[time-nuts] Can I get 1 millisecond accuracy with a USB GPS-18

Chris Albertson albertson.chris at gmail.com
Mon May 13 11:24:00 EDT 2013

Why not run an experiment?  Connect a serial and a USB GPS-18 to the
same NTP server. and run BOTH ref clocks at the same time.   Let it
run for a day or so them please, post the results for NTPQ here.

Windows is a poor platform for timing.  You are best off with Linux or
BSD UNIX and a real hardware serial port.  The combination of Windows
and USB makes 1 millisecond about the limit.  But you can get  2 - 5
uSec. with Linux or BSD and a serial port.   The world record (as far
as I know) broke 1 uSec but required external hardware and motherboard

An experiment would answer this.  But you'd need to run both GPSes on
the same NTP server.

If you don't have the port, it is worth the effort to do a Motherboard
swap.  Intel makes a few Atom boards with real serial rs-232 ports and
the Atom is so low powered that it can use a passive heat sync.   If
you run this 24x7 the power savings can pay for the new $90 board.
$90 includes the cpu chip, mother broad and cooling and the serial
port.   What is your cost per KWH?   With a hardware serial port and
Linux (or BSD Unitx) single digit uSec is to be expected.

A common mid tower PC can burn 100W.  THat works out to 876 KWH per
year.  I pay 27 cents for each KWH so the cost is $234 per year.  The
new board is only about 10W.  You can see a payback on only 6 months
and you make $100 in the next 6 months.  The power savings can also
pay for replacing the GPS receivers with serial units.

On Mon, May 13, 2013 at 7:43 AM, Jim Lux <jimlux at earthlink.net> wrote:
> rummaged through archives and couldn't find anything..
> I've got some GPS-18's with the RS232 and 1pps output.  BUT, I'm wondering
> if anyone has tried to get timing with the USB version (Linux or Windows..),
> and if so, is getting 1 millisecond absolute accuracy feasible.
> The underlying USB thing has 8kHz frame rate, but I suspect that the serial
> port emulation (which is probably what they use) might not hold it that
> well.  I can see there being some sort of fixed offset (going through hubs
> and such), but once the configuration is known, is that stable?
> Any practical advice?
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.


Chris Albertson
Redondo Beach, California

More information about the time-nuts mailing list