[ntp:questions] NTP on local networks

unruh unruh at invalid.ca
Wed Aug 1 14:46:27 UTC 2012

On 2012-08-01, Terje Mathisen <"terje.mathisen at tmsw.no"> wrote:
> David Taylor wrote:
>> I have information on my Web site on the easy-to-use Sure GPS, as well
>> as the low-cost Garmin GPS 18x LVC.
>>    http://www.satsignal.eu/ntp/Sure-GPS.htm
>>    http://www.satsignal.eu/ntp/FreeBSD-GPS-PPS.htm
>> Mine are using simple puck antennas, indoors, on the top floor of a
>> two-storey building.  Be aware that USB-connected devices will give far
>> less accuracy than serial-port connected ones, but may be adequate if
>> "half-a second" is all you need.
> This summer I received two more SURE boards, it took about an hour total 
> to solder on the two required patch wires on each of them.

I have found that simply soldering a wire from the PPS output jumper
hole on the board to the DCD pin on the serial port works, at least with
my serial ports on my computer. While it give more accurate timing
pulses, it is really irrelevant, since the interrupt service time is in
the usec to many usec range.

This means that the soldering is much easier. In David's you have to
sloder to the tiny legs of surface mount chips, which has its dangers if
you are not very good or steady with the soldering iron.

> (I used David's pictures to remind me where to put the patch wires!)
> Together with a pair of old laptops (with proper serial ports) this 
> gives me another set of Stratum1 servers, for a total outlay of less 
> than $100.
> I've seen reports here that one of the timing nuts have measured the 
> SURE board as giving a PPS signal that's in the 25-50 ns range?

The problem I have had with the sure boards is the antenna they send. Of
two Sure boards I got, one of the antennas was defective ( very
intermittent, and got worse over time). While I notified Sure and they
said they would seend me a new antenna, they never did. 

> According to the loopstats files my laptop is mostly in the 3-500 ns 
> range, with excursions up to a us or two, obviously due to interrupt 
> response time jitter and temperature variations.

Look at the refclock file instead at the input time variations. I get in
the 1-10usec range because if interrupt timing problems.
> Terje

More information about the questions mailing list