[time-nuts] Looking for advice to get a submillisecond setup

[Jim Lux]

On 2/19/15 9:11 AM, Matt wrote:
> Hi,
>
> My university would like to have a <1ms precise source of time to do
> some networking experiments (measure one way propagation delays
> etc...). So I wandered on the internet to find the best choice with a
> budget of ~1000€ (~1100 American dollars).
> I've been overwhelmed by the number of possibilities (atomic
> clocks/GPS signal etc...) and as no price appear on the seller
> websites, it's difficult to rule out options.
>

An inexpensive GPS receiver with a 1pps output will easily get you to 
much better than 1 millisecond.  The Garmin 18 is but one choice.  It 
has the advantage that it's already packaged, as opposed to, say, one of 
the little modules designed to be attached to a Arduino.

Configuring ntp to use it is just a matter of setting up the file 
properly. NTP will use the 1pps coming in on one of the modem control 
signals (DTR, DCD, RI, etc.)

(I use a USB cable to get 5V to run my GPS-18x-LVC, and wire 1pps to 
DCD, pin 1 on the 9 pin connector)


So really, it's a matter of finding a place to put your Garmin receiver 
and string a cable that's not too long to your *nix box running ntp.

People have done it with a Rpi, if you want to go that route.

> I also read good comments on Garmin 18 hardware but it is so cheap I
> wonder if it precise enough.
>
> I wonder if we should buy a specific box or if we could not plug the
> antenna to a linux box with gpsd/NTPd on it ?
>

That's exactly what you want to do.

http://www.satsignal.eu/ntp/FreeBSD-GPS-PPS.htm

http://www.rjsystems.nl/en/2100-ntpd-garmin-gps-18-lvc-gpsd.php

(everything there should work fine with the current GPS-18x-LVC, but I'm 
sure someone on the list has actually done it and can confirm.)

http://www.catb.org/gpsd/gpsd-time-service-howto.html


Yeah, the Garmin is cheap ($85 US), so you're not going to get 
nanosecond timing, just microsecond level.  Since you need milliseconds, 
it's plenty good enough.