Today I wrote a script to extract the useful bits from the log files recorded during last Sundays test
Bellow is a plot of Roll and Pitch for the second test (the first one in the video(one where its still light)). The sample rate was about 50Hz. When the IMU was shocked too severely it gives really high values so I filtered out any values higher/lower than +/-180. From what I can see the only times the IMU gave wrong values was when it was bouncing on its tether. The values didn't seem to be affected by any engine vibrations although I wasn't pulsing the engine (was on full). During the static tests the vibration while pulsing is much more severe. I have been doing some research on the vibration sensitivity on the Arduimu's MPU-600 (gyro and accelerometer) and the chip should be able to handle the low frequency (30Hz) vibrations from the solenoids fairly well as long as the amplitude is reasonable because the MEMS accelerometers work at much higher frequencies (in the KHz).
You can see the point of liftoff in the middle where the roll and pitch peaks. All the oscillations are when it is rocking around on its tether.
Looking at the graph you can see that the IMU has a preference for positive values. There is a problem with some newer ArduIMU's where the manufacturer of the MPU-600 (gyro and accelerometer) changed its gyro scaling values and as such rolls and pitches appear exaggerated and wrong. The firmware I was running during the flight was older and I don't think had the new value so I dont think this is the cause of the problem. I really need to do more playing with it.