[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[gpsd-dev] Raspberry Pi progress
From: |
Eric S. Raymond |
Subject: |
[gpsd-dev] Raspberry Pi progress |
Date: |
Wed, 20 Apr 2016 10:56:57 -0400 |
User-agent: |
Mutt/1.5.23 (2014-03-12) |
I can now see 1PPS with ppstest on the Pi. I have fully documented
the procedure I used to get there in the draft HOWTO. I'll enclose
the current (0.2) draft with this message.
Steps not yet done: (1) GPSD can't yet see the PPS, (2) No instructions
on configuring NTPsec yet.
Gary E. Miller <address@hidden>:
> Yup. We've needed a dedicated RasPi/GPS howto for a while.
That now about 85% exists. The only missing piece is beating GPSD on the Pi
into seeing the PPS signal. Can I delegate that to you while I work on
NTPsec configuration?
> > Evidently there are two steps: (1) Specifying the GPIO pin, and (2)
> > ensuring that the pps-gpio module is loaded.
>
> Uh, more than two. And separate sections for the RasPi3/wheezy,
> RasPi3/jessie, Raspi/wheezy and RasPi/jessie.
All I had to do was declare the logical pin in /boot/config.txt and
force the pps-gpio module to load. ppstest worked immediately after
that.
Policy decision: I'm not going to try to be exhaustive about old
Raspbian versions. Instead, the build will say to upgrade your
installation to Raspbian current as part of the build.
> You missed the part about the kernel patch needed to see both edges of
> the PPS. You have a 50/50 chance you can't see the right edge, and
> no good way of knowing you have the right edge.
I don't know what this means in practice. I gather this introduces an error
or inaccuracy into time reporting, but I don't know what kind of error
or how large it is. The HOWTO should be clear about this and what the
fix for it might be.
> > Since /dev/pps0 is referred to, I gather that doing this and rebooting
> > should enable an RFC2783-style kernel PPS interface watching the GPIO
> > pin. One of the experts should explicitly confirm or correct me if
> > I'm wrong about this, because it affects the explanation I'll write
> > for the HOWTO.
>
> Carefull, as the /dev/ppsX is NOT the /dev/ppsX from the ttyAMA0 or ttyS0.
> This is the ppsx from the gpio pin by way of the flawed pps_gpio module.
>
> This ppsX must be specified on the gpsd command line.
Running
gpsd -N -n -D 5 /dev/ttyAMA0 /dev/pps0
doesn't show PPS events, even though ppstest does. Is there something
else I need to do?
> > gpsd and gpsmon can use an RFC2783-style PPS interface.
>
> Are you sure about gpsmon? Can gpsmon accept TWO devices on the CLI?
> Like this:
> gpsmon /dev/ttyAMA0 /dev/pps
Now that I look, it can't. But I routinely see PPS bars in gpsmon on
snark. Are you telling me this only works when GPSD is using
TIOCMIWAIT rather than the RFC2783 interface? Because if that's so,
it's a bug in our stuff that needs fixing.
> > GPSD's
> > config/ppscheck cannot; instead it relies on a Linux-specific ioctl
> > that watches handshake lines on the serial device.
>
> But that is not needed anyway because you can do this:
> ppstest /dev/pps0
>
> Which will show you the one present, and one missing, edge.
I see that on the Pi, right now.
> > Gary refers to GPIO pin 4. His "dig deeper" source is David Taylor's,
> > which says GPIO pin 18. Examining the traces on my Pi 3 causes me to
> > suspect that it uses pin 4. Did this vary by Pi model? If so, we
> > need to be quite clear about which pin it is on which Pi.
>
> It certainly caries by Hat model. My two hats use pin 4.
Right, I understand this better now and have fully documented it in
the draft.
> >
> > $ sudo nano /boot/config.txt - Add dtoverlay=pps-gpio,gpiopin=18 on
> > a new line.
> >
> > If you previously added bcm2708.pps_gpio_pin=18 to the end of
> > cmdline.txt, remove it. Save and close.
> >
> > $ sudo nano /etc/modules – Add pps-gpio on a new line.
> > Save, close & reboot.
>
> The first is for when the pps_gpio module is compiled in, I always compile
> in my modules. I do not like loadable modules.
>
> The second is for when pps_gpio is a loadable module.
Right. That's the simple case. We are *not* going to assume the HOWTO
readers are building their own kernels on the Pi. If they are, they'll
know eniugh to skip the /etc/modules step.
--
<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>
signature.asc
Description: Digital signature
- Re: [gpsd-dev] 1PPS on the Pi - let's pool our knowledge., Frank, 2016/04/19
- [gpsd-dev] 1PPS on the Pi - let's pool our knowledge., Eric S. Raymond, 2016/04/19
- [gpsd-dev] Raspberry Pi progress,
Eric S. Raymond <=
- Re: [gpsd-dev] Raspberry Pi progress, Gary E. Miller, 2016/04/20
- Re: [gpsd-dev] Raspberry Pi progress, Eric S. Raymond, 2016/04/20
- Re: [gpsd-dev] Raspberry Pi progress, Gary E. Miller, 2016/04/20
- Re: [gpsd-dev] Raspberry Pi progress, Eric S. Raymond, 2016/04/20
- Re: [gpsd-dev] Raspberry Pi progress, Gary E. Miller, 2016/04/20
- Re: [gpsd-dev] Raspberry Pi progress, Gary E. Miller, 2016/04/20
Re: [gpsd-dev] [gpsd-users] 1PPS on the Pi - let's pool our knowledge., Fred Hillhouse, 2016/04/19
Re: [gpsd-dev] [gpsd-users] 1PPS on the Pi - let's pool our knowledge., Jon Brase, 2016/04/20
Re: [gpsd-dev] [gpsd-users] 1PPS on the Pi - let's pool our knowledge., David J Taylor, 2016/04/20