[AphileEarlyAdopter]

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

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.  So no, they're probably not all out to put one over on us, but in a competitive industry everyone is trying to differentiate their products.  Making claims (even measurable claims) about jitter is one way to do so.  The problem is that the difference between good and great in this industry can only ever be a matter of perception, and very subtle perception at that.  People always seem to underappreciate how complex the notion of reliably evaluating the impact/benefit of a component, or a component of a component, necessarily is.  The component that "fixes jitter" may be doing a host of other things well or poorly, and may have been heard -- will have been heard -- under conditions nearly impossible to duplicate by you.  And of course heard by ears other than yours...

All that said, I'm more inclined to listen to those who, rather than swapping components to fix the "problem", swapped out the crystal chip or whatnot, since that's obviously a localized change that leaves everything else in place.

Brian, you raise valid points. There are people who buy $10000 transports !!! I guess it mainly addresses jitter. I will not go that far. I am not sure I can easily appreciate its difference over a $2k transport. That said, I am way down in the value chain on these. I use either a Zu Ash digital cable or Sound Professionals  Glass toslink. I also run a pure digital playback chain - a digital modded SB3-> Panny XR55 -> biamping a Silverline Sonatina speakers.
In terms of jitter, I would quantify its affect as the difference between a crappy interconnect vs a well-constructed one (I use a rscables.com silver ic for eg), in simple terms. But the benefits are far reaching. I could easily tell the difference between a Philips 963sa vs a stock SB3. The stock sb3 was very soft sounding. The modded SB3 made a significant difference - everything is sharper - bass/treble. Low jitter also produces a more 'natural' sound. People call it less 'edgy'.
I dont know how much you are aware of the mod business. Quite a lot of modders change the clocks out in CD/DVD players. I really wanted the clock replaced in my SB3, but it looks like there is not enough space. Some DIYers have already done this. The benefits of a better oscillator is wellknown in the audiophile world, especially people who do  not want to spend the big bucks on big name transports, but understand the benefits of low jitter.

[Bob in St. Louis]

Could jitter be compaired to a strobe light effect? Or maybe a disco ball? "jerky instead of smooth"?
Not sure if I worded that well enough?

Bob

[AphileEarlyAdopter]

Could jitter be compaired to a strobe light effect? Or maybe a disco ball? "jerky instead of smooth"?
Not sure if I worded that well enough?

Bob

In simple terms,  for eg. a CD should be played back with a 44.1khz clock, instead, it gets played back at 44.0999999 or 44.1000011 Khz or something like that. This value is  not constant but keeps changing. Its like the way the DJ's move the record back and forth, but very minutely at a very very high rate (actually it is exactly the same mechanism).
It  is the same thing as FM broadcast - there it is a feature, here it is a bug (if you know the s/w industry jargon). Stereophile uses the same mechanism to detect jitter. They measure how a central (only) signal frequency gets distorted. If you had a 6Khz sinewave on a Cd, you played it back, in the output you should see only a 6Khz wave, but you see output at other frequencies too.

[genjamon]

This whole conversation deserves some discussion about how a string of bits becomes converted into an analog waveform.  Any effects of jitter would lie somewhere in this process.  I'm not the one to provide any authoritative discussion about this process, but let me outline my own thoughts a bit based on some very introductory knowledge.

As far as I'm aware, regular redbook waveforms are generated from 16 bit binary "words" sampled 44,100 times per second.  I don't know how a 16 bit word translates into what part of a waveform, nor how two channels are coded.  Is every other 16 bit string a portion of the left and then right channels, in an alternating fashion?  Or is half of each 16 bit string depicting the left, and the other half the right channels? 

Regardless of the specifics, jitter would effect bits on a relatively random basis.  A lost/smeared/confused bit due to jitter might occur on the first bit of the 16 bit word, on the fifth, on the last, etc.  Furthermore, it would not likely be a regular pattern of bits that are affected. 

My understanding of binary is that, just like in decimal counting, the further along in the word, the higher the order of magnitude of effect in the value of the word.  In decimal-based counting, a change in the sixth letter would be on the order of 100,000's, and a change in the second letter would be on the order of 10's.  In binary counting, a change in the second letter would be on the order of 4's, while the sixth letter would be on the order of 64's, and a change in the 16th letter would be on the order of 2^16 (65,536). 

What I'm getting at here is that the effect of jitter should be relatively variable, depending on where exactly lost bits are occurring in the actual 16 bit words that eventually constitute our waveforms. 

One implication in my mind is that since words are made of 16 bits, the picosecond measurements of jitter are misleading, as the actual impact is orders of magnitudes larger than the actual number of lost/smeared bits (since words take 16 bits of perfect data in order to construct accurately).  Surely our speaker magnets can't respond in picosecond timescales, but picosecond timescales in bit communication may lead to microsecond or nanosecond effects on fully constructed waveforms - something I bet many of us with highly resolving systems might discern.   

Another implication is that as the losses are likely not to be completely uniform, there will be some words or phrases (in bits) that have higher levels of loss than others.  There could be small sections of the waveform that are highly distorted by multiple errors in whole clusters of words.  However, other sections of the waveform could have relatively few errors from relatively intact words, or maybe the errors take place in the first few digits of words (where errors don't mean much in terms of their effect on the total value of the word). 

Anyone care to discuss any of this further, or clarify any misconceptions I may have?

[Bob in St. Louis]

it gets played back at 44.0999999 or 44.1000011 Khz or something like that.

Yea, but doesn't it rapidly oscillate between the two? Hence my "strobe light" example. Kind of a very rapid pulsing between "just a little bit too fast", and "just a little bit too slow".
And also, doesn't multiple digital devices in the chain worsen the effect since their all "strobe-lighting" at their own rate?

Bob

[miklorsmith]

What if the clock was always a little late or a little early?  Each cycle, the error would get a little worse but would reset.  Looking at the Stereophile article and the Lessloss demonstration make this look closer than the completely random model where each cycle is different in length but centered around 44.1.

Or, it might be the Stereophile and Lessloss graphics are simplifications.

[genjamon]

But isn't jitter where the rise time of the detector isn't in sync with the electrical pulse of the digital signal?  It IS a timing issue, but clocks just help to make sure the electrical pulse arrives at the detector right when the detector is ready for it - not when the detector is already rising from another pulse, or when the detector is still falling from the last one.  In the process, the detector fails to rise due to the signal arriving at the wrong time (which can be aided by having high quality clocks), and the bit gets lost, right? 

[art]

Pat, I read the Stereophile article, but at a certain point ADD kicks in.   In your opinion, is the explanation inaccurate or wrong?

Can you elaborate in more layman's terms what you see as the important issues pertaining to jitter?

Is it distortion of the waveform? Is it an aliasing? Noise that's being added? Is it across the entire FR spectrum? Where above the noisefloor is it, or is it throughout the dynamic range?

I'd like to learn more about how jitter manifests itself in audio recording and playback.

Cheers

I don't see anything wrong with it. I would probably have expressed it differently. My point is that he is saying pretty much what I have (since before that article was written in '93), but he does not appear to be using that to sell a product. Frequently, the naysayers point to potential commercial gains to dismiss concepts when they lack understanding of the subject. (Funny how some of the same folks buy into other, less well-researched ideas. Like plunking down $$$$$$$$ on an interconnect.)

There are 2 things that are hard to grasp about jitter.

First, is what kinds are harmful, and which ones are more benign. That requires a careful examination of all the possible causes, which may be too numerous to discuss here. So, let's try to stick to the concept of what it actually is.

The best analogy that I can give, for all you vinyl-philes, is the difference between wow and flutter. One is a very low frequency modulation of the playback material, the other a much higher frequency rate of modulation.

The wow is easy to pick out: you hear a note bending up and down, just like a Doppler effect. But you also have to make note of how much it bends. One, the rate at which the pitch changes (frequency modulation rate), and the other is how far the pitch bends up and down (the amplitude of the modulation rate.)

So, I hope you can understand that jitter has a similar effect. We are modulating a playback note, by a certain frequency rate and amount. You have to know both to quantify it!!!!!!!!!!!

A very slow wow rate might not be discernible, if the pitch bends very slowly. But if you increase the amount, then it may stand out. Same rate of change, just more of it.

Likewise, if the rate is a bit faster, it may stick out like a sore thumb. No matter how much the amplitude of the change.

So, in not-so-layman terms, we now have a form of frequency modulation. But, in order to quantify it, the amplitude and rate must both be known. You can not just say "Well, we have sidebands @ 1 kHz, because it is changing at that rate." Nope, you have to know how much it is changing.

So, the article may have simplified things, to a degree. (It is intended for layman, so I see nothing wrong with that.)

OK, so we have sidebands? So what?

Well.......it depends!

Is the modulation, and therefor the sidebands, random? Or some sort of data-correlated corrupted signal. SPDIF has both. Your CD player could have both. A lot of that depends on whether the oscillator is stuck into the same chip as the digital filter. Or if it shares an common power supply, with a very noisy chip, and there is no decoupling between the two. (As you have in the SB3.)

Random jitter is not that harmful. Depending on the rate and amplitude. Data correlated is always harmful.

This gets to problem #2 with understanding jitter.

Just exactly what does it do?

Tough question to answer.

I can measure it, quantify it, come up with jitter spectrum, and all other sorts of technical mumbo-jumbo measurements, but there is one thing that we probably can not do.

Precisely explain why it sounds the way it does. I'll be the first to admit that I am not an expert in psychoacoustics. I can show you how certain forms of jitter are more detectable.........well, the sidebands that they form are..........but why they sound the way they do........I don't know.

OK, to some of you that is an admission that I am a total dweeb, and everything that I say is hogwash. Sorry, ain't so.

I feel that part of the answer lies not only in psychoacoustics, but how the modulation takes place, and where. Mind you, we are no longer talking about a simple playback note on a rotating piece of plastic. We are modulating a digital waveform, probably operating at 352 kHz or so, that is being reconstructed to replicate that same playback note. The modulation is occurring (most likely) on a 11 MHz or so clock, that is responsible for making sure that the 352 kHz data bit occurs at precisely the right point in time.

So, I would not call it distortion: the playback note looks pretty much normal. I would not call it aliasing, as we are not creating an "image" at some higher frequency. You could say it is "noise" as it is some garbage buried way down in level, that mucks up the sound. (Sure, random noise modulation gets mapped into more random noise. Data correlation......not so.) Again, you have to know the amount and rate of change to know where to look for these modulation products.

And even when you..........why does it muck up the bass??? Why do the highs sound so rough. Well, in truth, I dunno. I just know that when I reduce jitter, from any cause, those 2 things go away. I can say how much I feel it needs to be reduced by, and some may have other opinions. But as to why........don't know.

Yes, I realise that some of you will not be satisfied with that response. Would you rather that I invent some nonsense, just to see if it gets you off my back? My bet is "no". If you think about it hard enough, you can see where jitter is in the same boat as distortion. I can measure distortion all day long. But I can not explain why a lot of listeners might easily pick an amp that has high distortion numbers over one with number so low that they can not be measured.

Maybe the easiest thing to do is accept it as one of those mysteries of audio. If audio was "strictly by the numbers", the whole world would own Sony and Bose.

But, we all know that is not the case.

Bottom line is jitter does exist. It is not that hard to eliminate. Eliminating it first requires a knowledge of what the cause is. Different forms of jitter require different means to correct. Yes, there are some "one size fits all" methods to eliminate it. There are other means that might be easier, and more cost-effective. And then you may still want to employ every possible approach there is. And still feel that that you have not done enough. It will never be totally eliminated. Just reduced to the point where it is no longer a concern.

Your CD player is probably 96% of the way there already. SPDIF.......needs lot of attention. USB.....I am not even going to attempt to open that can of worms. The jury has a long way to go on that subject.

Pat

[Bob in St. Louis]

This is interesting. Might help to clear some things up:

http://www.lessloss.com/types.html
(Be sure to click on the 6 different setup buttons, and move the slider on the left side of the screen)

Granted, they sell stuff to fix jitter...... This is not a recommendation, just a link to something I found interesting.

Bob

[DevillEars]

Pat,

Here's a suggestion as to one of nasty ways in which high levels of jitter can manifest themselves to the listener:

NOISE!?

If this is a likely manifestation, then aspects like low-level detail retrieval which helps with reproducing spatial cues (soundstaging and imaging) would be improved if the noise-floor was lowered as a result of reduced jitter...

Another likely scenario from clock-based jitter is reconstructed waveform shape distortion in D-to-A conversion caused by timing inconsistency errors (ie: when the "tables & troughs" of the digital signal are of irregular, randomly variable length. resulting in the vertical "plot-point" being displaced horizontally (too soon or too late).  The effects of such distortion would be more discernible at high frequencies than low (pulse "wavelength" being a higher percentage of signal wavelength).

Or is this sh*t I'm smoking affecting my brain?

 

[genjamon]

"Or is this sh*t I'm smoking affecting my brain?"

Why else would one do such a thing?

[darrenyeats]

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.

Agreed. I don't think anyone argues that reducing jitter (data-correlated or not) is a bad thing.

Same as reducing physical vibration of equipment is no bad thing. Same as filtering your mains is no bad thing.

The question is (a) are the differences audible at all and (b) if so how much bang-for-buck can I get by addressing it?

Either way, I applaud Pat for building equipment which is measurably better.

Again Pat is right that S/PDIF is crap. Personally, I've concluded that avoiding S/PDIF altogether is "a" way of avoiding the need to worry about it. I use the SB3 analogue outs at the mo' (runs for cover) and I would seek improvements by moving to a better performing all-in-one player e.g. Transporter. (Again, that's why I do my bass EQ using Inguz software, avoiding a S/PDIF connection to an EQ box.) Having said that, some external DAC manufacturers claim their DACs are 'jitter proof' as far as THD+N is concerned so it appears that jitter might not be the problem some people suggest? That may be sparking off a whole new argument.
Darren

[pbrstreetgang]

Pat is ART? If so I dont see any source components on his site.

[brj]

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.

Agreed. I don't think anyone argues that reducing jitter (data-correlated or not) is a bad thing.

Same as reducing physical vibration of equipment is no bad thing. Same as filtering your mains is no bad thing.

The question is (a) are the differences audible at all and (b) if so how much bang-for-back can I get by addressing it?

I think the key is to realize that the answer is different for everyone.  If you're listening to a boombox in a bomb shelter, jitter is probably not the weakest link in your audio chain and any effort spent worrying about it definitely leans toward the "gilding the lily" side of the scale.  If, however, you have a highly resolving system in a dedicated, well-treated room, then jitter may be the biggest problem you have left, and any effort spent addressing it will show significant returns.  As long as you're happy with your system, it really doesn't matter.  The problem shows up when the parties at either end of this spectrum don't acknowledge that their respective "weakest links" may be very different and then make generalized absolute statements that reflect their unique situation or perspective.

(And realize also that a how much "bang" you require from each "buck" is unique to every individual as well.  Some people are lucky enough to consider "bang" exclusive of "bucks"... and then there are those of us that squeeze each buck until it screams! )

Personally, I think that room acoustics are right up there at the top of the scale in terms of audio problems, if not the number one issue to resolve.  Jitter is definitely on my personal list of issues to address, however, and I keep that in mind whenever I consider components.  To each their own...

[art]

Either way, I applaud Pat for building equipment which is measurably better.

Again Pat is right that S/PDIF is crap. Personally, I've concluded that avoiding S/PDIF altogether is "a" way of avoiding the need to worry about it. I use the SB3 analogue outs at the mo' (runs for cover) and I would seek improvements by moving to a better performing all-in-one player e.g. Transporter. (Again, that's why I do my bass EQ using Inguz software, avoiding a S/PDIF connection to an EQ box.) Having said that, some external DAC manufacturers claim their DACs are 'jitter proof' as far as THD+N is concerned so it appears that jitter might not be the problem some people suggest? That may be sparking off a whole new argument.
Darren

Uh.....correction......we no longer build "source components". Used to, about 15 years ago. Nope, I am just trying to educate you guys in hopes that some day down the road it may lead to a better understanding, and maybe, just maybe, buy whatever crap we may be making at that time.

(I would rather do consulting than manufacture.)

Only, of course, it won't really be crap. Crap is just such a handy word............

Anyway, Mr. Brian:

Your SB3 does have quite a bit of jitter. Easy to solve. Just takes space. The SB3 has little to spare.

(Check back with me later next week.........!)

Trying to equate jitter to THD + N measurements........I don't think so. Some folks like the J-test that Stereophile uses. It has severe limitations, wrt to detectable level. Not to mention that it was  intended to only look for SPDIF artifacts. So, some look at it as "distortion measurement", just because some magazine has editors from the other side of The Pond, which have a penchant for stuff that comes from there, and so they use it. Hey, makes for great copy!. Meter-reading tweak-o-philes can compare meaningless data to justify whatever "dogma-of-the-month" they subscribe to.

(I don't know of any serious US manufacturers who place any faith in the J-test. Sure, may be some. I don't know them, and they are probably the better off.)

Pat

[Daygloworange]

There are 2 things that are hard to grasp about jitter.

First, is what kinds are harmful, and which ones are more benign. That requires a careful examination of all the possible causes, which may be too numerous to discuss here. So, let's try to stick to the concept of what it actually is.

The best analogy that I can give, for all you vinyl-philes, is the difference between wow and flutter. One is a very low frequency modulation of the playback material, the other a much higher frequency rate of modulation.

The wow is easy to pick out: you hear a note bending up and down, just like a Doppler effect. But you also have to make note of how much it bends. One, the rate at which the pitch changes (frequency modulation rate), and the other is how far the pitch bends up and down (the amplitude of the modulation rate.)

So, I hope you can understand that jitter has a similar effect. We are modulating a playback note, by a certain frequency rate and amount. You have to know both to quantify it!!!!!!!!!!!

A very slow wow rate might not be discernible, if the pitch bends very slowly. But if you increase the amount, then it may stand out. Same rate of change, just more of it.

Likewise, if the rate is a bit faster, it may stick out like a sore thumb. No matter how much the amplitude of the change.

So, in not-so-layman terms, we now have a form of frequency modulation. But, in order to quantify it, the amplitude and rate must both be known. You can not just say "Well, we have sidebands @ 1 kHz, because it is changing at that rate." Nope, you have to know how much it is changing.

So, the article may have simplified things, to a degree. (It is intended for layman, so I see nothing wrong with that.)

Thanks for elaborating Pat. I'm understanding it more now. The wow and flutter analogy I'm very familiar with from years of working with analog multi-track reel to reel recorders.

Also, in the analog (tape) domain you have the added phenomenon of high frequency aliasing, which can be quite degrading to sonics as well.

It seems there's no escaping the fact that in any transfer function, you have non linearities that appear, be it digital, or analog.

Random jitter is not that harmful. Depending on the rate and amplitude. Data correlated is always harmful.

If it is random, then that would make it way harder to detect. Same thing for rate and amplitude, small amplitude, long rate would also make it hard to hear. I agree.

The way I remember digital recording process being explained (back in the late 80's, early 90's when I first got heavily into digital sampling) was about bit rate and sample rate. The illustrations they used were about plot points on a sinewave. The higher the sample rate, and the bit rate (word length), the higher the resolution, and cleaner plotting of the waveform.

Based on that, one bit of (wrong) binary would be the equivalent of one pixel being the wrong shade in a high mega-pixel resolution digital picture. When you are ZOOMED right in on it, yeah, you can see the pixel in question, but ZOOM out to a normal size, and you'll never spot it. It's all about context.

I think that's where a lot of paranoia of digital being a harsh, zig zaggy thing stems from. Just a lack of understanding of bit rate, sample rate, aliasing, filtering, dithering, Nyquist etc...

Most people can't get the image of that stair step (3 bit illustration) image of a digital waveform out of their minds.  

If it is random, then it would be very hard to see. If there was a fixed or stable pattern of error, one could see the effects of it more easily.

It does leave me wonder, how much of this jitter issue is purely academic.    

I'm not saying it is, as I've heard at least one whopping example of a transport that was feeding a DAC that was the most horrible lifeless thing I've ever experienced. When the transport feeding the DAC was swapped, it sounded a lot better.

Night and Day, no 2 ways about it.

Personally, I think that room acoustics are right up there at the top of the scale in terms of audio problems, if not the number one issue to resolve.  Jitter is definitely on my personal list of issues to address, however, and I keep that in mind whenever I consider components.  To each their own...

Bingo.    Speakers and room acoustics go hand in hand, and are most definitely at the top of the list. Lowering the ambient noise floor is second. Everything comes after that if you want to properly evaluate anything in audio.

Jitter to me (at least at this point in my understanding of it) would be the equivalent of a transformer in my amp that I can only hear buzzing when there is no music on and my ear is 6" away from it.

The big yields I've had in my setup lately have been better power supplies to my front end components.

If I could eradicate jitter, yes, of course I would, but as it stands now, I feel that it might not be high on my priority list.

Cheers

[Daygloworange]

Your SB3 does have quite a bit of jitter. Easy to solve. Just takes space. The SB3 has little to spare.

I'd like to hear more about that Pat. I'll be yanking the guts of my modded SB 3 and putting them in a standalone chassis. So, no space constraints. 

Cheers

[art]

Personally, I think that room acoustics are right up there at the top of the scale in terms of audio problems, if not the number one issue to resolve.  Jitter is definitely on my personal list of issues to address, however, and I keep that in mind whenever I consider components.  To each their own...

As an end user, it will surely be easier for you to address that issue. Trying to eliminate jitter isn't something I would advise the average listener to attempt.

It is your listening room, and you should take steps to improve it. Don't ask me how..........I just design electronics.

Buying quality gear, that takes steps to fix it, yes, that should be encouraged.

Pat

[miklorsmith]

Lessloss thinks two-way clock signals (out from transport to DAC AND out from DAC to transport) riding the s/pdif cable is a major problem.  The clock frequencies do not mesh which further complicates things.  Their solution is to have the DAC clock output via separate cable to a transport with disabled clock.

Data stream only out the transport, clock stream only in.  It's an elegant solution which works quite well to my ears.

[art]

I'd like to hear more about that Pat. I'll be yanking the guts of my modded SB 3 and putting them in a standalone chassis. So, no space constraints. 

Cheers

No so fast!!!!!!!!!!

Fixing it would have been a lot easier in the design phase. There is a very small via that I need access to. That means taking out the inverter chip, where the oscillator is.

OK, I can take it out, and put it somewhere else. But where? The only way to move it would be to make the wires to and from it a bit too long. EMI City, Arizona. So, I need to see how bad.

Actually, what I have in mind (for the SPDIF crowd) leaves the chip in place. Yes, I will have to add some wires, but what I need to move and why hopefully will not be as bad as moving the entire chip. Like I said......if there was about a 1" x 1.5" space where the DAC chips are, from the git-go, no problema!

So, while space my not seem to be a problem, it is. But stay tuned anyway.

Pat