[time-nuts] GPS Weekly Rollover Fail

[Tom Van Baak]

Bob,

> is this just a known routine bug

Yes, known. Yes, routine. No, not a bug.

GPS system time is kept as a 10-bit week number and 20-bit second number. The seconds count from 0 to 604799 and the week counts from 0 to 1023. What this means is that every 1024 weeks (approx 19.62 years) there is a rollover. [1]

GPS time began Jan-1980 and the first rollover occurred in Aug-1999. The next one is April-2019 and Nov-2038 after that.

These rollovers tend to have zero impact on position and navigation, but it can affect the calculated UTC time in some receivers. This is not due to a flaw in GPS but a lack of rigorous design and testing by the authors of the receiver firmware. Even with simulation it's hard to fully test something that occurs only once every 20 years under all conditions.

We time nuts will be on full alert Saturday evening April 6 to Sunday morning April 7 to find and record any anomalies.

/tvb

[1] https://www.gps.gov/cgsic/meetings/2017/powers.pdf


----- Original Message ----- 
From: "Bob Betts" <rwbetts at sbcglobal.net>
To: <time-nuts at lists.febo.com>
Sent: Tuesday, February 05, 2019 10:11 AM
Subject: [time-nuts] GPS Weekly Rollover Fail


> Hello All:
> Just curious, but is this just a known routine bug that is methodically dealt with or is there a more serious problem in the system(s)? 
> 
> Here's the quote from a notice at www.gps.gov
> 
> Thanks,
> Bob, N1KPR
> 
> <https://www.gps.gov/governance/advisory/meetings/2017-11/powers.pdf>
> ----------------------------------------------------------------------------------------------------------
> 
> Failure: April 6/7, 2019
> 
> •UTC timing displayed and/or time tags of receiver data containing PNT information could jump by 19.7 years, resulting in system failures
> •Any month/year conversion could also fail•Navigation solution should be OK since GPS time is internally self consistent, but associated time tags could be incorrect thus corrupting navigation data at the system level•And the failure is not limited to April 6/7 2019
> •A common fix for week number ambiguity was to hard code new pivot date, which shifts event to unknown date/time in future.
> 
> –December 2014, older legacy USNO monitor receiver failed
> –Feb 14, 2016 Endruntechnology receivers using a Trimble GPS engine failed
> –Aug 14, 2016 Motorola OncoreUT+ older firmware failed
> –July 22, 2017 older Novatel GPS engine failed, notice was posted in Spring 2017 to upgrade firmware, but many did not check
> 
> 
> 
> 
> NOTE Pls add backup address to your phone book: rwbetts at yahoo.com
> 
> http://www.bobsamerica.com http://www.youtube.com/n1kpr 
> 
> Engineering: Where Enigma meets Paradox
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.
>