[time-nuts] week rollovers was Re: OT : New Horizons FEC

Magnus Danielson magnus at rubidium.dyndns.org
Thu Mar 3 20:14:00 EST 2016


Jim,

On 03/03/2016 05:28 AM, jimlux wrote:
> On 3/2/16 1:52 PM, Tom Holmes wrote:
>> While we are off topic, just wanted to note that Jim Lux was mentioned
>> and pictured in an article in the latest CQ Magazine for being the
>> keynote speaker at the Microwave Update  (MUD) conference last fall.
>>
>> Thought the hams on the list would be interested.
>>
>> Thanks, Jim, for being more than just a time nut!
>
> I think we're all more than just time-nuts...
>
> Folks on the list will get a kick out this:
>
> At JPL, I'm project manager for a cubesat that will do some measurements
> of HF signals, and it's all synchronized by GPS. That is, you tell it
> "at GPS time X, record Y seconds at frequency Z".
>
> We get GPS time from the spacecraft as week and milliseconds of week.
>
> There was a huge discussion at a design review (prompted by my section
> manager, Charley Dunn, who has had a big role in most of JPL's GPS
> receivers, blackjack, rogue, turbo rogue, etc., and who has had his
> share of synchronization and roll-over issues) about whether the week
> number was 0-1023 or since 1980 epoch, and how would we handle roll over
> (in 2019.. we might still be in orbit and functional).  (week 860 having
> recently occurred, of course)
>
> And then a whole bunch of questions about "how will you test that you
> are synchronized properly", since we claim to start the sampling within
> 1 microsecond of the specified time.
>
> All grist for the time-nuts list that has been covered over the last few
> years in one form or another.
>
> Most of the other review board members are going "why is this is an issue".
>
> Those of us who are time-nuts, though, know that this is the issue. Is
> the message before or after the 1pps tick? What happens if you miss a
> message or a tick? what if the message overlaps the tick? what if the
> week rolls over during a measurement?   We've all been caught.

Just as GPS itself uses scoped time to announce things, you can. A 
future order express it's time in GPS week and miliseconds of week, and 
it will be valid then. It will not be very hard to solve. It does 
however make some people go completely nuts as they are not sure.

You can upload through telemetry the year, from which GPS week 
unwrapping can be done if needed. This only to show it will sort itself out.

Cheers,
Magnus


More information about the time-nuts mailing list