Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Audibility of "typical" Digital Filters in a Hi-Fi Playback  (Read 324049 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #225 –
Quote from: amirm on
FYI, I passed all of my double blind ABX tests using my laptop and headphones.
You probably think you're getting a hard time here, but you should take it as a compliment that no one has challenged you on this. The assumption must be that, given your employment history, you must know what you're doing. Anyone else who arrived on HA and made this claim would get a really hard time, due to a long history of drivers, sound cards, and transducers that don't take kindly to ultrasonic content. ABXing hi-res vs 16/44.1 on such crappy equipment is easy, and it's not because hi-res is audibly better.

Cheers,
David.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #226 –
Are there some details on the system amir used? Which operating system/version, which headphones, which soundcard, which sound API, which DirectSound common sampling rate was configured (if applicable) ... ?
"I hear it when I see it."

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #227 –
Quote from: amirm on
Quote from: greynol on
We still have no idea whether any differences, supposing that they were detected reliably, were the result of the content or the hardware.

The thesis of the paper is neither.  Those were preconditions for the test.  The test hypothesizes that filtering from 192 khz down and conversion with or without dither may be audible.  And their listening test showed that to better than 95% confidence, they are.


As I read the paper there was no conversion of the data:

"We wanted to emulate the frequency and impulse responses of such converters while remaining in the 192 kHz domain. Both un ltered and ltered audio samples were sent to
the replay system at 192 kHz..."

Umm Amir due to this and other false claims like it I've read here lately I gotta ask - did you ever actually read the paper?

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #228 –
Quote from: amirm on
There are no obvious flaws.

Wishful think fallacy.
1) Test deliberately doctored with (RPDF) dither against BS own recommended best practices (TPDF) to fabricate possible positives.
2) Positives may have been the results of system artifacts, loudspeakers, switching software, etc., not just the pathological dither doctoring.

The test may have shown what it set out to do (fabricate positives with pathological dither). Or may it not have.

cheers,

AJ
Loudspeaker manufacturer

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #229 –
Quote from: amirm on
Quote from: ajinfla on
I'll remind you amir, there is not a whit of evidence to suggest 16/44 playback is non-transparent in home audio end use.

AJ, you are typing in a thread where we are discussing an award winning, peer-review paper showing such differences can be audible.  The time for making that statement has come and gone.  You should look for other talking points, battles to fight.  This one is done.
Quote
Dubious dither doctored tests and unsupervised online games are not evidence against this.

I hear you.  You are wrong.  But I  hear you.  Buy the paper, read it and try again.


Amir I've done exactly that, bought the paper, loaded it on my Kindle and several PCs, and read it any number of times, quoted the paper repeatedly in this thread and watched my quotes and comments be totally ignored by you.

Furthermore, I've posted evidence that suggests that you yourself don't properly grasp the content of the paper, particularly the provenance of the samples that the listeners listened to. You said that they were "Converted" but its quite clear from the paper's text that they were presented at the same sample rate and format as the 24/192 source material, and were never converted to any other sample rate or format.

For another example of well documented critiques that you have ignored Amir, the paper seems to almost completely obfuscate the listening environment. Over on AVS you harshly critiqued the Meyer-Moran tests for their non-compliance with BS 1116, but this paper has the same fault and yet you continue to praise it.  At least Meyer-Moran have published photographs of their listening environment (which you also trashed).

Finally Amir, I documented how this paper completely distorted ABX testing, and again I see no response from you.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #230 –
Quote from: amirm on
The test hypothesizes that filtering from 192 khz down and conversion with or without dither may be audible.

Well, I'm glad we all agree now. 
But did we really need this extensive BS paper effort to demonstrate what we all already knew?

cheers,

AJ
Loudspeaker manufacturer

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #231 –
Quote from: 2Bdecided on
Quote from: Arnold B. Krueger on
It is tragic that such high accolades are being heaped on a document that libels and/or ignores well known previous technology (ABX testing, BS1116 recommendations, etc.)
Come on Arny, it's a good paper. It accurately reports the lengths they went to in order to get some barely significant results.


Agreed.

BTW I don't know exactly how significant the results were because I can't find any clear numeric tally of the number of trials.  I do find the following:

"
2.4. Listeners
Eight listeners took part in the test, seven of whom
were male. Most of the listeners were audio engi-
neers, and their ages ranged from 25 to 65. All re-
ported normal hearing, although this was not tested
formally.

2.5. Procedure, trials and blocks
The procedure used was intended to combine high
statistical sensitivity and speci ficity with ease of use
by listeners.

In a given trial, listeners were presented with the
same extract of music in two intervals, and were
asked to decide whether or not the extracts had been
processed identically. The fi rst interval always con-
tained an un ltered extract, the so-called \reference
sound", and the second interval contained \sound
X", which was either identical to the reference sound
(un ltered), or had been ltered in some way. Lis-
teners were not limited in the number of times they
could listen to each extract before deciding. Visual
feedback was given as to whether the answer given
was correct or incorrect.

The extract presented in each trial was selected ran-
domly from the 17 sections into which the piece had
been divided based on musical phrases. Twelve trials
were presented within a block", with the results of
the last 10 being counted; the two uncounted trials
were included in order to familiarise listeners with
the task and processing before beginning the test.
For each block, the type of ltering used was the
same. Each listener completed 2 blocks for each con-
dition, giving a total of 12 blocks per subject.
"

I'm still trying to decode the above. Can anybody help me?


Quote
You can't blame the AES, or the paper's authors, for the frankly embarrassing way that Amir is lauding it about, or the way he's combining the results with some other seriously flawed tests to imply that hi-res audio is now easily and routinely ABXed.


Again agreed.  The overall percentage of correct trials was very low - near 50%. I know from personal experience that what most people consider to be audible effects and even many fairly subtle ones can usually be detected more than 90% of the time.  The glossed over asymmetries between the Meridian test environment and a normal or even exceptionally good residential listening system make any possible positive results irrelevant to the world of audiophilia.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #232 –
As I understand it 8 listeners completed 12 blocks each.

Each block consists of 12 A-X trials (where X=A, the original, or X=B, the processed A) but the first 2 trials were always (?) dismissed.
Visual feedback for each trial if listener chose correctly.
Each trial had a randomly assigned segment from a song.
Within each block, B would be processed the same way.

Now they say "2 blocks for each condition". What is a condition, are those the different processing setups (like different lowpass filters)? I guess so.
"I hear it when I see it."

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #233 –
Quote from: Kohlrabi on
Quote from: amirm on
Quote from: ajinfla on
There is not a whit of evidence that any audiophile, anywhere, can hear any "benefit" with bigger files, in their system, in their room.

There is no evidence that they are not either.  So?
Talk about shooting yourself in the foot with a rocket launcher. Your failing to understand how to formulate a valid null-hypothesis puts any analysis of the paper in question in serious doubt. I didn't read the paper, yet, but if the authors use the same "scientific" approach we're done here.

Thank you for your comment.  The day you learn to debate AJ is the day you give me advice.  He doesn't come from a logical point of view or interest in the topic.  He comes after you personally to get a rise  out of you and have fun doing it.  While I am answering him with a straight face, I am hoping you see the reality of this.  If not, then when you are dealing with him you can try to use the "scientific approach" and see how far that gets you.

To return the favor, you should know I have very little appreciation for people who enter a discussion about a paper but think they don't need to first read it before commenting on it, and instead focus on personal remarks of what the person knows or doesn't know.  See below on best example of why you don't want to be in that group:

Quote
To make it clear, the only meaningful, valid null-hypothesis is that all systems sound the same until proven otherwise.

And that is what the Stuart's study did: to better confidence than 95%, it showed that in 7 out of 8 tests of different filters and dither strategies, the difference was heard.  The null hypothesis is rejected.

These are differences that heretofore were said to be below JND.  If challenged, demand was made of a double blind ABX test.  That demand was met with running the tests offered.  That was not enough.  They wanted published tests.  Well we have a published test now in the form of Stuart's test.  The game of Calvinball then turned into, "oh, but it is not peer reviewed."  As I showed, not only was it peer reviewed but won an award among all the others presented at this year's AES conference.  An evidence that was visible on the first page of the paper which you say you have not read.

Seems like we have no interest in learning something new.  This is a thread about a new paper and listening test.  The remarks instead are judgemental about which party should be supported on principal and personal bias.  That is not how science works.  One of the core tenets of scientific method is being unbiased and neutral at all times.  Don't take sides and let that cloud your judgement.  Here is a great video from one of my favorite Physics professors explaining what the scientific method is:

https://www.youtube.com/watch?v=IX2NPyvYz3w

If you are too impatient to watch a 12 minute vide, jump to minute 6 and see the point I just made.

Amir
Retired Technology Insider
Founder, AudioScienceReview.com

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #234 –
Quote from: Arnold B. Krueger on
Amir I've done exactly that, bought the paper, loaded it on my Kindle and several PCs, and read it any number of times, quoted the paper repeatedly in this thread and watched my quotes and comments be totally ignored by you.

Yet you missed the fact it was peer reviewed and won an award before calling it and its author a bunch of names.  More below

Quote
Furthermore, I've posted evidence that suggests that you yourself don't properly grasp the content of the paper, particularly the provenance of the samples that the listeners listened to. You said that they were "Converted" but its quite clear from the paper's text that they were presented at the same sample rate and format as the 24/192 source material, and were never converted to any other sample rate or format.

You need to go and it again then Arny.  The sample rate was kept constant just like our tests on AVS Forum.  The signal processing in Matlab performed the filtering/dither while keeping the sample rate the same.  Just like you simulating 44.1/16 by giving us upsampled files to 24/96, they have done that using Matlab and 192 Khz sampling of the original tracks.

Quote
For another example of well documented critiques that you have ignored Amir, the paper seems to almost completely obfuscate the listening environment. Over on AVS you harshly critiqued the Meyer-Moran tests for their non-compliance with BS 1116, but this paper has the same fault and yet you continue to praise it.  At least Meyer-Moran have published photographs of their listening environment (which you also trashed).

And you learned what from that photo Arny? 

I have already compared and contrasted this work to that of Meyer and Moran with respect to BS1116.  I am happy to do a much more complete one.  You know Meyer and Moran will lose and lose big, yes?

Quote
Finally Amir, I documented how this paper completely distorted ABX testing, and again I see no response from you.

I intend to create a new thread on some of your comments there per TOS #5 and instructions from our moderator.
Amir
Retired Technology Insider
Founder, AudioScienceReview.com

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #235 –
Quote from: ajinfla on
Quote from: amirm on
There are no obvious flaws.

Wishful think fallacy.
1) Test deliberately doctored with (RPDF) dither against BS own recommended best practices (TPDF) to fabricate possible positives.

Explain how that is an obvious flaw AJ.  Can you do that?

Quote
2) Positives may have been the results of system artifacts, loudspeakers, switching software, etc., not just the pathological dither doctoring.

Sure, anything is possible.  Your job is to show it is probable and the way to do that is with data.  Otherwise, it is a FUD tactic to put a cloud on an award winning, peer reviewed, scientific paper, published and presented at AES conference, by an AES Life Fellow, with proper engineering and research experience.  Yes, I have learned to talk like you guys. 

You want to counter all of that with just your opinion AJ?

Quote
The test may have shown what it set out to do (fabricate positives with pathological dither). Or may it not have.

Rectangular dither is pathological dither?  Where would I read about that AJ?



Amir
Retired Technology Insider
Founder, AudioScienceReview.com

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #236 –
Quote from: amirm on
Explain how that is an obvious flaw AJ.  Can you do that?

Two words: noise modulation.

Now I don't think that the choice of dither usually matters, but in critical evaluations of the limitations of a format I would at least expect triangular dither.
"I hear it when I see it."

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #237 –
Quote from: amirm on
The day you learn to debate AJ is the day you give me advice.  He doesn't come from a logical point of view or interest in the topic.  He comes after you personally to get a rise  out of you and have fun doing it.  While I am answering him with a straight face, I am hoping you see the reality of this.  If not, then when you are dealing with him you can try to use the "scientific approach" and see how far that gets you.

In Amirworld this passed as logic. But out here in the real world of HA, that is classic fallacy of focusing on the arguer, rather than argument, which is the BS paper.
Amir, please keep your focus on discussion of the paper, not me. Thanks.
Loudspeaker manufacturer

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #238 –
Quote from: xnor on
Now I don't think that the choice of dither usually matters, but in critical evaluations of the limitations of a format I would at least expect triangular dither.
At least. Someone wanting to give 16-bits the best chance might use noise shaping, as advocated by a certain J.R.Stuart in his presentations at the 96th AES convention in Amsterdam (1994) and the 100th AES convention in Copenhagen (1996).

(preprints 3501 and 3871 respectively)




amirm, this really is getting embarrassing. The paper is very clear: they wanted to test bad downconversion. They intentionally introduced a flaw. Please stop pretending otherwise. They didn't feel the need to pretend that this was a state-of-the-art test; why do you?

Cheers,
David.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #239 –
Quote from: amirm on
Explain how that is an obvious flaw AJ.  Can you do that?
Rectangular dither is pathological dither?  Where would I read about that AJ?

Here: 

Quote from: amirm on
Quote
2) Positives may have been the results of system artifacts, loudspeakers, switching software, etc., not just the pathological dither doctoring.

Sure, anything is possible.  Your job is to show it is probable and the way to do that is with data.

Wrong. Your job is to learn logic and understand that the burden of proof lies squarely on those creating this BS paper and its claims. They need to show that system transparency - speakers, switching software, etc, etc....aren't the source of the positives....which would make them false (despite all the dither doctoring efforts).
Btw, that's in his previous paper also, warnings about false positives. Just like JJ cautioned...
You really should listen to those experts Amir. 

cheers,

AJ
Loudspeaker manufacturer

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #240 –
I thought my previous "Seriously? Doh." response was enough to show the absurdity of such a position, but I guess I was wrong.

There is no evidence that unicorns don't exist, so it's reasonable to assume that they do? That they are likely to exist?

And on the latter post: If I make an experiment that shows telepathic abilities, it is your job to show that the experiment did indeed show exactly that? It is your job to ensure that bias in the experiment was eliminated, variables were controlled ....? What?!
"I hear it when I see it."
 

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #241 –
Quote from: xnor on
As I understand it 8 listeners completed 12 blocks each.

Each block consists of 12 A-X trials (where X=A, the original, or X=B, the processed A) but the first 2 trials were always (?) dismissed.
Visual feedback for each trial if listener chose correctly.
Each trial had a randomly assigned segment from a song.
Within each block, B would be processed the same way.

Now they say "2 blocks for each condition". What is a condition, are those the different processing setups (like different lowpass filters)? I guess so.



A condition appears to be a test condition;;

Code: [Select]
Condition number   Filter cutoff frequency (Hz)     Further processing

1                         21591-22050                    None
2                         21591-22050                    16-bit quantization
3                         21591-22050                    16-bit quantization and rectangular dither
4                         23500-24000                    None
5                         23500-24000                    16-bit quantization
6                         23500-24000                    16-bit quantization and rectangular dither


Table 1: Details of each condition tested.



The above cutoff frequency numbers seem to relate to the low pass filter's transition frequency.

"
The frequencies of the transition bands were 23500-
24000 Hz and 21591-22050 Hz, corresponding to the
standard sample rates of 48 kHz and 44.1 kHz re-
spectively.4 Fig. 2 shows the amplitude and energy
of the impulse response for the 48-kHz fllter.
"

The following statement appears to raise more critical questions:

"These parameters were chosen to off er a reasonable match to the downsampling fi lters used in good- quality A/D converters or in the mastering process;"

I looked at the spec sheets for a number of modern good quality DACs, and found that the 500 Hz wide transition bands chosen by Meridian for their tests seem to vary considerably from the approximate 2 KHz or wider transition bands that I found in these commercial products operating with the same corner frequencies.

This is interesting because FIR filters with narrow transition bands were used for listener training, with the comment that "This fi lter was chosen as it would have been straightforward for most listeners to identify diff erences introduced by its application."

So, there you go - yet another asymmetry between the test conditions and the real world.

BTW, there is a lesson here. The Quality slider on software resamplers such as CoolEdit Pro/Audition is a tuning knob for transition band width. Higher quality means a narrower transition band.

If you push it to the far right (99%) the resulting transition band became  exceedingly narrow in my tests - just a few Hz, which puts it in the same realm as the signal that Meridian used for listener training.

If you push it to the far left the transition band is more like 1.6 KHz and better approximates a real world converter.

All of the sample rate ABX  test files I've circulated lately were made with quality set for the maximum. They are therefore not representative of real world circumstances. One word: Invalid.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #242 –
My guess is he ignores this and goes for low hanging fruit, or revert to jabbing at AJ, which he knows is not appropriate.

Or insist the paper is a slam dunk technical demonstration as to why people should feel justified in choosing hi-res because it won an award, which now seems to be the new default fallback position when he can't convince anyone that they shouldn't be concerned that the results may have been skewed by the hardware or choice of processing; let alone whether this is all that representative of real-world conditions beyond the world of selling compromised designs to the gullible eager to part with their cash.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #243 –
Quote from: amirm on
The sample rate was kept constant just like our tests on AVS Forum.


I clearly said that I converted those test files from 24/96 to 44/16 and then back again to 24/94. You can take that literally - the "4416" test files existed for a while as 44/16 files. The Meridian files never did.

By continuing to argue with me it is clear that you remain confused about the facts of the matter which is yet another win.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #244 –
Quote from: xnor on
Quote from: amirm on
Explain how that is an obvious flaw AJ.  Can you do that?

Two words: noise modulation.

And the data that shows it to be in effect in this test is where?

Quote
Now I don't think that the choice of dither usually matters, but in critical evaluations of the limitations of a format I would at least expect triangular dither.

If it usually doesn't matter then it is not an "obvious flaw."
Amir
Retired Technology Insider
Founder, AudioScienceReview.com

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #245 –
Quote from: Arnold B. Krueger on
Quote from: amirm on
The sample rate was kept constant just like our tests on AVS Forum.


I clearly said that I converted those test files from 24/96 to 44/16 and then back again to 24/94. You can take that literally - the "4416" test files existed for a while as 44/16 files. The Meridian files never did.

It is the same thing in Meridian test but using an advanced methodology (for lay people) using Matlab real-time processing.  David, do you want to explain this to Arny?  Clearly he doesn't want to listen to me.

Quote
By continuing to argue with me it is clear that you remain confused about the facts of the matter which is yet another win.

No Arny.  The paper has won an award from its peers.  Don't think of pedantic mistakes.  Any that you think you find is due to not understanding how professional audio tests are performed thee days.
Amir
Retired Technology Insider
Founder, AudioScienceReview.com

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #246 –
Quote from: amirm on
Quote from: xnor on
Quote from: amirm on
Explain how that is an obvious flaw AJ.  Can you do that?

Two words: noise modulation.

And the data that shows it to be in effect in this test is where?


Right here, Amir.

http://www.hydrogenaud.io/forums/index.php...st&p=881116

which links this excellent paper:

http://robertwannamaker.com/writings/ieee.pdf
Page 31 figure 5 (b) shows the noise modulation due to rectangular dither.
Page 34 figure 7 shows the lack of noise modulation with triangular dither.

The post was addressed to you Amir. You've responded to posts made since it went up. 

It is now obvious that you couldn't properly comprehend its meaning and relevance. Another win.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #247 –
Quote from: greynol on
My guess is he ignores this and goes for low hanging fruit, or revert to jabbing at AJ, which he knows is not appropriate.

Or insist the paper is a slam dunk technical demonstration as to why people should feel justified in choosing hi-res because it won an award, which now seems to be the new default fallback position when he can't convince anyone that they shouldn't be concerned that the results may have been skewed by the hardware or choice of processing; let alone whether this is all that representative of real-world conditions beyond the world of selling compromised designs to the gullible eager to part with their cash.

Disappointing to see yet another commentary devoid of any technical insight.  Would you like to say everything that is on your chest so that we can be done with it?
Amir
Retired Technology Insider
Founder, AudioScienceReview.com

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #248 –
Quote from: amirm on
Disappointing to see yet another commentary devoid of any technical insight.  Would you like to say everything that is on your chest so that we can be done with it?


You mean another commentary devoid of any technical insight like this one?

http://www.hydrogenaud.io/forums/index.php...st&p=881152

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #249 –
Quote from: 2Bdecided on
Quote from: amirm on
There are no obvious flaws.
http://robertwannamaker.com/writings/ieee.pdf
Page 31 figure 5 (b) shows the noise modulation due to rectangular dither.
Page 34 figure 7 shows the lack of noise modulation with triangular dither.

Wannamaker, Lipshitz, and Vanderkooy have spread this information, with varying levels of detail, across the world of audio since the 1980s. I am sure that the authors of the BS paper know and understand this information.

The BS paper accurately reports what they did, and tells you (at least partly) why they intentionally introduced this flaw into the test.

Saying "there are no obvious flaws" is just silly.

Cheers,
David.

Looks like the plot is lost.  You are giving me a tutorial on noise modulation and dither.  That has nothing to do with the topic at hand or the question I asked.

The test has been called to have a major flaw.  You need to demonstrate using the paper and data within, how it was that people heard noise modulation and not effect of filtering/dither.  Do you have that?

I also like to see people go on record and state that anytime a 24 bit file is converted to 16 bit using rectangular dither, the conversion is highly flawed and lossy.
Amir
Retired Technology Insider
Founder, AudioScienceReview.com