[art]

BTW, I can also show how digital cables affect jitter. Bring an extra $100.

I'm sure it can be measured if the cables are long enough. That's not the point. The point is if it can be heard and reliably identified by ear. We can easily measure 0.01 percent distortion versus 0.001 percent, but that can't be heard either.

--Ethan

No, you are incorrect. Shorter cables are more likely to have reflection problems than longer ones. My background is RF and microwaves, not room acoustics. I have to strongly disagree with you on that specific point. Agreed that 0.001% and 0.01% THD comparisons are probably meaningless. Sorry, I don't feel the comparison is entirely valid.

I think we are on 2 different wavelengths here. The methods that you are proposing may well yield different results than the ones that I am talking about. I strongly feel that jitter measurements, when it comes to SPDIF, are significantly different than the types you are eluding to.

As I have said repeatedly, jitter is a major problem in SPDIF. It is not a major problem in the typical stand alone CD player. It can have some of the same problems, if the design and execution is poor.

Ethan, I firmly believe that if you had the same "life" experiences as I have, both as a telecom R&D guy, and a high-end designer guy, who specialises in these sort of concerns, that you would understand why I feel we are on different planes. While your test would certainly be interesting from a scientific standpoint, I feel that it is not entirely valid to issues regarding SPDIF.

Pat

[audioengr]

If you understand the nature of real-time streaming audio, and you have some engineering background, you will quickly come to the conclusion that errors are very unusual and that jitter is the only real culprit in sound quality, given a particular sample-rate.  There are certainly the effects that the D/A conversion, digital and analog filtering has on the signal, but lets just talk about the digital signal itself here.

Jitter is timing inaccuracies in a real-time bit-stream.  Jitter has both amplitude and spectra characteristics.  In other words, it is possible to have a higher jitter system sound better than a lower jitter system if the spectra of the low-jitter system is more objectionable to the human ear.  Jitter can also have a filtering effect.  If your particular system for instance has objectionable sibilance due to a certain component or "edgey" silver cables etc., jitter can actually mask this, making the system sound better.  The real goal however, should be to identify the sibilance and remove it and then reduce jitter to inaudible levels.  Otherwise, the imaging and detail rendering of the system will never be world-class.  True 3-D imaging will never be achieved without this.

Jitter has different sounds, but usually high-frequency echoes or out-of-focus highs.  I have personally sat and listened to systems that I though were quite good and then when the jitter was removed, it is like cleaning the film from a pane of glass that you are looking through.  Until it's gone, you literally dont know what you are missing.

Another thing:  Many DAC manufacturers claim that their DAC's not only eliminate jitter, but they are jitter tolerant on the inputs.  I have yet to find a DAC that is insensitive to jitter on the inputs.  It is very difficult to do a design that is jitter insensitive.  It is also very difficult to make any clocking system that results in inaudible levels of jitter.  I have been working on this problem for many years, and I have a little experience with this - 30 years digital design as as EE.  I was a design-team lead on the Pentium 2 at Intel Corp..

Steve N.

[BrianM]

BrianM:

The clock should be "perfectly" quiet. The sound/music that you will hear is modulating the clock, because that is what you are hearing.

You asked where you can read up on SPDIF. I don't have a bibliography at my hands to show you where to look. Instead, I provided an example of why we came to realise it is so bad.

Thanks, I now have a better handle on your example. "So bad" is relative, of course. Depends on how audible it is through the speakers, and only one of us has an opinion on that.

The fact that you can pick out the music that you are listening to is an example of data correlated jitter. That is what makes SPDIF so bad. The method of extracting the clock contains modulation artifacts that is strongly dependent of the programme source. It shouldn't be.

I certainly can't disagree with that, as an ideal.

[audioengr]

And are you going to send me $100 for every one who can?

Heh, nice try. The $100 goes to those who can reliably identify jitter.

How are you going to reduce the jitter? Who sets the metrics for that? Who decides whether your "jitter reduction method" is valid.

I'll be glad to let you have a say. I tested this recently by playing 24-bit music files through my Delta 66 sound card using its internal clock and then clocked by a $6,000 Apogee A/D/A box. We also compared the same file played back through a $25 Soundblaster card. (The fellow who brought over his Apogee was unable to hear a difference between the three setups.) Even easier, and probably better controlled, is to use test files from Arny Kruger's PCABX site. We can play files with increasing amounts of jitter added artificially, and identify at what point the jitter is audible. If you have other suggestions, I'm all ears.

You are free to come to Texas. Bring your $100.

Sorry, the test will be done here in my two controlled and well-treated listening environments.

BTW, I can also show how digital cables affect jitter. Bring an extra $100.

I'm sure it can be measured if the cables are long enough. That's not the point. The point is if it can be heard and reliably identified by ear. We can easily measure 0.01 percent distortion versus 0.001 percent, but that can't be heard either.

--Ethan

Ethan - as has been pointed out already, shorter is not always better for a high-speed digital signal.  See this paper for an analysis:
http://www.positive-feedback.com/Issue14/spdif.htm

[BrianM]

I assumed Ethan meant the opposite - that he was sure reduction of jitter could be measured if the cable is long enough.  Since that finding gets bandied about a lot.  (Even I had heard about it!)

[art]

Well, you understand that you aren't actually hearing the jitter, per se, through your speakers. If you attempt my little test, you will hear the signal that is modulating the clock. In this case, the music modulates the clock.

Our one and only DAC thingey had some advantages over the CD players we were making at the time. Better DAC and filter chips. (Freed from the cursed use of whatever Philips was sticking in the platforms we were using.) While there was an immediate sense of better detail and resolution, we were appalled by how "squashed" everything sounded. I spent 6 months trying to get rid of that, without redesigning the whole works. I did everything except build a better reclocking circuit. (no room!) The jitter was obviously never eliminated, but it sounded better as it was lowered. We made a business decision to make as many as we good, as fast as we could, and sit back and wait for Stereophile to discover how much of a problem jitter was, and crater the market.

Which they did on both accounts.

Ah, I see Mr. N. has joined the discussion. Nope, can't trust him either. EE, selling a product......exotic circuits and what-not.......nope has to be in cahoots with me somehow.

Sorry, he ain't. Don't know the guy by anything other than name. Gee, funny.....he says a lot of the same stuff. Inquiring minds demand to know!

Pat

[Ethan Winer]

Shorter cables are more likely to have reflection problems than longer ones.

I was going on the theory that with longer cables, the higher capacitance can soften the rise times leading to more uncertainty as to where the transitions take place. I regularly use plain old RCA cables from 3 feet long from my DVD player to my receiver, to about 20 feet long between my DAT player and PC sound card. I've never had a problem passing digital audio through any of them.

Agreed that 0.001% and 0.01% THD comparisons are probably meaningless. Sorry, I don't feel the comparison is entirely valid.

Okay, now we're getting somewhere. If you agree that 0.01 percent distortion (sum of all artifacts are -80), why do you believe jitter that's even farther down below the music is audible?

The methods that you are proposing may well yield different results than the ones that I am talking about. I strongly feel that jitter measurements, when it comes to SPDIF, are significantly different than the types you are eluding to.

Okay, then you tell me how you'd have me test this with various listeners present. If jitter is a problem, why do you believe Arny's test files cannot be used to establish at what amounts jitter becomes audible?

Ethan, I firmly believe that if you had the same "life" experiences as I have

And I feel that if you had the same life experiences I've had, with deep interests in audio perception, principles of skepticism, and knowing how huge the affect of room acoustics are on what we hear, you'd come to the same conclusions I have.

This reminds me of a test I did a few years ago. Some guys were arguing that the nature of distortion affects how audible the distortion is, which of course is true. But "normal" analog circuits add harmonics such that the first few are loudest, and the rest become progressively softer the higher you go. To entertain their idea that an amplifier having mostly high-order harmonics far-removed from the fundamental (which will not generally happen) might sound different from another amplifier with similar THD, I made a test to see how audible that might be.

I chose 100 Hz and 3 KHz for the test, since 100 Hz is where our hearing falls off, and 3 KHz is where we're most sensitive. They're also so far apart that masking will not have an affect. I made a Wave file with 100 Hz at a high level, and added in 3 KHz at different levels to see how soft it could get and still be heard. I mixed in the 3 KHz tone as pulses - on for a second, off for a second, and so forth - to make it easier to hear. When the 3 KHz tone was 40 dB below the 100 Hz tone, it was very soft but I could clearly hear it. At -60 dB it was much harder to hear, but was still barely audible each time it started and stopped. But once the 3 KHz tone was 80 dB below the 100 Hz tone it was impossible to hear at all.

These are the types of tests more people should do for themselves. Only by experimenting like this will people learn what really matters with audio and what does not matter.

--Ethan

[BrianM]

Ethan, isn't the theory that the presence or absence of the 3Khz tone (to use your example) i.e. jitter, makes the 100Hz tone (i.e. the music) sound better or worse somehow?

[art]

Ethan,

The reason why I feel that arguing 0.01% and 0.001% is meaningless because a lot of folks like tube amps, and they certainly have more than those numbers.

Numbers don't always tell the entire story. How else can explain the popularity of tubes, despite 50 years of advancements in SS amps? Or why amps with numbers that low.....which, by your account, should not be detectable, but some listeners find objectionable?

I don't doubt all that you have learned, but I feel that we are coming at this from not only 2 different positions, but 2 entirely different problems.

Random jitter, in the levels you are talking about, probably is not a problem. But we are talking levels of jitter that are higher, and because of the amount of data-correlation are more objectionable. You can hear them, and it is important to quantify not only how much, but at what rate as well.

Somehow, I don't think our views well ever converge.

But let's see if I can get you to move on one other issue, same as I have on THD. (THD is a pretty much meaningless number, unless you are designing amps for a living.)

Cables:

You are only thinking in terms of HF signal loss, and whether it leads to inter-symbol interference, or data errors. If that is all that you are willing to accept, then this will go no further.

If you are willing to accept that cables can and do cause reflections, and that these reflections can and do change jitter levels and spectrum, then maybe there is room for common ground. If you think that the only criteria is passing data that is error free, then we will never converge.

Or, you could make 2 cables.......one 3' long......the other 20' long.......do your ABX test, and then tell me. Hell, invite all your buddies over. See what they think, as well.

I'll give you the address where to send your $100.

In a lighter vein, (and do this in your best McCarthy-esque voice):

"Are you now, or have you ever been, a member of the BAS?"

Pat

[*Scotty*]

Ethan,I think I may be misunderstanding your basic position on jitter. You appear to have made an argument that all jitter artifacts are at or below -120db and therefore cannot be heard.
This position appears to also be an argument against the requirement of clock circuits in digital audio applications in general.
Is this in fact the case.
Scotty

[AphileEarlyAdopter]

I didn't think jitter was a noise issue quantifiable as dB below signal, rather a timing error OF the signal.

My reference is Ken Pohlmann's Principles of Digital Audio which is considered by many audio engineers to be the bible of this stuff. His graphs show jitter typically 120 to 140 dB below peak level, and expressed that way as dB. Yes, you can express it as a timing error, which is what it really is, but as Pat explained it shows up as FM sidebands at some dB level.

This is how Stereophile explains what jitter is actually doing.

The problem with those charts is the jitter is exaggerated about a million times - literally - to make the point. I can show that eating broccoli is very bad for you if you'll agree to eat 35 pounds of it at one sitting.

I must point out that talking about jitter and what is and is not audible is meaningless without specifying amplitude, frequency spectrum, and level of data-corrleation. Random-occurring noise is much less noticeable than jitter that is data-correlated.

Yes Pat, but if the artifacts are 120 dB below the music it's simply not audible whether it's random or correlated.

A whole branch of industry (building transports, reviewers and DIYers) cannot be a bunch of charlatans or deluded.

Why do you believe that? If a normal amount of jitter is inaudible, what is your better explanation? Please be very specific!

I don't think anyone doubts the existence of jitter as a measurable phenomenon.  The disagreement is over how important, relevant, audible it is.  Building transports and DACs that attempt to address jitter doesn't have to be charlatanism, but it might be gilding a lily.  Lots of businesses promote "advancements" before there's universal agreement as to their benefits.

Give this man a ceegar.

Finally, I repeat yet again that human auditory perception is incredibly frail, and often we believe what we want to believe. That, plus comb filtering as explained in my article linked earlier. Not sure if any of you caught the recent news item about expensive placebos versus cheap placebos:

http://www.scienceagogo.com/news/20080204181613data_trunc_sys.shtml

Note that the reported improvements in both cases were due to placebos!

Those here who believe they can discern reduced jitter are most welcome to visit me and be tested blind. I have $100 for each and every one of you who can reliably pick out the difference. I am absolutely serious. Bob? Anyone else near me in Western CT?

--Ethan

Hi Ethan,
Yes, jitter can be expressed as side bands. But how does it sound in combination with the main signal ?
Lets take a standard music sample, do fourier analysis, and remove all bands below a certain db, will this sound exactly like the original signal ? (This is pretty much the mp3 compression logic and we know it is not the same as the original wav or do you think nobody tell the difference between mp3 and wav ?).

BTW, I do know that transport make a difference i.e jitter in sound quality. It is not a belief for me.

(I am a CS/EE person and I do temper my audiophile pursuits with technical validity).

[darrenyeats]

The reason why I feel that arguing 0.01% and 0.001% is meaningless because a lot of folks like tube amps, and they certainly have more than those numbers.

Folks eat doughnuts for breakfast every day too.

(This is pretty much the mp3 compression logic and we know it is not the same as the original wav or do you think nobody tell the difference between mp3 and wav ?).

We can prove mp3s aren't the same as lossless. We can prove 16/44 lossless isn't the same as 24/96 lossless too. As for people hearing the difference, isn't that a whole different discussion? Anyway, what mp3 bit rate are we talking about?

That is why the $100 prize is for hearing the difference blind. It's kind of easy being a golden ear when you know what's playing!
Darren

[Rx8man]

I Googled a summary link of Ken Pohlmann's Principles Of Digital Audio (good read Miklorsmith)

Like I suspected, the A/D conversion and power supply is the most critical area for the D/A to do its job more effectively.

My CEC transport could use improvements in this in area according to Joseph Chow at Component Plus USA (an expert at addressing low level noise)

I'll cross this bridge first, then wait for the onslaught of PC technology to iron out the final folds of its digital transmission(s)

[Ethan Winer]

Ethan, isn't the theory that the presence or absence of the 3Khz tone (to use your example) i.e. jitter, makes the 100Hz tone (i.e. the music) sound better or worse somehow?

I have no idea what the "theory" is because none of the theories I've read make any sense to me. From my perspective, we hear things only if they're loud enough to be heard. End of story. I do not accept that artifacts too soft to be heard on their own can somehow subliminally affect louder content.

Here's an easy test - sweep a sine wave from, say, 40 Hz up to around 300 Hz at a fairly loud volume. Sweep it very slowly listening for physical buzzes and rattles in your room due to loose windows, vibrating picture frames, and so forth. Unless your room is unusual, it's very likely you'll find several bass frequencies that excite obvious mechanical resonances. Now play normal music fairly loudly. Do you notice the rattles? Unless the rattles are very loud they are not usually objectionable.

Numbers don't always tell the entire story.

Of course they do! At least when the goal is to assess raw fidelity.

How else can explain the popularity of tubes, despite 50 years of advancements in SS amps?

I can see you did not read the Believe / Comb Filtering article I linked earlier. It's not a lot to read. Here it is again:

Why We Believe

See especially the last part where I talk about euphonic distortion. I agree that adding distortion can sound pleasant on some types of material. This is a matter of taste. However, I do not want my playback system adding its own distortion to everything I listen to!

Or why amps with numbers that low.....which, by your account, should not be detectable, but some listeners find objectionable?

If an amplifier has distortion low enough that all artifacts are 80 dB or more below the music, then the amp can be considered audibly transparent. So if a listener claims the sound is objectionable, I'm sure it's sighted bias and they'd never be able to identify that amp from another with similar distortion in a blind test. Ears are a terrible test instrument!

Random jitter, in the levels you are talking about, probably is not a problem. But we are talking levels of jitter that are higher, and because of the amount of data-correlation are more objectionable. You can hear them, and it is important to quantify not only how much, but at what rate as well.

Okay, so at what level below the music do you believe jitter can be noticed?

Somehow, I don't think our views well ever converge.

As a science-minded audiophile I am ready to change my opinion instantly when presented with compelling evidence. I'd hope you feel the same. So that means only that neither of us has found the right words to convince the other.

If you are willing to accept that cables can and do cause reflections, and that these reflections can and do change jitter levels and spectrum, then maybe there is room for common ground.

Yes, at 10 MHz cable reflections can change the signal in a way that matters. But at 20 KHz? With a digital stream? How is it possible for such reflections to be audible?

"Are you now, or have you ever been, a member of the BAS?"

What's the BAS?

You appear to have made an argument that all jitter artifacts are at or below -120db and therefore cannot be heard. This position appears to also be an argument against the requirement of clock circuits in digital audio applications in general. Is this in fact the case.

My understanding from reading Pohlmann, and seeing a few published jitter specs and measurements, is that -120 dB is typical. If jitter is at -60 dB I imagine it might be audible with some types of material. But how lame would a designer have to be to create a circuit with jitter that high? What situations do you know of where jitter is ever that high?

--Ethan

[Ethan Winer]

Lets take a standard music sample, do fourier analysis, and remove all bands below a certain db, will this sound exactly like the original signal ?

It depends entirely on where you set the threshold for removal!

--Ethan

[TheChairGuy]

I think what Ethan is trying to get across (mostly) is that tho jitter exists, it may have been minimized to a point of a relative non-issue now already......the problems inherent in most listening area/rooms is a far greater point to tackle.

What he hasn't said and I'll add is that we are butting up against the reality that Redbook may not be a sufficient digital format for our collective goals as audiophiles.  We are focusing on something as small as jitter when it may be the format itself insufficient. 

24/96 or 24/96 or 24/194 with MLP is a very much better technology for both recording and playback. A cheap DVD/DVD-A player and a purposeful recording in these formats will convince one and all of this fairly quickly.  You can hear the added resolution even on inexpensive speakers.

Not all folks care - in fact very few overall as the market for hi-quality DVD and DVD-A discs has dictated thus far - but we as audiophools should and do.

John

[Bob in St. Louis]

BAS = Boston Audio Society

[BrianM]

What he hasn't said and I'll add is that we are butting up against the reality that Redbook may not be a sufficient digital format for our collective goals as audiophiles.  We are focusing on something as small as jitter when it may be the format itself insufficient. 

24/96 or 24/96 or 24/194 with MLP is a very much better technology for both recording and playback. A cheap DVD/DVD-A player and a purposeful recording in these formats will convince one and all of this fairly quickly.  You can hear the added resolution even on inexpensive speakers.

A lot of people have yet to be convinced by the inherent audible superiority of 24/96.  Many 24/96 recordings arguably sound better because they're just recorded & mastered with extra attention to detail.  Many straight up redbook CDs sound just as good IME.  Right now I'm in mind of the Russell Dawkins CD I picked up recently, the 4 guys playin their geetars.  This is a redbook CD and it sounds every bit as good if not better than any 24-bit CD I've heard.  (And I've heard 24/96 on equipment designed for it.)  At the very least it's close enough to make a blind test very very difficult.  A lot of people's opinions about the limitations of redbook may be a result of the limitations of their own digital playback.  At any rate I'm almost completely sure that I would be unable to tell the difference in quality between an excellent 24/96 recording and an excellent 16/44 recording.  And I think there's even been a study (!) demonstrating just that.

[art]

My understanding from reading Pohlmann, and seeing a few published jitter specs and measurements, is that -120 dB is typical. If jitter is at -60 dB I imagine it might be audible with some types of material. But how lame would a designer have to be to create a circuit with jitter that high? What situations do you know of where jitter is ever that high?

--Ethan

Maybe not quite that high, but you have not spent much time analysing clock jitter in SPDIF receivers. I have. It isn't "lame", it is a product of a "lame" architecture.

Toobes........sorry, I did not have time to read the link, but you will not see arguing against you on the "why" part. Euphonic colouration aside, the reason why 0.01% vs 0.00001% doesn't work is that it leaves out the spectrum of the distortion. Pretty much the same as in jitter audibility predictions.

It isn't the level of distortion that is point, but the distribution. Some is euphonic, and some folks like it. Others (like 7th and 9th order) are especially nasty, and folks don't care to hear them.

As for level of audibility, I don't imagine that you ever measured capacitor distortion levels. They are typically down in that -120 dB range. I don't know of many folks here who won't agree that caps sound different. (Well, ok, I do know of one, but he isn't reading this. Yet) Now, I am not going to claim that caps sound different solely because of -120 THD. There are most likely other mechanisms at work. But, it may well enter into the equation. But if it is part of it, yet below what some feel is audible.........well, the mysteries of life and audio continue.

Point being: numbers by themselves don't tell the whole story. (Did I say that before?) But, if you are looking for a number, designers that I work with fell that clock jitter numbers <10 nSec are what the goal is. Close-in jitter is believed to be more detrimental than higher frequency jitter. Once again, you have to know the spectrum. Same as in THD. A simple number by itself is not sufficient.

For a stand-alone CD player, or a SPDIF DAC, with a good clock, right at the DAC chip,, that is not an insurmountable goal.

SPDIF without reclocking, then you are going to be hard pressed to get <200 nSec. The PLLs have that much built in. Add to that reflection-induced jitter, well you can see why -120 db is never going to happen.

Call it lame. I call it SPDIF. The scourge of the audio world.

Pat

[Daygloworange]

A lot of people have yet to be convinced by the inherent audible superiority of 24/96.  Many 24/96 recordings arguably sound better because they're just recorded & mastered with extra attention to detail.  Many straight up redbook CDs sound just as good IME.  Right now I'm in mind of the Russell Dawkins CD I picked up recently, the 4 guys playin their geetars.  This is a redbook CD and it sounds every bit as good if not better than any 24-bit CD I've heard.  (And I've heard 24/96 on equipment designed for it.)  At the very least it's close enough to make a blind test very very difficult.  A lot of people's opinions about the limitations of redbook may be a result of the limitations of their own digital playback.  At any rate I'm almost completely sure that I would be unable to tell the difference in quality between an excellent 24/96 recording and an excellent 16/44 recording.  And I think there's even been a study (!) demonstrating just that.

I'll second what BrianM says.

I have 24 bit capability in my recording studio. I use 16 bit. I've done many tests between 16 and 24 bit, and never found any difference that I could hear. I'm not saying there isn't one, but without having an ambient noise floor below 20db, a room heavily treated for reflections, and a super high resolution system, you wouldn't hear IME any difference anyways.

I've spent years and years and thousands of hours critically listening to audio reproduction (not music reproduction) of music, I think it's for all intents and purposes, a moot point. If 24 bit was audibly better to me, and being that I have the capability, I would use it.

I can tell you however that you don't need half the environmental requirements to hear the multitude of noise and other sonic distortions and aberrations of an analog multi-track recorder.

Digital is simply a more transparent and linear recording medium.

Cheers