Brickwall Digital Filters and Phase Deviations in Digital Audio

[gene]

Over the years controversy has raged back and forth on whether brick wall filters used in digital audio (namely for CD playback systems) exhibit phase shifts in the audio band (20Hz to 20kHz). Audio critics, pundits and assorted experts have gone back and forth on this issue with out much resolution as I can gather. Some companies have put out their so-called solutions to this problem. It is my desire that this article will put this issue to rest once and for all.

Brickwall Digital Filters & Phase Deviations in Digital Audio

 

[Rip Van Woofer]

Good stuff, Dan. The link to the digital audio site gives a 404 error; check that. I've seen the site before & it explains things well even for us non-techies.

A couple of tangental points that might be of further interest and/or amusement:

Even if there were phase shift, it's worth noting that work by authorities such as Stanley Lipschitz and others (as you know -- left out for space?) has demonstrated that even very large phase shifts are inaudible in all but certain types of test tones, and certainly not in music. So, ha! to the untutored pundits again. The exception being when phase shifts result in frequency response anomalies, and I think the latter problem mainly applies to speakers. Of course, that means we're not actually hearing the shift per se but a side effect of it.

An amusing development among the high-enders is a growing enthusiasm for (are you ready?) "zero-oversampling" DACs. Yep, after years of whingeing about the supposed inadequacies of the basic Nyquist-Shannon theorem to fully resolve the "air" and "layering" and "microdynamics" and being "harsh" and "edgy", suddenly they're singing hosannas to straight 16/144, no chaser. Some even say it's "almost as good as vinyl!"

Crazy world, huh?

Now, how about some straight talk on jitter?

 

[gene]

Rip;

The links have been fixed, thanks for noticing.

 

[16again]

Lineair phase in the pass-band is one good thing, but what about the Gibb's effect? Any brick wall filter (even the lineair phase FIR!) will exhibit ringing on impulsive signals.
This ringing effect gives you a kind of exponentially dampened 20kHz sine tone after any impuls present in the music.
Since this tone is too high for my ears, I don't hear the ill effects. However , I once tested using a 10kHz brick wall filter, and the ringing was really annoying.
So where audio bandwidth is to be reduced (bit saving) I prefer less steep filtering.

People with better ears than me might even be able to hear this 20kHz ringing.

 

[Rip Van Woofer]

Lineair phase in the pass-band is one good thing, but what about the Gibb's effect?

It only happens on BeeGees recordings.

(sorry, couldn't resist!)

 

[Dan Banquer]

Brick Wall Filters

Yes; the Gibbs effect is very real and yes the transient response is poor, but that has nothing to do with phase as far as anyone can tell. What "saves" the Redbook Format is that the great bulk of microphones used in recording roll off very quickly after 20 kHz and their are low pass filters in front of the ADC.
There is a white paper from Lavry Engineering on this subject that is in a pdf file. It's rather technical, but most appropriate.
Gene and all; can we have this pdf on file here at audioholics for people to refer to?
16again; if you send me an email address I will send the pdf to you.
d.b.
banquer@erols.com

 

[WmAx]

THis is all very interesting. Hoewver, let's all refer to one of the early perceptual tests determing audibility of a specific filter rate as could be emploed in DACs, that BTW used test signals tht maximized audibility:

Perception of Phase Distortion In Anti-Alias Filters
D. Pries, P.J. Bloom
AES, 74th Convention, New York, 0ctober 8-12, 1983

Additinally, filter slopes simulating what is found in typical RBCD DACs were used in the recent NHK Labs ultrasonic tests. This test used various musical passages. These tests used microphones/recording and playback systems fully capable of recording/reproducing the ultrasoinc information:

Perceptual Discrimination between Musical Sounds with and without Very High Frequency Components
Nishiguchi, Hamasaki, Iwaki, Ando
AES, Preprint 5876, 2003

Neither of the above tests could not obtain a signficance rate that concluded audibility.

Dan, please send me the pdf to which you refer: wmax@linaeum.com

TIA.

-Chris

Yes; the Gibbs effect is very real and yes the transient response is poor, but that has nothing to do with phase as far as anyone can tell. What "saves" the Redbook Format is that the great bulk of microphones used in recording roll off very quickly after 20 kHz and their are low pass filters in front of the ADC.
There is a white paper from Lavry Engineering on this subject that is in a pdf file. It's rather technical, but most appropriate.
Gene and all; can we have this pdf on file here at audioholics for people to refer to?
16again; if you send me an email address I will send the pdf to you.
d.b.
banquer@erols.com

 

[Dan Banquer]

Brick Wall Filters

I have sent the Lavry Engineering pdf file. I found it to be very enlightening. If you did not revceive it, ping me again.
d.b.

 

[WmAx]

I have sent the Lavry Engineering pdf file. I found it to be very enlightening. If you did not revceive it, ping me again.
d.b.

Thank you. It was an interesting paper.

I would like to make a correction to my last post: Disregard the NHK Labs paper(preprint 5876). I was not thinking of the correct paper.

I meant to refer to the peer-reviewed standards paper:

Which Bandwidth Is Nesecesarry for Optimal Sount Transmission?
Plenge, Jakubowski, Schone
JAES, 1980, March, Volume 28, Number 3, Pages 114-119

BTW, Dan, if you need a copy of one of the papers let me know and I'll email to you.

-Chris

 

[Dan Banquer]

Brick Wall Filters

Thanks; I'll take you up on that offer. Please send it to banquer@erols.com
d.b.

 

[WmAx]

Thanks; I'll take you up on that offer. Please send it to banquer@erols.com
d.b.

Sent.

For the reference of others: the standards paper mentioned uses the following filters and determined which ones were audible among a group of 43 test subjects, of which 21 were audio professionals with trained hearing:

LC, 13th Order - 15khz
Cauer,9thorder - 20 kHz
PCM-systemDBP* - 15kHz
Elliptical, 7thorder - 18kHz
Tscheby§cheff, 8thorder - 20 kHz
Tschebyscheff, 8th order - 15 kHz
LC, 2 x 7th order - 20kHz

The audibility of upper band near 20kHz filters never actually came up, as lower bandwidths demonstrated transparency in this test. This test used special test signals to maximize sensitivity to the filters.

Here is the main body of conclusion(ignore typgraphical errors, this was acrobat-grabbed text from an image scan of the pages which results in some errors):

3. CONCLUSIONS
Fromtheseresultsand'fromthoseof otherauthorsone
maydeducethefollowing:
1) A bandwidth limitation at 15 kHz is not discriminated
significantly more easily than a limitation at 20 kHz [4- 7].
2) If a limitation of the transmission range is recognized
in a fewexceptionalcases,this is not due to thephaseor
group delay distortion of the filter, but to the lack of the
extremelyhighfrequenciesin the individualaudiblerange
[2, 4, 8- 11].
Moreextensiverequirementswith referenceto band-
widthseemto be exaggeratedif theyare basedon the
properties of the sense of heating for the following reasons:
The question cannot be to propose an "absolutely
inaudible low-pass filter." There might still be subjects
exposed to a 20-kHz filterwho will recognize differences
when special test conditions are performed. Even with a
15-kHz filter it is accepted that differences will be heard
only in rare exceptional cases. Such conditions as extraor-
dinarily strong high-frequency signal partials, a sufficient broadband electroacoustic transducer, listeners with a par-
ticularly wide frequency range of hearing, and the possibil-
ity of direct signal comparison coincide extremely seldom,
even in recording studios.

Since this test may peak interest in ultrasonic content detection, I now refer to the NHK Labs test from earlier:

Perceptual Discrimination between Musical Sounds with and without Very High Frequency Components
Nishiguchi, Hamasaki, Iwaki, Ando
AES, Preprint 5876, 2003

This test, using over 30 audio experts as test subjects, did not result in positive results for audibility of ultrasonic content.

-Chris

 

[av_phile]

Sent.

This test, using over 30 audio experts as test subjects, did not result in positive results for audibility of ultrasonic content.

-Chris

Ofcourse... audio experts. Correct me if wrong, but they must be in the 35+ age group. People in that age group are not known to possess the ability to discern above 17khz. Why not use teenagers?

 

[WmAx]

Ofcourse... audio experts. Correct me if wrong, but they must be in the 35+ age group. People in that age group are not known to possess the ability to discern above 17khz. Why not use teenagers?

I can only speculate: They valued the impressions of people trained in audio ratehr then unexperienced people, who may have lesser ability to discern audio differences. Some teenagers were used in the test, though a small number.

Since you are interested in the ages, here is a summary excerpt about the ages from the article:

The subjects were 30 males and 6 females including 33 experts in audio engineering, 2 students, and the musician who had recorded the sound stimuli. They consisted of 3 teenagers, 12 in their twenties, 16 in their thirties, 3 in their forties and 2 in their fifties.

-Chris