[plonter]

HI all. I don't know so much about jitter and the whole process of digital data streaming, and from what i learned so far, jitter is when the digital bits don't arrive in time to the dac or something like that...

can you guys explain to me exactly what jitter is (correct me if i was wrong) and how does it work..?
If we take my rig for example...I have a dvd which connected digitally to my dac.. is the jitter created inside the dvd even before it outputs the stream to the dac? and how is jitter created anyway?

I searched on the net but with all the difficult terms and explanations i couldn't managed to understand exactly. If you can explain to me like i am a 5 years old i will appriciate it.

 

[mike1127]

Quote:

Originally Posted by plonter /img/forum/go_quote.gif HI all. I don't know so much about jitter and the whole process of digital data streaming, and from what i learned so far, jitter is when the digital bits don't arrive in time to the dac or something like that... can you guys explain to me exactly what jitter is (correct me if i was wrong) and how does it work..? If we take my rig for example...I have a dvd which connected digitally to my dac.. is the jitter created inside the dvd even before it outputs the stream to the dac? and how is jitter created anyway? I searched on the net but with all the difficult terms and explanations i couldn't managed to understand exactly. If you can explain to me like i am a 5 years old i will appriciate it.

It is very hard to explain this a 5-year-old. But you are smarter than one, I'm sure. Digital signal theory is a mathematical explanation that an analog (purely smooth) waveform can be perfectly reproduced by sampling it at regularly spaced points. So if the smooth analog waveform is this :
....**
...*..*........**
..*....*......*..*
........*....*....*
.........*..*
..........**

Then you can reproduce it by sampling it at just a few points:

.....*
.................
..*...........*..*
........*..........
.............
...........*

In theory, you can perfectly reproduce it. I know, it sounds impossible. It's hard to wrap your mind around, but it is a result of the mathematics.

But this is the theory. In practice, certain things go wrong.

The theory says that these samples have to perfectly spaced in time. They must be evenly spaced or else there will be distortion.

In practice, digital recorders and CD players don't put the samples exactly evenly spaced in time. That's called jitter. I hope it is obvious to you why that would cause distortion. The diagram might look something like:

......*
.................
..*..............*..*
.........*..........
.............
..............*

I hope you can easily imagine that when you connect these dots, which have moved in time, you won't get back the original smooth waveform. Note that "connecting the dots" is not just drawing straight lines, but using the mathematical theory to try to get back the original smooth line. Again don't worry if it sounds utterly mysterious or impossible. It's a result of the math. The math is hard to understand. It gave me headaches in college.

 

[Sherwood]

Mike, that is a really excellent description of jitter. I've never heard better. Bravo!

 

[Pio2001]

Jitter is when the DAC doesn't convert the data exactly at the right time. The weveform is a bit distorded horizontally.

In a CD player, the jitter depends on the clock, and the way its timing signals arrive to the internal DAC.

In a Drive / DAC combo, a given amount of jitter is present in the digital output, but it is small compared to the jitter caused by the cable.
In an optical cable, the electric / optic and optic / electric converters have jitter too.

But this jitter is then completely eliminated by the DAC (For All Practical Purposes, because some detectors may be able to see remainings of the cable jitter), which reclocks the incoming data. The remaining jitter is just the one of the DAC's stabilizer circuitry (called PLL).

I know about two blind tests done with a jitter generator in order to test the DAC sensitivity to the incoming jitter.
Here is the most recent one. Its french. I have no link to the other, that is japanese.
homecinema-fr.com • Voir le sujet - Kangourou ABX 5 - 27/6 - Le JITTER : Tout ce que vous ...
In both tests, no one could hear any effect, even with amounts of jitter more than one hundred times bigger than at a CD Player's output. In the second test, most listeners didn't even start blind testing. They directly agreed that no effect was audible, listening sighted.

In both tests, the jitter was raised until the DAC looses the signal. At about 250 ns in the first test, and about 150 ns in the second.
The first effect heard in the second test were sharp clicks because of data missed by the DAC, with about 100 ns of jitter.

 

[AtomikPi]

Quote:

Originally Posted by Pio2001 /img/forum/go_quote.gif snip

Very interesting tests, and now I will be even more suspicious when some individuals claim that the jitter in even cheap gear of less than 1 ns is significant.

Here's the translation according to google translate

Quote:

Essentially, the jitter generator was necessary for this experiment because we had to climb to 100 nanoseconds to hear something. En dessous (jusqu'à 90ns), on n'entendait strictement aucune différence (ABX impossible pour tous). Below (up to 90ns), it did absolutely no difference (ABX impossible for all). Au delà de 100ns par contre (valeur variant suivant le matériel), ça devenait évident et audible de tous (ABX facile pour tous). Beyond by 100ns against (value varies depending on equipment), it was becoming clear and audible to all (ABX easy for all). Bref, c'est un peu du tout ou rien : soit on entendait vraiment beaucoup de défauts, soit on n'entendait rien. In short, it's a bit all or nothing: either you really meant a lot of defects, or you did nothing. Pour référence, il semble que les valeurs de jitter des équipements courants soient au moins 100 fois plus faibles (à préciser), donc très en dessous de nos valeurs de tests... For reference, it seems that the values of jitter available equipment are at least 100 times lower (specify), so far below our values of tests ... :-? Deux types de défauts ont été entendus : Two types of defects were heard: 1) En poussant le jitter, des "bruits impulsionnels" ou "claquements" apparaissent, bien audibles, puis en poussant encore le DAC commence à vraiment perdre les pédales et on finit par ne plus rien entendre du son source. 1) By pushing the jitter, the "impulse noise" or "slamming" appear, though audible, and then pushing again the DAC begins to really lose the pedals and then not hear any more of its source. Ce type de défaut a été interprété (avis personnel et subjectif) comme une "perte de synchro" avec des valeurs de samples complètement discontinues. This type of fault has been interpreted (personal opinion and subjective) as a "loss of sync" with the values of samples completely broken. Je crois que ça s'appelle du "data jitter". I think it is called "data jitter. On a même entendu des sortes de "répétitions" ou des "ralentissements" de la source avant de la perdre complètement. We even heard some sort of "rehearsal" or "slow" the source before losing it completely. 2) Plus finement, dans la zone des 100 nanosecondes, des défauts de "distorsion" (j'ai bon ?) apparaissaient, toujours aussi évidents mais moins agressifs : ils sont clairement audibles quand la source est à un niveau élevé et beaucoup moins quand elle est à un niveau faible (donc corrélés avec le signal), diminuant nettement plus qu'une diminution classiquement "proportionnelle". 2) More precisely, in the 100 nanoseconds, defects of "distortion" (I am?) Appeared, still aggressive but less obvious: they are clearly audible when the source is at a high level and much less when it is low (thus correlated with the signal), decreasing significantly decrease a classically "proportional". Ce type de défaut a été interprété (toujours de façon personnelle) comme un défaut de timing (un peu en avance ou un peu en retard) de sortie des samples. This type of fault has been interpreted (always so personal) as a default timing (a little ahead or a little late) output samples. Je crois que ça s'appelle du "sampling jitter". I think it is called "sampling jitter.

 

[plonter]

thank you mike for the detailed explanation, and although i didn't quite get it all , i got some idea about it. but i still don't unserstand this: when i play a cd in a cd player right...the player is taking the information from the cd and do what? does it sample it in anyway (or any kind of process), or just outputing it through the digital outputs?

 

[Sherwood]

it just outputs it. The data was already sampled when it was converted to digital. Jitter is the problems with the accuracy with which CD players output ditial data to a DAC.

 

[KingStyles]

Heres a post on jitter.
jitter
Hope this helps.

 

[plonter]

Quote:

Originally Posted by Sherwood /img/forum/go_quote.gif it just outputs it. The data was already sampled when it was converted to digital. Jitter is the problems with the accuracy with which CD players output ditial data to a DAC.

ok...thanks ,that's one thing strait. now that i know that the cd player doesn't do anything to the digital stream i can move farward to my second question:
If a CDP is only outputing the data to the dac,what can be wrong with that?
you mean that the CDP doesn't output exactly the same data like on the cd?
not "bit perfect"? but that's another issue ...that's not what jitter means.
so where all the wrong timing is taking place? and by which side? the dac or the CDP?

 

[plonter]

Quote:

Originally Posted by KingStyles /img/forum/go_quote.gif Heres a post on jitter. jitter Hope this helps.

thanks for the article, I remember that i read it once,but it was in conjuction to digital cables issue, now it is another matter so it is interesting to read it again and give the focus to the jitter issue.


"Technically, playback jitter is the inaccuracy in the timing of the "ticks" of the clock that transfers the samples of digital data into the D/A converter chip. To move data in a digital system from one point to another, it is usually clocked. In order for the D/A conversion to work, a new data word must be presented to the D/A converter periodically, or at a fixed frequency. The system clock or clocks do this. "


this is a quote from this article....can you help me to clear this up? does the CDP "clocks" the digital information in order to send it out to the dac? where is this clock?

let me put it this way...when i transfer a data from one hard disk to another, it is not important that the data will arrive in a certain time, the important thing is that is DOES arrive in the end. What's different between this and a CDP streaming data to the dac? (that means audio)

 

[Currawong]

Ideally, the "on" and "off" digital signals should be as a perfect square wave:

However, in reality the signal is never perfectly on or off, but distorts due to many factors. The below picture is from some or other hi-fi maker's web site showing, supposedly, very low jitter. Notice though that despite that, the wave isn't perfectly square.

 

[WesMiaw]

mike1127's post is a nice explanation but unfortunately incorrect. His example refers to a signal that contains multiple values, but the vast majority of our digital signals are binary. Currawong's illustration is an accurate representation.

A digital receiver samples the input signal at a specific point in time. Usually dictated by its own clock. In Currawong's example if the clock "ticked" before the blue line rose to the required voltage, you'd read it as a zero instead of a one. And vice versa if the clock "ticked" before the blue line fell to the bottom.

Assuming the rise/fall transitions occur at regular intervals, within some tolerance, you can imagine synchronizing your clock and making it "tick" in the middle of those blue plateaus and valleys. In which case any jitter (falling within the tolerance) doesn't matter. Kind of like if you shine a strobe light on a spinning wheel, you can time it so the wheel looks like it isn't spinning.

Likewise, you can see that the voltage takes a bit of time to stabilize and wiggles a bit. So a digital one is usually something lower than the actual max signal voltage, and a digital zero usually higher than the actual min signal voltage. e.g. 3.3V is a one but the signal one is 5V and 1.2V is a zero but the signal zero is 0V.

How do you synchronize your receiving clock to the sending clock? You can do it by identifying the transitions. This is why maintaining transitions is very important in these sort of digital signals. Even if your data is all zeros, you usually use some sort of signal that uses both high and low voltage, with regular transitions, to represent all zeros to the receiver.

When a CD drive is reading the disc, it looks for pits which represent ones or zeros (I can't remember which). As described above, if there are errors in the pits due to manufacturing defects, or if the sample clock is off, or there is a fingerprint, you might not read the pit correctly. That's why CDs also include error correction. In case you read something wrong, you can detect and correct the error.

Once this data is read, a CD player usually converts the values to S/PDIF for digital output over coaxial or optical.

 

[mike1127]

Quote:

Originally Posted by WesMiaw /img/forum/go_quote.gif mike1127's post is a nice explanation but unfortunately incorrect.

"Unfortunately" for you, my explanation is correct.

Quote:

His example refers to a signal that contains multiple values, but the vast majority of our digital signals are binary. Currawong's illustration is an accurate representation.

I was talking about the forest---you and Currawong are talking about the trees. You are talking about a mechanism from which jitter arises. I'm talking about a broader view, in particular the digital theory itself and why jitter affects the analog waveform.

Why don't you learn some digital signal theory before you go "correcting" people?

Your post is in fact useful, but you are talking about a specific mechanism, which there is no theoretical reason we have to use, although it happens to be in common use.

EDIT: actually I just realized you are completely wrong about what jitter means, if I understand what you are trying to say. You seem to think jitter is a mechanism that causes 1's to be read as 0's or vice-versa. That is completely wrong. Jitter happens when the D/A converter places the samples unevenly in time - even though they are the correct samples (i.e. 1's and 0's are right). Also the samples themselves are in fact "multiple values"---specifically 65,536 values (in standard CD) and 1,677,216 values (in SACD).

 

[dvw]

Mike is correct in his explantion. The simple way to explain this. Let's assume you get a tick at every second. So each tick should happen at 1,2,3.... n sec. But if a tick happen at 2.5 sec, then you have a 0.5 second error and this error is called jitter.

 

[mike1127]

Quote:

Originally Posted by Sherwood /img/forum/go_quote.gif Mike, that is a really excellent description of jitter. I've never heard better. Bravo!

Thanks.

One thing that people are often confused about is they say "analog has infinite resolution" while "digital is a fragmented waveform." That's not correct, due to the magic of digital signal theory. D/A convertors produce a smooth waveform which is very, very close to the original. Except for one thing. The output is bandwidth-limited. For people who are interested in pursuing this, I can say simply that limiting the bandwidth of the digital signal to 22 KHz is the trick that makes the mathematical magic possible.