Re: [gpsd-users] 1PPS works on CH340G based usb-serial converter

[Chen Wei]

On Tue, Mar 24, 2015 at 11:48:19AM -0700, Gary E. Miller wrote:
> > The GPS is a ublox-5 module. The 1PPS wire is soldered to CH340G's DCD
> > https://dl.dropboxusercontent.com/u/509097/CH340G-usb-gps-1PPS.png
> I do not understand your setup.  Can you explain a bit more?

The ublox-5 module has 6 pins: 5V GND TX RX PPS BOOT. BOOT is unused.
"CH340G-usb-gps-1PPS.png" shows the USB-Serial converter, which has 5
pins: 5V 3.3V GND TX RX, but doesn't has PPS pin. The red wire is
directly soldered to CH340G's DCD pin, which is Pin 12 of the IC. This
red wire connects to GPS module's 1PPS output pin.

The complete assembly:
https://dl.dropboxusercontent.com/u/509097/Complete-CH340G-usb-gps-1PPS.jpg

The USB-Serial converter is covered with black tape in above picture.


> > The Type 22 reference clock of NTP is not working,
> That would be an ntpd problem, not a gpsd problem.

Turns out the problem is my ntp.conf. After "prefer" a peer other than
the clock from Type 22 driver:

server 127.127.28.0 minpoll 4 maxpoll 4 prefer

The NTP Type 22 clock is working.

> > probably because the 1PPS from this
> > USB-GPS combination is not good enough, which indicated by the kernel
> > error message "hardpps: PPSJITTER: bad pulse".
> Garbage in, garbage out.  Any idea why?  Have you put a scope on it?

No scope.

> With some limitations.  The pps-gpio driver only timestamps the
> asser edge, not the clear edge.  I jsut reported that bug to the

The rising/falling puzzles me. This ublox-5 modules is configured to
"synchronized at rising edge, pulse length 100ms", but if the ntp
capture on rising by set Type 22 clock flag2 to 0, the time is 100ms off
as compare to time from ntp poll servers.

> > The jitter of this USB-GPS is around 300-700 micro-seconds.
> Which jitter, the serial time jitter or the PPS jitter?  I assume you
> 

The jitter from ntpq output.

> How long did you let the system run?  I find it takes days for a RasPi
> clock to settle down.

It is connected to a laptop, which runs less than 12 hours a day. The
PPS jitter reaches 300-700us very quickly, less than 30min I think. In
the end of the day, the jitter doesn't improve much.  The jitter is
lower when this GPS module connect to RasPi's GPIO, I don't recall how
long it takes to settle down, but my impression is it doesn't need days.


-- 
Chen Wei