Re: [gpsd-dev] Where is your latest refclock_gpsdjson.c that I should be testing?

[Gary E. Miller]

Yo juergen!

On Fri, 03 Apr 2015 23:52:01 +0200
juergen perlinger <address@hidden> wrote:

> Thinks worked out a bit faster than initially expected... happens
> seldom enough. The latest shot is in
> 
> http://people.ntp.org/perlinger/for_test/refclock_gpsd/ntp-4.2.8p2-RC3-gpsdjson.tar.gz

Great, I'm grabbing a copy now.

> Gary, just one question: Would it have been easier to provide an 
> 'ept'-like precision estimation?

Not much point.  The precision is about -1 or -2.  Since ntpd seems to use
the precision as a floor, jjust hard code -2.

> Just parsing an integer is easier
> for me, of course, but I have to convert the ept value to a precision
> anyway for the serial data. Having one style of data for timing
> precision might be a tad more consistent on both sides.

ept has zero correlation to the quality of the TPV offset provided
to you.  There are two main components to GPS serial time:
        1. chip vendor
        2. serial port speed

For #1, I have latest generation GPS chips here, both running at
9600.  For one I get the TPV with a delay of about 100 mSec and a
jitter of about 50 mSec.  For another I get a delay of about 550 mSec
and a jitter of over 200 mSec.  Pretty much -1 or -2.

For #2, some people run their GPS at 4800 (NMEA standard), others at
9600, 38400 or much higher.  If the first TPV message is 50 bytes (500
bits w/ start & stop) then the transmission time differs by a factor
of 8, almost 10% at the assymptote.  So pretty much -1 or -2 again.

These erros are orders of magnitude worse than ept.

gpsd sometimes attempts to adjust the fudge depending on chip and speed
but is a likely as not to makes things worse instead of better.

So, in the present case ept is meaningless.  Just skip the ept/precision
and go with -2.

> Feel free to forward this to anyone who might be interested at your
> own discretion.

I'm copying gpsd-dev, there are several interested folks there.

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