Thanks for the answer. Now I'll try to describe what we get.
In our project we have a platform that needs to be stabilized. We have a central controller and 3 servos that control the platform's yaw, pitch and roll. All data transfers are done in digital and there are no components and connections that would be evidently succeptable to noise or temperature (at least we haven't discovered them yet). When we start working, the stabilization goes well and is nice and smooth. After working for 3-4 hours we can see that the platform starts giving a jerk from time to time. Also occasionally it starts moving from the stable point and goes quickly to the mechanical limits of the axes. We can also say that this effect is not a result of an accumulated error somewhere in software as a power reset does not change the situation.
After some analisys we concluded that the most probable candidate that's responsible for the issue is the IMU. So, here are some questions, that I'd appreciate to get explanations for:
1) You wrote:
Noise/stability does tend to increase with temperature
Did you wanted to say DECREASE? or not? What does it mean in terms of stability?
2) What data from ADIS could we expect most likely to contain mistakes?
3) Stepping a bit aside, if the sensor gets overheated, will it invoke an emergency powerdown or continues working until its death?
4) And one more question, just to avoid creating a new topic:
How can transportation with a plane to a long distance affect the IMU? I mean if we bring our system to another place, does it need to be recalibrated in some way (aside the magnetometers)?