[time-nuts] GPS week number rollover finally bites Garmin GPS 45XL
dave.martindale at gmail.com
Sun May 25 16:54:05 UTC 2014
A few days ago, I took my collection of obsolete handheld GPS
receivers outside and turned them on, to let them find themselves and
collect new almanac data. Most of them probably hadn't been turned on
for 2 years or more. All of them eventually acquired satellites and
started navigating. The old single-channel 45XL took 30+ minutes for
this, the little Sony GPS-CS1 position logger needed only 30 seconds,
and the others (all Garmin 12-channel handhelds) took perhaps 5
Once navigating, almost all of the units agreed that the date was May
22, 2014. All except the 45XL, which insisted on a date of October 6,
1994. Yes, exactly 1024 weeks early.
I bought the 45XL in 1996, my second GPS receiver. It survived the
actual week number rollover in 1999 (with a correctable problem; see
below), and continued to work (as well as a single-channel receiver
ever works) for many years afterward. Thus, the firmware was clever
enough to know that low week numbers actually dates after mid-1999,
not in early 1980. It probably did this with a "birth week" embedded
in the firmware somewhere.
But the "birth week" appears to be 1994 or earlier, since the 45XL
looks at the current GPS week and concludes that today is in 1994, not
2014. I doubt if there is anything to be done about this, since the
45XL predates the Garmin units with firmware in flash. I think the
12XL was the first unit that could be updated by the factory, and
user-updatable units came later yet.
It remains to be seen what the other Garmin units (II+, eMap, eTrex)
will do as they get near 20 years old; I wouldn't be suprised to see
them start reporting incorrect dates as well. They do have flashable
firmware, but I don't really expect Garmin to release firmware updates
for 20 year old units.
About the 45 XL in 1999: At the actual time of the WNRO in August
1999, the 45XL was operating in my car while driving on a highway,
logging the trip. Nothing happened at 00:00 UTC that I noticed. I
turned the 45XL off at about 00:30 UTC (17;30 local time) when we
stopped for dinner. On return to the car, the 45XL would not acquire
satellites, and we went without it for the rest of the trip.
Within a few days, Garmin released a fix - a PC program that
apparently used an undocumented interface command to clear the old
almanac. (I guess the firmware did not handle week number rollover in
the almanac correctly). But with the old almanac cleared, the 45XL
did a cold start, resumed navigating, and continued working for
another decade or more.
More information about the Time-nuts_lists.febo.com