time server ??

Discussion in 'Linux Networking' started by Andre, Aug 26, 2011.

  1. Andre

    Moe Trin Guest

    On Wed, 20 Mar 2013, in the Usenet newsgroup comp.os.linux.networking, in
    That temperature variation is the fly in the ointment. If you can
    keep the crystal temp consistent, something like NTP can compensate
    for the rest. But then you start flogging the computer...

    Much more than three quarters of the crystals manufactured are "X"
    cut, which while "easy" to cut also has about the worst temperature
    coefficient. There are several other cuts with substantially better
    characteristics, but they're harder to cut and thus more expensive.
    Somewhat after the transistor was invented, a rather simple way to
    improve the frequency stability was to (essentially) glue to crystal
    to a small power transistor along with a thermistor and two fixed
    resistors. The circuit was such that the transistor heated the crystal
    and thermistor, and the fixed resistors were chosen to bring the
    crystal to about 45-50C. If the transistor current gain was high
    enough, this stabilized the temperatures pretty well, and a relatively
    low cost in hardware.
    Bill, I'm old, but not THAT old ;-)
    Somewhat interesting that the 1934 (US) federal regulations for AM
    broadcast stations (540-1600 KHz) required a frequency tolerance of 10
    Hz (6.25 - 18.5 ppm) and we had to jump through moderate hoops to be
    able to measure that error (double ovened crystal oscillators). The
    "VHF" television transmitters had similar requirements, but later "UHF"
    channels had tighter specs (1.1 to 2.1 ppm).

    I've always found it interesting exploring older technology. The
    Harrison Number 4 Timekeeper is a mechanical marvel. The November 1761
    voyage to Jamaica showed an error of 9 seconds in the 60 day voyage,
    which is 1.74 ppm. Not bad for hardware, especially given the
    conditions at sea (ships motion, temperature problems and what-not).
    If they're disciplined, yes. But in the 1960s it wasn't that difficult
    to get stability. We used to cheat and use VLF radio transmissions as
    a timing reference - the LORAN-C chains (100 KHz) in the 1960s were
    controlled to parts per billion (10e12) or better (multiple cesium
    clocks averaged at each chain master site), and were receivable over
    large parts of country. For the most part, 0.1 ppm was almost trivial
    to obtain. The other part of the question is "what is the actual
    need?". I was working on terminal area aircraft flight tests, and the
    aircraft was at 300 KIAS or less, which translates to 500 ft/sec. max.
    Even with our spiffed-up laser tracking systems, knowing where the
    aircraft was to an accuracy of +/-4 feet (X, Y and Z) was pushing
    things - thus 2 msec timing accuracy was sufficient.

    Old guy
     
    Moe Trin, Mar 21, 2013
    #21
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.