[ntp:questions] Slightly OT: Looking for good GPS PCI time sources
Richard B. Gilbert
rgilbert88 at comcast.net
Fri Jul 6 20:39:34 UTC 2007
Perry E. Metzger wrote:
> I'm building a system for a client to measure network latencies in a
> distributed system very accurately -- microsecond accuracy is desired
> (though I suspect in practice we'll end up doing a bit worse). To do
> this, I need to get high quality clock sources for the machines doing
> the measurement. One obvious choice would be PCI cards that
> synchronize to GPS.
> I figure if any group has significant experience with such devices,
> this one would -- anyone have any recommendations? Also, can anyone
> think of another reasonable way to get this sort of accuracy on stock
> PC hardware?
> One consideration might be whether the cards in question allow readout
> of the time as a single atomic operation -- if they don't, one can't
> (for example) read the time without worrying about locking...
PCI cards that synchronize to GPS are available. Meinberg Funkuhren
sells one such. Another is the BC637PCI from Symmetricomm. They are
rather expensive; the Meinberg card sells for over $1000 US. I don't
know what list price for the BC637 is but I saw a used one sold for over
$400 US on e-Bay a few years ago. I believe that both of these cards
feature a GPS disciplined OCXO, at least as an option. If you buy one
and get tired of it, all donations are gratefully accepted! :-)
There is one GPS receiver, whose brand name I have forgotten, that will
give you the time stamp for the leading edge of a pulse you supply. The
name "Palisade" just bubbled up from the sludge at the bottom of my
mind. http://www.leapsecond.com/museum/palisade/ It does this trick via
a serial port so you can forget about time stamping two pulses a few
microseconds, or even milliseconds apart.
More information about the questions