Our lab has been using 3 ADIS16251 sensors mounted orthogonally to each other to provide 3-axis rate determination. The 3-axis unit consists of a main board with a microcontroller and the Z axis sensor to which two daughter boards carrying the X and Y axis sensors are connected as shown in the photo below. The boards are mounted to an aluminum enclosure with a set of screws to ensure orthogonality.
Recently, we've noticed that the bias of the X and Y sensors (the sensors on the two daughter boards) may change drastically after mounting into the enclosure. After some experimentation, I discovered that the bias change seems to be related to the torque on the two screws holding each board to the enclosure. Plots of average bias vs. screw torque for one of the units are shown below for the X and Y axes.
The biases shown here were calculated by averaging rate sensor measurements polled at 1 Hz over 10 minutes for each value of screw torque. The torques were only measured coarsely, but the plots show that the biases tend to change toward some direction (decreasing in this case) as screw torque is increased, with a total change of about 60 counts on the X axis and 100 counts on the Y axis.
Besides the bias, the units still function normally. Rates are still measured at the right scale factor. I also measured the temperature vs. bias curve, and though it looks different with the unit integrated inside the enclosure compared to without the enclosure, the shapes of both curves aren't unusual compared to other units.
We were able to rule out contact between any board components and the enclosure as the cause. The underside of all boards is isolated from the aluminum by a layer of Kapton tape, and cutouts are made for through-hole components like connectors. To my knowledge, this has not been observed for a previous batch of units we made two or three years ago, but has been observed for every sensor we've made recently. Nothing has changed in the design of the units - neither the boards nor the enclosures.
Does anyone know what the cause of this behavior could be? Also, is this is a concern for the future reliability of the sensor, and has this been observed before?
I would appreciate any insights you can provide.