Why 192kHz matters?

[davewantsmoore]

We don't have biological brickwall filters, per se.

 

Yes, I should add, another aspect I tried (and I have read others trying) was boosting the amplitude > 20khz to unrealistically loud levels.     Even adding in my own "fake" HF content.

 

 

 

NB - certainly I don't say my tests were 'conclusive' .... as they weren't robust enough ..... but it is interesting and fun to see what happens....   and TBH I seriously didn't want it to be true, so I had to try for myself.

[Guest rmpfyf]

People have obviously done these tests...  many times, over many decades.    It's a very attractive thing, to be able to increase audio quality by adding in content above 20khz.    So far the overwhelming result is that it is not audible.

 

I spent years trying it myself..... I was caught up in the lossy vs lossless, vs high res, vs ?, problem as a way to try and increase audio quality.    I so wanted it to be true, as it would be such a (relatively) easy way to get better sound (with exciting potential).

 

Long tests (weeks/months) .... or sitting there in the listening position with a button which toggles the HF in and out.    Can't hear it.

 

 

 

OTOH.   Changing the oversampling and filter settings in a DAC chip .... and feeding different rates and over sampled versions of the content to the DAC.    CAN BE audible (although not usually very)    What happens depends very strongly on the equipment.    When you look at how DACs work, this isn't super surprising or unexpected.

 

Don't disagree with any of this.

 

I'll leave a door open for a mega-optimised system being able to better linearise a waveform under 20kHz with greater accuracy. We're splitting hairs though, getting some way towards the 'PCM don't do square waves' argument...

[Newman]

"The threshold of time delay is discussed in isolation of the audible threshold of its amplitude. And it turns out that the amplitudes he uses are massively higher than anything that would occur in real life, including music. An analogy would be the bone conduction detection of ultrasonic frequencies, which proves we can detect ultrasonic frequencies but when looking at the amplitudes involved, trying to achieve that via airborne sound would deafen everyone in seconds."

There you go confusing me again.

How can you be deafened by an ultrasonic frequency you can't hear?

 

If you play an ultrasonic signal using a vibrator like a hammer drill pressed against the jawbone, the signal is 'detectable' ('heard' might be the wrong word). In that way humans can 'respond' to ultrasonic sound. (and this is useful knowledge because the profoundly deaf might be able to hear using translators that convert normal speech frequencies to higher frequencies and 'hammer drill' them into the bone, and the deaf may be able to 'hear' the spoken word. What is not useful is when audiophiles point to these experiments (and they are known to do so) as proof that we can hear ultrasonics and CD is inadequate. They are forgetting the threshold issue. Kunchur seems to have done the same.)

 

But if you play 'normal' music (that contains ultrasonic components) so loud that the jawbone threshold of detection is reached via airborne sound, the rest of the music will be so loud as to instantly deafen.

Edited January 23, 2016 by Newman

[Bilbo]

OK, now I understand your point.

[LHC]

If you play an ultrasonic signal using a vibrator like a hammer drill pressed against the jawbone, the signal is 'detectable' ('heard' might be the wrong word). In that way humans can 'respond' to ultrasonic sound. (and this is useful knowledge because the profoundly deaf might be able to hear using translators that convert normal speech frequencies to higher frequencies and 'hammer drill' them into the bone, and the deaf may be able to 'hear' the spoken word. What is not useful is when audiophiles point to these experiments (and they are known to do so) as proof that we can hear ultrasonics and CD is inadequate. They are forgetting the threshold issue. Kunchur seems to have done the same.)

 

But if you play 'normal' music (that contains ultrasonic components) so loud that the jawbone threshold of detection is reached via airborne sound, the rest of the music will be so loud as to instantly deafen.

 

I don't think the point of this thread or discussion is about 'hearing' ultrasonic. I already established in a much earlier post that we should all start from the same basis that we can't hear those frequencies beyond 21kHz, that hasn't change. No one is contesting that here.

 

What has changed is the following. If 21kHz is used as the bandwidth for CD recording and playback, then it implies the smallest time misalignment or differences that we could hear is around 20us (according that that Yamaha white paper). But the experiments of Kunchur have shown that human can discernment differences as low as 5-6us apart. This recent result is new, but it is not inconsistent with the threshold that you wrote about. 

 

Yamaha's white paper picture it this way.

 

Using the ISO226 loudness contour hearing thresholds, the 120dBSPL pain threshold, Kunchur’s 6 microsecond time coherence threshold and Sperling’s echoic memory limit of 20 seconds, we propose to define the level, frequency and time limits of the human auditory system to lie within the gray area in figure 410:

 

 

In that diagram the 6us limit is labelled on the time axis, it doesn't alter the other parts of the threshold diagram. There is no inconsistency.

 

The key question is the 'so what?'. How does the fact that human can hear timing differences so accurately influence the way we record and reproduce music? I accept that people will have different and opposing points of view. The engineers behind Yamaha, MQA technology, and other companies may view it as a legitimate requirement for using higher sampling rate and retaining the ultrasonic information; others may disagree and maintain faith in redbook CD. Its all good.

[Newman]

That post #228 of mine was not about Kunchur. I was trying to answer a question from Bilbo on a side point. 

 

My post on Kunchur was #220.

[Nada]

.....(Back to think a bit more about what else hires might offer over easier filters...)

 

After being a hires devotee my experience at this stage has found hires is good for

 

a. accessing good performances and recordings

      presumably  marketing logistics demand that hires acts as a quality filter as dud performance and lousy recordings would damage the product

 

b. from above it follows its a great marketing tool for the industry to once again get us to buy new music as past of a long rip off process that began I guess with 78 rpm records, then 45;s,  LPs, quad LPs, tape, cassette, CD, SACD, DVD-A, hi-res, MQA..... and they will keep coming

 

c. placebo effect - lets be clear that the placebo effect isnt imagination. Its a real experience. If playing 192/32 helps to really enjoy the music more that's just as effective as a hard gear upgrade

 

d. alleviating neurotic concerns that hires is free of contamination from down sampling with bad math eg

 

 

 

Edited January 23, 2016 by Nada

[Newman]

c. placebo effect - lets be clear that the placebo effect isnt imagination. Its a real experience. If playing 192/32 helps to really enjoy the music more that's just as effective as a hard gear upgrade

 

While I agree entirely with what you are saying, the bold bit could barely have been worded worse.

[Nada]

While I agree entirely with what you are saying, the bold bit could barely have been worded worse.

 

Why?

[Newman]

The imagination produces real experiences all the time. Placebo effect is one such.

[Guest rmpfyf]

What has changed is the following. If 21kHz is used as the bandwidth for CD recording and playback, then it implies the smallest time misalignment or differences that we could hear is around 20us (according that that Yamaha white paper). But the experiments of Kunchur have shown that human can discernment differences as low as 5-6us apart. This recent result is new, but it is not inconsistent with the threshold that you wrote about. 

 

LHC, a few points:

 

• CD is not and was not designed to be a recording format. It is â€‹explicitly designed for playback. 
• ​Kuncher's own experiments were completed within what's accepted to be psychoacoustically audible and replicable with Redbook audio.

 

I get that Kuncher has proven that 5us is an audible delay threshold. This is not a proof for hires audio.

 

Look at it another way - what's a realistic occurrence where a 5us delay is critical to replicating music accurately?

 

I can think of a few corner cases where hires has it over Redbook, which have nothing to do with replicating adjacent peaks at 5us delay in the same channel of data (which, BTW, is neither what Kuncher proved, nor an easy thing to get out of speakers and to ears accurately). What would you suggest for this?

[LHC]

 

LHC, that papers are published and peer reviewed does not make for their being absolutely correct. It just means the work was published.

 

Similarly there's good debate and criticism about his work.

 

It's published work, this doesn't make it an absolute. 

 

 

Agree that a published work doesn't mean it is absolutely correct. What it does mean is that it has been through a properly administrated scrutiny process and shouldn't be casually dismissed out of hand. The only legitimate way to refute or debunk a published work is through the same process, i.e. by peer-reviewed professional journals.

 

There is no doubt his work has attracted debate and criticism, he openly said that. That was the reason he wrote his FAQ to better explain his interpretation of the results and address some of the technical aspects.

 

I do note that Newman and Dave have presented their concerns with Kunchur's work, and I would be surprised if he hasn't heard of those before. But other's work have also attracted criticism. Take for example the often cited 2007 work of David Moran and Brad Meyer that seemingly show that people can't reliably tell the difference between redbook and hi res audio. Well that work also has its vocal critics like this http://www.realhd-audio.com/?p=1497 for example. I have also seem some criticisms of the recently published tests by Bob Stuart. 

 

There are no shortages of critics in this world, especially if the published results don't conform to their world view. But I take a more positive perspective, i.e. I think it shows how agonisingly difficult it must be to perform proper rigorous experiments on human hearing. 

Edited January 23, 2016 by LHC

[huxmut]

So HiRes may not matter, or it may matter.

But if it does matter, it doesn't matter a lot to those in a home listening environment ?

[zenelectro]

There is a basic misunderstanding between a/ HF bandwidth of Redbook and b/ L R channel timing accuracy.

Just because a/ is limited to 20kHz doesn't automatically imply that b/ is restricted to 20uS.

I believe the L R timing accuracy of Redbook is capable of much better than 20uS but other factors such as clocking and digital filters come into play.

Maybe this is one reason we can hear differences between different clocks that have jitter theoretically way below audibility in the low pS range.

cheers

Terry

[Guest rmpfyf]

@@LHC

 

Having both authored and reviewed published work, I can't agree with the 'absolute' view you take. Peer review is not a process to prove or disprove work, or to legitimise it as 'absolutely correct'. If learning and evolving from published experience had to involve additional peer-reviewed work, learning would be an inordinately long process - particularly when what's been published has blatant error, or is of limited relevance. That an audible misalignment criteria is 5us, that speaker placement to mm levels are important, that smaller fullrange drivers allow better spatial placement... are all known, the first is best proven by Kunchur's experiments. That conventional playback means can't capture near-ultrasonic harmonics of higher-frequency square waves is known... and of little relevance. His claim that this all means hires is necessary isn't substantiated by his proofs - because, and I can't stress this enough - what he tested that pertains to music can be replicated in Redbook just fine. Which has been explained my many professionals in the audio community.

 

His FAQ does delve - correctly - into what's tricky about reproducing Redbook accurately, which are valid points many have mentioned throughout this thread (filtering/quantisation error, jitter etc). These concerns are relaxed at higher frequencies, and make better playback an easier thing on many systems. 

 

The debate about hires vs Redbook will go on forever. As it should. Not least because we cannot blind test music played back in both mediums without introducing other factors, and because some of those factors are themselves limitations. Example? On my hires system I'll play the latest master of Jeff Buckley's 'Hallelujah'. I'll play it resampled Redbook, then at the native. People pick the difference. Then I'll play it on my Redbook system. They think it's even 'higher' res (the Redbook system is a very tuned). How does the hires sound on the Redbook system? Dunno, the DAC doesn't support it. Could you compare directly if you could? Well, no, they're from different masters, different equipment, different studios, different sound engineers... etc.

 

Don't get me wrong, this isn't a critique at you. I own some hires content and I do like it. It gets the most out of one of my rigs and for that particular music, I like the idea of owning a copy of the master. Critiquing that would be like telling vinyl people they're silly for investing in an inferior format... you could argue that it maybe is an inferior format, but some people just like the way it sounds, love the challenge of making it sound better and love the very deliberate sense of occasion it comes with. Whatever floats anyone's boat. 

 

As @@zenelectro correctly states, Redbook channel timing is way, way under 20us. Orders of magnitude under 5us. It's at the heart of what makes a stereo recording seem live. It is possible, with a compromised system, to have Redbook (or hires) sounding with a 'flat' soundstage. You're literally hearing time smearing here, the very filtering issues (usually in phase not amplitude) many have written about. Mind you, getting a 5us average difference out of two speakers and to your ears accurately is a significant challenge, just think about what's involved here (assuming it came faithfully out of a pair of speakers) - again, much like splitting hairs on square waves, we're playing at the edge of relevance with concerns like this. 

 

Can you do that with a single channel (testing for masking now, which is Kunchur's assertion to hires)? No, and that's not relevant from two perspectives:

 

• It doesn't provide realism any more than being able to see 3D with one eye open
• Adjacent peaks at 5us in the same channel constitute spectral content at ultrasonic, and you can't hear it anyway

 

The second point is the more contentious and for me forms the basis of hires vs not. A great example is a cymbal - nice high-energy content well beyond 20kHz. Redbook will not recreate this. It will approximate what can be heard using spectral (psychoacoustic) methods. It will use a limited-length sinc function to approximate what's recorded, where that sinc function is limited by Redbook considerations. So:

 

• Do we get adjacent peaks in the same signal 5us apart? No, and that's probably not important - we can't hear that.
• Do we get the shape of the waveform recorded as accurately as possible? Also no, because it takes higher-frequency spectral content to shape it appropriately, and that's beyond Redbook.
• Can hires just play back the damn cymbal as it was and let our ears and brains decide what we're actually hearing? Yes... assuming we've a system that you can stick a reference microphone in front of to actually validate that you've a system that can replicate what's played back accurately, otherwise we're back with the 'but everything's a low-pass filter' argument. Whilst that last part is an important caveat, it's relevant.

 

Which as ever brings us back to 'what do you listen to'. People with mega-midrange-beast fullrange drivers are unlikely to hear the difference, or want to. Full orchestral, some electronica (you can fill in here) that has relevant content... sure. There's a possibility. 

 

All of which is still a very long way from the corner-case argument Kunchur presents 

[davewantsmoore]

Take for example the often cited 2007 work of David Moran and Brad Meyer

 

A difference between the Kunchur text and Moran and Meyer is that ....

 

Kunchur presents some results which say X  (what they found people could hear) .... and then went on to draw a conclusion which those results didn't support.

 

Note that the reciprocal of 6 microseconds is 166kHz - indicating that an audio system should be able to process this frequency to satisfy this timing perception

 

He presents no justification or explanation for this conclusion....     and, when you examine it, it's incorrect.    You don't need to be able to "process this frequency" in order to facilitate a timing resolution of < 6us   (as it related to their experiment of the audibility of misalignment of two sources).

 

 

 

OTOH.   Meyer and Moran present results which showed people couldn't hear the difference between something .... and concluded that people couldn't hear the difference between something.

 

Of course, Meyer and Moran may have conducted a 'flawed' experiment....   and regardless of any flaws or not, their results may not be universal 'proof' of anything....   but they've only made conclusions in-line with what their results showed.

 

M&M shouldn't be used as "the end of the story".   It shouldn't be called "proof".    Their results may give us an idea of what to expect, and people should be encouraged to contradict their results if they can.

[davewantsmoore]

getting some way towards the 'PCM don't do square waves' argument...

 

No system can produce a (literally) square wave.

[davewantsmoore]

What has changed is the following. If 21kHz is used as the bandwidth for CD recording and playback, then it implies the smallest time misalignment or differences that we could hear is around 20us (according that that Yamaha white paper). But the experiments of Kunchur have shown that human can discernment differences as low as 5-6us apart. This recent result is new, but it is not inconsistent with the threshold that you wrote about.

 

They are incorrectly conflating the signal bandwidth with 'time resolution'.

 

PCM can time the signal (within the system bandwidth, which is set by the sampling rate), to essentially infinitely small 'resolution'.

Edited January 24, 2016 by davewantsmoore

[davewantsmoore]

There is a basic misunderstanding between a/ HF bandwidth of Redbook and b/ L R channel timing accuracy.

Just because a/ is limited to 20kHz doesn't automatically imply that b/ is restricted to 20uS.

[Nada]

The imagination produces real experiences all the time. Placebo effect is one such.

 

 

Your explanation of the placebo effect is dated and misleading. I suggest you look at the current literature. Its fascinating. I wont side track this thread going into the science but suggest you start a new thread if you need some help with it.

[davewantsmoore]

Your explanation of the placebo effect is dated and misleading

 

Even just 10 words to summarise how would be very helpful for people.   TIA

[Newman]

Or maybe you still misunderstand. I am not in dispute with you about the meaning of placebo effect, and I said as much.

 

My comment was about your writing with the presupposition that 'imagination' and 'real experience' are mutually exclusive domains. In truth, there is a big overlap.

Edited January 24, 2016 by Newman

[Guest rmpfyf]

He presents no justification or explanation for this conclusion....     and, when you examine it, it's incorrect.    You don't need to be able to "process this frequency" in order to facilitate a timing resolution of < 6us   (as it related to their experiment of the audibility of misalignment of two sources).

 

+1 (as said previously)

[Guest rmpfyf]

No system can produce a (literally) square wave.

 

As explained in post #53.

 

No system can... and (more importantly) no system should be able to, either.

[LHC]

A difference between the Kunchur text and Moran and Meyer is that ....

 

Kunchur presents some results which say X  (what they found people could hear) .... and then went on to draw a conclusion which those results didn't support.

 

 

He presents no justification or explanation for this conclusion....     and, when you examine it, it's incorrect.    You don't need to be able to "process this frequency" in order to facilitate a timing resolution of < 6us   (as it related to their experiment of the audibility of misalignment of two sources).

 

 

 

OTOH.   Meyer and Moran present results which showed people couldn't hear the difference between something .... and concluded that people couldn't hear the difference between something.

 

Of course, Meyer and Moran may have conducted a 'flawed' experiment....   and regardless of any flaws or not, their results may not be universal 'proof' of anything....   but they've only made conclusions in-line with what their results showed.

 

M&M shouldn't be used as "the end of the story".   It shouldn't be called "proof".    Their results may give us an idea of what to expect, and people should be encouraged to contradict their results if they can.

 

So what if the 'mistakes made by' Meyer and Moran, and by Kunchur, are of a different type? It is completely irrelevant to the point of discussion. (I am 'just saying' here, not asserting that anyone is mistaken)

 

What is relevant is that both are properly published work and should be accepted as the state of our knowledge to date. Until someone published a full rebuttal of their work then it remains that way. In the meantime, people like Roger Sanders could freely reference Meyer and Moran's paper to support his argument and not be criticised. Now, Mark Waldrep have written some good arguments against the Meyer and Moran experiment, but I believe they only exists as blog commentary and are not yet published. So it doesn't formally change anything.

 

I do 110% agree with you that those results (mistaken or not) should only be treated as the start of the conversation, and not be an end to the line of inquiry. So we could use Meyer and Moran's results to heighten our awareness that any differences between redbook and hi-res could be very subtle for many people, that is fine. What we shouldn't do is to use their result to tease others who claim to be able to hear such differences.

 

Certainly more rigorous research work in the same vine as Meyer, Moran and Kunchur is desperately needed. I noted that Waldrep wrote about his proposal entitled â€œA PROPOSAL TO INVESTIGATE HIGH-DEFINITION AND SURROUND AUDIO WITH REGARDS TO PERCEPTABILITY, PREFERENCE AND MARKETABILITY† to the CEA board in 2013. He wrote "The major thrust of the paper is to do a rigorous investigation into the anecdotal and unproven claims made by advocates of high-definition audio, which includes me.". I wonder whether that ever took off. 

Edited January 24, 2016 by LHC