Rob
The horizon drift is caused be the Calibration not being set up right. Alex has the procedure on his web site.
Cross axis interference.....
I've gone through all that. Either the instructions suck so bad that I can't figure it out, or the AHRS is crap. I think it's the latter. Every video I've seen where the gimbal is being dynamically moved, the horizon goes off. If anybody has evidence to the contrary, I'd love to see it. Most videos show people hovering, or moving slowly. I'm banking and yanking at 60 km/h.
I did the 6 axis calibration exactly as described. Did it many times since the official instructions were not clear, but finally figured it out. Sort of. Maybe. How would we know? The GUI gives us zero feedback.
Here's a couple TBS DP's with off-level horizons too:
http://www.youtube.com/watch?v=CKqsGK8p_H4
http://www.youtube.com/watch?v=xQOpWRIqweE
http://www.youtube.com/watch?v=eIYfn3zxa0A
http://www.youtube.com/watch?v=H9aJjMWJunc&feature=youtu.be
Maybe even Trappy gets it? Interesting his videos rarely show panning shots where you can see the horizon. Except for here and... ah, the world is tilting! Actually, if you watch from that point on, there are a couple shots which have this exact problem.
http://youtu.be/2W3CB4bi_ts?t=2m46s
Interesting thread here. Seems many people have the exact same problem with TBS DP, and nobody has an answer. Not even Trappy who basically says (paraphrasing) "nothing is perfect, and you'll have to edit it out".
http://fpvlab.com/forums/showthread.php?17304-The-TBS-DISCOVERY-PRO-brushless-gimbal-quadcopter&p=336280&highlight=horizon#post336280
Not really wanting to compare to TBS DP again, these machines are in completely different classes, but it's the only other machine I'm seeing flying like this. And they have the exact same problem, even good old Trappy.
I think it comes down to the fact that, it's absolutely irrefutable, that an AHRS cannot work properly without GPS guidance and a compass. There's no way around that.
I have converted my 2 axis gimbals to 3 with a simple heading lock gyro and a fast servo. It is working fine.
Now that's an interesting point that I have been thinking about. There's really no burning need for the pan axis to be on a brushless motor. Assuming the copter has relatively good yaw stability (and most are excellent) you don't need a super fast system. I've thought a servo would work just fine here. And certainly it simplifies the mechanics.
Anyway, I just wanted to share what I've got here. I'm pretty happy with the performance of my system. It's not a bathroom mirror test, or a simple hover test. I took the lens hood off and went out again last night, even better at speed. I'm not having any problems with dampers or "head banging" either. I have the lowest level of vibration of any airframe running Arducopter, but no pitch or roll bounce either on the camera, and that's without even bothering to balance props or even thing about that stuff.
Now I just need to find some interesting subjects to shoot.