Hi All,

Hopefully some clever people are able to shed some light / information about this one, as I'm having trouble finding useful information about how the sensors and components actually work.
I've recently been setting up a custom built gimbal. All seemed to be going quite well, managed to create a profile which had a nice smooth return to the origin etc. I tested this for a while, put it down, and came back to it later to find that it wouldn't sit straight anymore (sitting with roll on a lean sideways).

Presuming that a calibration had been messed up somewhere, I hooked it up to the GUI to see what was going on. The software kept showing me that the axis were all correctly 'level', but the 'current position' indicator arrow was constantly spinning for the roll axis. Eventually this became a problem with all the axis (Roll Pitch and Yaw), after leaving it for some time. Doing a basic calibration didn't seem to help, so I figured out how to perform a 6 axis accelerometer calibration with reasonable accuracy, but this had no effect.

After removing the main board and sensor from the frame to eliminate those variables, I eventually figured out that it seemed to be an issue with the Gyroscope. If I did a Gyro calibration with the sensor sitting still on an undisturbed surface, it seemed to correct the problem reasonably well, until I start to move the sensor around (then the 'slipping' effect gets worse).

Originally I had the IMU mounted on it's side, with the axis set to orient it correctly, so now with it on the desk, I've defaulted the orientation, and it's the Yaw that is slipping instead of the Roll. I know that Some 'Yaw Slipping' is normal behavior, and about a degree per minute is considered to be normal - according to the manual, but this is much worse, so I've recorded some timings to give you an idea what I mean.

After calibrating the Gyro (with sensor sitting flat on a stable surface, watching that the stability peak meter is within the 'green' range), the slipping slows down to the times as follows, where each time in seconds is a single degree slipped:

1. ‎38.22
2. ‎17.66
3. ‎16.27
4. ‎16.91
5. ‎15.68
6. ‎22.54
7. ‎29.30

So this corresponds to somewhere around 2-4 degrees of slippage per minute.
Now, I took the sensor, and moved it around a bit in the air, then put it back into it's place, and recorded these times:

1. 05.14
2. ‎04.99
3. 05.31
4. 05.32
5. 05.56
6. 05.32
7. 05.46
8. 05.34
9. ‎05.62

Which is about 12 degrees per minute. After sitting there for about 30 minutes, it eventually sped up to around 1 degree per second!

Then I noticed something odd, that if I held the sensor tight against the desk (by holding down two opposite corners) the slipping would stop. So my curiosity wondered what would happen if I calibrated the Gyro in that state, and sure enough, it flipped the issue. while held against the desk in that way, it would be slip at about 1 degree per second, but if I let it go, it would stop.

Obviously the take away from this, is that the sensor components are highly sensitive, as is mentioned by many, but does this mean that the gyro specifically has to be calibrated 'in the condition that it will be mounted'? It still feels like the gradual slipping, which is accelerated by moving the sensor, is kind of undesirable. Since I have two IMU's (other one's never been used), I tested this also, and it behaves in the same way, so I'm guessing it's more likely to be a configuration or set up issue that I'm missing rather than a hardware fault?

The other thing I tried was calibrating with the sensor on a non-hard surface (ie pillow), to see if it was anything to do with any vibrations that may have been affecting it while calibrating - but with no success.

If anyone has any ideas, or any in depth info about how these sensors work, or has already come across this issue before and knows the easy fix, I'd be super grateful to hear from you :)

Thanks,
Matt.