Re: [gpsd-users] Garmin 18X-5Hz

[Gary E. Miller]

Yo Robin!

On Tue, 16 Aug 2016 22:23:22 +0200
Robin Schwab <address@hidden> wrote:

> Dear Gary
> 
> Am 16.08.2016 um 20:39 schrieb Gary E. Miller:
> > On Tue, 16 Aug 2016 09:32:37 +0200
> > Miroslav Lichvar <address@hidden> wrote:  
> >> I mean frequency of the oscillator on which is based the system
> >> clock. As the temperature increases the clock slows down and
> >> ntpd/chronyd has to tell the kernel to speed up the system clock
> >> to compensate for that change.  
> >
> > The temp is no gonna take a steap jump for 10 seconds.  That is the
> > length of my spike.  
> 
> I think you're right when you say you're stubborn. ;-) You tell the
> list there is a random jump roughly every two weeks in your system.
> But when I tell you to try to correlate it with GPS signal quality
> (in particular TDOP) you tell me I must prove my theory to you.

Yeah, and I also said I was going to add some gpsd data graphs to
see that, but I have other data graphs to add first.

I got enough known bads to chase, then I'll add some more unknowns to
chase.

> I just tried to help.

Well, helping is not when you add to my todo list, but instead when you
take something off it.  Ideas are not in shoort supply, repeatable
data is.

This gpsd/ntpd community only just started to gather hard data, instead
of anecdotes, a lot of old wives tales to confirm or debunk.  I keep
finding things, and the get added as NTPsec bugs, so I'll just keep
at it.

> If there would be a strong correlation of TDOP
> with your problem (bird sitting on the antenna) then you could find a
> way to discard the bad TDOP data. Otherwise you would at least know
> that GPS signal quality in your case is not the problem and you could
> focus on software bugs and other possible theories.

For now, I know there is a strong correlation to a kernel stall.  So
that is higher on list.

> I was once doing a map survey with expensive Trimble DGPS equipment. 
> Even with the super expensive stuff we had a DOP forecast telling us 
> which time of the day we could trust the data most. As others said
> the Trimble will even give you a fix inside a metal enclosing. But
> precision will drop.

You'll be interested to know I now am getting centimeter level
accuracy from some $50 GPS.  Takes a master/slave setup, but it works.

>  From a signal theory point of view I would like to know wether I
> have the metal encasing/bird on the antenna or not. This is what DOP
> provides.

Yes.  My antennas are inside, so no birds.  I do see some solar day
effects and temperature effects, but the LF seems to correlate best.  I
can in fact force the failure by forcing high LF.  And since it happens
worse on the Pi2 than the Pi3 I guess I am just not quite fast enough on
the Pi2.

When I get failure, it is only for 3 or 4 seconds, the whole system gets
behind by a second.  ntpd freaks out, jumps the clock, then takes
many minutes to recover.  Removing the cron load removes the failure.

I have several things near the top of my queue to explore this area.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
        address@hidden  Tel:+1 541 382 8588

pgpgRfRmX1Jp4.pgp
Description: OpenPGP digital signature