GPS accuracy testing at 0 degrees East

…When East Isn’t East.
Where better to test the accuracy and resoution of your GPS devices than the agreed ‘Prime Meridian of the World’ at the Royal Observatory in Greenwich, where east meets west at zero degrees of longitude? I thought that if I was to place a GPS reciever on the handily marked meridian line, I would easily be able to see how close the reading would come to zero degrees in one axis. I could then compare the results of the built-in GPS receiver in the Nexus One handset (same chipset as the HTC Desire – a Qualcomm gpsOne)  with the Nokia LD-3W bluetooth GPS with EGNOS (see previous post), and use this as a baseline for further testing as improved GPS units become available. This would also be an opportunity to compare the results from the tried and trusted ‘GPS Status‘ application with the new ‘GPS Averaging‘ application.

Equipped with a warm coat, camera, Nexus One, HTC Desire and Nokia LD-3W, I climbed the hill to the observatory, set the Nexus One on the marked meridian line and turned on the GPS.

Placed nearby the intersection of the marked meridian line and the perimeter fence kerb, the Nexus One quickly acquired a GPS fix, displayed in GPS Status as N 51.47801 W 0.00152 with an approximate error of 3m and a fix from 4 of 12 satellites. The next test, using the Nexus One and ‘GPS Averaging’ at the same spot delivered a reading of 51 degrees 28.681 minutes north, 0 degrees 0.087 minutes west, over 151 samples with no option to show decimal degrees as in test #1. Setting the Nokia bluetooth GPS in the place of the Nexus One, turning off the  phone’s GPS receiver and connecting to the Nokia with the ‘Bluetooth GPS Provider’ application, I noted it reported 9 satellites in view and an accuracy of 1m. ‘GPS Status’ claimed it was fixed from 5 of 11 satellites, an error of 0m and a postion of N 51.47804 W 0.00143. ‘GPS Averaging’ then made 166 samples to complete the testing, showing 51 degrees 28.683 minutes north, 0 degrees 0.086 minutes west.

So, how come the results were not at E 0.0000 in any of our tests? It turns out that the widely advertised ‘Prime Meridian of the World’ is based on a definition of the Earth’s shape called the Airy Ellipsoid, derived from ground-based observation during the 1830s. The advent of satellite surveying led to a different model for the Earth’s shape being used, with the result that the Greenwich Meridian would be displaced eastwards by around 5 seconds, or around 102m. This was inconvenient for the Ordnance Survey, who continue to use the OSGB36 co-ordinate system, based on Airy’s observations, for their mapping. The satellite-derived WGS84 (and its European equivalent ETRS89) system with its shifted meridian is used by most GPS naviagation software, the Admiralty for its charts, the ICAO for international air navigation.

To show these readings on a map I needed to convert the degrees and decimal minutes notation produced by ‘GPS Averaging’ to degrees, minutes and seconds by multiplying the minute fraction by 60, and then using this tool to convert the minutes and seconds into decimal degrees:
Test #1 (Nexus 1 GPS / GPS Status): N 51.47801 W 0.00152
Test #2 (Nexus 1 GPS/GPS Averaging): N51.47801 W 0.00145
Test #3 (Nokia bluetooth GPS/GPS Status): N 51.47804 W 0.00143
Test #4 (Nokia bluetooth GPS/GPS Averaging): N 51.47805 W0.00143

View Meridian GPS test in a larger map

Update:

I was struggling to interpret these results. It seemed from the Google Maps that Test #1, which should have been the least accurate in theory, was closest to the estimated position on the Google photo and that the two results (Tests #3 &4) from the EGNOS-assisted Nokia GPS were clustered several metres away to the northeast. Could it be that Google’s imaging was inaccurate, or was the built-in Nexus One GPS closer with a ‘lucky fix’ than either an average of around 150 samples from either the Nexus or Nokia receivers? D.M. Stillman’s paper ‘Plotting Surveying Data in Google Earth‘ helpfully points out that distortions of the Google image data to fit known features may lead to a difference between a position and its representation on Google Earth overhead imaging. Leszek Pawlowicz has had similar issues with positioning on Google Earth with low-cost GPS devices, and has written some great articles on getting the most out of consumer-grade GPS (parts 1 & 2),  and made some comparative tests of consumer GPS devices , attempting to assess their Circular Error Probability.