the human ear can pick up extremely small amounts of jitter at extremely low levels, why can't we hear the effects of the dither?  That's not impossible.  And to me it is the same as adjusting an EQ.  I had no bias when I listened to the three types of POW-R. Ronny, use POW-R on something you know well.  Don't listen for the noise, listen for any timbral changes.  Tell me what you hear! Barry And oh by the way, The Town Halo, apparently this thread is no longer about you.  It is about *us*!  But I am glad you've found an answer on your own to your question!!

I'm not saying that I can't hear the effects of dither, I can versus non dithered truncations, but I can't hear a jump in your lap tonal difference between any of the popular dithers, when I'm working in my own environment on my own gear and with material that I've working on.

[Ged Leitch]

Ronny wrote on Sun, 14 August 2005 18:41

. I don't doubt that people hear it and I might also, if I took a more time to set up some tests that elimnate all factors but the dither applied to the music as we normally use it.

I think thats your answer man.

[Barry Hufker]

Let me clarify.  I don't hear the dither but its effects.  And I only hear the effects when the dither is noise-shaped.  TPDF is TPDF.  There are various forms of noise-shaping.  The paper I posted earlier shows the different curves being used.

I contend, because I can hear it, that "frequencies aren't altered" but that extra energy (from the noise shaped dither) is applied at various frequencies and it is *that* change that is audible.  It is the same (to me) as adjusting an EQ.

As I already showed (by linking to the attached paper) that much of the noise shaped dither is in the audbile range.  Therefore it is potentially audible energy.  I say that while the energy itself may not be audible, it changes the perceived quality of certain sounds or frequencies.  

And I just realized the connection to the "Trivia" I presented.  It was proven to the designers that "dither" applied to WWII bombsights made a visible difference in the accuracy of the sight.  I am saying here then that dither (especially noise shaped) has an audible effect on the music signal.

Barry

[bblackwood]

Ronny wrote on Sun, 14 August 2005 12:41

As I mentioned dither is applied before the quantizer, if you applied it at the 16 bit, wouldn't the noise cover the dynamic range between -144 and -96dB?

Yes.

Quote:

Although I've seen 2 LSB dither in the olden days, TPDF for example has a SNR loss of about 4.8dB. When I run dither with no audio playing at 24 bit, I can see the noise decibel level is at -141.4dB on Digichecks RMS +3dB meter. After the quantization the noise is still there but it's at the LSB of 16 bit and below -90dB.

So how do you think the noise level jumped from -141 to -90? The noise remains because you applied the dither at the LSB of the destination wordlength.

Since the thread has turned remedial, I recommend reading Pohlmann's 'Principles of Digital Audio'...

Quote:

I can't hear -90dB at my normal monitoring levels. I'd still like to know how dither tonally or physically effects any frequency that is above -90dB on the final 16 bit file. I don't doubt that people hear it and I might also, if I took a more time to set up some tests that elimnate all factors but the dither applied to the music as we normally use it. Anyhoo, I haven't noticed where the difference between say POW-R3 and IDR are significant to the material that I work on.

I don't know if I can reliably pass a blind test on flavors of dither, but I do know that flat TPDF has served me well for a long time. I'd certainly be interested in seeing results from the self proclaimed golden ears myself...

[Barry Hufker]

Curious... am I now falling into the "golden ears" and so have to produce proof?  I just want to know what's expected!  AND!! If so, then I am thrilled as this will be the first time I get to be one of the "genetically superior!"

Barry

[TotalSonic]

Barry Hufker wrote on Sun, 14 August 2005 19:55

Curious... am I now falling into the "golden ears" and so have to produce proof?  I just want to know what's expected!

Participation helping out in my dither shootout detailed on this thread - http://recforums.prosoundweb.com/index.php/t/6537/2475/?SQ=5 52e29de07f449975188fa2a9406042c
could be a start.

and - Guessing every single example correctly in a blind test would definitely qualify you as "golden ears"  

Best regards,
Steve Berson

[bblackwood]

Barry Hufker wrote on Sun, 14 August 2005 13:55

Curious... am I now falling into the "golden ears" and so have to produce proof?  I just want to know what's expected!  AND!! If so, then I am thrilled as this will be the first time I get to be one of the "genetically superior!"

Are you proclaiming yourself as golden-eared?

I'm fairly certain I hear the difference between different dithers, but remain unsure if I would actually pass a blind test on it...

[Barry Hufker]

Steve,

Your test looks very interesting.  But I don't think being able to identify every type of dither (with or without) noise shaping is the goal.  I think the goal here is whether there is a sonic difference at all or not.

I have just created 4 twenty-five second samples of music I know well.  I have the original 24 bit and all three grades of POW-R (with noise shaping).  I don't want to convert them to MP3 as I am afraid important differences will be lost or changed by the conversion.

But I don't have a site I can link you to, and the damn things are too large to upload...

And NO!  I am *not* proclaiming myself as golden eared.  I just didn't know if I was being shoved into that category (of dubious quality) by saying I can hear a tonal difference.

I am convinced anyone can hear the difference provided they know what to listen for and if their monitoring chain is good enough.

Mine is simple: RME soundcard to NAIT5i integrated amp to Avalon Professional Monitors (not to be confused with Avalon Design).  So if I can hear it through this, surely everyone else can hear the difference through better.


Barry

[TotalSonic]

Barry Hufker wrote on Sun, 14 August 2005 20:22

Steve, Your test looks very interesting.  But I don't think being able to identify every type of dither (with or without) noise shaping is the goal.  I think the goal here is whether there is a sonic difference at all or not.

Barry -
when you see a "   " next to a statement - what it means is that  I'm joking.  Sorry if the joke wasn't actually funny.  The actual goals of my test are detailed at the end of my post.

Quote:

I have just created 4 twenty-five second samples of music I know well.  I have the original 24 bit and all three grades of POW-R (with noise shaping).  I don't want to convert them to MP3 as I am afraid important differences will be lost or changed by the conversion.

I agree that lossy compression formats such as mp3 are useless for these kind of tests.

Quote:

But I don't have a site I can link you to, and the damn things are too large to upload...

No biggie.  I'll have a blind test available for download from my site sometime soon.  They will still require a fast connection to download though.

Best regards,
Steve Berson