LETS MAKE A MOLEHILL OUT OF AT LEAST
one MOUNTAIN!
12 November 1998
===============================================================
I asked one of our GPS engineering consultants who is well versed in GPS
technology theory and practice to tell us what spectacular event would
happen to a GPS that was NOT "Y2000 compliant". Not much
says he. Read on for further comments. Since he does
not have time to respond to questions from the newsgroup,
he has asked that his name not be published. Here are his comments.
================================================================
Joe, you wanted to post my comments about
the Y2k and EOW "controversy" (let's
see now, that means End of World in Year 2000, doesn't it,
like the television psychics want to predict). Well,
here is The Short Version (consider it an
abstract). Consider this oversimplified. But I note that
at least one of your readers won't
be satisfied with anything less than an official
manufacturer's certifying stamp on each unit, accompanied by
an international press conference and appearances on
Oprah, Nightline, and Larry King Live and attested to
by Bill Clinton (;->).
Hey, ya know what? As far as position and navigation goes,
Y2K (and any other artificial calendar counting time) doesn't matter
one bit to a GPS receiver. The position of the SV is found
from the orbital elements and the difference between the epoch of
the element set as transmitted in the navigation message and the time of
transmission of the message. It's only a _difference_ in time, not
the absolute time. And both the orbital element time and the message
time are contained in the message. The computer doesn't care whether
you are measuring on a Gregorian calendar, a Moslem calendar,
a Chinese calendar, a Jewish calendar or one you just made up. The
algorithm just takes the times given in the message and calculates
the current Mean Anomaly, from which you get the True Anomaly,
from which you ultimately get the position of the SV in
GPS coordinates (NOT lat/lon, UTM, or other geography
units). The only place the date appears in some sort of everyday
calendar format is on your display screen.
And as has been pointed out many times before, the
receiver's clock is, at most, only used to get the initial search configuration.
It is _not_ used for the PVT solution iteration (Position-Velocity-Time).
When you put your receiver in free search mode ("autolocate"),
it doesn't even use its own clock. The "message received" time is
determined as part of the solution in deriving the pseudoranges
for each satellite used. Sort of like the
lat/lon vs UTM in 100 different datums. The number shown
isn't what the computer in the GPS uses anyway - it's only
there for display purposes. The system coordinate
system is an Earth-centered Earth
fixed Euclidean system, no latitudes,
no Eastings. Hey, guess what? It's like GPS time vs UTC
vs your local standard or daylight time. The GPS uses GPS time
for its computations, then displays whatever you
want. In the vast majority of units (all the consumer
toys and virtually all the marine and air navigation units)
you can't display the GPS time, even if you want to. The time units
are 1/403200 of a week, which is about 1.5 seconds, not seconds or
nanoseconds (it's 1/806400 of a fortnight, though (8>D).
The one problem is the week rollover. And, ya know what?
That really is only a problem for a unit that doesn't have a
current ephemeris for a given SV for a short time around
the rollover date. If you are more than a few hours past the rollover,
all the SVs will have a new ephemeris, your internal
clock will be counting up mod 1024 weeks, and you are
fine. The problem comes when you have an ephemeris which is epoch
1023.xxx weeks and your clock has rolled over to 1024.yyy, which means
it reads 0.yyy. It will compute a negative time, unless
your unit has a way of catching that (as do all
units from the major manufacturers in the past 5 or 10
years). But, since the ephemeris is uploaded a couple times
a day, at worst your position computation will be wrong
for a day or so in August 1999. My understanding from the manufacturers
I have talked to (4 of the major
ones), or indirectly from ones Joe, Sam, and Jack have
had contact with, is that current units catch even that small
problem. Some older units will have a hard time locking on because
they are calculating visibilities from the canonical orbital elements
stored in ROM, but once they have locked on and updated
their ephemeris set, they will give the right location
and time of day (but not the right calendar day, just off by about 230
days).
Sigh! What it comes down to is that there are a bunch of
folks who have absolutely no concept of how orbital calculations
are done. Sorry folks, the Earth isn't flat, it isn't the center
of the universe, and even the Sun isn't the center of the universe.
I groan every time I see another of
these anthropocentric, ethnocentric, or worse yet, egocentric postings
that claims that the poster's way of viewing things is the only way,
and the rest of the universe can --- whoops, the universe can't
jump in the lake can it? (My calendar is the only one,
my foot ruler is the only way to measure, my view of the world
is the only correct one -so there!)
======================================================
I edited the above very slightly for clarity(?).
If anyone has questions/comments, please post directly
to the newsgroup.
Joe Mehaffey
Note: One DOD GPS analyst suggests that
the above is overly optimistic and that there are a lot of unknowns out
there in the Y2K arena. Industry people however believe that THE
UNITS THEY HAVE TESTED for EOW and Y2K compliance will work just fine.
I guess the prudent thing is to not have yourself in a life threatening
situation in case of GPS failure immediately before or immediately after
EOW or Y2K rollover. I personally trust manufacturer's engineers
and when they say they have tested a particular unit and all is well,
I fret no more.
Joe Mehaffey