My consumer grade accelerometer can theoretically measure less than 1° of tilt. Is it still possible to achieve such measurement accuracy under temperature changes and vibrations?
The answer is likely no. Questions about clear tilt accuracy values are always difficult to answer because of the many environmental factors that need to be considered when it comes to MEMS sensor performance. Typically, consumer-grade accelerometers have difficulty detecting tilts of less than 1° in dynamic environments. To demonstrate this, we compare a general-purpose consumer-grade accelerometer to a new generation of low-noise, low-drift, and low-power MEMS accelerometers. This comparison looks at the many sources of error that exist in tilt applications and which errors can be compensated or eliminated.
0 g bias accuracy, 0 g bias drift due to soldering, 0 g bias drift due to PCB housing alignment, 0 g bias tempco, sensitivity accuracy and tempco, non-linearity, and horizontal axis can be observed Sensitivity and other errors, and these errors can be reduced by a post-assembly calibration process. Hysteresis, 0 g bias drift over life, sensitivity drift over life, 0 g drift due to moisture, and PCB bending and twisting due to temperature changes over time, etc. These error terms cannot be accounted for by calibration or other methods, It will need to be reduced with some level of in-situ maintenance. In this comparison, it is assumed that the transverse axis sensitivity, nonlinearity, and sensitivity are compensated, since it takes much less effort to minimize these errors than temperature coefficient offset drift and vibration correction.
Table 1 lists the ideal performance specifications for the consumer ADXL345 accelerometer and estimates of the corresponding tilt error. When trying to achieve the best tilt accuracy, some form of temperature stabilization or compensation must be employed. In the examples below, a constant temperature of 25°C is assumed. The most important error contributors that cannot be fully compensated are temperature drift offset, offset drift, and noise. The bandwidth can be reduced to reduce noise, as tilt applications typically require less than 1 kHz bandwidth.
Table 1 ADXL345 Error Source Estimates
Table 2 lists the same standard that applies to the ADXL355. The short-term bias value is estimated from the Allan variance plot in the ADXL355 data sheet. At 25°C, the compensated tilt accuracy of the general-purpose ADXL345 is 0.1°, and the compensated tilt accuracy of the industrial-grade ADXL355 is 0.005°. Comparing the ADXL345 and ADXL355 shows that the errors due to major error contributors have been significantly reduced, such as the error due to noise from 0.05° to 0.0045° and the error due to offset drift from 0.057° to 0.00057°. This represents a huge leap in performance for MEMS capacitive accelerometers in terms of noise, temperature coefficient, offset, and bias drift, providing a higher level of tilt accuracy under dynamic conditions.
Table 2 ADXL355 Error Source Estimates
Choosing a higher-grade accelerometer is critical to achieving the desired performance, especially when the application requires tilt accuracy of less than 1°. Application accuracy depends on application conditions (large temperature fluctuations, vibration) and sensor selection (consumer vs industrial or tactical). In this case, the ADXL345 would require extensive compensation and calibration work to achieve a tilt accuracy of less than 1°, adding effort and cost to the overall system. Depending on the amount of vibration in the final ambient and temperature range, it is simply not possible to achieve the above accuracy. The temperature coefficient offset drift of 1.375° over the 25°C to 85°C range exceeds the requirement for less than 1° tilt accuracy.
The temperature coefficient offset drift of the ADXL355 from 25°C to 85°C is:
As shown in Table 3, Vibration Correction Error (VRE) is the offset error introduced when the accelerometer is exposed to broadband vibration. When the accelerometer is exposed to vibration, VRE can cause significant errors in tilt measurements compared to 0 g offset due to temperature drift and noise. This is one of the main reasons why data sheets are no longer used, as it is easy to obscure other major specs.
Table 3 Errors in terms of inclination
In environments with higher amplitudes, a higher g-range accelerometer must be used to minimize offsets due to clipping. Table 4 lists the ADXL35x family of accelerometers and their corresponding g ranges and bandwidths.
Table 4 Measurement ranges of ADXL354/ADXL355/ADXL356/ADXL357
Choosing the ADXL35x family of accelerometers for tilt applications will ensure high stability and repeatability, tolerate temperature fluctuations and broadband vibration, and require less compensation and calibration than lower cost accelerometers. This series of products is packaged in a hermetic package, which ensures that the repeatability and stability of the final product will always meet the specifications after leaving the factory. ADI’s new generation of accelerometers provide repeatable tilt measurements in all environments, and they achieve minimal tilt error in harsh environments without extensive calibration.
Chris Murphy is an Applications Engineer at the European Central Application Centre in Dublin, Ireland. He joined Analog Devices in 2012 to provide design support for motor control and industrial automation products. He holds a Master of Research in Electrical Engineering and a Bachelor of Science in Computer Engineering.