R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: [1]   Go Down

Author Topic: intersample peaks  (Read 8090 times)

zetterstroem

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 765
intersample peaks
« on: February 28, 2005, 08:21:12 PM »

is this true?

http://www.tcelectronic.com/media/nielsen_lund_2000_0dbfs_le .pdf
http://www.cadenzarecording.com/papers/Digitaldistortion.pdf

as i wrote elsewhere: (assuming of course that daw d/a's behave much like consumer ones)

i think this paper implies that all system not having oversampled peak meters do not accurately display the peaks.....

that means 98% of all DAW out there!!!

and it also means that if you mix in the analog domain and all of your signals are full code then you essentially have maybe 64 d/a's distorting at the the same time......

my conclusion is..... if this is true almost all digital audio up until now is useless!!  

by the way .... ammitsb
Logged
Noting the music industry's complaints that illegal downloading means people are getting their music for free, he said, "Well, why not? It ain't worth nothing anyway." (b.dylan)

RobertRandolph

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 54
Re: intersample peaks
« Reply #1 on: February 28, 2005, 08:52:23 PM »

It is true.

But good DA can handle it.
Logged
Now I know a disease that these doctors can’t treat
You contract on the day you accept all you see
Is a mirror, and a mirror is all it can be
A reflection of something we’re missing

ammitsboel

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1300
Re: intersample peaks
« Reply #2 on: March 01, 2005, 09:07:14 AM »

So a good DA is something like the one i have in my Tibook?
Logged
"The male brain is designed for ecstasy" -Dr. Harvey "Gizmo" Rosenberg

zetterstroem

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 765
Re: intersample peaks
« Reply #3 on: March 02, 2005, 04:21:35 AM »

any opinions???

should we go about and replace all d/a's in the world??

or should we just use oversampling meters on every output??

don't anyone see the problem here? or do you just don't care??

it would be nice if someone with more digital knowhow than me were to respond....

Logged
Noting the music industry's complaints that illegal downloading means people are getting their music for free, he said, "Well, why not? It ain't worth nothing anyway." (b.dylan)

David Satz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 661
Re: intersample peaks
« Reply #4 on: March 02, 2005, 08:07:32 AM »

Hello, Zetterstroem--I hope that you haven't really stayed awake all this time from anxiety.

Yes, it's true that in a digital audio system--let me say this as carefully as I know how--the analog reconstruction of a signal can occasionally contain peaks which briefly exceed the peak level of a full-scale sinusoid. These peak values are real parts of the audio signal and should be preserved as faithfully as possible.

However, any well-designed D/A converter will do this already.

This situation directly affects the working definition of "enough headroom" for a DAC. The clipping point of the analog circuitry should not be right at 0 dBFS, for example; a few dB of further headroom must be allowed. And digital filters must have adequate numeric range to prevent overflow or underflow. But none of that is especially mysterious or esoteric. It's important, but so is every other part of a designer's engineering responsibility.

There have been dynamic range issues that the audiophile public didn't understand ever since the early days of "oversampling" DACs (DACs with digital filters). It's one of the things which taught me years ago never to believe the golden ear publications too much when they go on their hysterical binges. Back in the mid-1980s when controversy over DAC quality was all the rage and the hot issue was linear vs. "oversampling" converters (e.g. early Sony vs. Philips CD players), the self-appointed arbiters of "what sounds good" all swung to the side of the systems that had the worst problems with numeric overflow and underflow, cycling noise, etc. Evidently those people really couldn't hear what was going on. But their self-confidence was so high that they influenced a whole generation of believers, who have regurgitated many of the same viewpoints and arguments ever since.

Many people have their simplified, visual/mental model of how digital audio works, and they cling to that instead of hearing what actually occurs--meanwhile bashing anyone who measures anything or observes the behavior of actual circuits or systems.

As an example: Anyone who criticizes digital audio because it has the problem of "intersample overload" but also claims that higher sampling rates are needed because digital audio is "deaf" to what goes on between the moments at which sampling occurs, has soiled his own pants. The problem of "intersample overload" (to the extent that it is a problem) exists precisely because digital audio is not "deaf" to what goes on between the moments at which sampling occurs.

But a firm belief in that "intersample deafness" is still very widespread among audiophiles and even some professional engineers. So to those people I say, fine: If you believe in "intersample deafness" then kindly ignore the issue of "intersample overload," because it can't possibly happen in your world.

--best regards
Logged

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: intersample peaks
« Reply #5 on: March 02, 2005, 05:49:41 PM »

ZETTERSTROEM wrote on Mon, 28 February 2005 20:21

is this true?

 http://www.tcelectronic.com/media/nielsen_lund_2000_0dbfs_le .pdf
 http://www.cadenzarecording.com/papers/Digitaldistortion.pdf






Yeah, it's true. Gerzon knew about it long ago and I believe it has been written up in Waves' literature as well.

While a zany signal (square wave) could be made to theoretically produce +3 dBFS on the analog side, with most practical musical sources I've found the dBFS+ signals can be prevented by dropping the ceiling on an L2 or any decent digital limiter, to -0.3 dBFS. TC designed their new brickwall limiter to prevent the intersample peaks and it's very effective so you don't need to drop the ceiling (on the TC it's just an output fader, which you can leave at 0 dB).

BK
Logged
There are two kinds of fools,
One says-this is old and therefore good.
The other says-this is new and therefore better."

No trees were killed in the sending of this message. However a large number of
electrons were terribly inconvenienced.

RobertRandolph

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 54
Re: intersample peaks
« Reply #6 on: March 02, 2005, 06:51:22 PM »

The voxengo and izotope limiters will not have trouble with intersample peaking either.

Voxengo limits on an oversampled signal and the izotope (ozone)'s uses an oversampled signal for level detection but performs amplitude changes on the original signal.

I beleive L3 is similiar.

For what it's worth I have not caught any of this phenomenon using the sawstudio levelizer... but Im not sure of it's inner workings.
Logged
Now I know a disease that these doctors can’t treat
You contract on the day you accept all you see
Is a mirror, and a mirror is all it can be
A reflection of something we’re missing

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: intersample peaks
« Reply #7 on: March 02, 2005, 07:03:13 PM »

bobkatz wrote on Wed, 02 March 2005 22:49

ZETTERSTROEM wrote on Mon, 28 February 2005 20:21

is this true?

   http://www.tcelectronic.com/media/nielsen_lund_2000_0dbfs_le .pdf
   http://www.cadenzarecording.com/papers/Digitaldistortion.pdf






Yeah, it's true. Gerzon knew about it long ago and I believe it has been written up in Waves' literature as well.

While a zany signal (square wave) could be made to theoretically produce +3 dBFS on the analog side, with most practical musical sources I've found the dBFS+ signals can be prevented by dropping the ceiling on an L2 or any decent digital limiter, to -0.3 dBFS. TC designed their new brickwall limiter to prevent the intersample peaks and it's very effective so you don't need to drop the ceiling (on the TC it's just an output fader, which you can leave at 0 dB).

BK




Yes there is a problem, and I agree with you that the best way to prevent it is to stay a couple of dB from clipping when doing AD, as I believe was stated by Bob a few times.

Regarding the mechanism of the problem and the reasoning behind it, I do not think it is well understood, so I will try to clarify a couple of points:

1. I heard people refer to it as an analog filter problems. I do not think it is correct. An analog filter, when fed a clipped sine wave, will track it down pretty well, and the exception is that the sharp corners on the wave (where the clip starts and stops) will be somewhat "rounded". Sharp corners require infinite bandwidth, which the analog filter does not have (thus the rounding).

I do not see any reason or mechanism by which the analog filter will try and "fill in" the clipped part. Analog filters do not have inertia and they do not resonate.

2. The problem occurs during the interpolating filter process. An unity gain FIR (LPF) will have the sum of its coefficients equal 1. That way, if one enters all 1 (full scale) DC signal, each coefficient is multiplied by 1 and if the sum of the coefficients is 1, the output will also be 1.
However, an FIR will have some of it's coefficients positive and the rest are negative (a sinc function envelop). The sum is 1 or less as long as the signal is "legal".

But what would happen if we could for example do the following: We line up the data (momentarily) so that all the positive coefficients are multiplying data values of +1, and all the negative coefficients are lined up with data values of -1?      
The sum will me greater then 1, because all the negative coefficients became positive for that specific moment. Some filters will output a value of over 2!

Of course such a worse case signal is impossible to come by - it is in fact the absolute value of the sync function of the FIR multiplied by a huge value (such as 2^24). But a clipped signal, especially with high frequency content can yield some serious "overs", as much as 3dB on rare occasions. The data does not need to be all clipped (1's and -1's). The problem has a lot to do with both clipping and alignment of the data cycles with the filter coefficient cycles. It works out to a bigger problem for higher frequencies (with a clip).

The reports of 12dB overs are plain wrong, and 6 dB is also wrong. But 3dB can happen in the real world.

The interpulator (upsampler) designer can not ignore that effect. They receive say a 44.1KHz signal that is, by definition, music to be played. You up sample and if the signal was clipped the outcome could exceed the limits of the compute engine (+/-1). That waveform will not even try to follow the original signal waveform. It will follow the FIR coefficient curve interaction with the signal. But one way or another, you run out of code (over 1 stays at 1, under -1 stays at -1). The analog part of the DA gets an upsampled clipped signal.

The best way to avoid it is to scale the data down by say 3dB before the interpolation. But given that many DA's in homes and cars are not scaled, it would be good if the recording and mastering stay away from full scale by a couple of dB.
     
This is an engineering comment. I really do not want to speak about mastering - I am not a mastering engineer. I know a lot of mastering people over drive their converters to make the music louder (on a relative scale). They probably don't think about the fact that the end customer may have up to 3dB more clipping then a studio with quality DA.

Regards
Dan Lavry
Logged

zmix

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2828
Re: intersample peaks
« Reply #8 on: March 03, 2005, 09:14:19 AM »

Here is another article:

http://www.airwindows.com/analysis/SlewRates.html


 "A new type of distortion in Audio CDs has been identified- it can be called 'Gibb effect', taking its name from the more abstract mathematical phenomenon which it exemplifies. This distortion works as follows- if you have two adjacent samples that are not themselves distorted, but they differ enough, the reconstruction filter of the CD player is forced to produce an output that goes beyond the DAC's clipping threshold- sometimes wildly beyond. The result is an intense burst of typically transistor distortion, of varying ugliness depending on the headroom of the DAC: even in the absence of DAC clipping, the overshoot produced by the necessary brick-wall filters can be a ringing of very high amplitude. Certain sounds are particularly suited to illustrating this effect, such as tambourine (which produces a striking change in timbre when it begins to distort with Gibb effect).

 There are varying ways to deal with this, including slew limiting, the use of certain classic compressors (including setting them for no actual compressing- using them as a peak clamper), and simply turning the volume down while mastering (in this day and age, this last option is seen as unworkable, as it can require the hottest samples to be more than 3 db down from full scale!).

 My purpose in writing these notes is to bring attention to the severity of this problem, which is poorly understood. Given that this affects all D/A converters, including fancy ones with elaborate digital oversampling (in some ways the problem affects these more harshly than old D/As with crude analog filtering!), including those in radio broadcast after the air chain, and basically every CD player out there that's ever existed, this problem needs attention if the CD isn't to be abandoned as a total failure...."



And it goes on for a few pages like this....

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: intersample peaks
« Reply #9 on: March 03, 2005, 01:14:58 PM »

zmix wrote on Thu, 03 March 2005 14:14

Here is another article:

http://www.airwindows.com/analysis/SlewRates.html


 "A new type of distortion in Audio CDs has been identified- it can be called 'Gibb effect'...
And it goes on for a few pages like this....



Well, I already said that Gibbs lived a long time ago, he died in 1903.

The paper you posted is mostly about listening to clipped material. The mention to Gibbs seems to be out of place.
My last post explains the mechanism of what is going on. It is very easy to verify and study. It can be done in hardware or with software simulation. I just use a computer math package for the simulation.
For starters you can take a full scale sine wave tone at 10KHz. Multiply is by say 1.5 or 2 and clip it. Take a set of LPF FIR and run the clipped signal through it. Look at the output wave, and it is all there right in front of your eyes.

Bob Katz said Gerzon knew it a long time ago. So did I, the first time I designed coefficients for an upsampler.

The reports of 12dB or even more "overs" must be exaggerated. Anyone with a set of FIR coefficients can see immediately what the absolute theoretical over is (make all the coefficients positive and sum them). That is probably not much beyond  6dB or so for most FIR's. Imposing a couple of other restrictions on what is possible will get you to near 3dB maximum.

But 3dB is a lot and should not be ignored.

Regards
Dan Lavry
www.lavryengineering.com


Logged
Pages: [1]   Go Up
 

Site Hosted By Ashdown Technologies, Inc.

Page created in 0.079 seconds with 20 queries.