No Script, No Fear, All Opinion
RSS icon Home icon
  • EEVblog #61 – Crystal Oscillator Drift

    Posted on February 14th, 2010 EEVblog 25 comments

    Dave explains the what how and why of measuring crystal clock oscillator drift over time.

    The Clock Drift PCBs

    Real data collected with the sine fit envelope.
    And results showing the same characteristic result over a slow temperature ramp.


    Be Sociable, Share!
     

    25 responses to “EEVblog #61 – Crystal Oscillator Drift” RSS icon

    • Fun. Some of that is due to contamination migrating around on the crystal surfaces–the mass loading causes the frequency to drift. That’s how quartz crystal microbalances work, e.g. the quartz monitors used in thin film deposition systems for measuring film thickness.

      Cheers

      Phil Hobbs

    • dave,

      some nitpicking: under the assumption of a drift linear to time, you just need the slope of the drift curve. in that sense, your drawing of a circular arc between the curve and the time axis is misleading – a straight line down to the axis would be correct.

      also, the drift is a random function (“random walk of a drunken sailor”). In that sense, it’s indeed a brave thing to do just trying to calibrate the error out of the curve.

      The drift could be reduced by using multiple oscillators, assuming that they drift in uncorrelated way (which is a bit of stretch, I admit),

      Also, I wonder how viable it might have been to use a rf signal source as a calibration signal. But then again, water should be much more lossy than air for frequencies >100 kHz.

      • Dogbert,
        It may be possible to encode the pings themselves to help synchronize. If the ping could carry some information about itself (order in sequence), the sensor would only need to measure on the scale of difference between pulse from ship, and reflected pulse. Or (at least) three pings could be used from three locations and make an underwater GPS.

        Of course I’m guessing. Is this possible?
        –Alex

        • Alex,

          the speed of sound in water is limited. with a radio wave, you can assume that it reaches all the autonomous devices instantaneously without having to worry too much about their location relative to the signal source (resp. the time it took for the signal to reach the device). with an ping (an acoustic wavefront in a classical sense), you are probably in a world of pain: the devices have to know their exact location as well as their depth in order to get a good approximation of the signal delay – the location of the sender could potentially be encoded in the signal. But a good engineer follows one golden rule: keep it simple and stupid. It gets complicated enough later on anyway.

          By the way, Dave, nice blog – I’m enjoying it thoroughly!

        • agreed with evan about ping encoding. dave just showed graph for just equal temp. if its in real application with temp. fluctuation, the graph could be more rampant. my conclusion is, for the sake of accuracy, the dave graph is simply a “random” characteristic for my interest, we cant get a perfect device even the Rb rubidium?. the solution might be to synch the timer at a predefined interval, by kiss principle, use the noise/ping generator (of course we can use other generator or GPS some sort like that that will interupt the timer/controller). maybe a 1 minute or several seconds silence for each day for the next 3 months period of generator will showed up in datalog (i remembered the dave’s LED temperature graph .hit by storm, and the graph goes down immediately). that will be able to time skewed later on in post processing. instead of silence signal, maybe the “boat” can generate silence-(very) noise signal, in other word, very distinct and different signal at predefined time for us to recognize later. thats assumed after we have made sound velocity analysis and time to reach each timer, if we have to take into account the nonlinearity of sound wave velocity, ocean ripple, fishes swimming by, then we’ll have alot of headache.

          dave just showed the technique of how manufacturer write their datasheet for their product IMO i think. great idea enhancement video.

    • Dave,

      very interesting topic, again. Thanks!

      I see some SGI keyboards on your PCB photo. Just curious, what kind of machines have you been working with?

    • Dave,

      I may have missed something but why not take the frequency normal, a standard counter with GPIB or USB-interface and do the rest on the PC?

      CU
      CJ

    • Dave,

      I was wondering if you could post manufacturer part numbers for the DTCXO’s you reviewed or used. If their phase noise is no worse than the OCXO we use in our synthesizer then maybe I could save a couple of watts.

      Best regards,
      Mark Stanley

    • Great blog! That gives me a lot of insight into the design for my quartz-crystal microbalance, I can see why it reminded you of this problem.

    • Dave, how old is the description of hydrophones physically placed on the seabed? I used to process seismic data; the hydrophones trailed for kilometers behind the seismic survey boat just below the ocean surface, the noise source use are air guns which also trailed just behind the boat. Seismic surveys are expensive enough but it would cost a tremendous amount of money and time to place and fix them under the seabed, I have never heard of this in 25 years working in the oil industry. maybe before my time

      • Ocean bottom arrays are a more modern technology than towed streamers (which I used to mostly work on).
        They are used mostly for existing reservoir monitoring, the movement of oil in an existing field over time for example. They are not used for initial surveying, as you say it would be too expensive and time consuming for that. Should have made that clearer in the blog, I didn’t suspect any seismic industry people would be watching!
        They are used in what is called 4C ocean bottom seismic surveys (3 axis geophones + a hydrophone + 3 axis tilt sensor)

        Dave.

        • After my time not before it seems, the reservoir guys where I worked used 3D software to simulate the flow of reservoirs but this gives them more accurate data to work with. I have just looked up some more info and it seems they can place these phones quite quickly.

    • awesome…

      :)

    • Hey Dave,

      This is not regarding your video but I had a question….When looking up data sheet, where would you most likely find the part number because on a PCB board there is all kinds of numbers?

    • Very interesting !
      Can you elaborate on the math for the drift correction ?

    • Great video, I was thinking of something similar last week,as I may need a precise 2400Hz, and just ended up with 32.728 xtal, DDS a temperature sensor and a LUT ( in scribble ware).
      I was wondering what happens if you run several xtals, of the same frequency, XOR them, LPF them and store the “differentials”. Just an idea.

    • Hi Dave,
      We have a similar thread in our forum.
      As I have asked on youtube, could you pretty please make your not-slope-corrected data available (as a picture or raw data is enough). Such info would be more useful for us.
      Thanks!
      A.

    • Hey Dave,
      Just out of curiosity – you may have mentioned this in your video and I didn’t catch it – what part did you end up using for your freq. standard? I’m curious because I can’t find any DTCXO’s with better than ~0.2ppm. I’m also a bit confused because the DTCXO’s I’ve been able to find aren’t any better than high-end TCXO’s. What are your thoughts on this?
      Thanks!

    • But you are all assuming that gravity is a constant at the macro level. It’s not.

    • Hi Dave,

      I’ve just seen this video, and it instantly reminded me on the ‘zero dead-time measurements’. I’m not sure, if you know details about those ‘modulation domain analyzers’, which use this measurement technique. There should be such devices, which are capable of this measurements. Still it might have been the right decision to create a own setup, instead of using a oversized high end measurement device. But this devices are quite neat and if you don’t know them yet, you might want to have a look into their build-up and the range of applications. They are really elegant for a wide range of modulation and noise measurements of RF signals and it would have been a good starting point for informations about measuring noise and drift of RF signals.

    • Heya i’m for the first time here. I came across this board and I find It really helpful & it helped me out much. I’m hoping to provide something again and help others such
      as you helped me.

    Leave a reply