On May 1, 2000 the intentional degradation of GPS signals (called SA) was terminated. With this change a GPS without a DGPS unit attached became almost as accurate as a GPS with a DGPS attached. What follows is the result of 2 separate 24 hour tests, one with DGPS and one without.
The tests were both done using a Garmin 12xl GPS (sw. version 3.53) outputting NMEA data which was logged to an attached NetBSD computer. The GPS antenna was a roof-mounted Trimble/Lowe antenna with 15 meters (50 ft.) of RG58A/U connecting the antenna to the GPS.
For the second test a Garmin GBR-21 DGPS radio tuned to the Point Blunt, CA Coast Guard station was used to provide the DGPS corrections. The Coast Guard station was located a mere 53 km (33 miles) away and provided the faster (thus lower delay) 200 baud corrections.
samples: 43478 (24.2 hrs)
ave lat: 37.55 degrees
ave lon: -121.94 degrees
ave alt: 39.474 meters
50.00% confidence: 2.5 meters
68.27% confidence: 3.8 meters
95.45% confidence: 7.0 meters
99.73% confidence: 9.8 meters
Alt
50.00% confidence: 5.6 meters
68.27% confidence: 7.4 meters
95.45% confidence: 14.4 meters
99.73% confidence: 21.3 meters
What this means is that a user taking a single non-DGPS correction can expect the reading to fall within 2.5 meters of the correct position 50% of the time. Similarly the reading should fall within 7 meters of the correct position 95% of the time.
The noise level recorded by this test is so low that one should be suspicious that the receiver firmware is pulling a fast one on us. The receiver is probably detecting that it is stationary and using that information to filter the heck out of the lat/lon/alt numbers. I would not expect to get this sort of an accuracy out of a non-stationary test.
samples: 43156 (24.0 hrs)
ave lat: 37.55 degrees
ave lon: -121.94 degrees
ave alt: 30.727 meters
50.00% confidence: 1.6 meters
68.27% confidence: 2.2 meters
95.45% confidence: 4.2 meters
99.73% confidence: 6.7 meters
Alt
50.00% confidence: 1.8 meters
68.27% confidence: 2.7 meters
95.45% confidence: 6.4 meters
99.73% confidence: 12.0 meters
What this means is that a user taking a single DGPS reading can expect the reading to fall within 1.6 meters of the correct position 50% of the time. Similarly the reading should fall within 4.2 meters of the correct position 95% of the time.
The exact position of the rooftop antenna is not currently known. The position of the rooftop antenna is derived as the average of the 24 hour position. If there are any biases in the system, or if the GPS were to have an error in calculating its position correctly then some undetected errors could sneak in.
The differentially determined position and the non-differentially determined position differ by 2.8 meters. Approximately 1.4 meters of offset were expected due to the fact that the Coast Guard adds an intentional offset of 1.4 meters to their reference station's position. The source of the other 1.4 meter error is unknown. In any case the two setups should be looked at as having a slightly different datum setting.
Similarly to the horizontal offset, there is a 8.75 meter offset between the two altitude averages. The source of this offset is also not known.
Garmin NMEA issues: Garmin's NMEA output is limited to 0.001 minutes of arc. This means that any position wandering within a 1.8 meter by 1.4 meter box could go unnoticed. The receiver's noise could well be +/- 0.7 to 0.9 meters more than the graphs have one believe.
Further tests using a Motorola Oncore receiverwith a much higher 1cm reporting resolution are currently underway.
The above statistics were gathered with a home grown unix program I call salog. Feel free to grab the above distribution and run your own statistics. It should compile up with little difficulty under netbsd, freebsd, openbsd and linux.
.
![[ Powered by Fedora Core ]](/icons/powered_by_fedora.png){kind=icon}
wolfgang.rupprecht+web@gmail.com
(Wolfgang S. Rupprecht)
WSRCC Home Page || Up One
Level
..