[ntp:questions] Re: Bought a GPS receiver--What next?

David Schwartz davids at webmaster.com
Tue Nov 2 02:25:13 UTC 2004

"David Schwartz" <davids at webmaster.com> wrote in message 
news:cm6qmk$29e$1 at nntp.webmaster.com...

>    One millisecond is pretty poor PPS accuracy for a GPS receiver intended 
> for timing applications. One microsecond is typical.

    Ahh, I get it now. The PPS output is accurate to 200nS or so, more than 
good enough for GPS timing. The confusion is in the time output from the 
serial port, which is the time at which the message was generated, not the 
time stamp of the last PPS output.

    This is a good compromise that allows you to use the PPS output alone or 
the serial port alone or the two together and get about the best possible 
accuracy in all configurations. NTP supports this configuration perfectly, 
using the serial data for rough acquisition and then switching to the PPS 
output once it's locked within a second. This will also work well without 
the PPS input or on platforms where PPS isn't supported.

    By the way, I saw one GPS receiver specifically sold for timing 
applications with no PPS output and no specified accuracy on its serial 
outputs. Do not *ever* buy a GPS clock without a PPS output and without a 
clearly defined relationship between the PPS output and the serial output. 
And shame on someone for selling a GPS receiver specifically for use in 
timing applications that not only has no PPS output but no defined 
relationship between the time in its serial output and GPS time.

    Worse, they quote a time accuracy of one microsecond. When I asked them, 
"one microsecond between what and what?", they ignored me twice. Clearly, 
it's one microsecond between GPS time and the clock's internal 
representation of time -- to which you have no access.


More information about the questions mailing list