THE GPS CLOCK'S FUNCTION and MORE ROLLOVER TRIVIA One newsgroup poster stated: Quote: ".GPS signals on 22 August 1999...will begin telling their systems that the current date is 6 January 1980." Is this a problem? ---------------------------------------------------------------- Well, yes, sort of. But that's like saying if you only look at the day of the week on your digital watch, after midnight Saturday, your watch is telling you it is Sunday of the first week the watch was made (or better, the first week of Creation, and Adam ain't here yet). Or like saying that a Rolex's big hand is telling you that it is one minute past midnight every time it passes the 12. The week count is, always has been, and always will be a 10-bit number, just like there are 7 days in a week, 24 hours in a day (split into 2 segments of 12, so how do you tell on a Rolex which half of the day you are in?), 60 minutes in an hour, etc, etc, etc. Tell me now, would you reject a top of the line Rolex because it has only 3 hands and maybe a day of the month count? Gee, Joe, it doesn't even have numbers on the face, only a few little glassy looking stones - what do they call them, diamonds or something like that, some kind of crunched charcoal, isn't it? Part of the idea of the GPS was to keep the signal as simple as possible and as robust as possible. So the data rate is only 50 baud. As it is, it takes minutes to transfer the entire message including security coding and error checking, the almanac for the whole constellation, and information to correct for atmospheric effects. If you went to 11 bits, you would lengthen the week count to 39 years or so, and then you would have a "crisis" in 2018 or 2019. You would also lengthen the message by a bit per subframe. I'll leave it as an exercise for the reader (my professoriality coming out again) to figure out how many subframes in a complete message, hence how much longer it would take to do the complete transfer. But, just as it is pretty trivial to figure out whether it is morning or afternoon on your 12 hour Rolex, or which week you are in, or even which month (except that the day count for a month has a variable maximum, depending on the month), it is pretty trivial for the USER to figure out which week count cycle it is in (the "User" is the GPS receiver logic). The original specifi- cation made it very clear that this was for the USER to implement in the receiver if it was important. Recall from our earlier discussions that the week count and message time is only used for the orbit calculation (plus the clock display). The orbit calculation depends on the time difference between the epoch and current time. The pseudorange depends on the difference between the receiver time and message time, which is actually found iteratively from the matrix solution. If there is no correction done, the receiver will be a little confused for a few hours until it receives a post-rollover ephemeris for all satellites used in the calculation. There were a few early receivers that would have that problem (you published a list of them a few times), but recent units (past 3 or 4 years) know about how to figure what cycle they are in. There is absolutely no need to change the week count, just as there is no need to have your Rolex modified to add a few more hands and windows. Now Robinson Crusoe had to notch sticks to count days, as I recall. I guess that must be a deficiency of all the units on the market - they don't automatically notch sticks. 