clock jitter

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

  1. djmukilteo

    djmukilteo Well-Known Member

    Joined:
    Nov 23, 2008
    Location:
    Rainy Roads WA USA
    Still just digital data being clocked into a digital converter chip nothing that special!
     
  2. Boswell

    Boswell Moderator Well-Known Member

    Joined:
    Apr 19, 2006
    Location:
    UK
    Home Page:
    You miss the point that Mr Ease is making: in audio A-D and D-A conversion, the exact timing of the conversion between analog and digital domains determines the sampling jitter and is a dominant factor in the accuracy of the process. Whether the serial digital data is transmitted and received without bit errors is a separate and less demanding issue in this context, but is what the web seminar was mainly concerned with, albeit at much higher frequencies that are found in audio devices.

    Audio converter designs usually employ a divided down version of the over-sampling clock as the data stream clock, but this is not an absolute requirement. Instrumentation converters commonly use independent clocks for sampling and for clocking the serial data stream.
     
  3. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    Thank you Boswell!
     
  4. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    BTW, if you think that high speed digital data transmission or low jitter clock recovery is just a matter of connecting a few IC's then you are hugely underestimating the engineering required.
     
  5. djmukilteo

    djmukilteo Well-Known Member

    Joined:
    Nov 23, 2008
    Location:
    Rainy Roads WA USA
    Circuits that involve digital audio/video systems involving clocking is still a digital transmission issue nothing more or less.
    jitter is just one of many slewing and timing issues....so I think you missed the point and audio is not even close to high speed digital data transmission trust me....in fact the problem is the speed at which it takes for the conversions to take place and the cable design used....and I wasn't looking for an argument over jitter just trying to bring the topic into perspective...
     
  6. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    This is just arguing questionable semantics and I would argue that clock recovery has no part in the transmit/receive process that both conveys and corrupts the clock! If all you are interested in is the data integrity over the transmit/receive link, then I would agree. However, at the receiving end of either an S/PDIF or ADAT transmission, we can use a proprietry clock recovery IC which will output a submultiple of the clock required for the local A-D converters. This clock will need to be cleaned up as much as possible and converted to the required frequency for the A-D converters. This is where the PLL comes in and is what I refer to, together with the IC, as the clock recovery circuit. The PLL performance has nothing at all to do with the integrity of the data transmission.


    Your last comment concerns me as it directly infers that my posts are lacking in perspective. I certainly cannot see why you think this. Let me explain the perspective I perceive I have used throughout this thread in trying to answer the original question, "What is clock jitter in detail?".

    The simple and complete answer would be that clock jitter is a total evaluation of the timing errors in a clock due to both random and systematic noise (pretty much what the seminar discussed!). My perspective is that this would be pretty uninformative for this forum. I have therefore attempted to explain in more detail what, exactly, any clock jitter will mean to the sampling of our precious analog audio. This is what I consider to be the most important aspect to everyone here. If that lacks perspective I do not know why.

    Now you originally commented in post #80 with this,

    Data transmission may be your topic here but it was never mine. What I have been talking about is how clock jitter occurs and the effect it has on the A-D process. Data transmission noise is just one element of the total clock jitter. I think I have already explained above that clock recovery is not part of the data transmission per se, nor is the A-D process which is the main focus of my posts.

    Thank you for pointing this out. This is EXACTLY why I said I thought that the Tektronix seminar was not particularly relevant to the main focus of my posts - although it was you who disagreed. I DID NOT say, though, that it had NO RELEVANCE to the topic. What point did I miss?

    I'm sorry but it is my opinion that it is you, not I, that is lacking perspective on what I have been trying to achieve here. Trust you? No offence but I would rather trust 30+ years of my own experience thank you.
     
  7. djmukilteo

    djmukilteo Well-Known Member

    Joined:
    Nov 23, 2008
    Location:
    Rainy Roads WA USA
    I thought we were talking about clock jitter not "clock recovery" circuits whatever those are....I guess maybe your describing PLL circuitry I don't know maybe something else...
     
  8. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    Whaaat! You don't know what a clock recovery circuit is and yet you have twice implored me to trust you!

    Sorry, I'm just jesting... :<) No offence intended.

    Seriously, I have been doing all of this to try and help understanding of what clock jitter really means to our audio and clock recovery is a very necessary and probably the most significant part of the overall clock jitter picture when a soundcard is externally referenced. It seems you have misunderstood the impact the data transmissions we use (i.e. S/PDIF, AES/EBU or ADAT links) will have. I did a quick check on Wikipedia. To save my time I suggest you look up both "S/PDIF" (look at the biphase modulation) and "Clock Recovery" and you should get the basics of why we need clock recovery.....

    So yes, I can confirm we are definitely talking about clock jitter and in particular its effects on the A-D process. If we are slaving a soundcard to an external clock then the "data transmission" aspect is only a small (almost negligible) part of the overall clock jitter picture. The quality of the original source clock and the performance of the clock recovery circuit (including the PLL) will have a far more significant effect.
     
  9. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    It occured to me that we already have an example of what can happen when there are problems with the logic of clock recovery circuits. In post #33 (on page 4) the last photo shows a blip in the clock jitter. I am quite convinced that this most likely to be due to timing or logic errors in the clock recovery circuits of the MOTU. Quite a clear example of the effects that clock recovery can have on clock jitter and certainly not something caused by the data link between the two soundcards.

    It is interesting to note that the MOTU does not appear to use an "off the shelf" clock recovery chip but use their own logic embedded in a Xilinx FPGA. I would hope they have fixed this in subsequent versions of their products. Sadly I doubt they will be offering a firmware update for my older gear to fix the problem.
     
  10. djmukilteo

    djmukilteo Well-Known Member

    Joined:
    Nov 23, 2008
    Location:
    Rainy Roads WA USA
    Sorry Mr Ease if I sounded like I was being arrogant or ignorant on your thread.
    Probably a bit of both.....this is a great topic...and I was not trying to derail your explanations in the least. Sometimes I get passionate about stuff I'm interested in and I know if we were sitting around a table discussing this face to face it would come across totally different...

    The thing I've been trying to understand goes back to my post regarding atomic clocks and clock sources that make the audio "sound better" and my skepticism on how or where clock generation, clock sync, word clock and timing systems being designed today affect the analog sound in the bandwidths were talking about.

    I feel from a basic perspective we are dealing with two locations at which analog (linear waveforms) are digitized....going into an A/D and coming out a D/A.
    I'm confused about how the performance of these conversion which is based solely on internal clock generation being derived from a crystal source. If jitter is a form of timing error offset and we employ clock recovery or clock stabilization circuitry to prevent or minimize this....then isn't that already taken care of as part of the A/D or D/A designs themselves?
    If we are talking about synchronizing multiple electronic circuit devices using a transmission medium (i.e. cable, fiber optic, coax) then I would think we are talking about something entirely different (maybe I'm wrong) and we are no longer talking about linear waveform conversions....we are talking about data transmission....Am I all wrong here and if so please feel free to beat up side the head...LOL
    Cheers
     
  11. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    LONG POST WARNING!

    Ok, let me try and address your questions as it seems my earlier explanations have not fully hit home. Say we have a multi-channel A-D and D-A soundcard , something like my MOTU828. Internally this will utilise several A-D and D-A IC's and these will all need the same clock reference. This is not part of the chip itself but is, if you like, an internal master oscillator. As this will usually be a crystal controlled oscillator it should be perfectly OK in stand alone mode (i.e. not externally sync'd). Of course it is possible to get even this wrong but let's assume the designer knows what he is doing. As a side note I have come across many instances over the years where this is not the case, mainly with predominantly "logic" designers.

    This clock is distributed to all the chips that need the reference. This is an area that must also be done carefully as it is possible for noise to be picked up on PCB tracks from adjacent tracks with either fast logic transients or high currents. Again we must assume that the designer has done his job properly. From the various companies that offer post market "improvements" we could perhaps deduce that the original designs are not always the best. Bear in mind that these companies also offer mod's to the analog side. What is quite easy to find out is the specification of the particular A-D and D-A IC's being used. The specifications of these chips determine the best we will ever achieve when we use them as there is no scope to improve on these figures. There is however plenty of scope to throw away this performance. Such is life, we choose the soundcard that we feel offers us the best bang for our buck.

    So in answer to your first question, if you regard the A-D and D-A designs as the entire soundcard then it is taken care of with the above caveats. If you regard the A-D and D-A design as the chip itself then no, it is entirely your responsibility to make sure the clocks are good and the analog design is top class.

    When we talk about synchronising multiple soundcards then we have to use one of the soundcards as master or use an external master clock. This is usually done via the usual digital interfaces (or wordclock) which all distribute the master clock reference at the much lower wordclock frequency rather than the master oscillator frequency itself. This will be done with either coaxial cable or lightpipe. Now what you need to grasp is that the timing explanation in the Tektronix seminar examines the data transmission at both the transmit and receive end. At the transmit end the slew limitations, timing jitter and reference levels are all generated by the transmitting equipment - irrespective of the bandwidth of the transmission. This is NOT something caused by the connection between the transmitter and receiver. When we examine the data at the receiving end then we are looking at the effect of the cable linking the two. For a long cable and very high speeds then we will see slower transients, ringing lower logic levels etc. depending on the cable characteristics and impedance matching at each end. Suffice it to say that at the speeds we use in digital audio and the typical cable lengths (a few metres) you would be hard pushed to spot the difference at each end of the link! What you will NOT see is an increase in clock jitter at the remote end as the cable characteristics do not change with time...

    Now if we use a wordclock distribution then the wordclock can be used directly for the slaved soundcards PLL but more normally we will be using S/PDIF, ADAT etc. In this case we will need to use a clock recovery circuit to decode the wordclock to be used as a reference for the PLL. The PLL in all cases has the job of reproducing a facsimile of the master unit clock which is locked to the master clock frequency.

    The fact that the master clock is a digital signal and has been distributed via a digital interface actually has no effect on all of the A-D IC's in either the master or slave soundcard. All they see is the clock they are fed with and care not a jot where they came from. What IS important is that the master and slave clocks do not suddenly have excessive jitter. Where might any such jitter arise then? Well there is no reason for the transmitted wordclock master to be degraded from the master oscillator and we will not see a significant increase in jitter from the cable connection. The key elements then are the clock recovery circuit in the slaved soundcard (when using S/PDIF etc.) and the PLL. It is these circuits that will have the most profound effect on clock jitter within the slave soundcard. This is the reason I discounted any sonic improvement being be realised by using an atomic frequency standard - they have no better jitter performance than just a simple, well designed, crystal oscillator and in some cases (i.e. Caesium) they are worse!

    So to finally answer your second question, yes the clock is transferred from soundcard A to soundcard B via a digital link. Is that digital link critical to our A-D and D-A performance? No. The only thing we are interested in is the jitter of the master and slave clocks in order to not degrade the performance of our A-D's and D-A's. The basic perspective here is that the A-D's being used in the master soundcard are being faithfully replicated by the slave soundcard with no degredation. Of course if you slave your PC's onboard sound system to a Linx then the onboard sound system is not suddenly going to sound like the Linx! :<).

    I hope that this lot has not bored the pants off you all and casts another incremental bit of light on this whole process. Once more I am writing this trying my best not to be overtechnical so that you can all understand. Do let me know if I have failed! I will read back through this at leisure and will try and edit any tatty english...
     
  12. rmburrow

    rmburrow Active Member

    Joined:
    Dec 26, 2012
    Location:
    western PA
    Mr. Ease: I enjoy the technical discussions here. Besides cymbals and percussion, there are Fourier components of violin strings up near 20 kHz and theoretically beyond. Ever read Prof. Philip Morse's "Vibration and Sound" and work out some of the examples? This text was written before computers but the proofs in the text are time invariant. The complete solution of the vibrating string is contained in the Morse text as well as other textbooks.
     
  13. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    I'm glad you enjoy the discussions. I am well aware that all "natural" instruments have varying degrees of harmonics and many extend to supersonic frequencies. What is important to appreciate though is that the fundamental notes never exceed a few kHz so all the energy in our recordings above 10kHz are actually harmonics. While these harmonics contribute to the overall sound of each instrument the biggest factor in being able to identify a particular instrument comes from the transients at the start of a note. This is easy to verify by removing the starting transients of notes on various instruments and slowly bringing up the level. When you do this it becomes much, much harder to identify individual instruments.

    As far as this topic goes, what is important is that the level of any harmonic is always considerably lower than the fundamental, even for the extreme cases of square and triangular waves. A far as clock jitter goes this, fortunately, plays in our favour as it limits the amplitude of the very highest frequencies that could cause audible artifacts. :<)
     
  14. audiokid

    audiokid Chris Staff

    Joined:
    Mar 20, 2000
    Location:
    BC, Canada
    Home Page:
    What a stellar thread. Best on the web for us guys.
    Thanks again for taking the time on this one.
     
    kmetal and DonnyThompson like this.
  15. apstrong

    apstrong Active Member

    Joined:
    Feb 13, 2009
    Location:
    Canada
    Don't know if this helps anyone, but in a more practical vein (i.e. what does that sound like?), the folks at Cranesong put together a couple of pages about jitter with downloadable audio samples:

    http://www.cranesong.com/jitter_1.html
     
  16. DonnyThompson

    DonnyThompson Well-Known Member

    Joined:
    Nov 25, 2012
    Location:
    Akron/Cleveland, OH
    Home Page:
    That's a very useful link, AP... thanks for posting it.

    After listening to the 4 recordings of the exact same performance, make sure to check out Page 2 of the link as well, which isolates the jitter of each sample file through the use of phase cancelation of the primary sound sources.

    The result is that you can actually hear what the various forms of jitter sound like. I think you'll be surprised at what you hear, and in how much different they all are from each other.

    Also... check out the "setup" page ( link below), which diagrams the routing used as the "control" foundation through which all the test samples were routed:

    http://www.cranesong.com/JITTER_TEST_SETUP.pdf

    The article is very upfront about the fact that its aim is to educate listeners to what jitter sounds like, as opposed to being a tightly controlled scientific test.
    There are situations where the gain of certain phase cancelled files was increased to make the jitter more audible.
    So, it's not as much a tightly controlled A/B test, as it is a way for people to become educated to what various forms of clock jitter sound like.

    I was amazed at what I heard on Page 2.

    FWIW

    -d.
     
  17. MrEase

    MrEase Active Member

    Joined:
    Jan 19, 2006
    Location:
    West Suss ex, UK
    Wow, it's nearly three years since this thread was active and I see it's now over 16000 views!

    AP, that is indeed an interesting test. However I don't think there will generally be a particular "sound" associated with jitter as it will depend very much on the jitter characteristics. Hence, I doubt if you'd ever be able to hear a particular "sound" and say conclusively "that is due to clock jitter!"

    The reason I say that, is that jitter can arise from many sources and the spectrum of the noise causing the jitter will change the jitter induced artifacts. As I think I mentioned back in the thread somewhere, white noise sourced jitter should be the least intrusive (which should sound like an elevated noise level) but any jitter arising due to a discrete modulation (which does happen) would affect the sound in different ways depending on the modulation frequency and would also be more noticeable.

    Having said all that, I don't have time (or a decent sound set up on this PC) to try the tests for myself. If I get the chance though I'll try to find out more about the jitter they have induced on the tests. A quick read through didn't make it clear to me whether discrete modulation frequencies were used or not.
     
  18. DonnyThompson

    DonnyThompson Well-Known Member

    Joined:
    Nov 25, 2012
    Location:
    Akron/Cleveland, OH
    Home Page:
    I agree. More to the point, and being totally honest, I wasn't able to hear it at all on the samples they provided - until I went to the second page and heard the various types isolated through phase cancellation.

    I think that this is kind of the point to the exercise - that it is hard to recognize, and that it never really "sounds" the same, either.

    It's not really an identifiable sound, something that you could point to and say unequivocally, "Oh, yeah... no doubt, that's " _________" , as would be the case with something like SMPTE Time Code chatter, which is immediately recognizable.

    So... I'm curious - because there was only the one instrument used for the various samples - I think that it begs the question, "so what happens when you have that jitter happening on 24 different tracks in a mix at the same time?"

    It leaves me wondering if - much like noisy preamps can do - it would become more obvious as the various jitter issues would accumulate/stack up, track by track? That would seem to be the case, but I don't know... which is why I'm asking. :)
     
  19. Chris Perra

    Chris Perra Active Member

    Joined:
    Jul 13, 2014
    Location:
    Toronto
    Home Page:
    I dunno listening the the tests I couldn't hear much if any difference between them, The phazed examples were nuts that's a big difference.
    I think in the practical world jitter is just going to become a part of your sound like everything else. Preamps, eq, mics etc.

    Is there any product out there with absolutely no jitter? Any eq or preamp with zero noise? At what point is it measurable but not noticeable?

    I loaded the files into my Daw B is the control the rest are out of phaze with B to show the difference. There's is definitely stuff there and a difference but the level is unhearable unless you crack your speakers or headphones to a crazy level. i can;t see how in the real world that's going to be a factor. Unless you are recording ants walking across the floor I can't see how you'd reach/need that level of gain boost or compression to actually hear it.
     
  20. audiokid

    audiokid Chris Staff

    Joined:
    Mar 20, 2000
    Location:
    BC, Canada
    Home Page:
    I think part of your comment in regards to not hearing anything to be concerned with, may also be something some of us know effects the greater of what isn't realized on the surface. Accumulating ?
    an interesting pov from both you and Donny .

    Could this fall into personal hearing. Examples: loss, ability, awareness on an individual level ... why some of us freak out when someone scratches a black-board, or why some of us can continue working when a baby is screaming.

    Tolerance to freq or less bothered by, does it really matter ? ...
     
  • 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