This is an old revision of the document!
This page is intended to introduce and discuss common characteristics of MEMS IMUs, how to measure those performance metrics and finally some applications where those metrics matter.
Measured in °/hour.
Offset error in MEMs output on a particular axis. Fixed error typically. Measured in °/hour.
An error in the gain of the response. Non-uniformity that affects measurement output.
A plot of noise versus averaging time.
This can be thought of as the sensor resolution in an “ideal case.” It is also found as the minimum of the AVAR plot.
Measured in °/hour/√Hz or °/√hour.
Applies to gyroscopes. It is the “close-in” (small τ) on the AVAR curve and is noise that comes from the quantization process. Taken at Τ = 1 second. In order to calculate the ARW, take the value of the square-root of the AllanVariance and divide by 60 [iMAR].
For an example consider the ADIS16153, with datasheet here: media/en/technical-documentation/data-sheets/ADIS16135.pdf#Page=08. The graph of the AllanVariance is this:
Additional notes: iMAR Research Background
Applies to gyroscopes and is the “far out” (large Τ) on the Allan Variance (AVAR) curve. This quantity represents long-term stability of the sensor measurement.
Similar to the angular random walk (ARW) for gyroscopes. Noise from quantization of the axes.
Similar to amplifier noise density. RMS noise per unit of BW. Can be approximated with the formula: ND ≅ ARW/60* √2
Measured in degrees/sec/g².
Unintended effect where rotation/acceleration from one axis is detected on another axis. Can be calculated from axis to axis misalignment error for accels/gyros: sin(axis to axis misalignment) in %
Example: For the ADIS16448, the accel axis to axis misalignment is 0.2 degrees. Therefore Cross-axis sensitivity accel is sin(0.2) = 0.197%. Gyro is sin(0.05) = 0.09%. Magnetometer: sin(0.25) = 0.247%
Unintended effect where linear acceleration is detected as rotation. Measured in degrees/sec/g.
* Information on Allan Variance for Gyros: http://www.alexandertrusov.com/uploads/pdf/2011-UCI-trusov-whitepaper-noise.pdf