Re: [gpsd-dev] SemPiTernal - Bounding PPS uncertainty

[Gary E. Miller]

Yo Hal!

On Fri, 22 Apr 2016 06:07:40 -0700
Hal Murray <address@hidden> wrote:

> address@hidden said:
> > If you recall, we had an issue with connecting the Garmin GPS
> > hockey puck to a serial port - even after converting the 0 to 5VDC
> > data lines to RS-232's -10 to +10VDC, we still couldn't read it,

RS-232 is +/- 12V.

> > because all the data was inverted. That's why I built that little
> > PC board which fits in the blue plastic case
> > - to convert NMEA GPS data lines to RS-232 with the PPS line
> > included.  It's got _two_ chips - an inverter, and a voltage level
> > shifter.   
> 
> It also works if you just wire it up without either chip.

It depends on your particular RS-232 port on your motherboard, and which
Garmin model.  But to acutally meet the specifications of the Garmin
and of RS-232 you need the level shifter.  On the bench I'll take
whatever works, but in public I want it done by the book.


> address@hidden said:
> > The issue here seems to be not so much what the GPS devices
> > themselves do, nor what the kernel does - they're both well
> > defined.  What's kind of random is how things get connected
> > together - does everyone who's making these hats, coats, shields,
> > etc. pay attention to making sure the PPS signal always hits the
> > CPU in such a way that the edge the GPS chip thinks is top of
> > second is the edge the kernel thinks is the top of second?  

The kernel does not care.  gpsd or ntpd cares.

> In general, all sorts of scrambling happens in there.
> 
> GPSD does the right thing automagically.  ntpd requires you to tell
> it which edge to use.  That assumes the kernel supports both edges.
> 
> The broken kernel is not a problem for ntpd with either of the 2 HATs
> that I know about because they got it right.

There is no 'right'.  Or, when there is, the 'right' way is ALWAYS
to make the clear the important edge.  The clear edge is always
as sharp, or sharper, the the assert edge.

gpsd user have both assert and clear, so any howto needs to explain
both.

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

pgpyBLfTPHphZ.pgp
Description: OpenPGP digital signature