[GPS_Standard] 1-PPS output also?

[Dave Platt]

On 03/21/2012 11:54 AM, Chris Howard w0ep wrote:
> 
> I'd like to know if I need to have some kind of
> circuit in my GPSDO in order to feed the 1pps signal
> to an output.
> 
> Somebody here (or somewhere...) said I should run
> the 1-pps to my computers for NTP.  Thought I would pursue
> that idea.
> 
> My GPS is a resolution-T.  I am currently running
> an NPN/PNP amplifier circuit to boost the 1-pps to 5 volts.
> 
> You know... what I'd really like is a low power single-board-computer
> which gathered my GPSDO stats and also acted as an NTP server
> to my network.
> 
> That and a good ham sandwich.

I did that, in my own VE2ZAZ-based system.

What I did, was buffer the PPS signal coming out of the GPS
receiver itself, using a two-stage fast-bipolar-switching-
transistor common-emitter circuit.  I tapped off one of these
two signals (I forget which) and fed it to one of the TX
inputs of the same MAX232-family TTL-to-RS232 converter
chips that handles the TX/RX communication for the GPS
receiver and the FLL board.

The RS-232-side output of this one line goes to one of the
handshaking lines on the serial port connector for the GPS-
to-PC feed.

I've got two serial-port cables going to the PC - one
monitoring the FLL reports every 16 seconds (and, in a
script, doing some tracking and adjustment) and another
reading the GPS output and PPS handshaking pulses.  A
slightly-hacked version of the Linux "gpsd" program
handles the GPS data and PPS, and acts as a dual time source
(Oncore-based serial, and PPS) to the "ntpd" timekeeper
daemon.

Performance seems good:

     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
 adsl-71-159-209 76.191.159.165   2 u   15   64  377   19.001   -2.148   0.230
-clock.team-cymr 172.16.65.22     2 u   53   64  377   63.892    2.014   0.370
+mailman.pittsbu 209.51.161.238   2 u   48   64  377   51.767    1.096   0.653
-ntp.silkytech.c 209.51.161.238   2 u    1   64  377   88.618    1.925   0.666
+isaachayes.khre 204.123.2.72     2 u    7   64  377   11.325    1.086   0.170
-SHM(0)          .MOTO.           5 l   31   64  377    0.000   -0.979   1.552
*SHM(1)          .PPS.            0 l   28   64  377    0.000    0.010   0.002

This is the lowest PPS jitter estimate I've seen so far.

Once Debian Linux goes through another major version upgrade, I believe
that its standard Linux kernel will have the kernel-based PPS support
in the serial-port line discipline (i.e. the system time will be sampled
in the kernel during the serial port handshake-line-changed interrupt,
rather than in GPSD).  This should reduce PPS jitter even further.

My system isn't quite an SBC - it's a Atom-based mini-ITX motherboard in a
minitower case - handles the GPS and FLL, incoming and outgoing email,
a small Web site, and an Asterisk VoIP server. Handles it all without
any strain.

No ham sandwiches, though.