Print

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

[danlavry]

Peter  Oxford wrote on Tue, 20 July 2004 08:10

http://www.dcsltd.co.uk/papers/aes97ny.pdf Best Wishes Peter Peter Poyser

You brought up the dCs paper by Mike Story so I will explain what is wrong with the white paper titled  "A Suggested Explanation For (Some Of) The Audible Differences Between High Sample Rate And Conventional Sample Rate Audio Material" found on the dCs website.

I would first recommend that you read my paper "Sampling Theory" that explains the immediate relationship between bandwidth and impulse response width. It is entirely incorrect to suggest than an impulse width and bandwidth are separate issues.

I have been bothered by that "Suggested Explanation" for quite some time. When doing analysis, appropriate tools must be used for each job.
For example:
If I borrow a million dollars I have to pay back at least a million dollars.  It would be great if I could choose a different method to pay such as to take the log of a million which is $6.  The lender will scream "bloody murder" because I used the "wrong tool." I used a logarithmic scale because it is advantageous to me.

Well, it is OK to talk about audio volume in log scale. The ear is nearly logarithmic, when it comes to relating POWER and volume, so the log scale is a good tool to use.  But it is wrong to use a logarithmic scale of a TIME DOMAIN WAVEFORM because that does NOT correspond to the ear perception of a waveform.

We do not take the log of the waves that make music. Doing that would create unbelievable distortions of both harmonics and volume perception.

The impulse response of a 96KHz bandwidth is half the width of 48KHz. The paper by Mike Story from dCs does take a log of a time domain function. This is very wrong, and leads to wrong conclusions. If you arbitrarily put me in front of a warped mirror to show me that I am very thin or fat, the picture is distorted.  This presentation takes liberties beyond engineering and science, and the use of a log scale with a totaly arbitaray -30dB referance distorts the picture totally. It states that:

"the 44.1 and 48KS/s filters spread the audible energy over 1msec or more. The 96KHz filter is much better, keeping the vast bulk of energy within 100uSec. The 192KS/s filter can be very good indeed, keeping the energy within 50uSec."

This statement may make one believe that the width ratio going from 48 KHz to 96 KHz is 10 to 1 instead of 2 to 1.

Also, any worthwile EE knows that there is NO POINT in offering 96KHz bandwidth when the mic and speakers are 20KHz or so. Any EE knows, or should know that the time domain behaviour and the frequency domain are one of the same, just a different way to present and view the same thing. It is in my paper, which is a good paper, not full of holes.

My guess is that they will continue to display that "paper" in their web. The honorable thing to do is to publicaly appolgise for diservice they have done. This very faulty paper was one of the rationals for doing 192KHz conversion. I learned to not hold my breath where marketing plays a role  

BR
Dan Lavry  

[ted nightshade]

Not sure you got my point, Johnny- there's this intangible thing of emotion in music, but intangible may not be the word, because I feel it alright. Unquantifiable may be it- Nika says everything in the world is quantifiable, I say bullshit. How to quantify beauty? In millihelens? The increment of beauty needed to launch a single ship?

I've run a process or turned a knob and had that emotional impact greatly reduced in a recording. Most unnerving...

And then sometimes it can come across through some crummy chains indeed, regardless.

That's something that's there live, that's not a frequency response issue, I don't think anyhow, that we won't get on a recording by capturing 100kHz or whatever, and it's got to be more important than whatever trumpet or cymbal harmonics are clear up there.

[bobkatz]

[quote title=fuze wrote on Mon, 19 July 2004 18:03]

Quote:

any of the other problems with digital such as zero headroom

Zero headroom?  Since "headroom" is strictly a matter of definition, headroom is ABSOLUTELY no problem in digital recording. It's the people who use it and don't understand digital audio that causes the overload or distortion problems.

For example, "soft clipping" circuits are totally unnecessary, do not help and hinder sound quality. In an attempt to emulate analog soft saturation, they end up neither here nor there, producing a kind of distortion that is not pleasant and really hides the fact that too many people are using too high of an average level.

Using an average level which is too high is what creates a "headroom" problem, that is all...

[Level]

Hi Bob! Remember in the 80's when -17dB was hitting it pretty hard with the average? I considered it 0 VU then.

[Paul Frindle]

[quote title=bobkatz wrote on Tue, 20 July 2004 23:49]

fuze wrote on Mon, 19 July 2004 18:03

Quote: any of the other problems with digital such as zero headroom Zero headroom?  Since "headroom" is strictly a matter of definition, headroom is ABSOLUTELY no problem in digital recording. It's the people who use it and don't understand digital audio that causes the overload or distortion problems. For example, "soft clipping" circuits are totally unnecessary, do not help and hinder sound quality. In an attempt to emulate analog soft saturation, they end up neither here nor there, producing a kind of distortion that is not pleasant and really hides the fact that too many people are using too high of an average level. Using an average level which is too high is what creates a "headroom" problem, that is all...

Yes I would agree with that sentiment theoretically.

However building mixing applications that contain signal processing without ANY headroom between processes or even the means to properly display the resulting reconstructed signal level, is a recipe for clipping internally, or even at the final result.
Whilst one is permanently fighting overs and red lights every time you do almost anything and adjusting gains and levels (and thus your whole balance) to avoid them - you are not concentrating on the artistic business at hand.

This is the single biggest pitfall of specifying (or at least encouraging) an operating level which is just one count out of 2^24 less than flat out(!) - where even that level may not be legitimately decoded on reconstruction, depending on the kind of processing applied to the signal before hand

[PP]

[Duardo]

Quote:

Just at the moment, I am in the process of some very complex and quite involving tasks so please forgive the extreme shortness of this response.

Can't speak for everyone, but I forgive you.

-Duardo

[bobkatz]

Level wrote on Tue, 20 July 2004 18:59

Hi Bob! Remember in the 80's when -17dB was hitting it pretty hard with the average? I considered it 0 VU then.

Hi, Bill...  That's the level of Glenn Meadows' excellent master of Lyle Lovett's Big Band album from 1989. How quick we forget how great it can sound!


BK

[bobkatz]

Paul Frindle wrote on Tue, 20 July 2004 19:36

However building mixing applications that contain signal processing without ANY headroom between processes or even the means to properly display the resulting reconstructed signal level, is a recipe for clipping internally, or even at the final result.

Oh yes oh yes. The poster was strictly referring to the end format where the concept of headroom is meaningless. However, as you say, when you deal with internal stages, dealing with internal headroom ABOVE 0 dBFS is absolutely required.  But that's a designer's consideration, and 9 times out of 10, the designer's work should be good enough that the end user doesn't even know the designer put in extra internal headroom above 0 dBFS.

BK

[Johnny B]

Thanks Paul for clarifying the "Zero Headroom" issue.

Perhaps both you and Bob K. can discuss it here a little more.
I've noticed that a great many people get easily confused, esp. those making the transition from analogue to digital. I sense it would be helpful.

 

[Paul Frindle]

bobkatz wrote on Wed, 21 July 2004 04:28

Paul Frindle wrote on Tue, 20 July 2004 19:36 However building mixing applications that contain signal processing without ANY headroom between processes or even the means to properly display the resulting reconstructed signal level, is a recipe for clipping internally, or even at the final result. Oh yes oh yes. The poster was strictly referring to the end format where the concept of headroom is meaningless. However, as you say, when you deal with internal stages, dealing with internal headroom ABOVE 0 dBFS is absolutely required.  But that's a designer's consideration, and 9 times out of 10, the designer's work should be good enough that the end user doesn't even know the designer put in extra internal headroom above 0 dBFS. BK

Whilst something can work theoretically - people all too readily ignor (or are unaware of) the secondary effects of using it and the effect it has on the USER who is operating it to get the right perceived result. A very common problem that is rarely discussed and is just as important (sometimes even more important) as the theoretical integrity of the application.

[PP]

[steve parker]

Quote:

Subsequently people feel over confident in pushing ever upwards toward the clipping line, for a louder track when if they left just a little more headroom it would often be a cleaner and better way to go.

too damn right peter!

somewhere (i can't find it) i read a paper nika wrote long ago on the possibility of reconstruction  overs, given that meters give sample value overs rather than waveform overs.
this is one reason why i give up my last bit and hit -6 for theatre playback cds.
and why i go no where near odb when recording with 24 bits.

perhaps nika, you can link it?

steve parker

[Level]

We all must remember the definition of insanity.

"Doing the same things over and over and expecting different results"

[Nika Aldrich]

Steve,

It is here:

http://www.tllabs.com/index.php?option=content&task=view &id=37

It looks like you need to register to get it now?  Hmm.

Nika.