yes, I know but can rotating landing gear (I guess we speak about yaw motion) affect the roll? If this is the case what would be the best idea to at least try to clean such relation up?
From my observation yaw seems to be rock solid ad the only issue I can see is roll.

I think filtering could help somehow but I have no idea on how to figure out the settings.

Did anyone try that new FW versions with 'analyse" options? I'm wondering how precisely we can narrow problematic areas of our setups?

If there is no suggestions, shall I get rid off that landing gear and use foldable legs instead?

while moving by hand I cant feel any resistance. everything seems to be absolutely OK. What I have noticed though is that when I knocking the roll gage I can easily take it out of balance - much easier than in pitch or yaw case were I need to use a bit more force to take it out of steady position. This made me think that the motor is not strong enough.
Later on today I will try to post some videos of my setup, BGC settings and then the video taken - maybe this will give you an idea.
I was already thinking to get a bigger motor for roll but I don't really know what motor will be good and strong enough.

I manage to get quite a decent results. After rebuilding whole gimbal's mechanical setup from scratch and perfect balancing the stabilization performance is where it should be .. almost. All seems to be fine except one case ... there is a bit of ROLL shake time to time .... it's most visible when camera is pointing down. I'm now wondering ... is it possible that air coming from main air frame motors is interrupting roll performance? From the logical point of view it make perfect sense. The roll bar is placed perpendicularly to the air flow coming from props therefore it's much more prone to wind (air flow) distortion than any other axis. When the camera is looking down then roll surface is even bigger which makes it even more roll sensitive - is my way of thinking correct ?

I currently fly 800 mm hexa frame with 14 inch props .. this gives me about 430 mm distance from prop to prop and I starts to think it's not enough. I'm thinking on arms extension to 900mm or even going to something like Tarot X6 where all arms are raised a bit causing by the same motors airflow goes a bit out side rather than directly down ... do you guys think this would help?

Thanks a lot Garug, as always valuable piece of information from you.
The gimbal is quite compact I would say (some pictures already included above) but the second point took my whole attention. Have you tested encoders on your gimbals? Is it really worth to get them on the gimbal? I was googling for some info regarding this topic but haven't find much to be honest. If you have any valuable links which can be shared I will be more than grateful.

Yes this is exactly the setup I have. I guess you might be right about the landing gear attached to vibration dampers. Need to test it somehow. Maybe I will attach small mobile with accelerometer to the gimbal and try to log shaking move while flying - not sure whether this will show anything but at least it's a good point to start with.

Coming back to encoders. So from what I have understand I need to get a gimbal with encoders already build in ... there is not really possible to get third party encoders and put them on the frame above - am I correct?

I will look around and maybe some usable solutions are out there ;)

Small cameras and gimbals are prone to vibrations and rolling shutter (jello effect), so they are a bit more demanding in terms of balance, stiffness, smoth and proper operation and fine-tuned settings than those bigger, heavier setups with GH4 or Red cameras...
Maybe the encoders would help you, I'm curious. Tell us how it goes!

unfortunately putting encoders on my gimbal will not be that easy as I initially thought. Yaw will cause a problem because of complicated mechanical build. I will need couple of days to figure this out.

I have already stiffer vibration dampening and it seems to be better already. Previously when I was panning the camera up or down almost every time I released a joystick command, the camera was getting a bit of jerk while trying to slow down. Now when the whole dampening system is a bit more stiff this strange jerk does not appear anymore. It seems that while fast slowing down (usually it appears in dynamic movement) the camera weight was influencing a dampening which later on can cause a bigger error on a IMUs. One thing bothers me a bit though. I can understand that camera IMU can be giving bigger error (more movement) thanks to lose vibration dampening system, BUT there is also another IMU placed directly on the gimbal frame (under YAW) which is also getting same movement information (shake caused by vibration dampening). Having data from both sensors, shouldn't this be somehow calculated to provide better stability?

So far the stability seems to be better now but I need to take it in to the ear to test it in real condition. Unfortunately it's raining since couple of days and I need to get a good weather window.

Interesting thing. Yesterday - just for test - I have replaced NEX 5(about 300 grams) with Nikon D3000 (about 700 grams) on my gimbal. After balancing it appeared that gimbal is working absolutely perfect without touching any of PID tuning or any other value. I was pretty surprised ;)

one question. How important is precision of 6 point IMU calibration? Today I was trying 2.5 GUI and firmware (previous 2.43 version was working fine for me). I went through 6 point temperature calibration and I have to say the results are much worst. Actually I can't get decent stabilization and I can't even say why. The only thing I can think of is 6 point IMU calibration which I did very fast and probably not accurate enough. Both sensors were packed in EPP foam to assure slowest possible temperature changes. I guess sensor might move inside the foam while calibration process.

I don't really understand why screwing the IMU back could destroy the calibration?
Anyway it shouldn't be an issue in my case. I use two sided tape for frame IMU and same for camera IMU. Actually the camera IMU is attached to a plate which is then screwed to the bottom of the camera plate.

Magnets should not affect much, at least when they are only temporally. The IMU is electro mechanical. Pending circuitboard can pend the chip and that affects the calibration. As you describe it installed, it should not pend.

More likely it is some setting that was not ok on 2.50. With some older FW I had some problems with IMU calibration after updating, Erase EEPROM solved those problems. Now I am at 2.56.

Note: you must use GUI that corresponds to the FW.

it is good.

There is something seriously wrong. It looks front heavy, but even if so it should not cause that. I the pitch well balanced?

Did you say it works ok with older FW?

There is the euler angle orders on new FW, but should be ok if you have not changed it and it should not cause that.

It looks like something wrong with RC. If you disable all RC from GUI, how does it work then?

Are all axis moving freely when unpowered?

The 2.50 is better than 2.43, it is something else. I had long time ago some serious issues you have, but that was after testing many beta FW. The solution was to load EEPROM from other card. But if yours is working with 2.43, strange. You could try the news beta, just enable checking Beta.

Are you always using the corresponding GUI (same number as the FW)?

One thing the picture above remind me that this is RLG. On the video it is standing on landing gear? have you Yaw disabled? If you have yaw activated, the hole thing must hang from the yaw so that it can rotate. Otherwise anything can happened. If 2.43 work better on that situation, it could be, but it is not meant to work like that. If yaw is activated, it must be able to rotate.

The strange thing is if it works with 2.43.

Otherwise I would suspect HW failure, especially would check Pitch motor.

All axis should work just perfectly, 2.43 or 2.50, the basic stabilisation works just great on both FW.

It could be that something is corrupted on EEPROM. But at this point I would say likely it is something else.

After you erased the EEPROM, did you do the complete basic setup again? I would advice to do so, make sure motor inverted is correct, set motor poles manually, make sure of IMU alignment and GUI configuration and calibration etc. It could be something like this causing the problems, but strange thing is if it works with 2.43.

It could be also that PID tuning is very wrong, did you auto tune it with good stability settings? Autotune the pitch axis only with good stability setting. You must tune PID when changing FW. PID scaling has changed on some FW versions.

The wrong inverted has happened for me too, and not so long time ago :)

I would recommend starting with normal Below Yaw, the +PID is more experimental, but it could help RLG, so it is worth of try after you have first tested normal Below Yaw.

Works well in windy condition:
damperzen.com
http://lensair.eu/wp2/wp-content/uploads/2015/07/P1160625.jpg