Re: [gpsd-dev] A detail in your ntpoffset script

[Gary E. Miller]

Yo Hal!

On Wed, 23 Oct 2013 12:52:38 -0700
Hal Murray <address@hidden> wrote:

> address@hidden said:
> > On one host I have two GPS/PPS.  They differ by a consistent 1.5
> > uSec. 
> 
> Is that just the expected synchronization?

My expectations are not relevant.  It is what I see.  Some units
agree more, some less.  SSince it is very constant I am left to 
conclude they are stable but do the agree.

> What would happen if both PPS pulses happened at exactly the same
> time?  The interrupts would get merged into one.  The interrupt
> processing would consistently check one first.  How much time would
> you expect between first check and the second?  1.5 uSec seems like a
> reasonable ballpark to me, depending on your CPU.

I see where you are going.  But I have gottent tow PPS to agree to better
than 1 uSec so I do not think it is a measurement error on my part.

The error is attached to the GPS, not to the port or host I conenct it
to.  So it is a characteristic of the GPS, not the measurement
techniwue.

> It's slightly more complicated with USB.

Acually.  It is much simpler.  There are USB 1.1 1024 time slots.  That
works about to a time slot every 1 mSec.  Each device is always polled
in the same time slot.  The time slots are derived from the system
clock.  So once the system clock is stable the sample instant is euqally
stable.  Once it settles I get an observed stability of one part in
about 1,000.

Sometimes, when the moon is in the seventh house, I see under 200 uSec
jitter!  But that is probably a self measurement and tends to sawtooth.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97701
        address@hidden  Tel:+1(541)382-8588

signature.asc
Description: PGP signature