MEMS gyroscopes and accelerometers are now smaller, lighter, and more capable than before. Modern chips have advanced significantly compared with a decade ago, enabling low-cost MEMS inertial measurement units (IMUs) to offer performance approaching tactical-grade systems that were previously found only in expensive, high-end applications.
Despite these improvements, MEMS IMUs have unique characteristics that require attention. By accounting for these characteristics and following good IMU data practices, you can achieve better measurement results for your application. The following are practical tips to improve inertial sensor performance.
Part 1 — Isolate the MEMS IMU from Vibration
Isolating the MEMS from unnecessary vibration is essential for accurate measurements. Users are typically interested in overall system motion, such as vehicle trajectory, not the high-frequency vibrations. Vibration sources include system components like motors, structural resonances caused by terrain, and even people walking nearby. Thoughtful mounting of the IMU on a vibration-isolated platform can minimize these effects and improve sensor performance. Vibration isolation also protects the device from high-impact events that could damage sensitive internal elements.
MEMS sensors can exhibit a noise component when subjected to vibration known as vibration rectification error. Oscillatory vibration signals can be rectified in the sensor output into an undesirable offset, producing a false DC component that degrades measurement accuracy. Minimizing vibration on the IMU reduces this error and improves overall system performance.
Part 2 — Capture Gyro Bias Regularly
All gyroscopes exhibit bias, a nonzero output when no rotation is present. This bias can drift over time and is often more pronounced in MEMS sensors. If you use an attitude and heading reference system (AHRS) or an inertial navigation system (INS), the internal filtering algorithms typically estimate this bias in real time, but the filter may require some time, sometimes minutes, to converge to an accurate value. If you are using a basic IMU to read angular rate measurements directly, it will not estimate this bias on its own. In either case, periodically capturing the gyro bias while the device is stationary optimizes performance.
Bias capture requires the device to remain stationary during the capture period and for any vibration sources, such as an engine, to be turned off. Some IMU vendors provide tools or SDK functions to initiate a bias capture; after a few seconds the device estimates the bias and stores it in internal memory. Regular captures help mitigate gyro aging effects.
When performing integration, use the integrated delta-theta (Δθ) and delta-velocity (ΔV) values rather than raw angular rates and accelerations. Modern MEMS IMUs can output data at 1 kHz or higher. Instead of integrating raw angular rate and acceleration values in your own filter, request the IMU's preintegrated outputs: delta-theta and delta-velocity. These preintegrated values account for coning and sculling effects that occur during real motion, producing more accurate integrated quantities even at high rates.
Using integrated outputs also reduces CPU load. Because the IMU performs the high-rate integration internally, your system can request these values at a lower rate. For example, if the IMU reports at 1,000 Hz but your filter only needs 50 Hz, using delta-theta and delta-velocity can reduce processing by a factor of 20 while improving accuracy since the IMU has already captured high-rate dynamics and accounted for coning and sculling.
Part 3 — Pay Attention to Time Synchronization
At some stage in IMU use you will find that timing matters. Accurate timestamps and time alignment are often overlooked until late in development, and unsynchronized clocks can cause subtle but significant errors. Clock drift between the system and the IMU can accumulate and behave like an incorrect scale factor on the IMU angular rate and acceleration outputs. As system dynamics increase, the impact of timing errors also increases.
Higher performance applications require tighter time alignment. Two approaches can mitigate timing issues: use a precise reference (hardware pulse or similar) to align clocks across components and compensate algorithmically for any misalignment, or use an IMU that supports event-driven operation. The first approach is common but can be complex in systems with many components. Some IMU models, such as the 3DM-CV7-AHRS, support event-driven operation to address strict timing requirements.
With an event-driven IMU, a hardware input pulse can trigger the IMU to generate corresponding integrated outputs, and vice versa, simplifying handling of external events. For visual odometry, for example, knowing the IMU integration that corresponds exactly to a camera frame is critical. By connecting a camera shutter output to the IMU and configuring the IMU to provide Δθ and ΔV when a pulse is received, you obtain precise integrations aligned with image timestamps. Conversely, an event-driven IMU can generate pulses to trigger capture hardware while maintaining strict time alignment.
Following these practical tips can improve IMU performance and measurement quality across a wide range of applications.
