[ntp:questions] Timing GPS recommendations

unruh unruh at invalid.ca
Mon Aug 20 19:17:58 UTC 2012


On 2012-08-20, Chris Albertson <albertson.chris at gmail.com> wrote:
> On Mon, Aug 20, 2012 at 9:13 AM, unruh <unruh at invalid.ca> wrote:
>
>> On 2012-08-20, Hahn, Ron <ron.hahn at fmr.com> wrote:
>> > Colleagues,
>> >
>> > Chris Albertson, made the following statement:
>> >
>> > --< begin >--
>> > I'm using the t-bolt.  It seems to work.   I guess there might be more
>> > functions the driver could implement.  Perhaps better support for
>> > self-survey or logging or whatever.  I'd like to see internal temperature
>> > logged.  But for normal NTP use it works fine.
>> >
>> > If you only have one GPS the t-bolt is the one to have.  But if you can
>> > afford two there are lower priced GPSes that use less power that work
>> well
>> > with NTP.   I have the t-bolt and two UT+ units.
>> >
>> > Whatever you get make sure it is a TIMING gps.  The Garmin units and the
>> > Sure Electronic board at not timing GPS although they do have PPS
>> outputs.
>> >  They work but they are "uSec" level devices.  I'd look for 100ns or
>> better.
>>
>> No The sure is 10s of ns device. Unfortunately this is useless as you
>> cannot get the time into your computer to better than usec. The
>> interrupts are not serviced fast enough on any PC to give better than a
>> few us and the interrupt routines get delayed on a working system by
>> more than that at times (eg due to disk priority, or things switching
>> off interrupt processing).
>>
>> You would have to build a special board for your computer to better than
>> usec.
>>
>>
> The errors add together.  If the GPS and the computer both had 1 uSec of
> uncertainty the total is 2uSec.  So there is an advantage to having a GPS
> that is in effect "zero".  I think this is the same ia when you build a
> circut with a bunch of 5% tolerance resisters.  You can get much larger
> than 5% error because of the ways resisters add.

Uh, but then you need to know what those two errors are. If the gps
error is .1us then nothing you do there will every improve anytbing. You
are trying to fix the wrong thing. Imagine that you have a wristwatch
that is accurate to the second but your send out messangers by foot to
deliver the time to your friend across town. Do you really think that
buying a new wrist watch is the way to improve the time your friend
gets?

>
> I think I would some day like to try and build an external PPS interface
> that works at nanosecond resolution.  It would not be hard.  A fast counter
> is PPL'd to a 10Mhz lab reference and then the PPS captures the counter and
> then interrupts the computer.  the computer then reads my external box
> rather then the CPU counter.  Only a few lines of C code need to change
> inside the Linux driver.
>
> Chris Albertson
> Redondo Beach, California



More information about the questions mailing list