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: two hi rez vs redbook train wrecks in progress (Read 86803 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

two hi rez vs redbook train wrecks in progress

AVS/AIX High-Resolution Audio Test: The Results So Far
http://www.avsforum.com/forum/91-audio-the...lts-so-far.html

6 participants reporting perfect 3/3 scores for files supplied by AIX.  All 6 have 'hi rez audio' capable systems  (typically meaning, extended frequency response)



debate-thread-scott-s-hi-res-audio-test
http://www.avsforum.com/forum/91-audio-the...audio-test.html

discussion of the above test, as well as addition of two more tests, involving 'jangling keys' files.  Again several participants reporting no problems ABXing 96/24 vs 44/16  (96/24-->44/16-->96/24)




remarkable rates of success, no? 

two hi rez vs redbook train wrecks in progress

Reply #1
Worthless IMHO. It's not a blind test, and they didn't control for the possibility of resampling during playback. A commenter on the first thread says they didn't control for level differences (although I'm not sure 0.2 dB is certain to matter).

two hi rez vs redbook train wrecks in progress

Reply #2
Apart from everything else, in the results thread, did they report results only after the level difference was corrected, or include results from when the level different was present? EDIT: Yes, seems so.

How are they defining a "hi-res" system? EDIT: Seems to be, as you say, extended frequency response, plus dynamic range of equipment.

(I have tried to find the answers some, but 85+ pages and 3 threads is too much for anyone! EDIT: found a fourth thread!)

EDIT: Sorry, I've just noticed its a sighted test and the zipped file sizes are different (though not necessarily in the expected way). Ignore me  ABX results are still interesting through, and I'm not sure it really counts as sighted unless you cheat. A/B preference when you don't know which is which is still double-blind, isn't it? The statistics are different, but it could still meet TOS8.

FWIW the 16-bit dither is dubious / non-existent, and one of the "high quality" originals clips, but I doubt either factor is audible.

You seem to be implying that people are cheating. Arny is saying it's people hearing distortion. It would be interesting to put a microphone in their listening position and capture the signal that reaches their ears. Apart from Arny's keys sample (which anyone with a bad soundcard can ABX), the other samples don't have that much high frequency content.

Cheers,
David.

P.S. Sorry for all the edits. Last one: I probably wouldn't pay for this music at any quality, but it's nice of the owner to make the full tracks available for free, even as an obvious marketing ploy. While I'd only ABX a few seconds, it's helpful to have the full track available to be able to pick which part of it you ABX, and to listen to the whole context.

two hi rez vs redbook train wrecks in progress

Reply #3
Worthless IMHO. It's not a blind test
Participants are encouraged to use an ABX application if possible (foobar is recommened). That looks pretty blind to me.
In short: 3 music samples in 24/96 format are compared to their lo-res versions which have converted to 16/44.1 and back to 24/96.
The thread has become so long and confusing that it's nearly impossible to find anything, even if you have followed most of it (I tried ).
Nevertheless, positive ABX results can not be denied and should be discussed IMO. If there are flaws in the test, they should be analyzed.
The purpose of the test was to have fun, not to be scientific. But as it goes, positive results will be claimed as proof for audible differences between hi-res and redbook.

@2Bdecided: "How are they defining a "hi-res" system?". AFAIK basically any playback system with >20kHz bandwidth.

two hi rez vs redbook train wrecks in progress

Reply #4
Re: blind/sighted, I only read the first post, which didn't mention ABX (and subsequent posts initially rejected the suggestion of using ABX) ... and I admit the title of this thread imparted some bias.  I stand corrected. (?)

two hi rez vs redbook train wrecks in progress

Reply #5
Re: blind/sighted, I only read the first post, which didn't mention ABX (and subsequent posts initially rejected the suggestion of using ABX) ... and I admit the title of this thread imparted some bias.  I stand corrected. (?)
You don't need to use ABX for it to be blind. It takes a little more effort to fake an ABX test though. It's trivial to fake an A/B test. It's beyond most people to fake ABC/HR.jar results.

Cheers,
David.

two hi rez vs redbook train wrecks in progress

Reply #6
You seem to be implying that people are cheating. Arny is saying it's people hearing distortion. It would be interesting to put a microphone in their listening position and capture the signal that reaches their ears. Apart from Arny's keys sample (which anyone with a bad soundcard can ABX), the other samples don't have that much high frequency content.

Cheers,
David.

P.S. Sorry for all the edits. Last one: I probably wouldn't pay for this music at any quality, but it's nice of the owner to make the full tracks available for free, even as an obvious marketing ploy. While I'd only ABX a few seconds, it's helpful to have the full track available to be able to pick which part of it you ABX, and to listen to the whole context.



I forgive you for the edits. ;> 

I'm not implying  specifically that cheating might have occurred -- though of course it might have -- I am expressing my view that both efforts are hopelessly muddled clusterf*cks at this point, and I am prematurely both bored and dreadful at the prospect of seeing the results touted as dispositive by the usual high rez cheerleaders.

Note that both Amirm and John Atkinson are on the 'debate' thread, both engaging in the usual bullfight and bloodletting with Arny, and both (I think, though honestly I can barely rouse myself to confirm anything in that mess at this point) claiming to have aced the hirez vs redbook jangly keys ABX , absent IM distortion.

If anyone *here* wants to review all the posts *there* and make some helpful suggestions *there*, I absolutely do encourage that, which is really the point of me starting this thread *here* (despite my biased subject line). 

two hi rez vs redbook train wrecks in progress

Reply #7
I'm over there, under my longer form name. [add "lch" to the end of mzil]

My understanding is that an "HRA" as they define it there, hi-res audio system, means you have an outboard DAC that accepts 96k. People with just AVRs, for example, don't qualify. There was no mention, I am aware of, which stipulated what kind of speakers were up to the task.

Oh wait, I now see this posted there:

"The system you specify certainly qualifies as HRA [an Oppo BD player using a thumb drive for the files], but only as a front end. You also need to consider the preamp, power amp, and speakers or headphones; they all need to be able to pass frequencies well above 20 kHz and a dynamic range beyond 93 dB. If any component in the system can't do that, it's a bottleneck that renders the system as a whole non-HRA."

I doubt a single person there [except the wealthy with dedicated theater/media rooms] has a living room with a 93 dB range; even dead quiet libraries have a background noise SPL of around what, 40 dBSPL? Add peaks above that by 93 dB?! Yikes.


two hi rez vs redbook train wrecks in progress

Reply #8
I would be cautious about simple abx results from someone making money with audio gear and hanging around in a dedicated forum.
For example Arny is asking uncomfortable question about some claims at a well known forum of that kind.
Alone for that i can imagine someone does some fake foobar abx to make Arny shut up (no pun intended  )
Just take the files to  test, fill one with silence, do abx, impress the world.
These simple foobar abx like we use here together with mostly known members should not be taken to serious elsewhere.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

two hi rez vs redbook train wrecks in progress

Reply #9
Maybe someone can clarify for me how WinABX/ fooABX work --

when you select two files for fooABX, does it always assign the file 1  to 'A' and the file 2 to 'B'?  And does 'A' stay '1' throughout the test (up to the point you hit 'end') or are 'A' and 'B' randomly reassigned to 1 and 2 during the test?



EDIT:  I played around with fooABX using two plainly different tracks, and answered my own questions.  In my hands, whatever track is first in order in the Columns UI/tracklist view  (whether it's the first selected, or not) is A, the the other is B.  And A stays the same throughout.


IOW the order in which you select the two tracks does not matter.  The position of the track in the tracklist you selected it from, does.  And the identities of A and B do not change during the test.



Does this sound right to you all?

two hi rez vs redbook train wrecks in progress

Reply #10
Yes, that's right, but I see you already have seen my parallel findings in the other thread at this point.

Whatever is on top in the list, at the time you select the two tracks for ABXing,  will be "A" for the entire test.
The order you select them, the order they were at first before you rearranged them, the order you brought them to the list, etc. is immaterial.

two hi rez vs redbook train wrecks in progress

Reply #11
AVS/AIX High-Resolution Audio Test: The Results So Far
http://www.avsforum.com/forum/91-audio-the...lts-so-far.html
6 participants reporting perfect 3/3 scores for files supplied by AIX.  All 6 have 'hi rez audio' capable systems  (typically meaning, extended frequency response)

debate-thread-scott-s-hi-res-audio-test
http://www.avsforum.com/forum/91-audio-the...audio-test.html

discussion of the above test, as well as addition of two more tests, involving 'jangling keys' files.  Again several participants reporting no problems ABXing 96/24 vs 44/16  (96/24-->44/16-->96/24)
remarkable rates of success, no? 


I'm concerned that people are gaming the tests, assisted by the fact that a lot of the monitoring systems they use are chock full of nonlinear distortion and/or are marginal for dynamic range as used.

The big surprise to me has been how much audio gear there is that falls apart under stress testing > 20 KHz.

To demonstrate this exposure, I gamed my own system and obtained positive results in an ABX test involving hi rez  audio.  Trust me the results I posted make $3 bills look like legal tender and of course I say as much as part of the so-called results.

http://www.avsforum.com/forum/91-audio-the...ml#post26095274

two hi rez vs redbook train wrecks in progress

Reply #12
I'm always concerned by the very words "High Resolution," and "High-Definition," applied to larger bit/sampling-rate numbers. Are they not marketing, rather than technical, terms, which feed upon the public confusion that what they can easily see in image formats also applies to audio (which it doesn't).

Confirm or deny please
The most important audio cables are the ones in the brain

two hi rez vs redbook train wrecks in progress

Reply #13
The files do not null.

I can only assume that Mark Waldrep does not know how to resample properly.


I also assume it is not explained how these files were created. How can someone make an open test if they do not openly explain in a detailed manner how the test files were processed/created? Right, if one has an agenda (like Linn did with a "high-res" vs. CD audio test, where they used different masters!).
"I hear it when I see it."

two hi rez vs redbook train wrecks in progress

Reply #14
The files do not null.

I can only assume that Mark Waldrep does not know how to resample properly.


I also assume it is not explained how these files were created. How can someone make an open test if they do not openly explain in a detailed manner how the test files were processed/created? Right, if one has an agenda (like Linn did with a "high-res" vs. CD audio test, where they used different masters!).


I had the same claim levied against some resampled files that I created.  I reran my procedures and found that they did null, unless I enabled my resampler's "Filter before and after" feature.

Further testing showed that the filter before and after feature was working well enough given what it did, which was among other things apply two more stages of digital filtering. The additional filtering did add a tiny amount of linear distortion, mostly in the form of a time shift. 

The so-called nulling failure IMO was due to the fact that I came up with yet another explanation (of 100s that I know to exist) as to why nulling is a very dangerous test because it flags alleged problems that are not really problems such as trivial time shifting.  Just what we need: more false positives.

two hi rez vs redbook train wrecks in progress

Reply #15
The files do not only show "trivial" (fractional) delay, but the frequency response show more energy at high frequencies too. Maybe too aggressive noise shaping? Maybe the resampler used has a weird frequency response?

Who the hell knows, and without any details on how to reproduce the files we can simply dismiss the test. It's as easy as that.
"I hear it when I see it."

two hi rez vs redbook train wrecks in progress

Reply #16
The files do not only show "trivial" (fractional) delay, but the frequency response show more energy at high frequencies too. Maybe too aggressive noise shaping? Maybe the resampler used has a weird frequency response?

Who the hell knows, and without any details on how to reproduce the files we can simply dismiss the test. It's as easy as that.



Which test?

To date I count *four* sets of test files being offered across both tests on those two thread (+ one set of IM distortion files).


That's as of about four days ago.  By now there could be more. 

two hi rez vs redbook train wrecks in progress

Reply #17
Which test?

Any format comparison using files that were processed in an unknown way.

Comparing different formats (such as 96/24 to 44.1/16) means comparing:
a) the resampler
b) quantization noise / dither / noise shaping

Choose some aggressive noise shaping with lots of HF energy and people with good hearing are more likely to successfully ABX the files, even though they would fail with simple dither.
Choose a bad resampler and you will get aliasing, frequency response distortion, ...
"I hear it when I see it."

two hi rez vs redbook train wrecks in progress

Reply #18
The files do not only show "trivial" (fractional) delay, but the frequency response show more energy at high frequencies too.


Again, which files? The ones I inspected ('On The Street' and Arny's keys from hell) look clear, with the subsampled versions nicely matching the original ones below 20kHz.


two hi rez vs redbook train wrecks in progress

Reply #19
It would be nice to re-run the test with unquestionably good down conversion, and a secure virtually tamper-proof ABX program. That would solve two problems.

I don't know how you solve the problem of knowing the amount of distortion in someone's system though. I know you can play undial.wav or better test signals through it, but you risk frying some very expensive speakers in the process. Any ideas?


Over a decade since I first heard it, I am still fascinated by the question of hi-rez audio. Claims of a night-and-day difference (which still appear!) have been proven false more times than I can count. The idea that it creates some improvement just on the threshold of audibility in some circumstances is still worth investigating IMO. If we ever get this properly understood, we'll either learn something "new" about human hearing, something "new" about audio systems, or something "new" about placebo.

(I put "new" in quotes because I think all possibilities have already been considered, no none is "new" - it would just be newly proven to be true.)

I must stress that I don't over-estimate the importance of all of this. I like it as an intellectual curiosity. There are far more important things even the in the world of audio, and there are more important things than audio (really!).


FWIW I think you only stand a chance of learning anything under HA type rules, and with good will on both sides. That avsforum discussion thread isn't hopeful.

Cheers,
David.

two hi rez vs redbook train wrecks in progress

Reply #20
Again, which files? The ones I inspected ('On The Street' and Arny's keys from hell) look clear, with the subsampled versions nicely matching the original ones below 20kHz.

I only looked at the first two files: Just_My_Imagination_A2.wav and Just_My_Imagination_B2.wav.
"I hear it when I see it."

two hi rez vs redbook train wrecks in progress

Reply #21
I just looked at the Mosaic files.

Running them concurrently through a spectrum analyser gives an impression of
large differences in the treble. But these are caused by the temporal offset
between both files. Fixing this the <22kHz spectra are most of the time identical,
except when the HF part of the signal gets close to the noise floor. Then the subsampled
version seems louder above 15kHz. This is because of a tall noise shaping
bump, as you can see in the quiet part before the track starts.

The only other anomaly I can see is a rising >22kHz noise floor in the downsampled track.
As if an inadvertent step of noise shaping was added during the upsampling.

two hi rez vs redbook train wrecks in progress

Reply #22
It would be nice to re-run the test with unquestionably good down conversion, and a secure virtually tamper-proof ABX program. That would solve two problems.

I was reading a bit on the AVS thread and to be honest i doubt a bit in armirm abx logs. That of course may only be me old fart loosing the faith in humanity more every day.
Since even encrypted abx solutions can be cheated by some real time analysis i only see one solution.
Arny has to pilgrimage to armirm and do some live testing.
I'd happily donate some bucks if needed.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

two hi rez vs redbook train wrecks in progress

Reply #23
I'd happily donate some bucks if needed.
to a former Corporate Vice President at Microsoft? I'm sure he could pay Arnie's air fair if he thought it would help.

While I'm disappointed with the level of debate, I think some of the skepticism over there is justified.

See an article on why normal rooms are too quiet for 16-bit audio to be sufficient...
http://www.madronadigital.com/Library/RoomDynamicRange.html

See also an article on why jitter needs to be much lower than you thought...
http://www.madronadigital.com/Library/Audi...istortions.html

There's lots of science in those articles, but I find them misleading.

Cheers,
David.

two hi rez vs redbook train wrecks in progress

Reply #24
Since even encrypted abx solutions can be cheated by some real time analysis i only see one solution.
Arny has to pilgrimage to armirm and do some live testing.


I strongly suspect that story would end just like this similar audio challenge did:

http://www.vxm.com/21R.64.html

Namely, a wasted trip.