The Skeptic Magazine article on Audiophilia

[jj]

Now, don't get me wrong, the article was right about the value of the extreme-audiophilic stuff, indeed.

I wonder, though, about the approach.  Perhaps I'm used to explaining things at a level that's far, far too high, but it seems to me that starting by explaining that "intensity" or "sound pressure" is what one measures in the atmosphere, and that "loudness" is the term for what people actually "hear" in their heads, thereby pointing out that even from the word go, it's necessary to consider the effects of human perception. From there, one can step to intermodal (intersense) effects (including knowledge and such of what is being heard), expectation effects, attentional focusing (the psychogists have a longer term for this), and the like, to explain why if you presume something MIGHT be different, and the test isn't blind, the chance is close to 1 that you WILL recall a difference, even when no such difference exists.

It seems to me that this well-established knowledge is a good place to launch any exploration of effects of, on, or to the auditory system.  It shows graphically the need for blind tests or cognates, and why tests that can not be falsified have no value at all (overdetection, false detection, etc).

From there, we can point out the fact that some things ("directional" couplers that only couple "energy" in "one direction" for instance) would also require a rewrite of physics that would make the present universe that we observe impossible, etc.

On to conservation of energy, information, etc, showing what the "Shannon Bound" is in laymans terms, and why an LP can not possibly contain "more information" than a 'CD'.  Etc, etc.

Those of you who frequent JREF have seen a set of links that various people, myself included, have listed on various audio hooraw that just can't be what it says.  Randi, himself, hit on perhaps the lamest to date, the "Intellegent Chip" (sp?), which is matched in its absurdity only by the "GigaHertz Tweeter" also sold by the same folks.   The "magic ring" could actually do something if it's a big ferrite bead...  Nothing useful, perhaps, but it could actually DO something ...

[Capthorne]

The matter of frequency range is still debatable. We can actually detect, and respond to, sounds that are over 20KHz even though they may be outside the range of detection by the ear. Similarly with low frequencies: any pipe organ, such as in churches, can be felt rather than heard and contributes to the music.
There is a company (Advanced Brain Technologies) that produces CDs which can be used to re-train the ear-system in certain cases of hearing loss. They require very high quality players to work properly.

[jj]

The matter of frequency range is still debatable.

Really?  Please elucidate!  I'm not aware of any serious conflicts with the actual understanding of the state of the art (as opposed to a lay interpretation or lay confusions).

We can actually detect, and respond to, sounds that are over 20KHz even though they may be outside the range of detection by the ear.

Very few people who I know dispute that children can hear above 20kHz. Nobody I deal with in the AES appears to be unaware that one can feel sonar pulses on the skin.

The intensity required for skin detection is very high.  In general, it's not particularly good for you to be exposed to such levels. Certainly, percussionists will have this kind of exposure, this is one reason that percussionists are advised to wear some serious ear protection in order to prolong their professional career.

For children, well, the question of when the ability to hear above 20kHz ceases is a hard question, it appears to be the case that it this often happens before the child is able to perform well as a listening-test subject.   There are many reports of adults who can hear above 20kHz. Some of these cases may be confirmed, many have been shown to demonstrate that the listening equipment in question created audible sub-20khz components via various nonlinear mechanisms.  What is clear is that the ability to hear much above 20kHz is very rare, and that such sensitivity only exists at very high levels in adults.

Similarly with low frequencies: any pipe organ, such as in churches, can be felt rather than heard and contributes to the music.

Why is this germane? What argument follows?  Present digital systems can go as close to DC as you want, although there are few loudspeakers that do well below 30Hz.  Most people don't want the ability to put DC through their loudspeakers, even if they don't know that.

There is a company (Advanced Brain Technologies) that produces CDs which can be used to re-train the ear-system in certain cases of hearing loss. They require very high quality players to work properly.

Really? Could you please post some information on this technology?  I can imagine some retraining in the case of mid-frequency loss, but that would not even concievably require a "high quality player".

Can you, perhaps, even explain what you mean by a "high-quality player"?  Most non-portable CD players nowdays have about 15+ bits manifested properly in their output, although some certainly have some analog electronic problems.

Certainly, there will be no meaningful signal at all above 22.05 kHz from any CD player, no matter what the quality, under any circumstances, and least any such signal that will have a substantial output below 22.05 kHz.  Furthermore, any signal arising from a CD player above 22.05 kHz constitutes clear evidence of maldesign or malfunction, no iff's, ands or buts.

There is a direct, but somewhat complex, relationship between the required (i.e. "not optional") anti-imaging filter in the CD player and time resolution of the reproduced signal.    This can be seen if you graph a sine wave at close to the limit, and then start adding its images in proper phase and amplitude.

[Capthorne]

jj:

AdvancedBrain.com

They have extended the methods that Dr. Alfred Tomatis developed into the ranges possible with modern technologies. He only had 78rpm equipment available.
This is mainstream; not some airy-fairy new-age waffle.
You might find it interesting; if so a Google on Tomatis would be a good place to start.

[jj]

jj:

AdvancedBrain.com

They have extended the methods that Dr. Alfred Tomatis developed into the ranges possible with modern technologies. He only had 78rpm equipment available.
This is mainstream; not some airy-fairy new-age waffle.
You might find it interesting; if so a Google on Tomatis would be a good place to start.

Why does it require a "high quality" CD player?

What was the significant points in your discussion of auditory and sensory bandwidth?

What, in short, are you arguing?

[Capthorne]

JJ:

Why does it require a "high quality" CD player?

What was the significant points in your discussion of auditory and sensory bandwidth?

What, in short, are you arguing?

First to avoid skipping and to deliver the maximum possible frequency range, second - you should look at what ABT is doing; there's too much to go into here. Three: I'm not argueing anything; just making the point that there are circumstances where an unusually wide frequency range is useful. I thought people might be interested.

[jj]

First to avoid skipping and to deliver the maximum possible frequency range,

My kid's portables don't skip.  As to maximum frequency range, since just about every player I've seen gets out to 18K, and most to 20K, what more is there to be had?  I haven't seen any players (certainly some may exist) that actually create some unnecessary frequency limiting.  In fact, most players use the same several sets of commercial output devices, all of which have remarkably similar characteristics.

If I can find my presentation on digital to analog conversion on the net (I've given it all over the country, I've given the .ppt's to a bunch of people to put up, but some of them are inside networks...) I'll put in a pointer.

I've looked at that ABT thing you pointed out. I haven't yet discovered a good, controlled study (i.e. a comparison to other listener training methods). If there's one there, I haven't stumbled on it.  Have you some guidance here, you're the one advocating the method.

[Capthorne]

I wasn't so much advocating it, 'though I have used it occasionally with excellent results. I claim no expertise on audio equipment. I merely threw in a comment about frequency ranges that I hoped might contribute to the general discussion.

On a different tack; in the early days of what were then known as "HiFi" systems (in the UK) a retail chain that I was with sold bell wire for use with loudspeakers. It worked fine with my bass-reflex speakers made from 1" particle board and 12" drivers and the ones a colleague made out of concrete.

[jj]

On a different tack; in the early days of what were then known as "HiFi" systems (in the UK) a retail chain that I was with sold bell wire for use with loudspeakers. It worked fine with my bass-reflex speakers made from 1" particle board and 12" drivers and the ones a colleague made out of concrete.

Well, I know of several very good listening rooms. I've designed and built two (one from scratch, with some help from collegues in the acoustic isolation design) myself.  One of them may (although it's mostly abandoned at a mostly-closed research lab ) may be the quietest listening room that is commercially available, it's less than 3dB noiser than the necessary molecular noise of the atmosphere at all but the very lowest (sub-audio) frequencies.  We're not sure exactly how quiet, the spectrum of the remaining noise could have been electrical noise. The measure was between 8dBc and 9dBc with nobody in the room.  It's quiet enough that you can hear people breathing at the far end, that a mouse click sounds loud, and that a cheap monitor screams like a siren.

That room was used for testing some very good audio codecs (as well as other things) and demonstrated sensitivity as high as any other place on the planet. The speaker wire was basically a low-inductance version of twisted pair zip, the other cables standard Canare' mic cable. Nothing fancy at all. No wire in the room (except perhaps the power mains feed, which was about OO) went for over $.075 a foot.

All of the rooms that I know of use something very akin to well-insulated zip cord for speaker cable, too.

Now, they use #12 or so, not #22 lamp cord, but it is 2-conductor stranded wire with no particularly special properties. There is no solid unobtanium,  $1000.00 per foot cable in use in these rooms.

[Capthorne]

I'm impressed, especially with hte bits I don't understand.
But, hey, what about concrete speakers in a small apartment livingroom, cast in place and complete with slotted shelf and lined with bits of those egg-carton-contoured foam pieces from shipping crates. When he moved, he had to pay someone to break them up. Martin something his name was.

[jj]

I'm impressed, especially with hte bits I don't understand.
But, hey, what about concrete speakers in a small apartment livingroom, cast in place and complete with slotted shelf and lined with bits of those egg-carton-contoured foam pieces from shipping crates. When he moved, he had to pay someone to break them up. Martin something his name was.

Sounds like the devil of a mess to me.

[marty]

On to conservation of energy, information, etc, showing what the "Shannon Bound" is in laymans terms, and why an LP can not possibly contain "more information" than a 'CD'.  Etc, etc.

I don't know exactly what you mean  here, but if you mean that a CD by design must have at least as much info as an LP, then I disagree.

CD's were designed by Philps (and Sony) engineers who are rumored to be less than audiophiles.  If there could be no improvement on the CD then the whole idea of the two "better" CD formats of SuperAudio and DVD Audio would be useless.

The best if LP's certainly contain more info than the worst LP's. Likewise I am sure we could agree that the best CD's contain more info than the worst CD's.

Am I misunderstanding your "more information" terminology or just missing your point?

P.S.  Still using zip cord here, too!

[jj]

I don't know exactly what you mean  here, but if you mean that a CD by design must have at least as much info as an LP, then I disagree.

"Information" is a formally defined, measurable quantity.  A CD has exacxtly 44100*2*16 bits/second of information rate, no more no less.  That is immutable for a redbook CD.

An LP has about the same bandwidth (yes, there can be higher frequency content) but the SNR (local SNR not dynamic range) is much lower than CD, especially above 3-5khz where distortion is the primary source of non-signal content, and above 5-7k where in addition to that, the channel independence goes to pot.

These calculations have been made over and over again, and CD has more information (and rate) in bits (bits/second) than an LP.  This is not a question of some interpretation, because the information in a signal is equal to log2(snr), and you can integrate over frequency (snr changing with frequency) to get bits/second., or then times time to get bits.

I've done demonstration calculations a few zillion times, they live back on netnews (USENET) in the various audio newsgroups under my old net name of jj@research.att.com, a no-longer valid address.  Dunno if you could find these or not.

CD's were designed by Philps (and Sony) engineers who are rumored to be less than audiophiles.  If there could be no improvement on the CD then the whole idea of the two "better" CD formats of SuperAudio and DVD Audio would be useless.

Who says that they mostly aren't? There is some cause to have a somewhat higher sampling rate, but more than that requires evidence that I am not aware of existing.

[marty]

I don't know exactly what you mean  here, but if you mean that a CD by design must have at least as much info as an LP, then I disagree.

"Information" is a formally defined, measurable quantity.  A CD has exacxtly 44100*2*16 bits/second of information rate, no more no less.  That is immutable for a redbook CD.

That's your definition of information, but not mine.

Take the following hastily graphed waves:

It would be nice if this image was working!?!


All three could be 20K/sec waves.  By your definition of information, if a CD can do any of these three (but not the other two or not a fourth more complex wave at 20K/sec) then it has just as much information as a medium that can do a, b, c and even more complex forms.

This is the basic knock on CDs and even if I haven't written the waves out in the precise way that CD critics have, I am sure that you must have at least paid some attention to this idea.

If I need to look for the exact problem with the waveform for you I guess I will, but really your definition of information is simplistic, isn't it?

[jj]

That's your definition of information, but not mine.

Then yours is simply wrong. "Information theory" is a rather large, complex branch of mathematics, and "information" is defined precisely. Look up the definition of "entropy" and its relationship to "information" in a good information theory introductory book.

Sorry.

If we could do formula here, I'd explain more, but a basic book on information theory, entropy, etc, would help you out here.  There's much more than can be communicated in a half-hour.

Take the following hastily graphed waves:

It would be nice if this image was working!?!


All three could be 20K/sec waves.

No, again you're mistaken. Assuming the rounded one is intended to be a sine wave, that one would have only a single component and that below 20khz.

The other two, assuming you mean square and triangular wave, have one component under 20khz, and the rest above.  Specifically the square wave would have  1/n * sin(n*2*pi*20000*t) components, and the amplitudes of the triangle would bre 1/(n^2).   N.B. there are phase components too, I'm leaving them out, the argument showing the frequency is enough, here.  (Note, for only odd 'n' for the symmetric cases assumed here, but not for the general cases.)

Within the bandwidth of 22.05 kHz, you'll only see one component and that will be the 20khz sine wave.

You're making the basic mistake of assuming that a periodic waveform at 20kHz has components only at 20khz, and that's dead wrong.  Its lowest component will be at 20khz.

CD's only reproduce up to 22.05 kHz, so in a CD all three of those waveforms will come out approximately the same.

Given a linear reproduction system, they will sound very close to the same, too.  As discussed above, the evidence for normal-level perceptions of supra-20khz sound is very limited, to say the least.

As I've said, I would prefer a slightly higher sampling frequency, and a slower filter cutoff, but for reasons not discussed here, yet, at least.

By your definition of information,

It's not my definition, it's the accepted one that I, among many other people, use.

if a CD can do any of these three (but not the other two or not a fourth more complex wave at 20K/sec) then it has just as much information as a medium that can do a, b, c and even more complex forms.

I have no idea what you mean.  Your statement relates neither to information nor to how the sampling theorem works, and I can't make head or tail of it. In short, I think it's mistaken, but I can't be sure.

This is the basic knock on CDs and even if I haven't written the waves out in the precise way that CD critics have, I am sure that you must have at least paid some attention to this idea.

Any one who argues that this is a mistake does not understand what the spectrum of a square or triangle wave is. It's that simple.

If I need to look for the exact problem with the waveform for you I guess I will, but really your definition of information is simplistic, isn't it?

There is no problem at all. None. You simply misunderstand what 20khz means.

I would, were I you, get Morrison's book "Fourier Analysis" or an equivalent text, and read it for starters, to see where you're going wrong with your "20khz" statement, then pick up a book on information theory to see what 'information" (i.e. entropy) is.  You obviously aren't willing to hear it from me, given your dismissal of an entire branch of mathematics that makes MP3, AAC, WMA, JPEG, H.261, ... work. (there are other mathematics involved as well, its not that simple)

Your meaning of information makes no sense to me, and includes basic mistakes in Fourier Analysis and sampling theory.  I'm sorry, but that won't convince anyone.

[marty]

Thanks for the response and I'll try to get the sources that I have read in the past who do a better job explaining what I mean.

I don't understand the Fourier stuff, although I was exposed to it over 30 years ago.

Your statement:

"CD's only reproduce up to 22.05 kHz, so in a CD all three of those waveforms will come out approximately the same."

seems to be exactly my point.  Is approximately the same equivalent to exactly the same?  If not then why isn't this germaine to the argument.

I tried to make this point and nothing more or less.

[jj]

"CD's only reproduce up to 22.05 kHz, so in a CD all three of those waveforms will come out approximately the same."

seems to be exactly my point.  Is approximately the same equivalent to exactly the same?  If not then why isn't this germaine to the argument.

All of the information below 22.05khz is captured. Period.  The information that you think is missing is at (for your example) 60khz and above.

[marty]

Can you explain in words what you mean by the 60K statement?  It's not that I don't believe you, but I sure don't understand.

[jj]

Can you explain in words what you mean by the 60K statement?  It's not that I don't believe you, but I sure don't understand.

The difference between a square wave and a sine wave is the harmonics.

The part at 20kHz is a pure sine wave.  (well, sine, cosine, some phase angle)

Since a square wave is symmetric, it has only odd harmonics. The next harmonic is at 60khz. If you add together all the harmonics, you get the square wave. If you just take the fundamental, you have a sine wave.

The other stuff that turns the 20kHz sine wave into a square wave starts to have frequency components at 60kHz.  That' well above what you can hear, or sense at any reasonable level. (Not talking full-power attack sonar here.)

[marty]

Perhaps I'm used to explaining things at a level that's far, far too high,.)

Finally got you down to my level.

Thanks.

So in essence you are saying that if we put a 20 kHz square wave and a sine wave through a perfect analog to digital converter, then the output would be the same.

I can believe you but I now wonder what happens when we put a real world analog to digital converter to work.

i.e. Will the square wave cause some distortion that will screw up the digital signal to the point that the resulting "music" doesn't sound the same as the sine wave and or sounds like distortion or noise.

[Electric Monk]

So in essence you are saying that if we put a 20 kHz square wave and a sine wave through a perfect analog to digital converter, then the output would be the same.

I think that what he is saying is that if you put a 20kHz sine wave, triangular wave or square wave through a filter that had a cut-off at about 22kHz, then the output would be the same, namely a 20kHz sine wave. This is because the square and triangular waves can be thought of as the sum a 20kHz sine wave plus of a bunch of different sine waves of even higher frequencies. (This is the Fourier analysis.)

His point would be that your ear is such a filter. It completely ignores anything above 22-ish kHz, and so there cannot be any percieved difference between the three waveforms at 20kHz. Now, if you had the same three waveforms at 4kHz, you would certainly hear the difference between them. But at these lower frequencies, the CD player would be able to reproduce the harmonics just fine all the way up to 22kHz, and you would get the square and triangular waves as output.

--James

[jj]

I can believe you but I now wonder what happens when we put a real world analog to digital converter to work.

i.e. Will the square wave cause some distortion that will screw up the digital signal to the point that the resulting "music" doesn't sound the same as the sine wave and or sounds like distortion or noise.

Real-world convertors have filters that remove, roughly speaking, everything above 20kHz.

So, in fact, they come out the same if they go through a competent convertor. I'm sure somebody has done one wrong, but most all of them these days are the one of 3 or 4 chips that are effectively the same except for part number.

Edited to add:

For a discussion of the "beating" effect that some high-end people are hung up on (that is a non-issue, but it's the same problem you had, which was not realizing the results of out-of-band signals) see
http://www.audioasylum.com/audio/genera ... 43134.html

I wonder if I should post that one in the links thread?

[Ethan Winer]

Folks,

Wow, I didn't even know this forum existed, let alone had a thread discussing my article, until someone sent me an email this morning.

> it's necessary to consider the effects of human perception. <

I agree completely. In fact, it seems to me that the frailty of human perception alone is why so many audiophiles believe that something has changed when it is impossible for a meaningful difference to exist. Such as oxygen free copper speaker cables sounding "better" than regular Zip cord of a similar thickness. Can you measure a difference? Sure, at 1 MHz. But can anyone actually hear a difference? Not likely.

> It shows graphically the need for blind tests <

Indeed, yet some audiophiles claim they can hear the improvement of their replacement power cords, and insist they don't need a double blind test to prove what they're convinced they can hear. This is the real problem. As is "My mind is made up, don't confuse me with the facts."

> Since a square wave is symmetric, it has only odd harmonics. The next harmonic is at 60khz. <

Exactly. And even if some people really could detect the presence of ultrasonic content, which I doubt, the more important question is if that content adds appreciably to the enjoyment of music.

While frequency response is important, the two most damaging artifacts are IM distortion and resonance. IM distortion (and likewise digital aliasing) contains frequencies that are not related musically to the source, and so add a dissonant quality even in relatively small amounts. Resonance is also detrimental in small amounts, because it causes certain frequencies to sustain audibly for longer than others. But even then, by "audible amounts" we're talking 40 dB or less below the music. This is equivalent to 1 percent, which is pretty high by today's standards. So when, for example, digital jitter is spec'd at 120 dB below the music, there simply is no way anyone could possibly hear that. Heck, that's 30 dB below the noise floor of a CD!

--Ethan

[jj]

I agree completely. In fact, it seems to me that the frailty of human perception alone is why so many audiophiles believe that something has changed when it is impossible for a meaningful difference to exist. Such as oxygen free copper speaker cables sounding "better" than regular Zip cord of a similar thickness.

Absolutely.  If you read down the thread, the problem of overdetection of events is very common in auditory perception. That's all it takes to "hear" something.

Can you measure a difference? Sure, at 1 MHz. But can anyone actually hear a difference? Not likely.

I disagree with your reasoning but your conclusion about not making a difference is spot on.

Oxygen Free Copper is what everybody uses nowdays. The only real reason for it is flexibility. It was invented for using for wire-wrap, because oxygen embrittles copper.  Yes, there is a teensy resistance change, one that is irrelevant to human perception, and the performance at 1MHz is not significantly affected any more or less than anything else is. So, no, you can mesure a difference at DC, and the same proportional difference at RF, but it is a difference that just doesn't matter.

Indeed, yet some audiophiles claim they can hear the improvement of their replacement power cords, and insist they don't need a double blind test to prove what they're convinced they can hear. This is the real problem. As is "My mind is made up, don't confuse me with the facts."

Indeed, to the extent that some boards, newsgroups, etc, ban the mention of double-blind tests. This is mildly excusable in some cases, there are some DBT proponents who are pretty obnoxious, and are used as an excuse.  The real reason, though, is "controversy", i.e. people who point to DBT's are regarded as somehow "lesser" because they "don't trust their ears'. It's simply a way of keeping one's head firmly stuck in the sand, I think.

Exactly. And even if some people really could detect the presence of ultrasonic content, which I doubt, the more important question is if that content adds appreciably to the enjoyment of music.

Check out the air attenuation (in excess of 1/r^2) at 50kHz in normal air at room temperature and humidity.  That creates some uncertainty as to what people could hear if they were high-frequency microphones...

Now, many devices, when stuffed with lots of supersonic frequencies, distorts like a (*&(*, and people can hear that difference.  It's just that they are hearing in-band (i.e. under 20kHz) distortion products, at least in reproduced cases that I know about. All this shows is that transducers, in particular, are not very good.

While frequency response is important, the two most damaging artifacts are IM distortion and resonance. IM distortion (and likewise digital aliasing) contains frequencies that are not related musically to the source, and so add a dissonant quality even in relatively small amounts. Resonance is also detrimental in small amounts, because it causes certain frequencies to sustain audibly for longer than others.

Resonances affect frequency and phase response, they are part of a complete frequency response analysis.  IM distortion (all nonlinear distortions) are a problem , especially if they are higher-order distortions, because they create components far removed from the original frequencies present in the signal.

But even then, by "audible amounts" we're talking 40 dB or less below the music. This is equivalent to 1 percent, which is pretty high by today's standards. So when, for example, digital jitter is spec'd at 120 dB below the music, there simply is no way anyone could possibly hear that. Heck, that's 30 dB below the noise floor of a CD!

--Ethan

Careful, there. Digital jitter is quite counterintuitive. The frequency spectrum of the JITTER (not the clock, but the error in the clock) aliases down into baseband. If the jitter is wideband it is both more perceptable and causes bigger errors.  Very low-frequency jitter is, on the other hand, several orders of magnitude less perceptable.

Now, spec'ing jitter is usually not donw "below the music", since the errors resulting from jitter grow with as f^2 in energy, where f is the frequency of the signal being distorted. It is very easy to oversimplify the effects of jitter, and there are some players and DAC's that have been caught inserting quite audible jitter.

I'll agree that it should NOT be audible, and that such problems represent a substantial design error.

Power cords are a great example of, well, curious marketing practices. Arguing about the bandwidth of a cord that transmits 60Hz is fairly odd, I agree.  Of course, it is possible, I suppose, for a power cord to deal with noise that isn't eliminated in the power supply. That says more about the power supply to me, though.

[Paulhoff]

Oxygen Free Copper is what everybody uses nowdays. The only real reason for it is flexibility. It was invented for using for wire-wrap, because oxygen embrittles copper.  Yes, there is a teensy resistance change, one that is irrelevant to human perception, and the performance at 1MHz is not significantly affected any more or less than anything else is. So, no, you can mesure a difference at DC, and the same proportional difference at RF, but it is a difference that just doesn't matter.
[quote]

Yes, as you say, everyone uses oxygen free copper. But the real point is, the so-called audiophile guys say that only the expansive speaker wires does and not the cheap zip wire.

About the power cables. A friend says that hears a difference when they are not used, and the reason that he has been handed is that the equipment power supply is just gives the bare minimum filtering. I do not buy into this, but unfortunately too many audiophile guys do.

This is a short reply because I am going out to breakfast.

[Ethan Winer]

JJ,

> I disagree with your reasoning but your conclusion about not making a difference is spot on. <

By "1 MHz" I didn't mean oxygen-free copper specifically, but rather the various expensive cables that tout extended bandwidth as a feature, such as those made using Litz wire or fancy insulation materials.

> some boards, newsgroups, etc, ban the mention of double-blind tests. <

As if censorship is a solution!

> Resonances affect frequency and phase response, they are part of a complete frequency response analysis. <

Of course, but it's the "time-extension" component of the ringing that's most objectionable, followed by the amount of boost at that frequency. I'm not convinced phase shift is particularly important, other than when it's shifted for only one channel of a stereo system.

> IM distortion (all nonlinear distortions) are a problem , especially if they are higher-order distortions, because they create components far removed from the original frequencies present in the signal. <

You bet. As the artifacts become farther away in frequency they are more damaging because they are masked less. But in most cases harmonics fall off linearly with frequency.

> Careful, there. Digital jitter is quite counterintuitive. The frequency spectrum of the JITTER (not the clock, but the error in the clock) aliases down into baseband. <

But if it's 120 dB below the music, how could it possibly be audible, no matter what frequency it is?

> spec'ing jitter is usually not donw "below the music" <

All of the specs and data I've seen - for example those in Ken Pohlmann's book - show it as some number of dB below the program. With 110 to 120 dB or even more being typical. Is this not correct?

--Ethan

[Ethan Winer]

Paul,

> About the power cables. <

I just came back from Stereophile magazine's big Home Entertainment show in NYC. At the Dolby booth they had one of those huge power distribution thingies, and all of the outlets used enormous "garden hose" power cables with gigantic mil-spec looking connectors to feed the power amps and CD players etc. But the power center's own power cord was a standard, run of the mill IEC cable!

I also noticed that all of the loudspeakers were connected using surprisingly thin wire. It looked like the 22 gauge stuff they use to make microphone cables.

--Ethan

[poohbear]

Audiophiles haven't changed much from a number of years ago when digital was first starting out, and they had to have separate CD readers and D-A converters. I remember reading in Audiophile Magazine about tests of three or four different cables at $300 to $600 each, to connect the CD reader with the D-A converter.

The reviewers could apparently detect subtle differences in the quality of sound from these cables. My feeling was then, and is now, that these cables are carrying a digital signal. Either they carry it accurately, or they don't. If they don't, they're garbage, and if they do, they should all put out the same signal at the D-A converter. I could probably get a $6 pair from Radio Shack that would do the same thing.

[jj]

Of course, but it's the "time-extension" component of the ringing that's most objectionable, followed by the amount of boost at that frequency. I'm not convinced phase shift is particularly important, other than when it's shifted for only one channel of a stereo system.

Again, I have to disagree.  The time extension is a problem if it's severe, but the precidence effect tends to somewhat mitigate such problems for a few dozen milliseconds, at least.

Now, phase response is very important, but it depends on the nature of the phase response and the signal.

Consider two signals, narrowband FM (carrier plus 2 sidebands, one negative, one positive) vs. AM (both sidebands in phase).  The envelope of the two is very different.

If the sidebands are clearly inside the same critical bandwidth, (i.e. the filter bandwidth of the ear centered at the carrier frequency) then phase shift can cause quite startling changes in perception.

This is also true for things like bells, glockenspiel, etc, where nonlinear effects swap energy between slightly different frequencies over time.

The old experiment using a low frequency and a high frequency was correct, those two tones do not have any point on the cochlea that they co-excite so the interaction is very limited, if it happens at all.  The problem is that signals are not like the experimental conditions.

There is no real doubt about this issue.

But in most cases harmonics fall off linearly with frequency.

Unfortunately, while that's true in analog, in digital harmonics and distortion products alias back into the passband. This is something that is very different between analog and digital signals.

> Careful, there. Digital jitter is quite counterintuitive. The frequency spectrum of the JITTER (not the clock, but the error in the clock) aliases down into baseband. <

But if it's 120 dB below the music, how could it possibly be audible, no matter what frequency it is?

Is what 120dB down? The jitter or the noise resulting from the jitter?

The jitter itself can be very, very low and still create troublesome sidebands, it depends on the spectrum of the jitter.

Also, to specify what the noise that results from the jitter is, one has to specify the signal being jittered. Jitter noise grows as f^2 (for the same jitter, with f being the frequency of the signal being jittered), so what is 120dB at 100 Hz is going to be 20dB  worse at 1kHz and 40dB worse at 10kHz base frequency.

> spec'ing jitter is usually not donw "below the music" <

All of the specs and data I've seen - for example those in Ken Pohlmann's book - show it as some number of dB below the program. With 110 to 120 dB or even more being typical. Is this not correct?

--Ethan

It all depends on the spectrum of both the source signal and the jitter. Jitter has been often oversimplified.

[Paulhoff]

JJ

With today’s equipment, I mean in the last 5 years (maybe more). Do you think that jitter is a problem anymore. I run my CD and DVD players straight into the receiver digital input, which of course goes thorough the receiver’s own wave shaper and buffer etc. We know that the digital information goes through a lot more then going straight to the DA converter. Unfortunately a lot of audiophile guys think it goes straight to the DA converter and have no concept of comes before DA, so they misquote the jitter problem. Jitter may be a problem, but I can’t believe it is anyway near the as bad as these so-called audiophile guys make out.


Paul

[jj]

JJ

With today’s equipment, I mean in the last 5 years (maybe more). Do you think that jitter is a problem anymore.

Jitter certainly should not be a problem.

Unfortunately, SPDIF and AES/EBU standards are not the nicest serial standards on the planet, and make it quite hard to recover a quiet clock.

It is possible, the methods were known at least from the 1950's in telcom, but they didn't penetrate aes/ebu receivers nearly as fast as one might wish for such old technology.

More unfortunately, in some SPDIF and AES/EBU DAC's, there is a data-dependent jitter problem. What's more, some DAC's are sensitive to the slew rate of the digital signal.  So, yes, jitter is still sometimes a problem.

There have been other errors in DAC's, both in separate-box DACS that use the serial line to pass the data, as well as in one-box units where the DAC reference is poorly filtered, and the CD servo throws an 400Hz, give or take, modulation on everything.

So, it is still sometimes a problem. It shouldn't be.

[Paulhoff]

JJ

Back to jitter. DVD Video would surely show up jitter, wouldn’t it? Especially on solid red, which the eye is very sensitive to, when it comes to anything off color. Red was always used in setting up color TV picture tubes for purity. I build two Heathkit 25” TVs and had to set up the purity and convergence.  

Paul  

[jj]

JJ

Back to jitter. DVD Video would surely show up jitter, wouldn’t it? Especially on solid red, which the eye is very sensitive to, when it comes to anything off color. Red was always used in setting up color TV picture tubes for purity. I build two Heathkit 25” TVs and had to set up the purity and convergence.  

Paul  

No, it's very different for video vs. audio, and you're confusing what "jitter" comes from as well. The source of the "jitter" in audio is the SPDIF interface, or a bad clock generator, not the "DVD" or the 'CD' or whatever.

In video, "jitter" sensitivity is very low, compared to audio, if you're talking about either spatial jitter or time jitter. If you're talking about time jitter (frame rate) the eye is very insensitive, compared to audio. If you're talking about spatial jitter, the sensitivity is a bit higher, but still well under the ear's sensitivity.

The reasons are simple...

The ear does (mechanically, due to its structure) a frequency analysis wherein the time-local dynamic range in frequency is somewhere in the neighborhood of 90dB. That means that sensation can arise from widely separated frequencies even if one is 90dB above the other, in some cases.

The eye has no frequency analyzer, it's primarily a spatial receptor, with some very simple frequency shaping due to MTF and inhibition. This is not a "strong" filter like the ear by any means.

So, the two must be treated quite differently when distortion products are considered.  The ear is very sensitiive to nonlinearity even though it's quite nonlinear itself, the eye is more linear in the short-term sense, but can accept quite some nonlinearities at its input, because it does not do frequency analysis, rather it does spatial analysis.

[Paulhoff]

JJ

You don’t consider color vision frequency analysis?

Paul

[Paulhoff]

JJ

I know were jitter comes in at. It is just will all the audiophile’s hype about it and how cheap intergrate circuits are now, that by the final stage before the DA, and the DA itself, the timing should be clean no matter want the SPDIF interface standard is.

How enough of that.

I would like to have information on to conservation of energy, information, etc, showing the "Shannon Bound" is in layman’s terms, and why an LP can not possibly contain "more information" than a 'CD'. Etc, etc.
Can you explain this a little more detail with a couple of examples with the math? I am not an LP fan or Tube fan. But I do go around and around with a friend who is. I would like to have more to throw at him that he would most likely not agree with.

Thank you for you time

Paul

[jj]

JJ

You don’t consider color vision frequency analysis?

Paul

No, I don't in this context. It certainly selects by frequency of light, but the "frequency analysis" in the case of jitter in pixels would be "spatial frequency", and the sensitivity of the eye to spatial frequency, although not flat, is not as intensely selective (by orders of magnitude) as the ear's.  A spatial prefiltering can pretty much describe the effects of spatial frequency in vision.

That's not at all the case in audio, where there is both passive and active, nonlinear filtering taking place on the cochlea. Check out http://www.aes.org/sections/pnw/aes2004.ppt for a talk I gave on the subject. It's an intro talk, there is obviously much more to be considered.

If you consider, your question shows another difference between hearing and vision. Hearing requires compression waves in the atmosphere, and there is one dimension of frequency.

Images have two dimensions, each of which can be analyzed (as can others that analyze in two dimensions at once) in frequency for two frequency axis (the two will be perpendicular, separable filters or not), and of course color, which is another pseudo-frequency axis( pseudo because monochromatic light of any given frequency can most often (but not always) simulated by a combination of 3 colors.), and then we have the difference in time, for a spacio-temporal axis. There are many axes to analyze video with, some of which (the temporal frequency axis, for instance) aren't that useful in some senses.

For audio there is only the pressure and a variety (no more less variety than for any one dimension of vision) of frequency analysis that can be made.

So they are very different animals.

[jj]

JJ

I know were jitter comes in at. It is just will all the audiophile’s hype about it and how cheap intergrate circuits are now, that by the final stage before the DA, and the DA itself, the timing should be clean no matter want the SPDIF interface standard is.

"Should be" is something that often applies to the audio industry.

Most of the audiophile's hype is just that, pseudoscientific hype, but there is, unfortunately, a thread of real problems with convertors, especially ones running from one of the serial standards. Just enough, I guess, to keep the hype flowing.

I would like to have information on to conservation of energy, information, etc, showing the "Shannon Bound" is in layman’s terms, and why an LP can not possibly contain "more information" than a 'CD'. Etc, etc.

Let's look at LP. It has a dynamic range of about 70dB in the best of times and cases (must count surface noise, an unavoidable part, as well as rumble, etc).

If it had two channels and the same bandwidth (and the same SNR at every frequency) that would work out to about 70/6 bits per Hz, that is to say about 12 bits/Hz (11 2/3). That by itself is inferior to the CD's 16 bits/Hz.

However, it is argued, and with some meaning, although how much is actually present in any real recording can be debated, that LP's can go to 30kHz.

So, multiplying by frequency, you see (one sided spectrum, for various reasons at the end we have to double all of this, but we'll compare apples to apples) 20*16 vs. 30 *12. That would seem to put LP with more information, HOWEVER...

Above 5-6kHz, the SNR of an LP goes down rapidly, instead of being 70dB, the actual SNR is more like 20-30 dB.    This is now 4-5 bits/Hz. Tihs applies over at least half of the bandwidth, in reality, so now we can treat that as (I'm approximating, but not terribly harmfully so)

lp bits = 12 * 5 +5*25 or about 185 compared to the cd's 320
(note both measures are for one-sided spectrum and one channel)

But that's not all. The LP has distortion. We can't send as much information at lower frequencies, so ew'd better use 50dB (.3 % distortion) as an upper figure instead of 70dB.

No, but wait, at high frequencies, channel independence goes to pot, and we only have 10-15dB of channel independence. So really at high frequencies (supra 15kHz for sure, maybe lower) we don't really have 5 bits/Hz/channel, we have more like (5+2,5) bits/Hz/2channels.

In other words, the LP isn't close in terms of total information content.

Point out to your friend that LP's do have some very euphonic kinds of distortion, i.e. distortions (l-r enhancement, noise floor, out of phase rumble, interchannel intermod, etc) that often sound better than the linear system.

Note: the above analysis is very, very rough.  It is, however, not too far off the mark.

You may hear people claim 90dB from turntables, ask them to turn off their A-weighting if they do.  You will have people who claim less distortion at high frequencies, or higher frequencies, but then we can have several discussions if you'd like, on the realities of getting 30kHz through either a new or old recording chain, as well as the atmosphere.

[Paulhoff]

JJ

You said,

"Point out to your friend that LP's do have some very euphonic kinds of distortion, i.e. distortions (l-r enhancement, noise floor, out of phase rumble, interchannel intermod, etc) that often sound better than the linear system."

IF I am not wrong, this is basically what I have been trying to tell my friend about LPs. He often complains that (want he hears) CDs as being flat with no depth. I have told him it is the channels being out the phase and giving him false audio clues. In producing an LP, there must be at lest 6 steps that I can come up with.

1) The original multiple channel tape.
2) The master mix down to two channels
3) The re-mastering for the LP’s (to make up for low and high frequency short comings)
4) The first metal master cut
5) The negative metal copy for pressing the LP
6) And finally the pressing of the LP itself.

All this in itself must be adding a lot the distortion to the LP sound.

I remember years ago taking the output from an amp to an oscilloscope, left channel to the vertical and right the horizontal. I played a record with a test tone on both channels. I should have (as you know) a nice 45 degree line. No, it was very wavy, for lack of a better word. Looking with just the horizontal at the sine wave, it was anything but steady.
Years go by. Now doing the same thing with a CD. What a big deference there was. The 45 degree line, rock steady. The sine wave, again rock steady. Not only that, but I tried 19 kHz, a nice steady sine wave.


I am not even going to start with the tube amp that he likes.

Thank-you again

Paul
 

[jj]

JJ

IF I am not wrong, this is basically what I have been trying to tell my friend about LPs. He often complains that (want he hears) CDs as being flat with no depth. I have told him it is the channels being out the phase and giving him false audio clues. In producing an LP, there must be at lest 6 steps that I can come up with.

1) The original multiple channel tape.
2) The master mix down to two channels
3) The re-mastering for the LP’s (to make up for low and high frequency short comings)
4) The first metal master cut
5) The negative metal copy for pressing the LP
6) And finally the pressing of the LP itself.

All this in itself must be adding a lot the distortion to the LP sound.

Well, all of the steps you describe create distortions and your case can be made on those, but what I'm referring to is the distortions due to the LP reading mechanics.

So, your case can be made on what you mention.  One can make the same case, perhaps more easily, by pointing out the various distortions in LP's.

[Paulhoff]

JJ

Me again, and the last on LPs. My friends favorite rugument about CD Vs LP is LP’s have information between the bits because they are analog like the real world. I try and explain to him that the ear is not truly analog and that the information is electrical pluses in the nerves and chemical between the nerves to the brain.

Can you go a little more into your quote below, I am sure some of the things you will say are the ones I have been telling my friend.

“I'm referring to is the distortions due to the LP reading mechanics”.


Thank you again,

Paul