[ntp:questions] Consumer GPS question

Michael L. Semon mlsemon33 at verizon.net
Wed Feb 7 01:07:28 UTC 2007

shane-dated-1173382499.fffb68 at cm.nu wrote:
> Hey all,
> I am looking at the possibility of using a GPS receiver for NTP time sync
> either via the nmea refclock driver or gpsd, just a little home project. 
> From my googling, I've read that the only way to get decent results is to go
> with an rs232 sync and stay clear of usb or bluetooth.  Looking at the
> mainstream gps sites, I've found the Globalsat br-355 unit and I'm wondering
> whether this is suitable for NTP sync and what kind of accuracy I should
> expect.
> http://www.buygpsnow.com/globalsat-br-355-water-proof-serial-gps-receiver-sirf-star-iii-br355-v3-1-1-free-arkon-mount-432.aspx
> The device seems to use a ps2 cable which converts via a y-cable into a db9
> serial connection and gets power from the PS2.  What I can't seem to find in
> the specs or on the globalsat site is whether the db9 connection provides a
> PPS output.  If not, are there any PPS capable receivers out there that
> don't require electronics knowhow (soldering etc.)?  That's all I've found
> so far.
> Tia,
> Shane

I'm agreeing with the others here:  Get the Garmin.  If you can find a 
Garmin dealer in town, and that dealer does car stereo installation and 
such, maybe the dealer would be willing to wire it up for you.  Note 
that these stores will usually sell only the "PC" (serial) or USB 
versions of the GPS 18 and not the LVC, so the LVC may be a special order.

I have a different take than the others, as a GPS 18 LVC owner.  I'm 
looking at the specs for the GlobalSat unit that is listed on this page:


The Garmin offers more output sentences, allows you to shut more of them 
off, and offers faster baud rates than does the GlobalSat unit.  The 
"Time" specification for the GlobalSat unit is probably how well the 
unit itself keeps time, not how well it's going to give that time to 
your PC.  You could ask them to clarify that, though...

The difference is this:  You can get stable time from a GPS with a PPS 
output, or you can get unstable time from a regular GPS but have very 
good long-term stability.  Here are the last 10 time and jitter values 
from my Garmin GPS 60, over USB, using ntpd's SHM driver with gpsd:

root at mls:~# tail -n 200 /var/log/ntpd/peer | grep '127\.28\.0' | tail -n 
10 | cut -d' ' -f 5,8
0.008913160 0.012008424
0.000041161 0.008891706
0.011150160 0.013180617
-0.001717840 0.010252016
0.007174160 0.011083976
0.007284160 0.010323697
-0.001586839 0.009591604
0.008517160 0.006426530
-0.006348839 0.011837580
0.001754160 0.006356728

I love how it goes from 0.00004 seconds on one reading to 0.01 seconds 
on the next one.  This is more the fault of the GPS and not so much the 
USB or gpsd.

Now here are the last 10 time and jitter values from the PPS on the 
Garmin unit, and here, I'm using ntpd's ATOM PPS driver and the LinuxPPS 
kernel patch for kernel 2.6.19:

root at mls:~# tail -n 200 /var/log/ntpd/peer | grep '127\.22\.0' | tail -n 
10 | cut -d' ' -f 5,8
-0.000002926 0.000000954
-0.000002807 0.000000954
-0.000000876 0.000001547
-0.000001053 0.000001355
-0.000001744 0.000000954
-0.000001706 0.000000954
-0.000001773 0.000000954
-0.000002260 0.000000954
-0.000002247 0.000000954
-0.000002448 0.000000954

That's the difference, and it's an important one.


More information about the questions mailing list