[gregorio]

[1] I also need to make an aside comment about how "tests whose results we agree with may have a few unresolved issues" but "tests whose results we disagree with get debunked".
[2] That meta-test was based on the same data, from the same flawed tests, that everyone wants to use as "proof" of the opposite claim.
Therefore, unless there are actual flaws in their statistical methodology, it is no more or less valid than the tests on which it was based.

1. True to an extent but the tests I disagree with are those whose results are contrary to known and demonstrated scientific facts.
2. Again true to an extent. The Theiss study for example demonstrated no ability to discriminate but in a supplementary non-ABX experiment with 3 of the subjects, a very high ability to discriminate was demonstrated. There are several reasons to strongly dispute this supplementary experiment, not least that it wasn't double blind but the meta-analysis only includes the supplementary experiment results, not the main ABX experiments. And, if we're going to eliminate the M&M study on the grounds that some of the samples were not Hi-Res, as the meta-analysis does, then how come there are other studies included which had no Hi-Res samples at all because they were well before hi-res was even invented?

The problem we have here is also one of time/history. Is it possible to discriminate 96/24 from 44/16? Yes, it is! BUT, using the same criteria it's possible to discriminate 44/16 from 44/16! In it's early days, 44/16 had it's issues, today's 44/16 is quite different. If it were possible to do a direct comparison, I'd be very surprised if we could not fairly reliably discriminate early 44/16 from today's 44/16. The same principle applies to 24/96. There was a period of 5 or so years when it was entirely possible to differentiate between 24/96 and 16/44. A large number of plugin processors operated significantly better at 96kHz that at 44.1kHz (easily determined in DBTs) and in some cases this is still the case, some: Compressors, limiters, soft-synths, guitar amp/cab and other modelling plugins for example. So, 24/96 can be audibly different/better than 16/44 then, case closed! Hang on, not so fast! Filter implementation has improved significantly over the last 20 years, so has upsampling/downsampling conversion, due to far more available processing power and improved software algorithms. Today (and for nearly a decade) a plugin which benefits from a higher sample rate can simply up-sample to that rate, process and then down-sample again, with no audible loss/artefacts. There's no longer any audible benefit to actually recording or outputting the audio files at 96kHz, and 24/96 can no longer be discriminated from 16/44, unless a DAC/DAP manufacturer deliberately (or inadvertently) builds a difference into their design.

G

 

[frodeni]

...We all know what the envelope of a bell-hit waveform looks like (a sine wave that jumps to full volume within one or two cycles, then dies down gradually).
We also agree that, once we band limit it, it will be spread out in time. ..

Signal processing is simply not my field of expertise. I simply do not know of the effect of this processing. Sorry, it is probably just me, but I am at least one guy, that do not know much about this effect. Do you have some simple explanation of this, with illustrations, that could help med get an idea of how this works, without the need for a ton of math? Also, what makes 16/44.1 "band limiting"? Does this affect all freq ranges?

If 16/44.1 is the limit of most peoples hearing, lets just say for simplicity sake. Is "band limiting" and issue if recording at 16/44.1?

For me personally, my range of hearing is being affected by my age. All the tests given, I simply fail to brake any records, and if doing any listening tests, my results are about where to expect them to be. I cannot hear sounds that high any more, to name one.

Still, using my PC or laptop as a USB source, is like night and day, for my setup. Which makes hardly any sense, but which is the case. The dynamic range of 16bit is far wider than my listening level, and given the sampling rate, it is closing at twice what is supposed to be needed. This makes no sense to me, and something is lacking by our understanding of this.

..And, if our brains use the relative arrival times between the sharply rising beginnings of those envelopes to calculate position, then isn't making them more gradual (by adding a pre-ringing "slope") going to make those calculations less precise?
Could this be why a lot of people claim to find signals reconstructed by filters that introduce pre-ringing to "sound less natural"? ...

If you are talking about degraded signals, far below 16/44.1 level of quality, how does that translate into above 16/44.1 quality? If this is an distortion of the signal, how does it manifest itself, as distinct and unique sonic traits?

I know you use this bell as a sample, but how will this translate for bell-less recordings? What is the subjective part of it. Is it possible to find a definition that people can agree upon, as to describe this shared experience, and still talk about the same sonic trait?

... Or they are socially encouraged to claim that's what they hear. Once again, how about these people do some blind tests?

Which is a valid argument, with a flaw. A shared understanding of how this manifest itself as clearly detectable sonic traits, cannot be achieved by blind tests. Hardly any understanding may be reached by a blind test.

If someone passes or fails a blind test, so what? For instance, if you have a lossy compressed signal, and then distorts is even more, what causes the sonic difference? The distortion of the lossy signal, as to emphasis its fails, or is it fair to claim that the extra distortion should not degrade the signal in any unfair way? There is no simple and straight forward answer, as it would depend on a lot of factors.

 

[gregorio]

[1] Still, using my PC or laptop as a USB source, is like night and day, for my setup. Which makes hardly any sense, but which is the case. The dynamic range of 16bit is far wider than my listening level, and given the sampling rate, it is closing at twice what is supposed to be needed. This makes no sense to me, and something is lacking by our understanding of this.
[2] A shared understanding of how this manifest itself as clearly detectable sonic traits, cannot be achieved by blind tests. Hardly any understanding may be reached by a blind test.
[2a] If someone passes or fails a blind test, so what?

1. So if something makes no sense to you then mankind is lacking understanding? If something doesn't make sense to you maybe it's because you simply don't know or understand something which mankind/science does? There are at least two explanations I can think of which make plenty of sense and do not contradict science.
2. Of course it can. A blind (preferably DB) test can tell us if that/those "clearly detectable sonic traits" are in fact audibly detectable in the first place and if not, the shared understanding is effectively a delusion, if it is still detectable, then we've eliminated a whole set of potential causes and significantly narrowed down the search for the real answer.
2a. No big deal but if everyone who takes it does, as the sample size increases so does our confidence in the result accurately extrapolating to everyone else.

G

 

[KeithEmo]

I agree..... with some qualifications.

We have plenty of scientific data about the audibility of continuous sine waves by human beings.....
However, I'm not convinced that the data for continuous sine waves can necessarily be generalized to other types of content.

I think you also have an excellent point about the history factor.
The technology has progressed a lot in recent years.
Perhaps it makes sense to ignore or discard ALL results from the past and perform some new tests using current technology.

Another flaw in many previous tests was that they were limited to certain specific pieces of equipment.
With that limitation, all you can claim to have proven is that the differences were or were not audible using that particular equipment.
Perhaps the differences are clearly audible with a certain speaker and totally obscured by another; perhaps they're only audible when a certain filter type is used by the DAC, or when a certain ADC is used to produce the test content.
(And, no, allowing a bunch of audiophiles to select their own equipment, based on their own preconceived notions, does not prove that equipment has been chosen that will actually reveal any differences present.)
If we're attempting to prove or deny a negative, then at least a good cross section of equipment must be tested.

I've found the attention to detail disappointing in ALL of the previous tests I've read about....

Here would be my minimum requirements:
- present detailed information about the provenance of all test tracks used
- present detailed information about how the different test versions were derived
(I would want both samples to be derived from the same master copy, of higher resolution, down-sampled using the same software and filter options for both samples)
- present a spectrum analysis of each track to show that it does in fact contain content above 20 kHz
- present a similar spectrum analysis, taken with a calibrated microphone at the listening position, to confirm that the "extended content" is actually making it to the listener's position
- present details about the equipment used (DAC type and brand, filter options chosen, speakers, amplifiers, etc.)
- repeat the test using a variety of equipment (to avoid situations where one particular piece of equipment masks the differences between different sample rates - or treats them differently).
- present details about the test environment (quiet, noisy, absorptive room, reflective room)
- perform the test with several different filter options (maybe you can hear the difference with an apodizing filter, but not with a flat-phase brick wall filter, or vice versa)
- I would want to include both headphones and speakers (I find electrostatic headphones especially revealing of small differences that may not otherwise be obvious)

This is all pretty basic stuff "when designing a credible scientific test"....
If we want to claim a general conclusion, then we need to include enough variations to reasonably do so.

1. True to an extent but the tests I disagree with are those whose results are contrary to known and demonstrated scientific facts.
2. Again true to an extent. The Theiss study for example demonstrated no ability to discriminate but in a supplementary non-ABX experiment with 3 of the subjects, a very high ability to discriminate was demonstrated. There are several reasons to strongly dispute this supplementary experiment, not least that it wasn't double blind but the meta-analysis only includes the supplementary experiment results, not the main ABX experiments. And, if we're going to eliminate the M&M study on the grounds that some of the samples were not Hi-Res, as the meta-analysis does, then how come there are other studies included which had no Hi-Res samples at all because they were well before hi-res was even invented?

The problem we have here is also one of time/history. Is it possible to discriminate 96/24 from 44/16? Yes, it is! BUT, using the same criteria it's possible to discriminate 44/16 from 44/16! In it's early days, 44/16 had it's issues, today's 44/16 is quite different. If it were possible to do a direct comparison, I'd be very surprised if we could not fairly reliably discriminate early 44/16 from today's 44/16. The same principle applies to 24/96. There was a period of 5 or so years when it was entirely possible to differentiate between 24/96 and 16/44. A large number of plugin processors operated significantly better at 96kHz that at 44.1kHz (easily determined in DBTs) and in some cases this is still the case, some: Compressors, limiters, soft-synths, guitar amp/cab and other modelling plugins for example. So, 24/96 can be audibly different/better than 16/44 then, case closed! Hang on, not so fast! Filter implementation has improved significantly over the last 20 years, so has upsampling/downsampling conversion, due to far more available processing power and improved software algorithms. Today (and for nearly a decade) a plugin which benefits from a higher sample rate can simply up-sample to that rate, process and then down-sample again, with no audible loss/artefacts. There's no longer any audible benefit to actually recording or outputting the audio files at 96kHz, and 24/96 can no longer be discriminated from 16/44, unless a DAC/DAP manufacturer deliberately (or inadvertently) builds a difference into their design.

G

 

[KeithEmo]

My point there was simply that, if our brains use characteristics of the envelope to determine location, and we then change the envelope, then we are likely to alter the results.
If we add pre-ringing to that recorded bell sound, we will alter its original envelope, which begins rather sharply, into one that ramps up more gradually.
I personally don't know how much that would affect how our brain interprets the location of the "real bell" - but maybe we need to find out.
(In analogous electronic test equipment, spreading out the envelope of a pulse does in fact make it more difficult to resolve the beginning time accurately.)

Recent research has shown that a lot of the analyses performed by our brains is remarkably complex - and often involves multiple parallel processes.
For example, when you see something, and move your eyes to look directly at it, it "seems" as if you recognize something and move your eyes towards it (as a single process).
However, according to recent research, your brain senses motion and "starts your eyes moving in the right direction" BEFORE identifying the target or even precisely locating it.
Once your eyeball is already moving, results from other sections of the brain, which are more precise but take longer to complete, either refine the motion command, or countermand it.

Therefore, it's not unreasonable to wonder if, while your brain is identifying the pitch you're hearing by using your ear like a spectrum analyzer...
another section has already tentatively identified the approximate location of the sound based on the beginning edges of the sound envelopes.
(Which might suggest that altering the shape of those envelope leading edges might affect that part of the result.)

Note that I said MIGHT.......
I don't know if this will turn out to be true or not; but I'm not willing to rule it out until it's actually been tested.
(Note that this wouldn't be especially difficult to test - I just don't think it's been done yet.)

And, yes, this would call for a lot of testing.
For example, we could ask people to locate a bunch of objects in the sound field, and rate their accuracy ("point to where it sounds like the violin is coming from").
We could then ask them to repeat the test with test samples recorded at various sample rates and see if their accuracy is the same for each - or not.
If a given person can locate various instruments with greater accuracy when higher sample rates are used, then it would prove that information is lost when reducing the sample rate.
If not, then it would prove that, at least under those particular conditions, it really doesn't matter.

...................................

It's already spread out in time, and the amplitude of the ringing will depend upon the samples with which the filter is convolved. Masking and other audibility issues must be considered here.

...................................

 

[bigshot]

I never cease to be amazed at how much time people spend thinking about the inaudible. I guess it's fun for them.

 

[RRod]

@KeithEmo If you get me a free set of HRTF measurements out of it, I'll participate in that test.

 

[frodeni]

1. So if something makes no sense to you then mankind is lacking understanding? If something doesn't make sense to you maybe it's because you simply don't know or understand something which mankind/science does? There are at least two explanations I can think of which make plenty of sense and do not contradict science.
2. Of course it can. A blind (preferably DB) test can tell us if that/those "clearly detectable sonic traits" are in fact audibly detectable in the first place and if not, the shared understanding is effectively a delusion, if it is still detectable, then we've eliminated a whole set of potential causes and significantly narrowed down the search for the real answer.
2a. No big deal but if everyone who takes it does, as the sample size increases so does our confidence in the result accurately extrapolating to everyone else.

G

You need to read carefully what you just wrote. If want to get a paper rejected, this is all you need to write. This will get you flunked, on its own. People do get the blunder, right?

 

[castleofargh]

I never cease to be amazed at how much time people spend thinking about the inaudible. I guess it's fun for them.

I also feel like I'm an alien sometimes. I hate loud music(that's on me), so realistically when I end up with a noise floor 35 to 40dB below music(computer, street, annoying people...), that's pretty much the hifi moment of my day. and distortions tend to be in that area too, not that I can tell unless it really goes high. then band limiting, 16khz is but a quiet remnant of sound, 17khz is gone at my listening level(in fact I also have a very local drop around 7.4khz). but just to be able to tell that 16khz is my limit, I have to be careful not to use most of my IEMs as they roll off even faster than I do.
I feel boxed inside a small frequency range and a small dynamic range for what I struggle to call fidelity with a straight face. and that's before looking at all the issues related to transducers and HRTF with headphones.
so when I read comments about the better soundstage with a gazillion taps, people crying about the lack of dynamic "caused" by redbook, or time smearing, temporal blur and other "hi I'm Barry Allen and I'm the fastest audiophile alive" references, they all feel divorced from my audio reality or the very concept of magnitude.

 

[bigshot]

More dynamics and frequencies isn't audiophile... More BALANCED dynamics and frequencies are.

 

[castleofargh]

I thought "more" was always audiophile. ^_^

 

[frodeni]

I also feel like I'm an alien sometimes. ...

I feel human.

 

[jimmers]

I also feel like I'm an alien sometimes. I hate loud music(that's on me), so realistically when I end up with a noise floor 35 to 40dB below music(computer, street, annoying people...), that's pretty much the hifi moment of my day. and distortions tend to be in that area too, not that I can tell unless it really goes high. then band limiting, 16khz is but a quiet remnant of sound, 17khz is gone at my listening level(in fact I also have a very local drop around 7.4khz). but just to be able to tell that 16khz is my limit, I have to be careful not to use most of my IEMs as they roll off even faster than I do.
I feel boxed inside a small frequency range and a small dynamic range for what I struggle to call fidelity with a straight face. and that's before looking at all the issues related to transducers and HRTF with headphones.
so when I read comments about the better soundstage with a gazillion taps, people crying about the lack of dynamic "caused" by redbook, or time smearing, temporal blur and other "hi I'm Barry Allen and I'm the fastest audiophile alive" references, they all feel divorced from my audio reality or the very concept of magnitude.

Me too,
My "Audiophile" What moment was when I first read a DAC spec, 32bit at 384k samples per second.
So I went and bought a 16 bit ladder (mainly R2R) DAC for my Redbook rips and gave up on stupid numbers and went back to enjoying my music.

 

[theveterans]

Me too,
My "Audiophile" What moment was when I first read a DAC spec, 32bit at 384k samples per second.
So I went and bought a 16 bit ladder (mainly R2R) DAC for my Redbook rips and gave up on stupid numbers and went back to enjoying my music.

Same here. Redbook and 320 kbps MP3 sounded amazing on the 16-bit R2R ladder-on-a-chip DAC that I have.

 

[bigshot]

Ears are what matters. Not theories.