GPS Week Roll Over

I guess many of the activators use a GPS for navigation. If you have an old one this may be worth reading up on and watching for from the 6th April 2019.

A few GPS vendors wrote their software to be aware of rollover of the week number. I believe (and will find out soon) that my Garmin should be OK. Its software dates from 2009 and when released it knew the week number. It knows then that week number > releasedate week number means up to April 6 2019 and week number < releasedate week number means add 1023 to the week and get a date after April 6th 2019.

We shall see…

Latest GPS data format has a 13bit week number which gives 8191 weeks or 157 years to rollover compared with the old 10bit number that gives 1023 weeks or 19.6 years.

I hope my 1965 Ordnance Survey Sheet 72 (Girvan) will still be OK, hasn’t failed me yet. :wink:
It has so many new marks on it, fossil sites, locations for gold, lead, copper, agates, etc. It has served well for so long.


We’ll all have died a firey death by then in the 2038 32 bit time_t rollover

So I have 19 years to update from Windows XP. I do like to get value for money. :laughing:

1 Like

Wrong OS.

I still have, but no longer actively use, a Garmin GPS12 which must date from before the previous rollover in 1999 (I have on file waypoints dated in 1998). I must try firing it up so see how it behaves. Of course it might have died for unrelated reasons.

My understanding is that the GPS timing system is self-consistent, and that the worst that should happen is the displayed date and time being wrong. Positioning accuracy should be unaffected, I think.


That’s what I believe will be problem will be Martyn, position OK, dates wrong. There could be problems with emphemeris downloading etc.

Please translate into English.

Walt (G3NYY)

OK, let’s try that in English! (My VPN makes Google think I’m either in Vienna or Chisinau, Moldova and I get German results sometimes.)


You got me worried now…my GPS is a Garmin Etrex Vista C. I’ve got saved tracks from as early as 2005!

This could be the end…


It won’t be hard to take a GPX file and add 1024 weeks to all of the timestamps.

The way the GPS time handling works suggests to me that GPS receivers should get the time of day display right, even to the handling of leap seconds, since the cumulative leap second count is transmitted in the data stream. What is likely to go wrong is the display of the date. Ancillary calculations such as sunrise/sunset times are likely to go awry in consequence.

Whilst there is always scope for catastrophic failure when something unusual happens, unless a device totally bricks itself it should be possible to recover it and have nothing worse than an incorrect date display.


Just to report:

My ancient GPS12 (firmware date 2003) is currently fine. I eagerly await the rollover to see what it does then.

While thinking about this issue, it struck me that only the dullest of firmware writers could fail to realise that it only takes a tiny amount of non-volatile memory to keep track of the current GPS epoch and thereby follow the date correctly through an indefinite number of rollovers, provided that the device does not remain switched off for the entirety of an epoch. You would only need to rely on the firmware date if you had to do a full factory reset.


Well, the rollover has happened: today is cycle 2 week 0 day 0.

My ancient GPS12 seems to have got the date correct.

However my slightly newer Vista Cx, which no longer gets out much because it has lost its waterproofing, has gone wrong. On first start up, it set the date to 2038-11-21, and then next time it chose 2078-02-20 which it now seems to have settled on. Definitely not what you might expect. As predicted, the positioning and time display is fine, so the device would still be OK for navigation.

My newer devices seem fine. Of course things can still go wrong at a later date.

Martyn M1MAJ

I just dug out my original Garmin “Yellow” Etrex and put some batteries in it.

It took about 30 minutes on the window sill before it got a good lock but now appears to be working correctly including the date and time.

1 Like

I tried both of my Vista HCx’s on Friday and everything was fine and have tried them both now. The main one was off but had batteries in it, the backup (a £35 buy from eBay) was on a shelf without batteries for months and months. Powered up both now and both think it’s the 7th April 2019. Not sure why your Cx gets the date wrong as it’s the same vintage and model range as my HCx, IIRC yours uses a SiRF III chipset and mine use MediTek chips. Maybe yours did a real full cold start?

I powered it up again (under cover) with the intention of checking whether I’d missed any firmware updates. I immediately noticed that it had the correct date! I let it see the sky, and the satellite fix took ages; maybe it had previously discarded the almanac. But eventually it got a fix and the date is still OK.

The evidence of the glitch remains as a couple of track files on the flash card:


Not surprisingly, there aren’t any firmware updates, but it seems that the device lives on!

Martyn M1MAJ

My Garmin Oregon 450t which is about 15 years old no problems when I put batteries in it yesterday. Took a long time for it to lock on.
Date and time are correct.


My Garmin Etrex Vista C from 2005 has time, coordinates, and date all correct yesterday and today, April 6 and 7. It finally went over to Daylight Savings Time, about 2 weeks late, but normal with the original firmware - does this every year.

Perhaps it will do something bad later, but so far, so good!