Jitter Limit

Hi AudioMasters.  

I've read that high resolution audio player such as 24-bit - 96k sampling rate etc.. have to have a low jitter DAC clock or else it could only be equivilent to a 16-bit player.

Does anyone know what sort of jitter specifications are we talking about before HI-RES Audio is no longer HI-RES Audio. (I'm sure there's someone)

Hi AudioMasters.  

I've read that high resolution audio player such as 24-bit - 96k sampling rate etc.. have to have a low jitter DAC clock or else it could only be equivilent to a 16-bit player.

Does anyone know what sort of jitter specifications are we talking about before HI-RES Audio is no longer HI-RES Audio. (I'm sure there's someone)

Since it's not really true, then there isn't really an answer to the question!

24-bit audio will always have a lower noise floor than 16-bit - jitter doesn't come into it. And the situation with the clocks isn't as simple as stated anyway. The really critical bit is the original sampling jitter - what ever you do with the output, you'll never be able to improve on this - if a signal is digitised through a converter with a relatively high random jitter rate, and played back through a converter with a lower rate, then there will always be statistical timing discrepancies at any instant during playback, about which you can do nothing.

It also follows that for any given signal, you can degrade the jitter, and if my some complete miracle (this doesn't happen - it's just a theoretical argument) you could reproduce exactly the same jitter pattern in the DAC as there was in the ADC at the digitising instant, however bad it was, then you'd get the best possible timing performance from the system, regardless of the actual jitter level.

In practice, the actual levels are irrelevant, because you can do nothing about them. What's perhaps more important is to look at the clock generation method, because it is well known that some generators produce lower close-in sideband nose than others, and that's what jitter actually is. I can tell you also that it's bloody difficult to measure acurately, too, requiring a carefully constructed test rig, and a digital RF spectrum analyser becfore you even get close. That's why there's a lot of people around who will talk the talk, but most of them actually have no idea of what this is really about at all, or what difference it actually makes to anything. They certainly can't give you meaningful levels, because they haven't got a clue about what they really are. For a start, the distribution pattern could well be significant, and you've also got to recognise that you are superimposing two different levels onto each other, something which  you will always have to compensate for.

If you increase the jitter level of a signal, what happens is that initially, you tend to get a very slight 'veiling' of the sound of things like reverb. And you'll only hear this on good monitors - very few domestic systems can reproduce this consistently. Also, you'll only notice this in an immediate AB comparison - even if you know what you're listening for, you won't be able to arbitrate with any more than 50% (AKA a guess) accuracy after a few seconds, or in fact with an ABX test. If you progressively increase the jitter, you will get to the point where the signal becomes unlistenable to - but it can always be restored by reclocking, something that the people who are always slamming S/PDIF don't seem capable of understanding.

So it's not possible to give you absolute values of when this becomes audible, because it depends on the program material, and the clock stability of the original digitising process. But if you want an indication, then Martin Walker (SOS reviewer) says that if you listen really carefully with the right material under good conditions, then you can just detect the difference between the internal clock jitter on an Echo Mia, and clocking it with an external low sideband noise external one. But he's not over-confident about it - and I don't blame him for this. Yes, the Mia has a pretty good internal clock, but so do most hi-res playback devices these days. And if you can get an external low-jitter clock to any DAC and reclock it, that's the best you will do - the jitter results you get then can only be ADC clock dependant.

There are far more important factors at work in music reproduction than jitter. When you compare the effects of jitter to the artefacts that are introduced into rather imperfect listening environments by most people, you start to get this in slightly better perspective. Jitter is generally only picked up as an 'issue' by people who have nothing better to do with their time than mouth off about it. Perhaps if they spent a little longer learing about what it's really about, instead of making obfuscational and incorrect statements about it, the world would be a better place.

Or, to put it ever so slightly more simply, don't believe everything you read

Or, to put it ever so slightly more simply, don't believe everything you read

Yes, but if I'd just said that, he would have said 'Why not?', and we'd have wasted even more time...

Or, to put it ever so slightly more simply, don't believe everything you read

Certainly nothing printed in a glossy brochure.

Steve I had a hard time really understanding what you said.

What I think you said is that the ADC and DAC both must have good clocks to appreciate low jitter in DAC.

Is that what you were saying?
Could you please give me a more simple answer since I am a simple man.  

THe book I saw this mentioned is in John Watkinson "An introduction to digital Audio"

[contributes]

IN regard to my first question
I think from the top of my head that John Watkinson was talking about
High Jitter reducing the resolution in ADC's and not in DAC's so I might have been wrong.

So what you were saying steve about no matter how good the DAC is, it don't mean a thing if the ADC clock has poor jitter in the the first place relates to what John said.

Anyway overall what would you say contributes to good soundquality in DAC's or soundcards - is it voltage stability etc...
please list ( I'd like to know all of them)  

Steve I had a hard time really understanding what you said.

What I think you said is that the ADC and DAC both must have good clocks to appreciate low jitter in DAC.

Is that what you were saying?
Could you please give me a more simple answer since I am a simple man.  

THe book I saw this mentioned is in John Watkinson "An introduction to digital Audio"

Yes, this is basically true. You won't get less jitter than that introduced at the ADC stage. And just how simple do you want it? Watkinson's explanations are a lot more detailed than mine! But go back and read page 8 very carefully - I know that he bangs on about interconnects and mixes ADCs and DACs up in the same section without clearly delineating between them, but the early part certainly implies (quite correctly) that it is noise on the clock signal that is effectively jitter. What he doesn't say is that there is a theoretical limit to the amount of noise you can remove, even with a PLL, and this is ultimately what you are left with in the system. Modern ADCs have the PLL built into the converter, but even this approach still has a jitter floor, and this is what ultimately limits the performance.

It is unfortunate, but true, that a lot of early equipment took a few short-cuts with clock signals, and that this has ultimately led to its downfall, and digital audio getting a bad reputation in some quarters, especially with Luddite audiophiles, who pounced on this with glee. But since clock correction techniques have been integrated into more and more kit, this situation has improved dramatically. But ultimately, the noise factor is still determined by whatever provides the PLL reference clock, and there's no getting away from that. Now, low-noise clock generation and PLLs is a seriously complex subject, and you don't need to go into it, but take a careful note of what Wilkinson says at the end of the last paragraph of section 2.8, p40. It is true, and ignored by virtually all audiophiles. (For those without the book, he says that if you can hear a audible difference between different digital interconnects, your equipment is badly designed, and this is absolutely true!)

[contributes]

Anyway overall what would you say contributes to good soundquality in DAC's or soundcards - is it voltage stability etc...
please list ( I'd like to know all of them)  

You'll have to wait for a list - there's quite a lot of factors!

Another observation:

In places, Wilkinson is surprisingly economical with the truth, and this I think may have contributed to your assumption that jitter was a significant issue as far as resolution was concerned. On page 39 he says that "even  small amounts of jitter can degrade the noise performance of a 20-bit converter to a good 16-bit unit". There are a couple of observations to be made about this; the first is that the amounts of jitter on the accompanying graph aren't small at all - they are bloody enourmous! 1nS of jitter isn't even a good figure, never mind the 10nS that he seems to think might be typical of a 16-bit converter. And secondly, he ignores the fact that the resolution of a 20+ bit converter will inevitably be better than a 16-bit one anyway. But you must note that strictly, he's right - but only when talking about noise, not any other considerations.

One thing he doesn't get into at all is related to the fact that PLLs and noise are a seriously complicated issue. I'm not going into detail at all about this, except to say but one thing; and that is that the phase noise of a PLL varies considerably according to how far away from lock it is at any given moment, and that this is a dynamic consideration, and cannot be considered realistically from a linear point of view at all... yes, they will considerably improve things, but absolutely quantitive statements about precisely how much really are a bit of a no-no. The maths behind phase noise and locking performance is, quite frankly, terrifying. And to purchase the literature about it, of which there is very little, is also terrifying - on your wallet. The standard reference work will set you back £150+, and that's just for one bog-standard textbook. I've got it, and fortunately, the maths doesn't change, so I won't ever need a newer version...

But as far as what makes a difference to audio performance goes, there are a number of key issues: Everything in section 2.21 as far as ADC and DAC is concerned, although what  Watkinson emphasises may not be the same factors that  people judging with their ears would choose; also, if they are used, the quality of the anti-aliasing filters is going to make a difference; and the quality of the analog components will clearly become more significant as the bit rate and depth increase. Ultimately it is the analog components, whose behaviour is limited by the Laws of Physics, that will finally determine the ultimate performance, though.

But it is true to say now, designers have realised that rather more of the factors that they thought didn't matter, and that they could get away with design short-cuts in, are actually significant. So over the last few years, most of the arguments about hi-res digital sound quality haven't really been about this at all - they've actually been about analog design, because by and large, the digital problems have been fixed as best they can be. But since analog design is all about compromises anyway, there will always be arguments about this!

Steve, what book are you refering to? I took my "The art of digital audio" by Watkinson and 2.21 is about variable-ratio conversion. Don't think it is that one....

Steve, what book are you refering to? I took my "The art of digital audio" by Watkinson and 2.21 is about variable-ratio conversion. Don't think it is that one....

The book I saw this mentioned is in John Watkinson "An introduction to digital Audio"

Confusing, isn't it? He wrote two books that are not dissimilar, although I think the one that you've got is the more expensive one.

Oh! That may explain a lot. So probably, it is just another chapter.