Shop OBEX P1 Docs P2 Docs Learn Events
More accurate clock? — Parallax Forums

More accurate clock?

goober-headgoober-head Posts: 14
edited 2011-08-16 22:01 in Propeller 1
Can I replace the crystal on the professional board with a phase locked loop and increase the clock accuracy?

Comments

  • TubularTubular Posts: 4,717
    edited 2011-06-26 21:59
    Hello

    You can drive the XI pin with a more accurate clock source, such as a temperature compensated crystal module, yes.

    I don't see how a PLL will improve anything, since it locks onto the crystal (plus its associated drift)

    What level of accuracy do you need, over the course of a year?

    Lots of people are using a GPS module to provide an accurate timebase. Is that an option for your application

    cheers
    tubular
  • nestor73nestor73 Posts: 33
    edited 2011-06-27 01:45
    Why not use this :

    http://www.sparkfun.com/products/99

    I use it and it works great
  • AleAle Posts: 2,363
    edited 2011-06-27 02:22
    Why not using a RTC chip if what you need is a real time clock (using the propeller to keep it may not be the most accurate way). I see that nestor73 posted a great rtc module :), if you have access to sparkfun of course :)
    There are also quite precise and affordable (some 100s) rubidium-based oscillators ...
  • goober-headgoober-head Posts: 14
    edited 2011-06-27 12:07
    nestor73 wrote: »
    Why not use this :

    http://www.sparkfun.com/products/99

    I use it and it works great


    I meant a clock as for a CPU, not a calander clock. I want a great 80 MHz input. I'm looking at TXCO's and OXCO's. So far I've found a module that will get me to +/- 10 ppm for $28 bucks. I'm still looking but this is the best so far.
  • goober-headgoober-head Posts: 14
    edited 2011-06-27 12:08
    Tubular wrote: »
    Hello

    You can drive the XI pin with a more accurate clock source, such as a temperature compensated crystal module, yes.

    I don't see how a PLL will improve anything, since it locks onto the crystal (plus its associated drift)

    What level of accuracy do you need, over the course of a year?

    Lots of people are using a GPS module to provide an accurate timebase. Is that an option for your application

    cheers
    tubular

    Are you refering to the GPS that Parallax sells?
  • Kevin WoodKevin Wood Posts: 1,266
    edited 2011-06-27 12:29
    I don't know how well it would work, but you can find rubidium oscillators on ebay...

    http://shop.ebay.com/i.html?_nkw=rubidium&_armrs=1&_from=&_ipg=
  • goober-headgoober-head Posts: 14
    edited 2011-06-27 13:15
    Kevin Wood wrote: »
    I don't know how well it would work, but you can find rubidium oscillators on ebay...

    http://shop.ebay.com/i.html?_nkw=rubidium&_armrs=1&_from=&_ipg=

    In a good Rubidium module, one can achieve +/- 0.01 ppm. I looked through your E-Bay list. Most of the modules listed are older technology. I'm going to try the OXCO first. Thanks for your input. I'll try it next if the OXCO does not pan out. The module I found was a 3.3 volt unit. Most of the Rubidium's have been 5 or 12 volts.
  • Toby SeckshundToby Seckshund Posts: 2,027
    edited 2011-06-27 15:33
    The Rubidium drives that we had (on "The Home Service" 200KHz Droitwich) were the size of a medium room. But then that was back 30+ years ago, the TXs were like a set from "Boris Karloff rides again".

    We used to use the "off air" of BBC2 as a calibration for the subcarrier (4.43MHz) for OBs, as that was ultimately referenced back to a NPL source. Now with digital transmission that little trick has been scuppered.
  • jmgjmg Posts: 15,185
    edited 2011-06-27 16:12
    goober-head wrote: »
    I meant a clock as for a CPU, not a calander clock. I want a great 80 MHz input. I'm looking at TXCO's and OXCO's. So far I've found a module that will get me to +/- 10 ppm for $28 bucks. I'm still looking but this is the best so far.

    If you need high precision, you could also look at SiLabs SI5351A (MSOP10) - these could allow almost any source frequency to generate the 80MHz, which might open up more choices ?
  • jmgjmg Posts: 15,185
    edited 2011-06-28 18:58
    goober-head wrote: »
    So far I've found a module that will get me to +/- 10 ppm for $28 bucks. I'm still looking but this is the best so far.

    If you open the Freqency specs a little, the mass-volume products come into scope, and examples from Digikey are
    2.5ppm FOX-924 - 27MHz for 4.76/1+ and $1.78/1000+ for 27MHz, or the NT3225SA-26.000000MHZ for $4.21/1
    You then scale these low cost TCXOs using the SiLabs SI5351A.
  • goober-headgoober-head Posts: 14
    edited 2011-07-28 10:19
    Tubular wrote: »
    Hello

    You can drive the XI pin with a more accurate clock source, such as a temperature compensated crystal module, yes.

    I don't see how a PLL will improve anything, since it locks onto the crystal (plus its associated drift)

    What level of accuracy do you need, over the course of a year?

    Lots of people are using a GPS module to provide an accurate timebase. Is that an option for your application

    cheers
    tubular

    How is the GPS clock hooked in?
  • prof_brainoprof_braino Posts: 4,313
    edited 2011-07-28 17:38
    May we ask what accuracy you have in mind, and what your target application is, and possibly why this extra accuracy is needed?

    I don't mean to pry, but I am curious. It might have some brearing on the solutions offered.
  • goober-headgoober-head Posts: 14
    edited 2011-07-29 04:59
    I have circuits that put out square waves whose width represents the quantity measured, so the more accurate the better ( within reasonable cost).
  • DoofusDoofus Posts: 28
    edited 2011-08-08 10:31
    goober-head wrote: »
    How is the GPS clock hooked in?

    Does anyone have a circuit on this one?
  • LoopyBytelooseLoopyByteloose Posts: 12,537
    edited 2011-08-08 11:02
    The problem with getting a more accurate source for a clock is generally thermal stability. It doesn't matter how fast a crystal you use, you get about the same error results UNLESS you get a crystal that has a thermal oven to stabilize drift.

    Better than that can be achieved by an atomic clock. I believe I saw recently in Nuts and Volts that there is actually a chip that can provide this level of accuracy now. But of course, you have to pay a premium for it by replacing the $1 USD clock crystal with something that is much more expensive. I believe it was priced in the thousands of dollars.

    Crystals in thermal ovens are not cheap either - about $105 USD for +/-0.25ppm and 10Mhz seems to be the slowest. Search crystal oven. The 3.3volt ones draw 450ma of current (after all is is an oven).
    http://search.digikey.com/scripts/DkSearch/dksus.dll

    There is an adujustible 1-20mhz Rubidium oscillator here. 15 to 18 volts at 700ma.
    http://www.freqelec.com/rb_oscillators.html

    I guess the advantage with a GPS clock is that it is reset by the satellite's reference which can be a very expensive clock.
  • DoofusDoofus Posts: 28
    edited 2011-08-08 12:33
    Loopy Byteloose wrote: »
    The problem with getting a more accurate source for a clock is generally thermal stability. It doesn't matter how fast a crystal you use, you get about the same error results UNLESS you get a crystal that has a thermal oven to stabilize drift.

    Better than that can be achieved by an atomic clock. I believe I saw recently in Nuts and Volts that there is actually a chip that can provide this level of accuracy now. But of course, you have to pay a premium for it by replacing the $1 USD clock crystal with something that is much more expensive. I believe it was priced in the thousands of dollars.

    Crystals in thermal ovens are not cheap either - about $105 USD for +/-0.25ppm and 10Mhz seems to be the slowest. Search crystal oven. The 3.3volt ones draw 450ma of current (after all is is an oven).
    http://search.digikey.com/scripts/DkSearch/dksus.dll

    There is an adujustible 1-20mhz Rubidium oscillator here. 15 to 18 volts at 700ma.
    http://www.freqelec.com/rb_oscillators.html

    I guess the advantage with a GPS clock is that it is reset by the satellite's reference which can be a very expensive clock.


    Someone in a previous post talked like this was do-able. I don't know if this was actual "We done it" information or just some more BMC (blog maybe conjecture).

    When you ask for a circuit, all of a sudden the rubber hits the road, and either it's been done and can be proven, or it reels in more blog blubber.
  • goober-headgoober-head Posts: 14
    edited 2011-08-16 19:42
    Doofus wrote: »
    Someone in a previous post talked like this was do-able. I don't know if this was actual "We done it" information or just some more BMC (blog maybe conjecture).

    When you ask for a circuit, all of a sudden the rubber hits the road, and either it's been done and can be proven, or it reels in more blog blubber.


    When counting in the 500,000 range, the standard deviation is about 100. I'd like to get it down to 10
  • DogPDogP Posts: 168
    edited 2011-08-16 22:01
    If you just need one order of magnitude better, simply going to a TCXO should do it, for a reasonable cost. The original crystal is rated at 30ppm... something like a Valpey Fisher TCXO can be as low as 0.1ppm, and the price isn't too bad: http://www.valpeyfisher.com/products-frequency-control-oscillators-tcxo.htm . You'd need to check what the lead time is on a lot of that stuff though. They also have really good OCXOs, but they're kinda power hungry, bigger, and more expensive with long lead times. A Rb oscillator seems like extreme overkill for this application, and really, if you're needing short term stability, not long term, then an OCXO is better anyway. Most expensive Rb modules actually incorporate an OCXO on the output (locked to the Rb) because of the better short term stability of the OCXO.

    Even something like this gives you an order of magnitude better, in stock, and pretty cheap: http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=631-1067-1-ND . Or go another order of magnitude in stability (and price ;-)): http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=CW648-ND .

    I remember seeing an article in Nuts and Volts a while back where someone made their own ovenized oscillator w/ resistors to heat the crystal... kinda interesting, but I don't remember how much power it ended up drawing, and what the specs ended up being after being heated.

    DogP
Sign In or Register to comment.