Re: Sun Solaris Time Servers?

KJ wrote:
Dan Foster wrote:

$60,000 clocks is nothing when it's billions to trillions of
dollars/euros moving *every single day*. They really can't afford
fraud/thefts, inaccuracies, or outages, so they spend what it takes to
prevent these problems.

<snip>

The cesium and rubidium clocks generally starts at about $30,000 to
$60,000 and up; I've even seen $100,000 units, so obviously, only buy it
when there's an extremely compelling need for it.

One issue now is I believe there is only one vendor (Symmetricom) of the cesium clocks. How do banks feels about that?? I don't suppose any price hikes would worry them, but there are other issues with only one supplier.

Agilent no longer sell them, although I believe Symmetricom have made all Agilents cesium tubes for a while anyway.

These atomic clocks also have special engineering built in them that
avoids fluctuations due to temperature or voltage variations, and does
have a 'warm-up' period upon power on.

[This is also one of the reasons why NTP has an initial delay period
before it will accept a newly init'd time source's time info for NTP
time source ranking calculations.]

There are some nice non-atomic NTP clocks at the low end, for about
$3,000 off eBay for used units, that can synchronize via WWV radio
signals, NIST (i.e. phoning a public NIST NTP number), GPS, TCP/IP-based
NTP, and a few other common means of updating.

You can make an NTP server sync via GPS for a *lot* less than $3,000. I have a Motorola receiver, which is optimised for time and not position, which was a few hundred $. There are some micros that are better than general purpose computers and make really good NTP servers.

It would be very easy to turn an old Sun into a decent NTP server just by putting a GPS receiver on the serial port and using the 1 pps output from one. I would however add a rubidium to it for short term outages of GPS. A rubidium source is pretty cheap - I have one on my desk. It only takes about 6 minutes to warm up and has all the bits to lock to GPS inside it. It litterly just needs a 1 pps input from a GPS.

You can buy cesium clocks on eBay quite easily, but the life of the tube is limited and are expensive to replace.

It seems to me, there is a huge jump in cost from rubidium to cesium.

-Dan


Wow. Thanks for such a thorough and thoughtful reply Dan!

Yes, it was very informative.

If I've ever thought I was pretty particular about my time accuracy before, I now stand much closer to the populace majority that's apparently not-so-serious about exact time. =)

For someone who is *serious* about time, you really should take a look at leapsecond.com. The site is well worth a look, especially if you have an engineering background.

http://www.leapsecond.com/

Tom, who owns that site, has clocks that are almost certainly more accurate than any financial institution. He claims his time lab is better than anything outside of a government standards lab.

He even has a hydrogen maser, which keeps more precise time than the cesium clock. It is probably only a matter of time (excuse the pun) before the second is redefined in terms of the hydrogen maser and the reference to cesium dropped.

His wrist watch

http://www.leapsecond.com/pages/atomic-bill/

is pretty accurate too, although it is a shame the battery life is not very long, and it is a bit on the heavy side.

The site is well worth a look. I quite like the BNC chess myself, but many will not appreciate that one.

BTW, there is a paper I scanned which is quite interesting.

http://www.g8wrb.org/useful-stuff/time/pendulum+quartz.pdf

It shows how to make an old antique pendulum clock very accurate. I am considering doing this, but hooking it up to an oven stabilised crystal.

A rubidium would be better than crystal, but more money, for no useful increase in accuracy. Thee is little point having a clock with only a minute hand keep to < 10 seconds per year, and that can be achieved without an atomic source.

--
Dave K

Minefield Consultant and Solitaire Expert (MCSE).

Please note my email address changes periodically to avoid spam.
It is always of the form: month-year@domain. Hitting reply will work
for a couple of months only. Later set it manually.
.

Relevant Pages

  • Re: Sun Solaris Time Servers?
    ... before it will accept a newly init'd time source's time info for NTP ... There are some nice non-atomic NTP clocks at the low end, ... You can make an NTP server sync via GPS for a *lot* less than $3,000. ... You can buy cesium clocks on eBay quite easily, but the life of the tube ...
    (comp.unix.solaris)
  • Re: Looking for ntp/PPS setup guide
    ... :> We've got a GPS receiver to make an NTP server, and it has PPS output. ... I use CDMA clocks and they my time is stable within 10 microseconds. ...
    (freebsd-stable)
  • Re: Myth of GPS Employing the Nonsense of SR and GR
    ... I wouldn't have thought GR was relevant to GPS? ... People often ask me "What good is Relativity?" ... GPS satellites are not in geosynchronous or geostationary orbits). ... their clocks ticking more slowly (see the Special Relativity ...
    (sci.physics.relativity)
  • Re: Successful remote AES key extraction
    ... >Dan's attack doesn't require synchronized clocks. ... to within microseconds. ... The reported AES timing attack requires the measurement of variations ... NTP clock synchronization is also affected by variations ...
    (sci.crypt)
  • Re: I Owe Einstein an Apology. Sorry Albert!
    ... Clocks moving in a frame of reference will not ... This is the case in the GPS. ... calculation of the rate of the GPS satellites. ... "GPS time" is a coordinated time where the coordinate ...
    (sci.physics.relativity)