Sampling rate vs output question

[MLXXX]

9 hours ago, LHC said:

In some published experiments that generated heated debates in the past, two identical speakers were set up to play the same waveform to a listener. The positions of the speakers could be set up (one on top of another) such that the wavefronts are separated by 10us in arrival time (to the listener's ears).

I'm not sure whether this 2007 paper Audibility of temporal smearing and time misalignment of acoustic signals is an example of that type of experiment you refer to, but it seems to be. 

 

I haven't had a chance to read through and digest this paper, but it seems likely to be relevant to recent discussion in this thread.

[Grant Slack]

9 hours ago, LHC said:

In some published experiments that generated heated debates in the past, two identical speakers were set up to play the same waveform to a listener. The positions of the speakers could be set up (one on top of another) such that the wavefronts are separated by 10us in arrival time (to the listener's ears). The individual speakers could represent your channel 1 and 2 as discussed earlier. When they arrive at the ear they get mixed and superimposed - the experiment tests whether the listener could hear two distinct sound signals, or only one. Putting aside the various sources of possible errors, such experiments found that tested listeners could hear two signals. The interpretation of such tests has always been hotly debated. Some claim this is no more than just a demonstration of 'timing resolution' (what was said above about shifting one waveform back or forth), which can be anything, and therefore there is nothing to see here. 

 

 

Hello LHC,

 

how would this setup be different in principle to playing a Dirac pulse over one's stereo speakers, and moving one's head a few mm to each side of the exact centre? It would be easy to achieve a 10 µs delay in arrival time. In my home setup, moving my head about 4mm to one side of the centreline would do it.

 

I just tried it a few minutes ago, and it sounds like the same, one click, listening head in the centre or off centre, two ears or one ear. No sense of 2 clicks, or any different sound quality to the one click. In fact, never mind 4mm, I can move hundreds of mm to one side and it makes no difference to the perception of one click or more than one, or any difference to the sound quality of the one click.

 

Does this 10 µs test result have any relevance whatsoever to home listening to music? I am experiencing strong doubts.

 

Regards,

Grant

[LHC]

31 minutes ago, MLXXX said:

I'm not sure whether this 2007 paper Audibility of temporal smearing and time misalignment of acoustic signals is an example of that type of experiment you refer to, but it seems to be. 

 

I haven't had a chance to read through and digest this paper, but it seems likely to be relevant to recent discussion in this thread.

it is. This has been debated here on several threads so I don't think it serves any purpose to repeat that exercise. There are plenty of criticism of the paper that could be found on-line, some claiming that it has been discredited or debunked etc. But curiously there are no scientific peer reviewed journal paper that actually rebut that 2007 paper, as far as I know, or was cited by Stuart and Craven. So from an academic point of view, that paper and its finding is still valid.

 

As to whether the result of that 2007 paper could be explained by timing resolution, my proposal above was to repeat such experiments using only one ear, or with listeners who is deaf on one ear. I think that is a way to falsify or support such hypothesis. 

Edited September 14, 2019 by LHC

[MLXXX]

2 minutes ago, LHC said:

There are plenty of criticism of the paper that could be found on-line, some claiming that it has been discredited or debunked etc

I've yet to form a personal opinion about the conclusions expressed in the paper.

 

However I can say that I'm finding the English expression in it exceptionally elegant and precise, making the paper a great pleasure to read, merely from that viewpoint.

 

I baulked though at the word "underceeds". Definitely a word I'd not encountered before!

[LHC]

38 minutes ago, Grant Slack said:

I just tried it a few minutes ago, and it sounds like the same, one click, listening head in the centre or off centre, two ears or one ear. No sense of 2 clicks, or any different sound quality to the one click. In fact, never mind 4mm, I can move hundreds of mm to one side and it makes no difference to the perception of one click or more than one, or any difference to the sound quality of the one click.

That is a sighted listening test  but I like the spirit.

 

44 minutes ago, Grant Slack said:

Does this 10 µs test result have any relevance whatsoever to home listening to music? I am experiencing strong doubts.

I think from a practical perspective the very short 10us would drive the system performance closer to one that could reproduce a square wave. Whether one needs to do this is the subject of the OP. People have already expressed different opinions on this thread. 

[MLXXX]

49 minutes ago, MLXXX said:

I've yet to form a personal opinion about the conclusions expressed in the paper.

I'm disappointed that the paper does not make any reference to feeding the closer speaker a time delayed version of the signal, to match and cancel out the time delay of hearing sound from the more distant speaker.  If that had been done, it could have served as a control that the differences perceived were most likely in relation to signal timing, and not to other possible complicating effects.

 

I would suggest that the decision not to use electronic manipulation of the test signals, and in that way enjoy the benefit of avoiding possible unintended artefacts in that electronic manipulation, whilst laudable at first blush, was probably unnecessary, as such electronic manipulation could routinely be done to an exceptionally high standard circa 2007.

 

And unfortunately, the mechanical technique of having a person listen to two speakers positioned at different distances, not far from the listener, as well as causing an amplitude difference could have caused other complicating effects that I won't attempt to identify, but which no doubt others have identified over the years since the paper was published.  I note that certain potential complicating effects are addressed in the paper, and addressed very well, with great rigour.  I suspect though that there'd be some further effects arising from using two physically separated speakers, not commented on in the paper.

 

For me, the brief  "post-publication note" might well refer to a  more persuasive experimental technique. It states: "In a closely related experiment [42], signals were temporally smeared by passing them through a low-pass filter (to simulate the effect of bandwidth restriction in an audio component) instead of by spatially displacing speakers. In that experiment, the stimulus was conveyed to the subjects through supra-aural headphones instead of speakers. The threshold low-pass time constant that could barely be discriminated had a value of 5 μs, which is comparable to what was found in the present work."

 

The footnote reference is to: "[42] M. N. Kunchur, “Temporal resolution of hearing probed by bandwidth restriction”, Acta Acustica united with Acustica 94, 594-603 (2008). (Preprint can be downloaded from http://www.physics.sc.edu/kunchur/temporal.pdf)"

 

I might try to look at that later this weekend.

 

I must say though that the 2007 paper is I believe the most detailed I have ever read for a listening test, in terms of its descriptions of the experimental technique and its rigorous presentation of related mathematical expressions such as the effects of the acoustics of the listening room. It is in my opinion an exceptionally well presented paper, a real tour de force. 

[March Audio]

Nice though this thought experiment is, wouldnt some further reading around the subject help before leaping?

 

How about the simple phenomenon of temporal masking?

 

https://www.sciencedirect.com/topics/medicine-and-dentistry/temporal-masking

 

https://en.wikipedia.org/wiki/Auditory_masking

 

 

[MLXXX]

6 minutes ago, March Audio said:

Nice though this thought experiment is,

Which thought experiment are you referring to, exactly?

[March Audio]

Really?

[MLXXX]

3 minutes ago, March Audio said:

Really?

Yes, please specify which thought experiment you are referring to.

Edited September 14, 2019 by MLXXX

[MLXXX]

On 14/09/2019 at 1:11 PM, MLXXX said:

The footnote reference is to: "[42] M. N. Kunchur, “Temporal resolution of hearing probed by bandwidth restriction”, Acta Acustica united with Acustica 94, 594-603 (2008). (Preprint can be downloaded from http://www.physics.sc.edu/kunchur/temporal.pdf)"

 

I might try to look at that later this weekend.

That was an experiment reported by the same author. It can be found at Temporal resolution of hearing probed by bandwidth restriction and is written up with a similar admirable attention to detail.

 

I see that a lot of effort went into the design of the electronics.

 

I personally am much more comfortable with this experiment based on use of electronics and high performance earphones than the experiment based on the use of two non-equidistant loudspeakers.

 

I have not read this paper as carefully as I read the first. However I have noticed a reference to a presumed lack of consequence of certain small changes in amplitude once a capacitor was switched into circuit to provide a simple first order roll off of higher frequencies:

At the discrimination threshold of τ=4.7 μs, the drop in the first harmonic level is 0.18 dB (a 4.1% decrease in intensity) and the drop in the total rms value (across the whole spectrum) is ΔLp= 0.23 dB (a 5.2% decrease in intensity). The just noticeable difference (JND) for the conditions in the experiment (f ≥ 7 kHz and Lp=69 dB) is known (from Jesteadt, Wier, and Green, 1977) to be 0.7 dB (a 15% decrease in intensity). Even the 3 standard-error lower limit of this JND is 0.5 dB (an 11% decrease in intensity). Thus in the present experiment, differences in levels and spectral weights between the threshold and control stimuli seem too small for the discrimination to arise solely from direct spectral amplitude changes.

 

If I were to dare to criticise the technique after having only quickly skimmed through the paper this afternoon, it would be to say that although achieving very high standards in certain respects (e.g. a very  fast rise time for the unfiltered version of the 7kHz squarewave), it is in my view unfortunate that the amplitude of the fundamental deviated by as much as 0.18dB.

 

I don't know what the experimental conditions were for finding a JND of 0.7dB in the 1977  publication cited above, but I seem to recall  achieving good ABX results for a 0.5dB change in amplitude for my own hearing of a test tone (I can't remember at what frequency) a few years ago.  Yes 0.18dB should be undetectable but it's just a bit too close to a detectable level for comfort for my taste, in terms of a very formal experiment, as this one was.

 

In 2019, using earphones or headphones with a very extended response, and a DAC capable of operating at a high sample rate, it ought to be possible to generate all of the test signals using Digital Sound Processing, and repeat the experiment with the assistance of a pc used as a Digital Audio Workstation or simply as an automated ABX playback device. The "unfiltered" square wave wouldn't have as fast a rise time as the setup used in Professor Kunchur's experiment but I don't think that would preclude a listener being afforded an ample opportunity to hear a difference.

 

On 14/09/2019 at 11:25 AM, LHC said:

This has been debated here on several threads so I don't think it serves any purpose to repeat that exercise.

Hi LHC,

I've missed out on a lot of previous debates because I was posting on DTV Forum mostly about video matters, and not posting here at SNA.

 

Out of curiosity, I might try to find some of those old threads and see what criticisms and comments people came up with regarding Prof Kunchur's papers, and any other similar papers.

 

Edit:

Actually I think I now remember having seen this second paper before (involving soldering in of resistors for an RC low pass filter), and perhaps even commenting on it. Woops!

Edited September 15, 2019 by MLXXX

[davewantsmoore]

16 hours ago, MLXXX said:

Yes 0.18dB should be undetectable

It can depend quite a lot on the spectrum..... so it is potentially flawed to say there is an "absolute" JND which can be applied to all circumstances.

 

[Guest rmpfyf]

On 19/08/2019 at 9:07 PM, davewantsmoore said:

We CAN hear the difference in where the 45us long click occurs in time down to very fine precision..... especially when it is slightly difference between multiple channels.

 

..... but we cannot hear the high(er) frequency content of the "click".

 

People routinely mix up the difference.

 

+1 million

 

Seriously. It's mixed up by prominent types too.

[MLXXX]

In this thread there's been discussion of various matters including:

• the occurrence and audibility of different types of phase shifts
• the ability of the ears operating in conjunction with each other to detect very small timing differences as between signals reaching one ear and the other ear (interaural time differences, or ITDs)
• the ability of one ear alone to hear a timing difference; and, 
• whether the upper limit for audibility of continuous tones occurring in isolation can be applied as a theoretical model for predicting that ultrasonic frequencies will make no perceptible difference when presented in conjunction with lower, audible, frequencies.

 

I'd suggest that SNA is not the best kind of forum for these sorts of technical discussions, as it has more a hobbyist than technical focus.  HydrogenAudio forum has a much more technical focus. Its members include a relatively high proportion of academics and engineers. If anyone really wants to pursue technical audiophile matters in depth that would be a good place to go.

 

 

A simple procedure for testing perception of an ultrasonic harmonic

 

On the question of determining whether a harmonic in the ultrasonic range can make a perceptible difference when presented to the ears at the same time as an audible fundamental, I devised a simple procedure back in 2008. It involved the following:

 

1. Playing an 8333Hz sinewave through a speaker.

2. Sometimes playing a synchronised 24999Hz sinewave (the 3rd harmonic of 8333Hz)  through a separate amplifier and speaker.

3. Listening for whether there was a difference when the third harmonic was played.

 

I've  gone back to the relevant 2008 post on HydrogenAudio forum. At the time I reported:

 

Stereo file A created with file 1 (8333Hz) as the left channel and file 2 (24999Hz) as the right channel. 
Stereo file B created with file 1 (8333Hz) as the left channel and zero signal for the right channel.

Playback volume of the left speaker was tested with the microphone 1 metre in front of the tweeter and feeding the oscilloscope.  Amplitude of the waveform from the left speaker remained constant whether or not the right channel was playing, i.e. whether file A or file B was played.

At a reasonable listening distance, A and B sounded different (file A seemed louder and a little richer).

 

The people at HydrogenAudio (HA) are sticklers for objective proof, so I set up an automated ABX procedure and then in accordance with HA requirements dutifully uploaded the successful ABX details. 

 

Of course the 24999Hz tone was not audible if played in isolation.

 

I didn't try to vary the phase of the third harmonic to see whether that would have made it easier or harder to detect. As phase has been an important subject in the current thread perhaps a very keen individual might wish to experiment with varying the phase of the harmonic relative to the fundamental (noting that that would be affected by the relative distance of the ears from each of the two speakers).

 

To be frank, the fact that an ultrasonic tone might be just detectable in an experimental setting such as I have just described is not very exciting. It isn't very exciting because the effect is so weak and subtle. It seems hardly worth worrying about.  There is a considerable doubt it would be detectable with music.

Edited September 15, 2019 by MLXXX
Results now more accurately described, after referring to what I wrote in 2008.

[davewantsmoore]

16 hours ago, MLXXX said:

HydrogenAudio forum has a much more technical focus. Its members include a relatively high proportion of academics and engineers. If anyone really wants to pursue technical audiophile matters in depth that would be a good place to go.

I am a frequenter of HA.....  are you?   (Best of luck to you....)

 

16 hours ago, MLXXX said:

1. Playing an 8333Hz sinewave through a speaker.

2. Sometimes playing a synchronised 24999Hz sinewave (the 3rd harmonic of 8333Hz)  through a separate amplifier and speaker.

3. Listening for whether there was a difference when the third harmonic was played.

I would not be surprised at all if this was audible.

 

.... but it is not the ultrasonic tone being detected, per se.    It is the alteration of the 8333hz wave.

 

[MLXXX]

17 minutes ago, davewantsmoore said:

 

I would not be surprised at all if this was audible.

 

.... but it is not the ultrasonic tone being detected, per se.    It is the alteration of the 8333hz wave.

 

With the procedure I described, the 3rd harmonic is independently amped and goes to a separate speaker system.

 

But yes the ultrasonic tone in isolation is not audible.

 

Edit:

The use of a separate speaker removes any suggestion that the loudspeaker reproducing the 8333Hz wave is being compromised or altered in its performance when the 3rd harmonic is added to the listening room.

Edited September 16, 2019 by MLXXX

[davewantsmoore]

10 minutes ago, MLXXX said:

But yes the ultrasonic tone in isolation is not audible.

Of course not.

10 minutes ago, MLXXX said:

 With the procedure I described, the 3rd harmonic is independently amped and goes to a separate speaker system.

Yes, I understand that.

[MLXXX]

18 minutes ago, davewantsmoore said:

I would not be surprised at all if this was audible.

Many people would be surprised, as conventional thinking is that it should make no difference to the perception of sound if frequencies above around 18kHz or 20kHz are not reproduced.

[March Audio]

17 hours ago, MLXXX said:

In this thread there's been discussion of various matters including:

• the occurrence and audibility of different types of phase shifts
• the ability of the ears operating in conjunction with each other to detect very small timing differences as between signals reaching one ear and the other ear (interaural time differences, or ITDs)
• the ability of one ear alone to hear a timing difference; and, 
• whether the upper limit for audibility of continuous tones occurring in isolation can be applied as a theoretical model for predicting that ultrasonic frequencies will make no perceptible difference when presented in conjunction with lower, audible, frequencies.

 

I'd suggest that SNA is not the best kind of forum for these sorts of technical discussions, as it has more a hobbyist than technical focus.  HydrogenAudio forum has a much more technical focus. Its members include a relatively high proportion of academics and engineers. If anyone really wants to pursue technical audiophile matters in depth that would be a good place to go.

 

 

A simple procedure for testing perception of an ultrasonic harmonic

 

On the question of determining whether a harmonic in the ultrasonic range can make a perceptible difference when presented to the ears at the same time as an audible fundamental, I devised a simple procedure back in 2008. It involved the following:

 

1. Playing an 8333Hz sinewave through a speaker.

2. Sometimes playing a synchronised 24999Hz sinewave (the 3rd harmonic of 8333Hz)  through a separate amplifier and speaker.

3. Listening for whether there was a difference when the third harmonic was played.

 

I've  gone back to the relevant 2008 post on HydrogenAudio forum. At the time I reported:

 

Stereo file A created with file 1 (8333Hz) as the left channel and file 2 (24999Hz) as the right channel. 
Stereo file B created with file 1 (8333Hz) as the left channel and zero signal for the right channel.

Playback volume of the left speaker was tested with the microphone 1 metre in front of the tweeter and feeding the oscilloscope.  Amplitude of the waveform from the left speaker remained constant whether or not the right channel was playing, i.e. whether file A or file B was played.

At a reasonable listening distance, A and B sounded different (file A seemed louder and a little richer).

 

The people at HydrogenAudio (HA) are sticklers for objective proof, so I set up an automated ABX procedure and then in accordance with HA requirements dutifully uploaded the successful ABX details. 

 

Of course the 24999Hz tone was not audible if played in isolation.

 

I didn't try to vary the phase of the third harmonic to see whether that would have made it easier or harder to detect. As phase has been an important subject in the current thread perhaps a very keen individual might wish to experiment with varying the phase of the harmonic relative to the fundamental (noting that that would be affected by the relative distance of the ears from each of the two speakers).

 

To be frank, the fact that an ultrasonic tone might be just detectable in an experimental setting such as I have just described is not very exciting. It isn't very exciting because the effect is so weak and subtle. It seems hardly worth worrying about.  There is a considerable doubt it would be detectable with music.

 

Did you do this test blind?  If not, epic fail. 

 

Also music only has very low level HF content.  Try the test again blind with the HF tone 70dB or more quieter that th LF tone. 

 

Finally you also need to demonstrate that that either tones were creating no significant spurious tones at other frequencies. 

 

Your speakers need to be super low distortion to successfully perform this test. 

Edited September 16, 2019 by March Audio

[davewantsmoore]

33 minutes ago, MLXXX said:

The use of a separate speaker removes any suggestion that the loudspeaker reproducing the 8333Hz wave is being compromised or altered in its performance when the 3rd harmonic is added to the listening room.

The high and low frequency waves mix together .... and are then "low pass filtered" by the ear.

[davewantsmoore]

7 minutes ago, March Audio said:

Also music only has very low level HF content.

Indeed... but for purposes of analysis, you could begin with waves of equal amplitude....  as we are interested in the question "could this ever be audible" .....  as opposed to "would this really be significant in practise".

 

 

This is the reverse of previous "doppler" discussions .... where of course, it is possible that this distortion could be audible .... but in practise it is quite unlikely to be.

[davewantsmoore]

9 minutes ago, March Audio said:

Finally you also need to demonstrate that that either tones were creating no significant spurious tones at other frequencies which could lead to IMD. 

Well of course they were.... they mixed together in the air.

[MLXXX]

43 minutes ago, March Audio said:

Did you do this test blind?

Of course. How else could the Hydrogen Audio requirements for objective confirmation have been met?

 

As I said in the post of mine you have quoted:-

17 hours ago, MLXXX said:

The people at HydrogenAudio (HA) are sticklers for objective proof, so I set up an automated ABX procedure and then in accordance with HA requirements dutifully uploaded the successful ABX details. 

 

[MLXXX]

42 minutes ago, davewantsmoore said:

51 minutes ago, March Audio said:

Also music only has very low level HF content.

Indeed... but for purposes of analysis, you could begin with waves of equal amplitude....  as we are interested in the question "could this ever be audible" .....  as opposed to "would this really be significant in practise".

Quite so.  The test was to explore the question "could it ever be audible?", not the question "will it be audible or significant in practice?". 

 

As I noted at the end of my post describing the test procedure:

To be frank, the fact that an ultrasonic tone might be just detectable in an experimental setting such as I have just described is not very exciting. It isn't very exciting because the effect is so weak and subtle. It seems hardly worth worrying about.  There is a considerable doubt it would be detectable with music.

[Ittaku]

1 hour ago, MLXXX said:

Many people would be surprised, as conventional thinking is that it should make no difference to the perception of sound if frequencies above around 18kHz or 20kHz are not reproduced.

I'm surprised, since I did a similar experiment and came up with nothing. Perhaps my hearing had already started failing at the time...

Edited September 16, 2019 by Ittaku