The sensor output with our EKF implementation works fine in normal conditions. However when the motor is started and real vibration comes into play, results are all over the place. The data just isn't good enough. The gas engine is attached to the firewall with rubber vibration dampeners.
What are your experiences/thoughts on the following:
1. How should sensors be attached to to the airframe (hard mounts, soft mounts)?
2. What should the data acquisition rate be in high vibration scenarios (max available or happy medium)?
3. What are the "9" DOF IMUs good for? Are the magnetometer signals useful as extra input into fusion algorithms or is it on the board purely for compass purposes?
4. Has anybody fused thermopile results with gyro/accelerometer results?
5. How tolerant are gyros and accelerometers of RF interference?
6. What indications should be used for measuring quality of sensor data when on level plane?
Any extra recommendations for making a significant leap in the quality of results?
p.s. Another issue we faced was that Arduino keeps restarting with vibration. Apparently the USB connection with an external computer is not tight enough. Based on preliminary tests, applying some tension on the connector fixed the issue.
Replies
2. For these sensors, it almost doesn't matter with respect to vibration. You don't want to propagate any engine vibration into the sensor chassis at all if you can avoid it. With a plane that big, you should have no trouble.
6. See Tuning ArudIMU and the Allan Variance threads under ArduIMU.
It is possible. This plane got a working MEMS IMU with a similar engine. We also had issues early on with engine vibration buggering up the IMU. Rigidly mounting the IMU in a heavy metal case that was itself mounted to the vehicle frame with soft rubber mounts solved the problem for us.