Hello,
I did a lot of testing in the purpose to solve those big GPS error jumping 10 or 20 meters away from the real position.
I arrived to the conclusion that the M8N GPS is not the most accurate, at least for my location (France). I got the best result with an old NEO-6 GPS.
The last test I did was on the "next" release on a revolution board version but I got similar results with the 15.09 version. The INS13 algorithm was used.
As the NEO-6 GPS unit I have does not have a compass I used a HMC5883 module connected to the I2C Flexi port.
I got a position within +/-3m as expected after replacing the patch antenna by a homemade backfire helix antenna, details can be found here :
https://librepilot.atlassian.net/wiki/display/LPDOC/Homemade+Helix+GPS+antennaThe satellite reception level is slightly better but what makes the difference is the better indirect path rejection.
When the tricopter is in the air the accuracy is even better, this might be due to the dynamic behavior of the INS13 algorithm or a better satellite reception.
Joe.