Mems Accelerometer Temperature Stability

[nickloss]

I am using a Mems accelerometer in a design to sense the tilt angle of a vehicle. When I thermal cycle the unit from +25C to +85C slowly(1 day) and then rapidly drop the temperature from 85C back to 25C the accelerometer output changes 4-5 counts. When the unit sits at 25C for 1 day the 4-5 count error goes away. What is causing this error? For example the 8 bit output will start at 127 at 25C, the temperature is ramped up to 85C slowly and then dropped rapidly to 25C and the output is now 132. If the unit sits at 25C for a day the output will return to 127. The main errors that I am compensating for with the micro are output sensitivity and 0G offset. The other errors of the accelerometer have not been compensated for with the micro since they are insignificant. I have guessed that the CTE differences between the accelerometer and the pcb are causing stress in the solder joints(leaded solder). These stresses are relieved when the solder cold flows or relaxes sitting at 25C. The manufacturer of the part has not been helpful when I ask about this issue.

 

[VEBill]

Are you assuming the problem is in the sensor itself? What about the analog signal conditioning circuit that must be embedded in the device between the sensor element and the digital circuitry?

Perhaps they failed to design the analog circuitry to compensate for value drift with temperature. Seems just as likely as the sensor itself.

If you're stuck with this sensor, could you characterize the offset and adjust it out against a temperature sensor?

Or, take advantage of Earth's 1G calibration service and periodically adjust the offset. A 3-axis sensor would be helpful.

Also. there's a company called Fredericks (not sure of the spelling) that makes direct-measure tilt sensors.

 

[MacGyverS2000]

VE is on the right track... you can't ignore temperature if you want long-term stability with MEMS devices.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[itsmoked]

Your supposition seems reasonable. Test it.

De-solder the MEMs and re-solder it via the thinnest wire you can lay hands on. Like a single strand of fine stranded wire. That way the board/joint stress can't be transmitted to the part.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[toybuilder]

Or, for that matter, the mounting hardware itself... If the product packaging and/or the mounting is also being heat cycled, it's possible that they are "walking" too.