How to measure gyro sensitivity with extremely high accuracy?

[mattgrant345]

Hi,

My company uses the ASM330LHH for dead-reckoning applications and we are trying to improve the performace of the gyro with our own calibration of rotation sensitivity. After calibration, we want the gyro sensitivity accuracy to be 0.1 % or better over -20 to 70 C. I know this may not be easy or feasible, but we still want to get as close as possible. Our current calibration procedure uses an Ideal Aerosmith rate table that we try to set to 100 dps +/- 0.01 %, but we're running into a ton of issues getting an accurate enough speed. I'm hoping someone can help show us a better way!

Right now, our rate table often misses the set rate by much more than 0.01 % or the rate will oscillate around the set point by much more than 0.01 %. We have tried using the rotation rate readout of the rate table, but that is showing its own set of difficulties. Is this what folks normally do for high-accuracy sensitivity calibration? Or is an ultra-high accuracy gyro normally used as reference for the rotation rate? If so, could anyone recommend a gyro part number? We would be okay spending several thousands on a FOG or RLG, but I was hoping to get feedback from an expert before making such a large purchase.

Thank you!

Matt

[Andrea VITALI]

Assuming the turn table has a constant rate but different with respect to the nominal selected value: just measure the time T needed to complete N turns and compute the actual angular velocity by N*360/T. The effect of any error on T can be easily evaluated. Accuracy of the computation can be increased by increasing N. This is assuming the turn table has constant rate in the interval T and you can count the number of turns N. 

Assuming you have a robotic arm that can execute a given rotation around a given axis, you can configure the robot to rotate around one of the axis, of the gyroscope (which can be difficult because of installation errors of the sensor with respect to the PCB). Then program the robot to execute a 90 deg rotation and verify that gyroscope integration does match the amount of rotation.

Talking about the installation error, there are a lot of contributions: the MEMS with respect to the package, the package with respect to the PCB, the PCB with respect to the device housing, and finally the device with respect to the turn table/robot. 

All this needs to be repeated at at least two different temperatures, assuming the compensation parameters can then be interpolated. Example: calibrate at -20, calibrate at +70, then interpolate the compensation parameters for the actual temperature. Unfortunately, there can be and indeed there is hysteresis. And there can be temperature gradients accross the sensor which may compensation very difficult if not impossible. 

If you plan for maximum accuracy, the easiest solution is to keep the sensor at the calibration temperature using a peltier cell to cool/heat as needed.