Hash: SHA1

Somebody in the thread at some point said:

Thanks for your work Stacy.

|> About the 600sec thing, you can get the GPS chip to report 4 times a
|> second instead of once a second, maybe that can be to do with it.  But
|> it is strange, noticing the heavy filtering on our results I wonder if a
|> lot of those samples coming so quickly were actually synthetic.
| It would seem that they are synthetic and have to do with how GPSD
| parses and interprets NMEA sentences. The other two reference receivers
| are put in SiRF mode and don't seem to suffer this effect. Has anyone
| figured out how to kick the GPS receiver into UBX mode? I would be
| curious to see if that changes anything.

It's actually documented by Ublox in here (Okular eats it)


Here's a UBX packet generator in Bash


and here is how to inject the packets from Ash

echo 0 > /sys/class/i2c-adapter/i2c-0/0-0073/neo1973-pm-gps.0/pwron &&
sleep 1s && echo 1 >
/sys/class/i2c-adapter/i2c-0/0-0073/neo1973-pm-gps.0/pwron && stty -F
/dev/ttySAC1 -echo && cat MY-UBX-PACKET >/dev/ttySAC1 && cat -u
/dev/ttySAC1 | grep -v ^$

|> For the last result, it seems to show only 2 / 2500 results from GPS are
|> within 20 Million meters of the location.  It's hard to square that with
|> the other decent results that have been reported for tracking, including
|> ones pulling maps from SD Card, although I guess you are spamming the
|> card as hard as you can for this test.
| It seems that gpsprof gets a value of 10,000,000,000 for NaN and doesn't
| check if lat/long values are in range before calculating the average. It
| doesn't take very many 10 Billion values to really screw up the average
| :-) I modified gpsprof to toss out invalid lat/long values.

OK makes a lot of sense :-)

| Additional test results (from 20080802) are up at
| http://www.millions.ca/~stacy/neoFreeRunner/gpstale.html
| I did some more testing yesterday after compiling the latest version of
| GPSD for the FreeRunner (it no longer dies when I spam the SDCard...
| that's progress :-) but have not yet got the results pulled together.

Doesn't look a million miles away from being right... there are two
outliers shown and I guess the filtered  NaN(s), but basically all the
other samples are in a decent place.

- -Andy
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org


Openmoko community mailing list

Reply via email to