gpsd-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gpsd-dev] Refclock cleanup


From: Gary E. Miller
Subject: Re: [gpsd-dev] Refclock cleanup
Date: Thu, 5 May 2016 14:39:44 -0700

Yo Hal!

On Thu, 05 May 2016 14:26:39 -0700
Hal Murray <address@hidden> wrote:

> address@hidden said:
> > gpsd can generate a huge amout of raw logging.  Seeing the
> > subsystem you want take hacks like this:
> >     gpsd -nND 9 /dev/ttyS0 |& egrep '(PPS|NTP}'   
> 
> I don't call that stuff "logging".  It's debugging cruft.

Well, debugging yes, cruft no.
 
> By logging, I mean stuff you can feed to gnuplot.  Things like how
> many satellites are in view or the parameters from loopstats and
> friends.  It's a compromise between saving everything and nothing.

Well, that you can do now.  Check out gpsprof.  It is a nice little program
to read gpsd JSON, and send it to gnuplot.

See attached.  30,000 points in a scatter plot, with CEP()'s.

I would love some Python guy to improve gpsprof to do real time plots.

We recently had some guys write a small C program to read the JSON, 
convert to an snmp style key/value pair of the sat bire, and feed
that to mrtd/rrd.

Something in my inbox that I need to add to gpsd/contrib.

> > Which of course perturbs the PPS timing you are looking at.  
> 
> Why would that perturb PPS?  Just do the printout after you have
> captured the time stamp.

gpsd logging can consume vastly more CPU than gpsd with no logging.
Each of those outputs is a system call, and some of them are in the timing
path between data reception and data time stamping.

Plus all that data either goes out short packets on the ethernet, or
data bloks on a disk.  Either one creates heat, so the XO perturns, and
lots of interrupts and syscalls.

On an idle server I can approach 1 microSec jitter from PPS.  If I do
much of anything at all that blows up to 20 microSec, or more.  Splitting
hairs, yes, but I like to see how fine I can split that hair.  And nowhere
near the theoretical 15 nanoSec PPS I'd like to see.  :-)

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

Attachment: ns-raw-rover-3.png
Description: PNG image

Attachment: pgpsJZoJCw6bl.pgp
Description: OpenPGP digital signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]