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

[Gary E. Miller]

Yo Miroslav!

On Wed, 17 Aug 2016 08:36:06 +0200
Miroslav Lichvar <address@hidden> wrote:

> On Tue, Aug 16, 2016 at 11:39:30AM -0700, Gary E. Miller wrote:
> > On Tue, 16 Aug 2016 09:32:37 +0200
> > Miroslav Lichvar <address@hidden> wrote:  
> > > No, it's an Intel machine with active cooling. On the graph you
> > > can see the frequency changes rapidly in both directions. On a Pi
> > > with no cooler the temperature goes up quickly, but it cools down
> > > slowly, so the offset swings only to one direction.  
> > 
> > My take is totally different, the ints on the Pi are slow, so when
> > the LF goes up, the PPS pulses get delayed.  I can easily force
> > this condition and see it.  The time spike is way to short to have
> > any relation to cooling.  
> 
> If the error was due to a delay in capturing of PPS timestamp, I think
> the offset would change in steps in both direction. In your graphs the
> offset is changing relatively slowly. To me it looks like a frequency
> error, most likely due to the temperature change.

I'm not sure we are talking about the same thing.  I am seeing 10 second
steps every 30 mnis when the graphs run.

Maybe you are looking at the daily swing?

In any case, I have a server in a cool basement, with much smaller temp
swings.  That will help disambiguate temp effects.

> Also, in your graphs I can see the spikes are smaller when the
> temperature is highest.

Odd, I'm not seeing that?  But if you look at today's graphs, you see
the LF effect.  I'm now doing the weekly stats every two hours, so that
spike is big, and the daily stats every thirty mins.  Daily spike much
smaller, so perfect correlation with LF.


> > > Well, yes, that's the point I was making. You need to keep the
> > > temperature stable (e.g. not running any CPU-intensive tasks) if
> > > you want ntpd to work well.   
> > 
> > Once again, not related to temps.  The time constant of the temps is
> > much longer than 10 seconds.  If you notice the thermometer at the
> > bottom you can see that.  others have used the onboard Pi temp and
> > got the same results.  
> 
> It seems the temperature in your graph is collected by the cron job
> itself (every 30 minutes),

It was, but I changed that yesterday.  The TEMPer TC is so long the
effect not very noticeable.

> so it won't show the short-term variations
> that ntpd is chasing.

It is room temp, no way to get CPU temp on a RasPi.  I could on an
Intel, but I do not see this on an Intel.

> The temperature needs to be logged independently
> from the cron job and at a higher rate. You know what Nyquist said...

Sadly, it takes longer to take temperature than the timestamp spikes I
am seeing.  And the TC of the TEMPer is pretty long.  No way this
room temp USB will ever catch 5 second temp spikes.

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

pgpLU9Sk3mnFH.pgp
Description: OpenPGP digital signature