Chris Albertson wrote:
On Thu, Mar 31, 2011 at 11:03 PM, Terje Mathisen<"terje.mathisen at
tmsw.no"@ntp.org> wrote:
I don't think any GPS comes even close to 2 ns, given that the entire GPS
constellation is only required to be within 14 ns or so of UTC.
First off this is a "timing gps". That means the receiver "knows" the
antenna is not moving and is fixed to a surveyed location..
I think if the receiver does not meet their specs you could ask for a
refund. They do place that 2ns spec near the top of their data sheet.
I think they do two things
1) The error is predictable, one of the outputs from the device is an
error prediction that looks like a sawtooth function if you plot it.
The 2ns only applies if yo use the error prediction.
2) One can get much better than even 2ns if you average many seconds of signal.
That is a totally different problem!
Getting a GPS to be UTC-aligned on average is far simpler than what
we're usually discussing here, which is getting the jitter value (really
rms timing error) as low as possible.
OK, I think I may have confused two email lists. This one is for NTP
where most people don't need even microsecond accuracy. But other do
use GPS for calibrating lab instruments and are needing oscillators
with error at the 10E-10 or even better level.
Chris, I soldered together my first UT+ many years ago, based on the
TAPR board. I do know the difference between timing and position GPSs.
My corporate setup has 6 distributed servers (spread over 3 locations),
they use the 12-channel Oncore receivers. (With the Oncore driver the
sawtooth corrections are of course included.)
Terje
--
- <Terje.Mathisen at tmsw.no>
"almost all programming can be viewed as an exercise in caching"
_______________________________________________
questions mailing list
questions@lists.ntp.org
http://lists.ntp.org/listinfo/questions