The SA Watch screen clip above represents more than five hours of data for the period ending at 2000 UTC on Sunday, 20 April, 1997. While the plot is rather boring the data is fascinating.
At a sample interval of two seconds this is more than 9,000 samples. And yet, the maximum error at the 95th percentile is a scant 4 meters - considerably better than the specified 100 meter accuracy with SA on. During this period, and for several hours before, the 95th percentile maximum variation of the trailing 100 samples rarely exceeded the 2 meter figure you see here. The 67% share of the samples from fixes with an HDOP value of 1.0 or less is not unusual for my antenna site - if anything, it is a bit low. Data was collected using a Garmin 12XL, a Garmin GA27 antenna and SA Watch.
So what was going on, when did it start and when did it end?
The next three charts fill in much of the detail and provide
grist for much speculation. These charts were prepared by
capturing and exporting GPS data using SA Watch for the period of
0000-2100 UTC and analyzing and charting the data in MS Excel 7.0.
Each chart spans a two hour period of time, centered on 0100,
0400 and 2000 UTC. Each charts plots the moving average of four
calculations:
The sum of the standard deviations of Latitude and Longitude of
each fix from my fixed antenna site.
The "error" of each fix, or distance between the fix
position and the actual location of the antenna
The HDOP of each fix, and
The number of satellite vehicles (SVs) used to (over)determine
each fix.
Here we see the onset of the event occurring at 0100. The smoothing effect of the moving average calculation obscures the fact, clearly evident in the data, that an abrupt change occurred at exactly 0100 UTC. Two added SVs come into view for a total of 9, and the HDOP drops from 1.2 to 0.8, both of which are confirmed by the Navy's Interactive GPS Satellite Prediction service. At the same time the Stdev and Error take a dramatic drop. Are these huge drops simply the expected improvement caused by the addition of two more SVs into the G-12XL's over-determination algorithm?
The plot thickens at exactly 0400 UTC. This time there is no great change in the locally visible satellite constellation to attribute the drop in Stdev and Error, and these very low levels of position variation hold consistently throughout the next 16 hours! At 0400 UTC the SVs used is eight - a very common figure for this location.
Finally, with the tension mounting, all hell breaks loose at exactly 2000 UTC! But first, for a brief period around 1937 UTC, there are as many as eleven SVs in use, with a resulting HDOP of 0.6, again confirmed by the Navy. Then, accompanied by only a modest rise in HDOP, the all-too-familiar wandering of SA makes a sudden, permanent and ugly return with the Stdev and Error figures shooting back to normal levels.
So what do we have to thank for all this unexpected entertainment? The first correlation the author made was to the news flash that the wreckage of the missing A-10 had been sighted in the mountains just 133km West of my location. That announcement, along with the decision that further efforts on the mountain had been suspended due to deteriorating weather conditions, came shortly after 2000 UTC. Had SA been suspended in support of the mixed military and civilian crews making that big, last, desperate push to locate the missing airman? Or was this just a routine system management event that might have gone unnoticed had I not just completed the implementation of SA Watch?
Return to the SA Watch Home
Page.