"Eric S. Raymond" <address@hidden> wrote:
Sanjeev Gupta <address@hidden>:
I built gpsd from git, and ran it on my production box, which has a device:
http://www.sureelectronics.net/goods.php?id=99
I believe it is a MTK chip. It is detected as "Generic NMEA" currently.
Running the git gpsd, and the git gpsmon, does not show any PPS lines
Please see:
1. gpsd-3.10.txt : This is the output of gpsd. Even at D 5, I see no
PPS messages
2. gpsmon-3.10.txt : This is the output of gpsmon. Note no PPS lines
3. gpsd-ntp.txt : And yet, PPS _is_ working.
There is reason to doubt 3.
Again, this is on a device
connected via Serial port, no USB.
gpsd:PROG: KPPS assert 1383888516.000032181, sequence: 1 - clear 0.000000000, sequence: 0
gpsd:PROG: KPPS data: using assert
gpsd:INFO: KPPS cycle: -2147483 uSec, duration: -2147483 uSec @ 1383888516.000032181
gpsd:INFO: PPS cycle: -2147483 uSec, duration: -2147483 uSec @ 1383888516.000052222
gpsd:PROG: KPPS assert 1383888516.000032181, sequence: 1 - clear 1383888516.100033998, sequence: 1
gpsd:PROG: KPPS data: using clear
gpsd:INFO: KPPS cycle: -2147483 uSec, duration: 100001 uSec @ 1383888516.100033998
gpsd:INFO: PPS cycle: -2147483 uSec, duration: 99996 uSec @ 1383888516.100048735
Those messages show that gpsd is picking up at least some features of the
PPS pulse. But the absence of any messages that include the string "PPS edge"
is consistent with no PPS events appearing in gpsmon - the hook function that's
supposed to be triggered by a PPS edge transition
never fires.
I think what must be happening here is that the particular shape and
duration of the MTK's pulse is defeating Gary's edge-detection logic.
Thus, unit 0 information is being passed to NTP *but not unit 1* - it
could still look like it's working if the device's latency between
top of second and start of message transmission has little jitter.