|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
24.179.64.196
Hello JA,
I have been following the uproar and subsequent search to root out the heretics and bring back the status quo over at AVS. I'm posting here because if I posted there I'd be stoned as well :)
I really enjoyed your discussion on perception and models. I found it very interesting and after I thought about it the concepts seemed both intuitive and obvious.
Best regards,
Marty
Follow Ups:
> I have been following the uproar and subsequent search to root out the
> heretics and bring back the status quo over at AVS. I'm posting here
> because if I posted there I'd be stoned as well :)For decades, so-called "objectivists" have demanded that audiophiles
perform double-blind ABX tests to "prove" that they hear what they claim to
hear. Now that some people have done that and have heard the difference
between hi-rez files and CD-quality versions of the same to a statistically
significant extent, the objectivists are making increasingly absurd
arguments why statistically significant positive results of double-blind
ABX tests don't actually mean anything. :-)> I really enjoyed your discussion on perception and models. I found it
> very interesting and after I thought about it the concepts seemed both
> intuitive and obvious.Thank you.
The main point I was making is that what matters to us as listeners is the
internal states of the internal models our brains construct as a result of
soundwaves hitting our ears, and those states are by definition not
accessible to measurement.But in my study of the science of perception, what astonished me was that
we don't experience sound in real-time but with a latency of around 100ms.
There is a _lot_ of processing going on in our brains before we experience
"sound," meaning we actually experience "reality" more than a tenth of a
second in the past. The latency is even greater with vision.
John Atkinson
Editor, Stereophile
Edits: 08/01/14
The only recent cd quality vs. hi-rez tests I've read about end up with something like 49/51% difference.
Or, just about chance.
"The problem with quotes from the internet is that many of them are just made up."
-Abraham Lincoln
I saw that one somewhere. Studio engineers hit the 51% as I recall, and self-described audiophiles were at the low end.
Some time ago we transferred several early recordings of Suzanne Vega from 2-inch/24-track to files, and spent some time listening. Using a UAD 2192 converter, we could tell pretty reliably which was 44.1 and which was higher (48/96/192) even if we couldn't pin down *what* the difference was. By "pretty reliably" I mean about two out of three times. Neither of us could discern the difference between 48kHz and anything higher. The gear we used was pretty much the same gear used to track and mix the sessions.
YMMV.
WW
"A man need merely light the filaments of his receiving set and the world's greatest artists will perform for him." Alfred N. Goldsmith, RCA, 1922
Any chance that the 44.1 files were 16bit while 48 and higher were 24bit?
Nope, we just played with 24-bit that day.
WW
"A man need merely light the filaments of his receiving set and the world's greatest artists will perform for him." Alfred N. Goldsmith, RCA, 1922
> The only recent cd quality vs. hi-rez tests I've read about end up with
> something like 49/51% difference.
Check out the ABX test results included in the message on the AVS Forum
link below.
John Atkinson
Editor, Stereophile
I used dBPoweramp to downsample my 24/192 Norah Jones to 16/44. I then used the Foobar comparator to test myself. I only picked correctly once out of my first three attempts but then latched onto what to listen for. I correctly identified the next ten consecutive and then quit because I was very confident that I could go on forever.
When I tried to distinguish 24/194 from 24/44, however, my success rate was pretty much no different from chance - and i had no confidence that I could tell the difference.
In my experience increased bit depth makes a far greater difference than a higher sample rate.
You will now have to start measuring brain activity. I see a whole new dimension opening up for you.
But it is expensive.
But but they use those machines on pertain'er every every TV show, everybody got um
Never trust an Atom, they Make Up everything!
"You out there, Golden Ears, the person who couldn't care less about present technical measurements but thinks of sound as a holistic experience. You're right, you know."
The technical side may still ask why, but the empirical side of music has nothing to do with gilt and flash or a filigree wordsmith description that falls short of the musical experience.
The Mind has No Firewall~ U.S. Army War College.
100 milisecond latency in hearing?
and greater with vision?
How can you drive?
If you are going 60mph, that is OVER 9 feet! Just in 'latency'.
At Racing speeds, approaching the 150 to 180mph region, latency of 100ms would result in the
racer moving nearly 30 feet from the event to its perception, not counting 'reflex' time and such.
Too much is never enough
There is a woman in Austria who is very popular with neuro-scientists because she had a stroke which destroyed a very specific part of the brain and this means she now can not see any movement at all.
What happens is that she sees a static scene which gets updated every once or twice per second.
We all see like that but the brain fills in the gaps so we think we perceive continuous motion. when we really don't.
> 100 millisecond latency in hearing?
Yes. That's why quiet sound can be obliterated by a subsequent loud one,
a phenomenon called "backward masking."
> and greater with vision?
> How can you drive?
Lay people refer to it as "reaction time." The latency is only for
conscious perception where the brain is fully involved. There are
mechanisms that allow for faster reaction to stimuli but they bypass
conscious perception
John Atkinson
Editor, Stereophile
...I've always found that some of what we hear as audiophiles in critical listening is processed subconsciously and we become aware of it later.
Another reason why short A/B comparisons tend to mask differences.
When judging whether one hears a tone or hears a difference, an objective decision is what we are after. When judging whether one speaker (or amp or....) is sounds better (more satisfying), a subjective decision incorporating multiple and inchoate parameters is called for.
Extended subconscious processing permits non-stimulus-related information to influence conscious perception and that would make the judgement less objective. Short interval A/B testing might force the subjects to "pre-conscious" and more objective assessment.
I've read--sorry, no citations and no time to look them up--that tests have shown that even these "deep brain," preconscious aspects of hearing are influenced by cultural--well, influences (i.e., influence of certain brands, like Coke). This has led me to conclude that hearing is intrinsically subjective, or, to put it another way, sometimes you really do hear differences you think you hear; it's not in the sound, but you hear it. Is this wrong?
Jim
All perceptions are subjective to some degree. Limiting the latency between stimulus and response serves to limit how much.
Of course, one could record from auditory afferent fibers or the auditory nuclei and, to a great degree, eliminate the subjectivity but that raises another issue: We need that to hear but we don't really "listen" with those. Cortical recordings might represent an objective measurement but it would be of subjectively-processed information.
So, it is a matter of refining the question before determining how we might answer it. If you want to find out what our auditory resolution (in any dimension) is, you want to eliminate subjectivity as much as possible. If you want to find out what we are hearing or how we are assessing what we hear, that subjectivity is an essential part and, perhaps, the major player.
...from what I've seen, short interval A/B testing with music, not tones, seems best suited to identify differences in gross frequency response, noise, distortion and loudness.Not much else.
With long term observational listening comparisons, all of the differences (and sometimes non-existent ones) can be identified, both objectively and then subjectively including things you are aware of subconsciously.
Edits: 08/03/14
the former is less likely to be influenced by memories and associations while, for the latter, those are essential.
...the first is more objective because if controlled and blind, it can remove some biases.But with music it's pretty worthless for identifying differences other than the ones mentioned.
Hmmm, except in the hi-rez recording trials John linked.
Edits: 08/04/14
. . . regarding the audibility of differences between crossover capacitors was conducted by some speaker building hobbyists at a DIY meet a couple of years ago, and reported on the PE Tech Talk board. The speaker used was a well-conceived design with excellent drivers. Several different caps matched to 0.5% tolerance were switched between in a second-order electrical highpass filter -- a good quality nonpolar electrolytic, and some different film and film-foil types ranging from basic Dayton polypropylenes to some exotic and costly ones.
Basically, nearly every member of the listening panel was able to consistently hear the difference between the NPE and the film caps -- comments ranged from "slightly worse" to "gritty and awful" -- but nobody could consistently identify one film type over another.
These guys were "objectivists" to a man, not of the hostile sort out to prove that very small measurable differences are inaudible, but skilled and practiced listeners with the genuine scientific curiosity to test what they could and could not hear.
Hello Brian,
When I lived in the area I attended the annual PE Summer DIY session. They are always good fun.
Years ago at one of the gatherings there was a fun test very similar to what you describe. The difference was we were listening for the difference between low cost mylar caps and pricier polypropylene caps. We were in a hotel room so only about 5-6 folks at a time could participate. So we took the test and when it was over we admitted we could not tell a difference, but there was one fellow who got it right. He nailed it. When asked he said over the years he had been trained to listen better. It turned out he was a local midwest manufacturer.
I found that experience an eye opener.
Regards,
Marty
Marty
I suspect there are a LOT of auditory phenomena which the ordinary music lover cannot readily discern (maybe especially under the stress of ABX conditions), but which can be readily heard and identified by those who have spent LOTS of time listening for the subtler effects and training themselves to recognize said effects.
I've known technicians who can listen to pink noise and instantly pinpoint to within a few Hertz any peaks or dips in the system response, and can hear and identify distortions at levels below 1%. Probably some innate talent there (like perfect pitch), but certainly cultivated through much practice in critical listening. My own hearing is pretty good, but not THAT good -- I can recognize problems with phase and frequency response when they are present -- something just sounds "wrong" -- but have to trot out the test gear to pinpoint them.
I had never heard that there is latency in human vision or hearing. That news is definitely today's "Wow!" moment and sent me to Google to research it. Thanks for sharing!
It takes half and hour for the stomach to tell the brain that's it has been fed. Now that's latency!
OK, well, Louis CK did a very funny bit on this.
___
"If you are the owner of a new stereophonic system, this record will play with even more brilliant true-to-life fidelity. In short, you can purchase this record with no fear of its becoming obsolete in the future."
Now that some people have done that and have heard the difference
between hi-rez files and CD-quality versions of the same to a statistically
significant extent...
When/where was this done?
> When/where was this done?
See the linked article below.
John Atkinson
Editor, Stereophile
For decades, so-called "objectivists" have demanded that audiophiles
perform double-blind ABX tests to "prove" that they hear what they claim to
hear. Now that some people have done that and have heard the difference
between hi-rez files and CD-quality versions of the same to a statistically
significant extent, the objectivists are making increasingly absurd
arguments why statistically significant positive results of double-blind
ABX tests don't actually mean anything. :-)
Ok, after reading the link you provided, I have to say this statement is very misleading.
The test was neither double blind nor ABX.
> The test was neither double blind nor ABX.
See the ABX results from one listener at the link below, obtained with the
Foobar ABX plug-in.
John Atkinson
Editor, Stereophile
Ok, so you've got something to start with. Though I see some valid arguments about IMD possibly being a culprit (shades of Oohasi). So it seems there's some work to do yet.
> see some valid arguments about IMD possibly being a culprit (shades of
> Oohasi). So it seems there's some work to do yet.I think this is a red herring raised by some who would like to dismiss the
positive ABX results. I tested 3 USB D/A headphone amplifiers - AudioQuest
Dragonfly, Meridian Explorer, Meridian Prime - with a 96k test signal
consisting of tones at 39kHz and 41kHz. There was no audio-band intermodulation
with the AC-powered Prime - see graphs at the link below - and I only got
audioband products with the other two bus-powered DACs when they were
driven into clipping. As all music doesn't have ultraonic content anywhere
near the level that would lead to clipping, I don't see that audible IM
with the 96kHz file would be an issue.
John Atkinson
Editor, Stereophile
Edits: 08/02/14 08/02/14
I was thinking of the transducers.
> I was thinking of the transducers.
Good point. But Arny Kruger made available files with high-level ultrasonic
tones spaced 4kHz apart, headed by a low-level 4kHz tone for training. If
there was audible intermodulation from the transducers, there would have
been a readily identifiable 4kHz tone. However, there wasn't such a tone
with the listener who identified an audible difference between the files.
John Atkinson
Editor, Stereophile
Arny Kruger, the guy who never accepts defeat regardless how hair brained his ideas, one upon the other.
Has he ever gotten anything right, I mean even C Johnson has his woodie.
;)
Never trust an Atom, they Make Up everything!
So he says. ;-)
illustrate his point, Steve. :)
I'm not in any sort of denial. I said he's got a good starting point. But at this point it's still just a starting point. As jj used to say, getting positive results is easy. What he meant by this is that there are many things that can conspire to give false positives. And I'm not saying that is the case here. Only that I think we need to go a bit further down the road before any safe conclusions can be made. I have no vested interest in either outcome. I just don't want to see this turn into "Someone passed a double blind test! It's been proved!"
Who on earth could possibly care?
.
> I just don't want to see this turn into "Someone passed a double blind
> test! It's been proved!"
But it _has_ been proved. That's what the statistical analysis of these
double-blind tests states. That of the three pairs of files that were
subject to the testing, it can be stated that the possibility of there
not being an audible difference was statistically insignificant. Remember
your statistics classes: that if the identification was not statistically
significant, it doesn't mean there was no difference, only that if there
was a difference, it could not be detected under the specific conditions
of that test. Thus one positive test outweighs a million tests that
produced a null result.
The next question is to examine why a difference was detected between
these pairs of files. But the fact that there is a difference is now
incontrovertible.
John Atkinson
Editor, Stereophile
Hi John:
So the subjectivist position could be summed up thus: If there is no audible difference the fault lies with the methodology, not the listening panel. However, if there is an audible difference then the methodology is sound. Is that about it?
The question arises: Can you envision a scenario in which proper methodology is employed, but yields no audible difference?
In the interest of intellectual honesty, it would be refreshing if both camps could agree the methodology to be employed is sound **prior** to beginning the test, and then let the chips fall where they may. That way, subjectivists are not given the convenient luxury of an escape clause. Perhaps sometimes there’s just no there there when it comes to demonstrable differences regardless of the methodology being utilized. In any event, it’s refreshing to see you so enamored with the efficacy of a double blind test. :)
> So the subjectivist position could be summed up thus: If there is no
> audible difference the fault lies with the methodology, not the listening
> panel. However, if there is an audible difference then the methodology is
> sound. Is that about it?
No. All I am saying is that in a test where you need to apply formal
statistical analysis, it is impossible to prove a negative. Of course,
repeated tests where the results indicate no difference from what would
be achieved by chance could be interpreted as circumstantial evidence that
there is no audible difference. But that is a subjective judgment, not a
scientific one.
John Atkinson
Editor, Stereophile
Suppose the experiment shows sufficiently strong results that one doesn't need not worry about the statistics. This will only show that the two files sounded different when played back on available equipment. This does not show that the lower resolution format is audibly inferior, because it is possible that the difference was caused by a poor format conversion. In addition, and this is a little harder to take, it may be that the differences that people hear are not actually on either of the two files, but are caused by differential artifacts in the playback chains and with better equipment the actual differences would not have been audible.
I will be very surprised if the usual die-hard "everything sounds the same" people change their tune. They are on record as sticking to their guns even when their own study shows differences exist (as was the case with amplifiers).
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
> Suppose the experiment shows sufficiently strong results that one doesn't
> need not worry about the statistics.
An ABX test mandates statistical analysis of the results when it is used to
examine the audibility of a small difference.
> This will only show that the two files sounded different when played back
> on available equipment. This does not show that the lower resolution
> format is audibly inferior...
An ABX test says nothing about quality or preference, only that a
difference was detected.
> it is possible that the difference was caused by a poor format conversion.
That is possible, or even probable, according to some engineers.
John Atkinson
Editor, Stereophile
Big whoop.
Thanks!
getting STONED. do i just go to AVS?
...regards...tr
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: