Audio Science: Does it explain everything about how something sounds?

[amirm]

I don't recall that AES paper and I'm a member. Do you know what number the paper was? When was it submitted?

I'll have to assume no one who ignore/hate science (per the above poll), would use any form of blind/controlled test results, such as ABX, to bolster their beliefs.

He is talking about Stuart paper on test of high-resolution audio.

As a side note, I don't want to derail Peter's thread by rehashing that topic which we have extensively discussed. Just answering Orb's question, sharp filters used to resample create time domain ringing. The ear is not a spectrum analyzer and it has been thought that it can hear this ringing as these samples arrive, instead of waiting for all of them to come and then say, "oh that is a perfect filter." So "measurements" do show a deviation here, albeit in time domain, not frequency.

 

[Phelonious Ponk]

I spent a lot of pages with data (with a lot of help from JA) showing why it would not be IMD, surprised you forgot that.
Your first point, that is also not plausible because the measurements are very small and one aspect objectevists arguing with the results said would be inaudible, so therefore you are concluding Amir skewed his listening to cause a flawed test
I did not see you argue that one back in the day Tim
And before you say I am putting words in your mouth, look at your context; you say you doubt is is something outside of the audible range and yet measurements-analysis done by other objectivists as I keep saying showed this is inaudible due to being very small, hence the controversial nature and continued arguing between objectivists on that matter, which never was concluded.
However going by scientific logic; we had a few listener pass a DBT ABX test where traditional measurements-analysis suggest there is no meaningful measurement difference between the files.
Those who initially said it was impossible to pass and no-one had in the past went on to say Amir and others (excluding those that deliberately did this with the earlier files that had additional ultrasonic info) skewed it listening too loud and causing IMD (although this was not an issue strangely enough in the past) where several of us spent a lot of time going through various data and logic-correlation-trend points and other measurements to put that one to bed, including and importantly finally the test using real music...
But again you come back to IMD like they did and with no data-correlation, which is kinda disappointing considering how much effort a few of us put into looking into that hypothesis...
Cheers
Orb

Sorry, Orb, maybe I bailed out of that thread or stopped reading long posts by the time you got to the stuff you're talking about. So, OK, not IMD. I'd still like to know what it is.

Tim

 

[Phelonious Ponk]

He is talking about Stuart paper on test of high-resolution audio.

As a side note, I don't want to derail Peter's thread by rehashing that topic which we have extensively discussed. Just answering Orb's question, sharp filters used to resample create time domain ringing. The ear is not a spectrum analyzer and it has been thought that it can hear this ringing as these samples arrive, instead of waiting for all of them to come and then say, "oh that is a perfect filter." So "measurements" do show a deviation here, albeit in time domain, not frequency.

Are you saying this is what you're hearing when you differentiate hi-rez from Redbook in ABX testing?

Tim

 

[ddk]

What amp?

Tim

Its up there with heckle & jeckle, you wouldn't have to ask if you were familiar with Proceed.

david

 

[amirm]

Are you saying this is what you're hearing when you differentiate hi-rez from Redbook in ABX testing?

Tim

No, i don't know what I am hearing. All I was looking for where differences in unknown samples. I have no idea which is which. What I showed is one of the theories of why audibility could be there.

The only exception was conversion to 32 Khz where I did observe audible differences that I could identify to be a degradation and reported them. Again, let's move on guys.

 

[ddk]

And.....we're back in the same well-worn rut

You can't squeeze blood out of stone, there's nothing there to get no matter how hard you squeeze.

david

 

[Orb]

No, i don't know what I am hearing. All I was looking for where differences in unknown samples. I have no idea which is which. What I showed is one of the theories of why audibility could be there.

The only exception was conversion to 32 Khz where I did observe audible differences that I could identify to be a degradation and reported them. Again, let's move on guys.

And I think that is what is interesting, you and a couple of others (ignoring those that went silly and turned it way up, especially on the earlier files) managed to identify a very subtle difference using a DBT ABX, which you agree Arny and several others said would be impossible, eventually doing this with music.
I agree no-one knows what you picked up on terms of the variable involved hence why I asked the question
But it is very important to this debate and not something we can just put to the side.
People in this thread have been arguing is it audible in an ABX test/meaningless small measurements can be dismissed as they have no indication-correlation/etc, that test raises some unfortunate points in that several of you did pass an DBT ABX test (as some keep raising in this thread) but with measurements that are too small to actually identify what variable you picked up on (which goes against the other aspect they raise).
It is the crux of this thread in some ways as it shows a divergence between the test result and what quite a few objectivists keep on about that if the measurements are not notable in context of audibility (unfortunately even this can be debated and has many times) they should be ignored.
This will bring me on to my next post and relates to Stan Curtis and when he did Discrimination Function Analysis involving two amplifiers with small differences (meant to be below audibility).
Cheers
Orb

 

[Tony Lauck]

I've been waiting for that answer for months. I don't doubt that there is a difference between hi-rez and Redbook. What I doubt is that it is something outside of the audible range that some audiophiles are hearing. Someone did mention IMD in the conversation, but my recollection is that they were talking about noise in the supersonic range creating IMD within the sonic range. Has that been ruled out? Has the difference Amir and others are hearing been isolated? I'd love to know what it is myself.

Tim

As I understand it, the argument is that what Amir heard was not the presence of high frequencies (above 20 kHz) because established tests showed that sine waves at the frequencies involved would be inaudible. Logically, this appears to be a contradiction. There are some possibilities:

1. Amir didn't actually hear anything, just made some lucky guesses
2. The test was somehow flawed (due to some defect in the test signals or the playback thereof)
3. Models of human hearing that predict inaudibility of complex sounds based in audibility of sine waves are only approximate and fail with trained observers.
4. Amir cheated.

I believe these are all the possibilities that are logically possible, but I will be happy to stand corrected if additional ones are pointed out. Here is my opinion of each of these:

1. Unlikely. Amir appears consistently able to get a high score. I have little doubt he could repeat these tests 100 times over and get huge statistical confidence that he wasn't guessing.
2. There was extensive discussion about this. I don't believe it was complete, but the biggest question concerned distortion in the playback and, at least in the key jangling test, this was adequately addressed. (Note that "distortion" in Amir's ears would not count as an error in the test files, provided that the volume levels were reasonable, which I believe they were.)
3. Models of human hearing, such as the F.M. curves and variants thereof are based on linear systems theory, which is a good first order approximation to reality that happens to greatly simplify the mathematics. However, such models may not capture all that is known about human hearing, such as the effects of non-linearity in human hearing and the possibility that beat tones between ultrasonic tones can be audible.
4. Any disagreements I have with Amir (and there are a few) have done nothing to cause me to doubt Amir's honesty in the slightest. I was shocked (but not surprised) that he would be accused of cheating. This episode calls into question the character of the accuser to the point where I will no longer place any credence in that person's remarks nor have anything do do with him, at least not until I see a public apology for his behavior.

In my opinion, the most likely explanation for Amir's passing the test is some combination of 2 and 3. Further investigation is warranted, because knowledge in both areas offers the promise of obtaining more transparency in audio playback. Either of these, or some combination thereof, represents a potential failure of audio science if the explanation is ultimately found in some portion of the record - playback chain that doesn't include the specific test equipment, e.g. the PC ABX software.

 

[amirm]

- vs the B&W, power hungry and an extremely difficult load fed by an inadequate amplifier. I don't about know how it rates as a mono speaker but I do know that its very setup sensitive and its dispersion abilities can't compete with a easy to locate wide dispersion theater horn designed for the task.

I'm trying to figure out if Amir is simply in disagreement that the test has flaws or actually saying that our hands on unscientific experiences worthless because he thinks so or is the subjective assessment of the speakers, the electronics and the methodology is incorrect. If its the latter then he can easily bring up the data and scientifically prove that I'm wrong. If its the former, well...

Let's discuss this in the context of Peter's thread and topic, namely whether measurements are useful in this regard.

The loudspeaker in question has a rated sensitivity of 90 dB. What this means is that at 1 meter, it generates 90 dB SPL sound pressure ("loudness") with 1 watt of input power. Now, this is a marketing number and could be a cheat. Fortunately stereophile reviewed this loudspeaker and measured it to have a sensitivity of 89 dB SPL, which is close enough. So good job B&W for reporting this correctly.

The listening level was calibrated across all the loudspeakers for 80 db SPL. I estimate that where I was sitting was probably 10 feet away or 3 meters to use whole numbers.

With this information we now can compute how much power is required to generate 80 db SPL using that B&W loudspeaker. Plugging that into my dusty SPL calculator in Excel, I get 1.1 watt! Yes, you read that right. It only takes one watt or so to generate 80 db SPL at that listening position because we started with 90 db and moving back happen to almost coincide with the level we were aiming at.

We don't want to cut this so close to the bone so let's put in 3 db of headroom and we arrive 2.3 watts of amplification power required to give us 80 db SPL at listening position.

Now let's look at how much power we have available to us. The amplifier that was used for testing is the Proceed AMP3. I own the Proceed AMP5 that I use to power my theater so it is a product I am familiar with. It is rated at 125 watts into 8 ohms and 250 watts into 4 ohms. Which number do we use? Well, we resort to measurements again. From stereophile review we see this measurement of B&W's impedance versus frequency: http://www.stereophile.com/content/bw-802d-loudspeaker-measurements#TlZzBOfBFAvB56L4.97

Ignore the dashed lines for this conversation. The solid line tells us what the impedance/load is at each frequency. Since most power in music is required in low frequencies, that is where we focus our attention. There, we see impedance dips down to 4.2 ohms at 20 Hz or so, and 3.5 between 70 to 200 Hz. This means that our amp will be operating using those loads and hence the applicable number of watts we have available is 250 watts (roughly speaking).

At the risk of stating the obvious, 250 watts is a heck of a lot more power than 2.3 watts. So there is no trace of amplifier running out of power whatsoever. Indeed, we can compute how loud it could get by working in reverse, and computing the SPL at listening position while using those 250 watts. That give us in excess of 103 db SPL which I can assure you, would have been far, far louder than anything presented.

So while you are right David that I would not hook up the B&W to a tube amp with high impedance given the dips in the response of B&W to less than 4 ohms, a powerful solid state amplifier has no trouble at all driving this test.

I can also share that subjectively, I did not at all hear any clipping distortion or the amplifier running out of power. Nor did any of the other people in the listening panel in the two times I have attended this test. Nor do I know of any objections to the paper/research in the grounds of amplification.

 

[amirm]

But it is very important to this debate and not something we can just put to the side.

OK, you guys can discuss it if it is OK with the thread owner Peter. But please don't ask me. I spent weeks on AVS discussing the topic, weeks more here, and then weeks more yet again HA forum. I have no more energy or anything new to share on it. Hope you understand .

 

[AJ Soundfield]

He is talking about Stuart paper on test of high-resolution audio.

Actually it's pretty clear he's not talking Stuart, but rather your home test of Arny's files. I've been following the AES comments on Stuarts paper, including their own, with quite some interest.

No, i don't know what I am hearing. All I was looking for where differences in unknown samples. I have no idea which is which. What I showed is one of the theories of why audibility could be there.

In my view, what you've shown, is you were wise to hire JJ for his world renown expertise in DSP and perceptual testing, as opposed to Arny. All files/test results follow.

Just answering Orb's question, sharp filters used to resample create time domain ringing. The ear is not a spectrum analyzer and it has been thought that it can hear this ringing as these samples arrive, instead of waiting for all of them to come and then say, "oh that is a perfect filter." So "measurements" do show a deviation here, albeit in time domain, not frequency.

As JJ might say, pray tell what that has to do with benefits of "Hi-Rez" vs pathological resampling...
I assume by now you've discussed Stuarts paper with him? Such a wonderful resource you have.

As a side note, I don't want to derail Peter's thread by rehashing that topic which we have extensively discussed.

I agree 100%. If there are thread discussing all these deliberate detours, that's where they should occur.
Now if Peter or anyone else could tell us about these audio N-rays emitted by their stereos, that were designed, engineered and manufactured in, but completely unknown to current science....

cheers,

AJ

 

[Tony Lauck]

There is no other explanation from the ABX mafia other than anybody who has provided a positive result was cheating!

I did give a recent example of mzil (one of the mafia) on HA reporting that he hears noise differences (it could only be dither) in new files supplied by ArnyK (the Don) which were a claimed to be new test files for ABX testing high-res Vs RB. He supplied other files to mazil, at mzil's request to specifically test this noise that mazil was picking up on - to rue out IMD - the files were of digital silence, so the only noise was shaped dither. The RMAA analysis of these specific files showed no noise above -130dB

Noise perception thresholds, anyone ?

I can hear 16 bit dither noise at volume settings that I occasionally use for playback (e.g. of Mahler symphonies). This was TPDF dither (no noise shaping). I can hear harshness when applying noise shaping at 16 bits at the 44 kHz sampling rate, at least on some recordings, but this harshness is not consistent and varies from recording to recording. Accordingly, I have banished the use of noise shaping at 44 kHz, as I believe the cure is often worse than the disease.

If you plot flat 16 bit dither noise you will find that it is around -130 dB on most spectrum plots, depending on FFT gain. What this shows is that differences on spectrum plots at the -130 dB level can not be assumed to be inaudible. I have found it helpful to generate test signals and put them through an FFT. This shows insight on how to interpret these plots.

 

[AJ Soundfield]

I can hear 16 bit dither noise at volume settings

This is a well known possibility and exactly as M&M did. But then it begs the question....

that I occasionally use for playback (e.g. of Mahler symphonies).

Numbers please. Exactly how loud? What are your measured full spectrum ambient noise levels?
Is there a thread you should post this in to avoid constantly derailing this one and repeatedly bringing up topics that completely contradict the thread topic/title? TIA

 

[AJ Soundfield]

I can also share that subjectively, I did not at all hear any clipping distortion or the amplifier running out of power. Nor did any of the other people in the listening panel in the two times I have attended this test. Nor do I know of any objections to the paper/research in the grounds of amplification.

Any idea what the peak levels are during the tests?

cheers,

AJ

 

[Orb]

I thought it was about time to bring this to the discussion.
Stan Curtis did an interesting article in Hifi Critic issue 1 Volume 7 (Jan-Mar 2013) where he decided to investigate small differences using the Discriminant Function Analysis.
The reason to use DFA is when measurements involve tiny signals-measurements and to identify consistent differences sample to sample.
If there is a trend it is easier to see and look at that from the perspective of correlation.
Anyway a segment from the article:

I proceeded to measure ten samples of an unmodified amplifier. Various measurements were made of frequency response, the harmonic structure of distortion at various frequencies and power levels; the harmonic structure with different output loadings and so forth.....
I then made modifications to the design that led to a clear improvement in the reproduction of recorded sound (at least to my ears).I then repeated the measurements and yes there were differences but they were what some people might term insignificant, or what my colleagues might describe as "two tenths of b****r all".
But now I undertook the DFA and plotted the results onto a chart. And guess what?
The two sets of results each formed a cluster so that the difference between the two versions of the amplifier was there; a difference based upon objective measurements.
Note, however, that this form of analysis takes note of the measurement in the round. You cannot say sounds better than because it has lower distortion, or more extended frequency, or a faster slew rate. The analysis is far more holistic, and in doing ties in with our subjective analysis which sums the parts to reach a view.
.....
I think it is also time to question those early arguments about the insignificance of measurements. Such statements find their work done with test tones....
Most of the existing literature on listening tends to follow a similar pattern, and simply has little relationship to what I term the holistic experience listening to a musical performance

.

This sort of comes back to some of what I was saying earlier, where it is a correlation-trend involving those small measurements with say what a listener reports (such as our example of cold start to warm-up where basic measurements indicate a change does happen with the amplifier behaviour but measurements show it is small, this is further compounded that only one measurement is done - in this example Stan measured multiple variables to create the DFA plots).
Yes I agree many cases of say cold-start is anecdotal posters/reviewers (and there will be bias related ones as well but it cannot be applied to all, which as I keep pointing out many have also mentioned some kind of perceived loudness shift) or engineers' experience but the point is looking to see if a correlation exists between that experience and measurements, where maybe they should use some kind of DFA test framework if looking at this only from measurements.
And before anyone says it is expectation bias that he made a change and so expected it to be better, his modification did actually influence the measurements to an extent a correlation-model could be seen from the DFA result (need to buy the issue if really interested or see if online).
I am just continuing with the cold start to warm-up example because it is something that has measurements and more solid anecdotal and engineering experience-comments.
Just as applicable is the ABX test with the hirez vs CD, jitter at say 500ps, the Melco review I recently posted where the reviewer deemed the ethernet side had only marginal benefits and can be done cheaper (this had low jitter value differences) while he was enthusiastic using it as a direct USB device to DAC (had notable noise improvements for tested DACs but like the jitter the noise was already at an 'insignificant' S/N ratio of 95dB for the worst one before the improvement) compared to a MacBook,etc.
Cheers
Orb

 

[Orb]

OK, you guys can discuss it if it is OK with the thread owner Peter. But please don't ask me. I spent weeks on AVS discussing the topic, weeks more here, and then weeks more yet again HA forum. I have no more energy or anything new to share on it. Hope you understand .

I know and remember I was heavily involved on here when it came to discussing some of those 'potential' hypothesis they kept throwing up
IMD on its own was 20 pages, with JA patiently posting test measurements I kept asking for to help investigate those theories.

You may remember I was pretty sure you would not change anyone's thoughts on that subject as well
Cheers
Orb

 

[jkeny]

I can hear 16 bit dither noise at volume settings that I occasionally use for playback (e.g. of Mahler symphonies). This was TPDF dither (no noise shaping). I can hear harshness when applying noise shaping at 16 bits at the 44 kHz sampling rate, at least on some recordings, but this harshness is not consistent and varies from recording to recording. Accordingly, I have banished the use of noise shaping at 44 kHz, as I believe the cure is often worse than the disease.

Is this harshness only perceptible at elevated volumes?

If you plot flat 16 bit dither noise you will find that it is around -130 dB on most spectrum plots, depending on FFT gain. What this shows is that differences on spectrum plots at the -130 dB level can not be assumed to be inaudible. I have found it helpful to generate test signals and put them through an FFT. This shows insight on how to interpret these plots.

I wonder would it be a useful exercise, in another thread - if you had the time & inclination, to give some guidelines regarding the generation of these test signals, the FFTs of such signals & their audibility? This would help everyone get a practical handle on just what it is that is being talked about here.

 

[Orb]

I can hear 16 bit dither noise at volume settings that I occasionally use for playback (e.g. of Mahler symphonies). This was TPDF dither (no noise shaping). I can hear harshness when applying noise shaping at 16 bits at the 44 kHz sampling rate, at least on some recordings, but this harshness is not consistent and varies from recording to recording. Accordingly, I have banished the use of noise shaping at 44 kHz, as I believe the cure is often worse than the disease.

If you plot flat 16 bit dither noise you will find that it is around -130 dB on most spectrum plots, depending on FFT gain. What this shows is that differences on spectrum plots at the -130 dB level can not be assumed to be inaudible. I have found it helpful to generate test signals and put them through an FFT. This shows insight on how to interpret these plots.

Keith Howard did an interesting article for Stereophile back in 2005 on dither, unfortunately he revisited this at a later date but for Hi-fi News and so it is not on the web, still supports much of what he raised back then.

he first thing you notice when comparing the piano tracks with the different noise signatures—again with the original as a reference—is that TPDF noise is more irritating than the more Gaussian-like alternatives. And it has an effect on the sound of the piano, which is more "clangy" than in the original, or in those tracks with a more natural noise PDF. Of course, this experiment was conducted with an amplitude of noise getting on toward 30dB higher than 16-bit TPDF dither, but it may be that even at this much lower amplitude, and even when the dither is added to more naturally distributed noise, the ear can still detect it as something unnatural.

Thinking along these lines, I have also experimented with how best to "hide" TPDF dither within a typical audio signal. It is normal practice in two-channel or multichannel recordings to use uncorrelated dither in each channel. This has the advantage that summing two uncorrelated noise signals increases the noise amplitude by 3dB, whereas summing two correlated noise signals increases the amplitude by 6dB. So uncorrelated dither promises a superior S/N ratio in normal listening, and if the channels are summed to mono or a multichannel signal is downmixed to stereo.

But there are other ways to look at this. Uncorrelated dither becomes part of the S (difference) component of a stereo signal, which is usually of much lower amplitude than the M (sum) component because the audio signals in each channel are typically quite highly correlated. This raises the possibility that dither noise might be easier to hide if it is identical in both channels rather than uncorrelated. It is also feasible that having the dither noise precisely located in the soundstage makes it easier to "tune out," whereas the diffuse nature of uncorrelated dither is less easy to ignore.

I have tested this idea only briefly, using the same piano track and the same high noise level. The experiment would need to be repeated at more representative noise amplitudes and with a range of source material to reach any firm conclusions, but what I've heard encourages me to believe this is an idea worth pursuing. Tony Faulkner put me on to this and suggested that this issue was probably the subject of research by Sony and others in the early years of digital audio, although I can't recall ever seeing a reference to it.

Contingent Dither: http://www.stereophile.com/features/705dither/index.html

Cheers
Orb

 

[Tony Lauck]

Is this harshness only perceptible at elevated volumes?

I wonder would it be a useful exercise, in another thread - if you had the time & inclination, to give some guidelines regarding the generation of these test signals, the FFTs of such signals & their audibility? This would help everyone get a practical handle on just what it is that is being talked about here.

We heard harshness when mastering a recording that had originally been tracked digitally at 16 bits. This was a "triple blind" situation, where we both heard differences from one day to the next when playing the last file the night before as the first file of the next day and it sounded different, even though it was supposed to be the "identical file". It turned out it wasn't, that the dither setting used to produce a 16 bit version had changed due to "cockpit error" on my part. It could be that the harshness came from starting with a recording that had already been subjected to noise shaping as part of the ADC algorithms, but I didn't have visibility into this part of the situation. Levels were those typically used for mixing and mastering, e.g. loud portions at around 85 dB at the listening position. There was no amplifier clipping in this playback volume (due to calibrated tests that guarantee this).

Note that harshness, per se, is not necessarily an indication of sonic distortion in a recording. I've heard it in live choral concerts where I was 10 feet from the singers in the church, so this was a case of my ears distorting. I've occasionally noticed distortion in orchestral concerts in Symphony Hall in Boston, when sitting in row T in the center section of the floor, but only briefly in fortissimo peaks.

I use SoundForge 10c to generate test signals and often use its spectrum plot capability and statistics functions. For sample rate conversions, I use the 64 bit iZotope SRC, presently the version in RX4. This version of the SRC does not add any offset to the sampled waveforms when the linear phase option is selected. Previous versions of this SRC, for example those in RX2 Advanced, have a sub-sample off-set. Versions of this SRC also comes integrated with various software programs, such as SoundForge 10c, but some of these use older versions of the software. (If there is offset in a sample rate conversion then null tests created by mixing out of phase signals will show differences that may not be audible, and variable click artifacts when using defective ABX test software.)

My advice to people is to play with tools and also study the theory of digital signal processing. The two approaches complement each other. I have found plots as appear in technical papers and internet postings can be suspect, because there are too many ways to rig them to show what the creater wished (either mistakenly or intentionally). The only way to recognize these possibilities is hands-on experience backed with some knowledge of theory. None of this came easy for me. I spent quite some time learning the necessary math and even more time playing around with audio editing tools and learning how to hear subtle differences these tools can create. Of course if one just wants to listen to music there is no need for any of this, but if one wants their understanding of the audio hobby or profession to be science based, I don't know of any easy way to get this.

 

[Tony Lauck]

Keith Howard did an interesting article for Stereophile back in 2005 on dither, unfortunately he revisited this at a later date but for Hi-fi News and so it is not on the web, still supports much of what he raised back then.


Contingent Dither: http://www.stereophile.com/features/705dither/index.html

Cheers
Orb

It is known that using the identical dither noise in both channels collapses the sound stage, something that may not be undesirable for a solo instrument recording. This was a problem with early digital work stations where there was insufficient processing power to generate many high quality channels of random noise. This is no longer a problem with modern processors. Mixing in Gaussian noise can mask dither distortion if done at a high level, but TPDF dither can eliminate distortion and noise modulation while adding much less noise.

I don't know why Keith Howard's piano recording sounded harsh. It could be the converters, microphone preamplifiers, microphones, microphone positions, venue acoustics, instrument voicing, or even the performer. (Many pianists are "bangers" and produce a harsh tone.)

The proper test for harshness of dithering is to compare a 24 bit recording with a 16 bit conversion with suitable dither. Adding additional noise (as in Gaussian noise) may serve to mask problems in the original recording and reach incorrect conclusions as to the source of harshness. I was comparing two different dithers and hearing a difference in the harshness, whatever the cause. If one works with digital audio files, or with line level analog signals if one has state of the art converters, one can do these kinds of comparison tests and expect some validity. If one gets involved with speakers and microphones, then one's in an entirely different space of "audio science".