clock jitter

Discussion in 'Converters / Interfaces' started by audiokid, Apr 18, 2010.

  • AT5047

    The New AT5047 Premier Studio Microphone Purity Transformed

  1. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    Sorry I haven't been around for a while. I'm sure you realise that we all get busy from time to time.

    So, the story so far is that we can see that the absolute worst case (when correctly calculated!) with the example I have given is
    -56dBFS. This is pretty awful but let's look a bit closer at my assumptions.

    1. With a full scale 20kHz signal the -56 dBFS level will only occur at zero crossing. Using the same formula the noise level will actually be zero (minus infinity dBFS) at the peaks (as Cos90 = 0). Clearly the "real" average figure would be somewhere between the two so things will not be as bad as it seems.

    2. A full scale 20 kHz signal is not going to occur in "real" music. There is no instrument that produces a fundamental at this frequency, only harmonics. By nature, harmonics are lower in level than the fundamental and hence the chances of seeing this "worst case" are significantly reduced. Where we will get an effect is with fast transients (most common with instruments such as cymbals and some other percussion). By their nature these fast transients are almost one off events hence the effect of jitter will not be a constant noise on the signal but more accurately a tiny timing error on the transient more akin to shifting a mic a fraction of an inch!

    So what we can deduce is that we are never going to get a "real" jitter noise problem that produces a continuous -56 dBFS problem. We can also deduce that in real terms any "problem" is going to be small but possibly audible. Audibility will certainly depend on the source material. We can also see that this problem will only occur on any soundcard slaved to an external timing reference. If you only use a single sound card, you really should not see a problem.

    For this reason with my own set up I normally have my 01V as reference and only use the 8 ADAT channels to the MOTU, so all inputs use a none slaved reference. On the odd occasions I need more input channels I am forced to use both sound cards. In these instances I usually set up all the drums and overheads on the 01V and carefully select less harmonically intense instruments on the MOTU. Things like bass guitar and mic'd rhythm and lead guitars (bearing in mind a 10 or 12 inch speaker is never going to reproduce much above 5 or 10 kHz).

    Using these methods I have never had any noticeable problem from jitter noise. I'm not saying it's not there and it may be possible to get better results but I have never had anyone point out anything other than "what a clean recording".

    As Boswell pointed out my (huge) 25ns jitter, I thought I'd also comment about this as it goes back to my points about the useless way most jitter spec's are presented. I also feel that these figures will be quite typical of many of the units of similar age. What I have not tried using is the S/PDIF I/O for clocks but I doubt this would help as both ADAT and S/PDIF clocking requires a clock recovery circuit to drive the PLL's. These are not always 100% reliable (as evidenced by the odd errors with the MOTU - see the earlier photo's). It would no doubt be better to use a wordclock interface but unfortunately the 01V has no such interface. The advantage of the wordclock system is that there is no need for any clock recovery circuit and we have a "direct" signal to lock the PLL. If you can use wordclock I/O with your system, I would always recommend it's use over ADAT or S/PDIF.

    Without accessing the internals of my soundcards I cannot actually measure the bandwidth of the jitter noise but as I pointed out with the photo's, the MOTU appears to have predominantly low frequency noise whereas the 01V appears to have a much wider bandwidth. This alone tells me something of the nature of the PLL designs and consequently I will now always lock the MOTU to the 01V. I had always done this previously but for no good reason.

    Why will I do this? Well, as indicated in the spec's I gave for soapfloats system, the bandwidth of the jitter noise will only affect our system when the noise is within the audible range. This is because the modulation due to jitter will create noise related to the jitter bandwidth. If the jitter noise bandwidth is outside the audible range, then any modulation will also be outside the audible range.

    Finally it is also worth producing some "worst case" figures for systems with better jitter noise. Say we have a system with jitter noise of 2.5 ns. This is 10 times better than my system and I would seriously doubt if any soundcards exceed this figure. The same "worst case" analysis gives us:

    125000 V/s * 2.5 ns = .3142 uV which is -76 dBFS. This 10 times improvement in jitter gives us 20 dB improvement in worst case performance. Naturally this is well worth having but it is still not perfect, a further tenfold improvement would give us -96 dBFS with 250 ps of total jitter. I cannot believe this is going to be achieved so the bandwidth of the jitter noise is always going to be a factor.

    The bottom line is that the PLL of the slaved soundcard needs to have a very low bandwidth (less than 20 Hz to keep it inaudible) and this is not easy to achieve when a clock recovery circuit is in use. Using wordclock's should certainly improve on this situation but whether the PLL's of such soundcards are optimised when using wordclock rather than ADAT or S/PDIF is another question.

    I hope this has covered various aspects of clock jitter sufficiently. I do not intend to add to this thread unless of course you have more questions. If anything is left unexplained please let me know and also if anything is incomprehensible with the way I have presented this, again, let me know.

    I will keep an eye on the thread to see if I can help more but for now I hope you have an answer to the original question!
     
  2. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    It has occured to me that I have not really done any sort of summary to these posts. As they have happened over a month or two not all my thoughts have made it into my posts. So I'll try for more of a summary....

    First of all it is very important to realise that the amplitude of any noise arising from clock jitter is entirely dependent on the source material as the noise arises from modulation of the recorded signal. Thus it is definitely NOT a fixed level problem. If there is no signal there is no noise caused by jitter. This is crucial to understand as I have already presented calculations infering worst case noise at -56 dBFS on my own system. More accurately that maximum noise is -56 dB relative to a 20 kHz signal of whatever amplitude. Now I do not know of any musical instruments that produce fundamentals above 5 kHz (with the one exception of some instances of pipe organs) - although I stand to be corrected on this. This would reduce the figure of -56 dB to -68 dB relative to a 5 kHz sine wave. Now for a real instrument there may well be harmonics that slightly increase the rise time which is why I went for the absolute worst case scenario of 20 kHz which is highly unlikely to be produced by any instrument in a none transient manner!

    The problem with this noise is that it is not at all harmonically related to the recorded signal (only the amplitude is) but is directly linked to the characteristics of the jitter noise itself. Just suppose we had jitter noise that comprised solely of 1 kHz sinusoidal phase noise (this is never going to happen - this is just an illustration). In the presence of my full scale 20kHz signal and with the jitter noise at 25ns p-p, then the jitter noise would be a 1 kHz sine wave at -56 dB. Clearly not very good but also a completely contrived example. In reality the jitter noise would be predominantly low frequency (with my MOTU) or essentially pink noise (as with my 01V). In neither case would this be particularly noticeable except that my ageing ears would not be able to hear the 20 kHz signal so I would only hear the noise! The good thing is that with "real" music being presumably audible to mere humans and therefore jitter noise would be a) lower in level and b) significantly masked by the "real" signal. In essence, the annoyance factor could possibly be equated to an amplifier with very poor intermod distortion creating a similar none musically related low level "mush".

    Also it is important to understand that it will only be any slaved sound cards that are going to contribut any significant jitter noise, as the internal clocks of almost any sound card should have inherently low jitter. This may be interesting to those who use an external reference such as a big ben. I would only use this when using multiple soundcards and even then maybe not, depending on the jitter performance of the various soundcards when slaved.

    It is also clear that most modern soundcards/converters are paying more attention to jitter noise so they will perform better than my set up but in fact they will not be so much better as to ignore jitter noise completely. These newer cards also generally seem to focus the jitter outside the audible band. They can only do this by focusing the noise as much as possible in the sub audible band. Thus it is more important than ever to make sure that you use high pass filters set appropriately to the recording. This is standard good practise but will also avoid any build up of low frequency mush arising from jitter noise as well as the other more usual sources.

    I hope this puts a lid on this with a few practical tips on how to minimise any effects you may get from jitter noise.

    As usual, please feel free to comment, criticise, question (or even acclaim!!! LOL) whatever these ramblings mean to you!
     
  3. Boswell

    Boswell Moderator Distinguished Member

    Joined:
    Apr 19, 2006
    Location:
    UK
    Home Page:
    I wanted to say thanks, Mr Ease, for all the time you have spent explaining this tricky subject to a wider audience. It's contributions like yours that make RO stand out among the many audio forum websites.
     
  4. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    My pleasure Boswell. I quickly realised that I had taken on quite a task... Having started though it was important to finish.

    To that end I would appreciate your input if you think there are any issues I have not addressed or not explained properly. As always it is difficult to target this type of thing for the relative layman and I only hope that I have achieved some sort of understanding for such. I thought this was important given audiokid's original question and occasional other commentary (not necessarily on RO) that clearly does not fully understand the subject! I hate misinformation!

    I just wish we could convince the marketing dept's to present jitter data in a meaningful way - I find cycle to cycle measurements are effectively useless in this respect.

    Again, if anyone finds parts of my explanations incomprehensible, please comment and I will try to elaborate - or at least explain more clearly....
     
  5. audiokid

    audiokid Chris Staff

    Joined:
    Mar 20, 2000
    Location:
    BC, Canada
    Home Page:
    You are not only brilliant but so generous to have taken the time to put this all together. Wow. Thank you .
     
  6. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    audiokid, you should never confuse experience with brilliance! :<)

    I'm just happy the effort is appreciated.
     
  7. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    I was just reading back through my ramblings and realised I had never responded to this post. In my "other" life I have had quite a lot of experience with atomic frequency standards. Although atomic standards have extremely high long term stability (hence their use in GPS and many navigation applications), this does not infer any particular advantage when it comes to phase noise (which is directly related to jitter). Both Caesium and Rubidium standards work in the same way. A small phial of gas is externally excited by RF which is within a control loop which maintains a plasma within the gas. This plasma emits light which is detected to complete the loop - if the light goes out you might as well leave the room! This control loop contains an oven stabilised crystal oscillator which provides the usual 10 MHz output.

    In essence what we have is a crystal oscillator locked to an external (but in this case atomic) frequency. Sound familiar? The problem with these loops is that they are dependent on an externally excited and detected phenomenon which depends on the plasma characteristic of the gas and also how well the electronics are designed. What I can say is that, with the measurements I made (many years ago) was that several different Rubidium standards had consistently better phase noise (and hence lower jitter) than that of any of the Caesium standards I measured. With the systems I was involved with back then this actually meant we used Caesium standards for one system (which required the best long term accuracy possible) and Rubidium for another as phase noise was more important than absolute accuracy (although the absolute accuracy of an ovened crystal was quite inadequate).

    Whatever these results is really of little importance when we use one of these standards as an input to our soundcard's PLL as the outcome (in terms of jitter) will be more a function of the soundcard PLL. The only difference will be that your 24 hour track will play back consistently within a few picoseconds rather than a few nanoseconds! How this could possibly affect sound quality is quite beyond me as it is little more than microtuning gone mad!

    My advice is spend the $6000.00 on something useful!
     
  8. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    I honestly can't believe this! Over 1900 views and no questions....

    Either (a) I have explained this far better than I thought (yea right!), (b) I have have been far too technical or (c) no one can be bothered. With so many views I can hardly believe it's the last option and it's certainly not the first so I'm thinking I've failed in trying to put this all into layman's terms.

    Any other thoughts?
     
  9. audiokid

    audiokid Chris Staff

    Joined:
    Mar 20, 2000
    Location:
    BC, Canada
    Home Page:
    Hey... I think I can speak for the majority

    You've cover it like a book and put a lot of time into this. We love you. Personally, my next task is to get this into our wiki so it doesn't get missed when ever clock jitter is mentioned.

    We will be able to read over this for years to come. Whenever I am wondering about anything to do with this topic.... " sun shine!" RO / including all of the web will know where to look for your generous insight into the world of digital audio. Some may eventually get it all sorted to a point where we could ask in depth questions which I'm sure will come.

    In all honesty, I haven't had a moment to sit down and study this so I can't break it down to questions yet. I know I will be inserting the introduction and then go from there. It will be added to the wiki this way Example:
    Starting with an autolink on the forums, followed by an approx 400 character intro and then into the meat. That's the next step for me. At that point, I will then be able to enjoy, absorb more and get it out to all the readers here via the wiki autolink.

    big hug from me, thats for sure!!!

    "your" entire contribution needs to replace this brief blurb into a full wiki form:

    http://recording.org/showwiki.php?title=clock+jitter
     
  10. soapfloats

    soapfloats Well-Known Member

    Joined:
    Aug 28, 2008
    Location:
    Cincinnati, OH
    Home Page:
    I echo audiokid's thoughts:

    You've covered this quite well - so well in fact, it's going to take some time for us laymen to digest it all.
    FWIW, I've learned to be less concerned w/ jitter given your treatise and my (new) setup.

    Your time and effort spent explaining this topic are much appreciated, even if under-appreciated.
    I've followed along from beginning to end, and while some has been over my head, it's been a fantastic read, and well worth everyone's time, IMHO.

    In the end, I too will refer back to this thread whenever clock jitter is a concern, or even a topic of discussion.
     
  11. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    Thanks for the responses and appreciation. I didn't post to "feel the love" as appreciation has already been expressed but I was honestly surprised that with so many views there had been no questions. I guess I was right in that it got a bit too technical!

    I hope that with a bit of time or study it should become clear but please let me know which bits went "over your head" and I'll try to explain again.
     
  12. audiokid

    audiokid Chris Staff

    Joined:
    Mar 20, 2000
    Location:
    BC, Canada
    Home Page:
    I have one for you.

    Does poor power effect clock jitter in a mastering environment, or high end audio?
     
  13. Link555

    Link555 Well-Known Member

    Joined:
    Mar 31, 2007
    Location:
    North Vancouver
    The devices power supply has to be designed properly to handle a range of power issues, under voltage, over voltage, transient suppression, surge suppression etc...
    As for clock jitter, if your power supply can not maintain proper power rails to your clock circuitry then yes. But this is true for any circuit on those power rails. The power supply design is often under appreciated, but in high end audio a robust power supply is essential.... my two bits...
     
  14. Boswell

    Boswell Moderator Distinguished Member

    Joined:
    Apr 19, 2006
    Location:
    UK
    Home Page:
    I would back what Link has said. I would also add that a recent bane of the lives of location sound recordists is the design and quality of the new wave of laptop power supplies. These horror devices are built to the absolute minimum requirements, and spray switching noise throughout the environment they are connected to, both through the laptop to the audio interface and also back out though the mains. In the majority of cases, the output voltage/current/connector is proprietary, and the manufacturers do not offer any way to spend a little more on a power unit to get a proper one.

    It's one of the reasons that I use a rack of Alesis HD24XRs for location recording rather than a laptop with interface.

    I know this appears to have little to do directly with clock jitter, but it illustrates the point that you are not going to get high-quality clocking where power units pollute the precision electrical circuits with un-related garbage.
     
  15. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    It is much as Link and Boswell have said. I pointed out in post #14 that it was important that good power supplies were essential to good oscillator performance. "Dirty" mains supplies can have all sorts of noise on them and while a PSU in a soundcard could perform well for general noise it could also be awful at handling "spikes" on the mains. A good PSU should cater for all that the mains can send to your power outlet!

    Basically anything that can disturb the normal operation of the oscillator or PLL is bad news and can certainly cause problems with clock jitter. However this only applies if the converters PSU is ill conceived. As Boswell stated much earlier in the thread, it is impossible to provide a general rule for jitter performance and the performance of individual units with respect to jitter would have to be seperately analysed.
     
  16. Boswell

    Boswell Moderator Distinguished Member

    Joined:
    Apr 19, 2006
    Location:
    UK
    Home Page:
    I have for some time been meaning to give a link to this application note from the chip manufacturer Maxim Random Noise Contribution to Timing Jitter—Theory and Practice
     
  17. audiokid

    audiokid Chris Staff

    Joined:
    Mar 20, 2000
    Location:
    BC, Canada
    Home Page:
    man, deep topic. I see this is a very important when we enter very high end audio. I also see the importance of very good power and conditioners that work. Most of this is over my head but I get it. Clean power is critical area in high end digital audio and mastering?

    Good converters vs poor converters must deal with how they filter and are effected with the power.
     
  18. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    Boswell, I haven't seen that particular app note before and it is certainly one of the better ones. Thanks for posting it here, very worthwhile. For the more mathematically adept readers it is nice for me that it covers almost exactly what I tried to in the non technical way - which I suppose is to be expected! The best bit to me though is that it does not talk of cycle to cycle jitter, probably because it is written from the RF perspective. Sorry folks if I'm like a dog with a bone on this! ;)

    audiokid, yes of course, good clean power is always the best starting point and should be a must in a studio. Unfortunately you cannot always record "in house" (or maybe you can!) so it should always be incumbent on the designer (of whatever equipment) to try and deal with any mains borne noise within the box. This should be regarded as "de rigeur" on top end gear but, as usual, things like this adds an expense that becomes an easy saving on budget gear when they do little for the spec. sheets. While that may be a good general rule there will always be exceptions in either case. Caveat Emptor as usual...
     
  19. jasonthomas

    jasonthomas Guest

    You hear that expression used in top of the line Blu-ray players now. That they have improved clocks to eliminate clock jitter.
     
  20. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    Clock jitter affects any product that uses either A-D or D-A converters so of course it affects Blu-ray just as it does DVD, CD etc. etc. While they may claim to have improved clocks, it certainly will not ELIMINATE clock jitter. Read through my posts and you will see why.
     
  • AT5047

    The New AT5047 Premier Studio Microphone Purity Transformed

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice