Acceptance Sampling Methods for Microelectromechanical Inertial Measurement Units
Variable acceptance sampling under ISO 3951-1 verifies continuous MEMS IMU drift profiles while protecting production lines from latent silicon wafer defects.
Non-Orthogonality
Angle Deviation
Geometrical error occurs when the intended ninety degree relationship between measurement axes is not perfectly maintained. In multi-axis sensors like three axis accelerometers, non-orthogonality quantifies the deviation of each sensing channel from its ideal Cartesian position. This displacement means that force on one axis creates a small false signal in the channel supposed to monitor a perpendicular direction.
It arises from misalignment during die bonding or from slight twists in the internal structures during the manufacturing sequence. The value is measured in degrees or milliradians and sets the limit on the precision of orientation calculations. If the angles are skewed too far, the navigation math will accumulate errors that grow over time.
This error serves as a boundary on the accuracy of three dimensional motion tracking systems.
Calibration Matrix
Mathematical corrections recover the true acceleration vector from the skewed raw data streams. To compensate for non-orthogonality, test engineers place the sensor on a high precision indexing head that can rotate in two planes. They measure the output at positions where each axis should read zero to find the exact tilt of the sensing elements.
If the table is not precisely calibrated, the software matrix will carry an offset that misrepresents the actual physical hardware. Sourcing high performance inertial units involves looking for low baseline values before mathematical fixes are applied. Measurement drift happens if the sensor housing or the internal beams settle over time into a new configuration.
Certificates of calibration include the final matrix used to align the data during real time operation. Regular checks ensure that these coefficients remain within the tolerance limits established during initial qualification.
Interface Drift
Physical stresses from the mounting environment introduce mechanical distortions that change the apparent alignment of the axes. When non-orthogonality increases after installation, it is often due to uneven torque on the screws or a curved mounting surface. These external forces twist the package enough to shift the internal proof masses off their normal center.
This interference creates a mismatch between the factory measurement data and the field behavior of the system. Monitoring this effect requires analyzing the correlations between the channels during a known circular motion test. If the axes are correctly compensated, the data points should map to a perfect circle in three dimensions.
Any oval shape indicates that the installation effect has degraded the sensor alignment beyond its expected range. Careful layout of the printed circuit board helps isolate the sensing component from these external mechanical loads.
Operational Consequence
Directional accuracy improves significantly when the tilt errors are identified and minimized. Use of non-orthogonality data allows navigation filters to assign weight correctly between multiple sensor sources in a vehicle. These findings are checked during the end of line testing to classify sensors into precision bins.
If a unit shows more than one or two degrees of skew, it is typically excluded from critical aerospace or stabilization applications. Sourcing procedures for top tier grades require that the vendor provide verified angle data for each individual serial number. Once the item is in place, verification involves zeroing the bias in a known stationary state to lock in the baseline.
The final logic ensures that a tilt in the device does not look like linear acceleration to the host computer. Every successful verification confirms that the vector stays in the intended geometric frame within the allowed margin.