So I had a very little time for test it, because a quite heavy storm hit my flying place. BUT i had some results.
First, here is a small video....http://www.youtube.com/watch?v=Uo_v6tBXqNU
Do not be missleaded by the altitude changes... the position stayed within a 2meterx2meter box during the video.
The wind was 17-22km/h and increased during the test.
Here is the onboard datalog (20Hz update rate) for the whole poshold period (about 5 min)
Red line is the distance from hold position in cm. the blue and green lines are the lat and lon rate I terms, you can clearly see the increase of the wind during the poshold..
(video is made during the last minute of this log)
And for the uBlox binary... It is awesome. Absolutly no need for GPS filtering anymore, which means that the latency that introduced by the moving average filter could be avoided.
Here is a log from NMEA mode :
Red line is the unfitered position, green is the filtered one. You can clearly see the steps on the unfiltered gps coordinates. These steps cause twiches during poshold, so this is why I introduced filtering.
And this is the output from the same uBlox gps, but using binary protocol.:
Clean and smooth.
So the conclusion : Forget MTK, get a uBlox with a decent patch antenna (25mmx25mm or bigger).
I know that MultiWii 2.1 is feature freezed, but I think it worth to commit the uBlox serial parser. I'll test it some more, and I'll do it during the weekend...