Re: [gpsd-dev] SHM only getting 2/3 of samples

[Gary E. Miller]

Yo Hal!

On Thu, 07 May 2015 03:46:28 -0700
Hal Murray <address@hidden> wrote:

> 
> address@hidden said:
> > Could be a Nyquist failure.  To check something that happens every
> > second you need to test at least twice a second.  Could be gpsd
> > withheld a sample for some reason too. 
> 
> Nyquist applies to bandwidth.  This isn't a bandwidth issue.  We are
> looking for events.

Same thing.  You can't measure at exactly 1 second intervals.  So
if you measure at X.99999999, the get a SHM at x.0 and another at x+1.0
then meaasure at x+1.000000001 you will see one event not two.

> If you are looking for events, and the clocks are not locked, the
> sampling clock has to run just a little bit faster than the source of
> the data.  The "little bit" has to include the jitter on both clocks
> as well as the actual frequency error.

Yes, ideally, but nothin ideal about this case.

> If the clocks are synchronized they can run at the same speed.  In
> the nasty case, they will line up and what happens then depends on
> the details of the handoff, but it's very unlikely to be consistently
> a pattern like 2 out of 3. I've been watching this sort of stuff for
> a long time.  I don't remember seeing anything like this before.

Remember ntpd can slew the clock over 8%.  Maybe it will not get you
to 1/3 missed, but you'd miss a lot.  Best to remove the measurement
uncertainy first.

> What's the timing on when the SHM data gets updated?  How much jitter
> is there?

SHM gets updated when gpsd has data.  SHM0 is the serial line time.  So
add the +/-8% clock slew (120 mSec) and the 200 mSec NMEA jitter and you
could easily see >300 mSec.

> Does that happen at the same time as the TOFF gets sent over a JSON
> pipe?  If so, I can measure it.

TOFF send directly after the NMEA sentence decoded and sent to SHM0
 
> I tried ntpshmmon, but it prints a line every poll try.  I was
> expecting it to print one each time it saw new data - that is
> suppress duplicates.

Patches welcome.

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