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

[Eric S. Raymond]

Hal Murray <address@hidden>:
> > I hope you're home this weekend, Phil, because I have a workout in mind for
> > that fancy diagnostic scope of yours.  We need to measure the PPS pulse
> > width on the Adafruit HAT.  If possible we need to figure out whether it's
> > he leading or trailing edge the kernel presents to the RFC2783 interface.
> 
> It doesn't take a fancy scope.

I was pretty sure it didn't.  That was me gently teasing Phil about
how much he enjoyed showing off his newest toy last Saturday when my
wife and I were over at his and his wife's place for boardgaming.
"Betrayal at the House on the Hill", if that means anything to you.

To be fair, it's a damned impressive piece of kit even to a duffer
like me who knows very little about oscilloscopes - very capable,
portable, ruggedized, obviously designed for serious industrial
troubleshooting in field conditions (which is what Phil does).

(Mark, you'll get to meet Phil at Penguicon. He's a friend of the
project and a close associate of mine and Susan Sons's.  He is the guy
who taught me enough soldering that I could assemble the HAT.)

> > We need the same information for the Uputronics HAT. 
> 
> Both are triggering on the rising edge.

I want to be very sure I know what you're asserting here.  Are you
saying that you *know* the pps-gpio driver reports the leading edge in
both cases?  While this is certainly plausible I'd like to know your
authority for it.  It's a crucial, delicate point on which I would
hate for us to inadvertently spread misinformation.

Once verified it would be excellent news for the composition of the
HOWTO.  It would mean I could exile the complications about edge
detection to an appendix that becomes relevant only when we have to
qualify a new HAT and check that the pps-gpio driver doesn't trigger
on trailing edge.

> The default pulse width should be in the data sheets.

I see one for the module in the AdaFruit HAT here:

https://cdn-shop.adafruit.com/datasheets/GlobalTop-FGPMMOPA6H-Datasheet-V0A.pdf

There's a functional description of 1PPS on Page 7, but the description of
the 1PPS pin on page 13 doesn't give pulse timing.  It does claim 10ns
typical accuracy,

> High end gear (GPSDOs) use narrow pulses.  Low cost gear uses wide pulses.  
> Anything over a few ms you can see with NTP.

Makes sense.

> Lab gear that uses PPS all agree to trigger on the rising edge.  I can't 
> think of any exceptions.  (But I'm sure there is lots of gear using PPS that 
> I don't know about.)  Where you get into wrong edge problems is if you add 
> TTL/CMOS to RS-232 level converters to connect low cost GPS toys to your PC.  
> Connecting a TTL signal directly to a RS-232 modem control input pin ends up 
> triggering on

This paragraph seems incomplete.

> [I haven't confirmed things with a scope.  I will if you think it's 
> important.]

I'll let Phil deploy his this weekend and have an educational experience
watching him do it.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>