I would like to share some test results that, to me, say quite a bit about the effects on positional accuracy and repeatability of RTK positions when using Glonass satellites in combination with GPS.
In summary they show that mixing Glonass with GPS makes no significant difference in positioning repeatability, accuracy relative to positions of known control points or initialization times.
The tests consist of several thousand RTK positions determined over a nine day period from April 22 thru May 14, 2008 using the Albuquerque Real Time GNSS Network (ARTGN). The primary reason for taking the observations was to test some new network software before installing it on our system. So actually the tests began as a test to ensure that initialization times were satisfactory and that positions from individual network base stations were comparable in terms of repeatability and accuracy relative to positions derived from raw phase and code observations generated by our network software for a nearby virtual base station (VBS). (I hesitate to use the term VRS as it has now been rather presumptuously trademarked by Trimble Navigation). So initially I was simply alternating between VBS positions and positions from an individual base station. (I use the term 1BASE for individual base station positions). I was taking all positions with both GPS and Glonass. I would take a series of RTK positions using a VBS method then, reinitialize and take a subsequent series using 1BASE. I repeated this for several hours then moved to a different control point. I ultimately went through a set of five existing Albuquerque control points scattered throughout the City. After several days it became apparent that there were no problems with initialization times and positional differences between VBS and 1BASE. Nintey-five percent of the time, initialization times were 10-20 seconds. I was using a Topcon GR-3 rover unit with TopSurv controller software. I broadened the testing to evaluate the differences between GPSOnly positions and 1BASE positions. I did this by switching between GPSOnly positions and GPSnGLonass positions while I continued to switch between VBS and 1BASE.
It was interesting. I couldn’t find any difference. Adding Glonass to the mix had no impact on positional quality and it had only a very minor impact on initialization times. Perhaps 98% of the time, the addition of Glonass made no difference on initialization times. On the very rare occasions when the GPS geometry was extremely poor, the addition of Glonass would usually improve the geometry and generate initialization times between 10-20 seconds instead of 1 – 2 minutes. Incidentally, all the points I used were at least 90% unobstructed.
Frankly I think most of the disagreement about this matter derives more from emotion than reason. So in closing I would like to add that these tests simply show empirically derived results and are not related to political, economic or philosophical considerations about Russia vs. United States, Topcon vs. Trimble or Free Market Capitalism vs. Communism.
If anyone is interested in looking at the details, send me an email and I’ll send several spreadsheets with the results.