So, what exactly is jitter? Where does it come from, and why is it so bad? How do you tell when you have too much?
Well, let's deal with the last bit first. As an RF engineer, I know how I would measure it. But since I don't publish a pretentious magazine, designed to create confusion amoung the laymen (and extolling myself), I have a hard time imaging why someone would publish something that is nowt more than gibberish. I have no idea what the magazines are measuring, and claiming that it is a metric of jitter. They appear to be using a test that was originally designed to measure the effects of jitter in SPDIF clocks. It makes use of the 292 Hz (or is it 262 Hz? I forget........) sub-code in the SPDIF signal. This technique is flawed in more than one way. First off, its limitations are too high to be of real use in a well-designed system. On top of that, using it to measure what comes out the back-end of a CD player makes no sense at all. -There is no SPDIF signal in a CD player, so there is no sub-code to muck things up-. Yet, its proponents see nothing wrong with that. It gives some print-out on some measuring gizmo, and since it looks impressive to layman, it supposed to be valid. About as valid as saying "I have a solid-state amp, with nested differential feedback loops, and it measures 0.00383% distortion!" Only to find out that people like the tube amp with only 0.8943% distortion more. So, whatever they are publishing: forget it. Useless numbers designed to impress some and confuse the rest. "OK, bub, how do you measure jitter?" Well, I am an RF engineer. We like to muck about with concepts as phase noise (of oscillators). "What is that? 'Splain it to us so that we can understand." I was fixin' to........don't be so impatient. Hobbyists............oy vey. OK, imagine the ideal oscillator as being some magical device that produces a pure signal at one, and only one frequency. And that if you looked at it on a spectrum analyser, you would find that it has a signal that is infinitely thin. Which it isn't. If you look, you will see that if you go out, say.......10 Hz to either side, that you will find some energy that is, say..........90 dB down from the main signal. Go out to 100 Hz, you may find that it drops off to -120 dB down. We would use the expression as "-90 dBc" or "-120 dBc" Which means that is down 90 (or 120) dB down from the *c*arrier. So, that is how dBc came about; db down from the carrier. "So, how do you really measure this stuff?" Hard to 'splain in 25 words or less, but......... Go to the website of an outfit called Wenzel. They make oscillators, and have lots of good info on stuff like this. Including how to build a phase-noise measuring thingamajig. (No, I am not going to dig up the URL. I have work to do. If you are too lazy........well.......fill in the blanks yourself.) Anyway, once you get this *spectral distribution* of the phase noise, you can progress. You can use some math (or s/w) to give you the amount of jitter, expressed in pSec or nSec (if it is really noisy). Now, here is what a lot of folks fail to grasp: *You have to know the carrier frequency in order to know what the jitter amount is.* This is critical, as 10 pSec at 10 MHz may be decent, but 10 pSec at 150 MHz isn't. "Uh, and why is this so important?" Because you have to know that in order to know if something is truly low jitter or not! Look, I got into a food fight with some self-professed know-it-all on some other forum (who may actually be knowledgeable his field, I dunno), but had no clue about jitter. He swore up and down that no one can hear jitter less that 2 uSec. Well, I know that he is wrong. But he had an AES paper to show it! Well, something to look at and see where his head was at. Oh, look-ee here.............they are measuring the word clock jitter. Oh, jitter at Fs, or 44.1 kHz. Gee. Manipulating their 2 uSec into the world of 11.2896 MHz (or 256 Fs), I get.....well, I forget, but it was a few pSec or so. Gee.......supported my claim that you need a clock that had jitter numbers in the single pSec range. So, trying to compare apples to oranges, and calling them bananas doesn't work. (Of course, once I pointed this all out to him, he ignored me and continued on his misguided rant. And some of you guys wonder why folks from industry hate these forums?) Anyway, whenever I speak of jitter, I am referring to jitter at the bit rate. Usually 256 Fs. Could be 192, 384, or even 512 Fs. But let's assume 256 Fs unless specified otherwise. "So, how did you decide that the jitter had to be that low?" Years of futzing around. Making various measurements, and comparing them to other data. Talking to others in industry who have done similar work. "So, how do we tell if something is low jitter or not?" Well, unless you are using a well-known and accepted metric, and know exactly what it is that is being measured, then you are best off ignoring it totally. -Especially what you see in the magazines.- I already told you what I think of that gibberish. "Well, suppose that I want to upgrade my (insert name of device you want to open up and risk ruining) by putting a better clock in. How can I tell if it really will be better?" Hard to say. The guys who make clocks will state "Jitter: < 5 pSec." Look in the fine print, and you will see "at frequency >1 kHz." Well, that is of little use. How much jitter does it have below 1 kHz? In terms of Gaussian noise (centred about the carrier), you should be more concerned about how much noise there is closer to the carrier. But, they usually don't tell you. Because it will be a much higher number, and some bone-headed purchasing agent in some failing telecom outfit will automatically buy the one with the lower number (regardless of how it is spec'ed), the guys who make these always fudge what they publish and what they omit. Do you blame them? Besides, how low it measures is also a function of the supply. Yep, ya gotta have a low-noise supply to get low jitter numbers. And since they have *absolutely no control over what you will use*, and will likely get worse numbers...............well, imagine that purchasing agent who lets the incoming QC guys test them with a '7805 and the jitter numbers are worse.........ya, they gotta omit that stuff as well. "So, are you telling us that......." I'm tellin' you guys that it won't matter if you buy a high-$ clock (sold by someone that I might know) or something from Digi-key if all you are going to do is stick a 3-terminal regulator on it. In order to have low-jitter, you must have the following: A good crystal. (Most of the ones Digi-key sells are as good as what you have now. Or what Brand X uses in theirs.) A good oscillator circuit. Not hard to do, but easy to muck up. A really quiet supply to power it, and lastly...... A good cock distribution scheme. Look, if all that you are going to do is shove it into the crystal spigot of the filter chip in your CD player (like some people who may sell high-$ clocks say and who may be friends of mine), then save your time. If doing that only degraded its performance by only 10 dB, that would be good. Probably more like 20 dB. Sticking the clock into the filter chip is perhaps the worse thing that you can possibly do. "More jitter?" Yes, substantially more jitter. "Why?" Well........one supply pin......one ground pin.......lots of stuff going on inside....recipe for lots of jitter. "Who does it this way?" Oh, about 99% of the CD players on the market. The 1% or so who don't are the expensive ones that the Japanese make, with pretensions of being of being high-end. Instead, they just stick into a hex inverter. That does more than one function. Same problem, only much less so. "Hey, isn't that what they do on the Duet.........." Sssh. Quiet. Don't want to upset The Boss. He may be reading........... "But........." Hey, what do you guys want for $150? C'mon, ya gotta draw the line somewhere. Or else make something that costs $2000. Your choice. "Well, aren't you going to give us some pointers? No! This is about what jitter is, not how to get rid of it. I charge $$ for that sort of info. (Gotta make a living somehow.) Maybe some day in the DIY section. Uh, oh.........post is too long......have to make 2 posts........ -- ar-t http://www.analogresearch-technology.net ------------------------------------------------------------------------ ar-t's Profile: http://forums.slimdevices.com/member.php?userid=13619 View this thread: http://forums.slimdevices.com/showthread.php?t=56425 _______________________________________________ audiophiles mailing list [email protected] http://lists.slimdevices.com/lists/listinfo/audiophiles
