KNOWLEDGE · IMU SELECTION GUIDE
How to choose the right IMU
An inertial measurement unit combines three gyroscopes and three accelerometers. Which one fits depends less on the longest data sheet and more on what your system has to do. This guide walks through the decisions in the order engineers make them.
By the AMCORIS application team · Updated October 2026 · 8 min read
QUICK ANSWER
Choose by task first, then by two gyro numbers, then by interface.
1
Define the job
Stabilization, navigation, motion control or monitoring.
2
Read two gyro numbers
Angular random walk for smoothness, bias instability for drift.
3
Pick interface and housing
UART/SPI inside a device, RS-422, CANopen or J1939 with IP67 outside.
4
Test with your own data
Borrow an evaluation kit and compare two candidates in your setup.
STEP 1
Start with the job the IMU has to do
The same sensor can be ideal for one application and oversized for another. Name the main task first, because it tells you which specification to optimize.
Stabilization
Gimbals, EO/IR payloads and antennas. Angular random walk decides how smooth the image or beam stays. Output rate and latency limit the control loop.
Navigation and dead reckoning
Robots, AGVs, drones and vehicles. Bias instability decides how fast heading drifts when GNSS or other aiding drops out.
Motion control
Robot arms, AMRs and machines. Output rate, latency, size and an optional functional safety rating matter most.
Monitoring and tilt
Structures, cranes and solar trackers. Here a low-noise accelerometer is often the better and simpler choice than an IMU.
More detail per field: stabilization · navigation · robotics · tilt
STEP 2
Read the two gyro numbers that matter
Angular random walk (ARW), in °/√h, is the short-term noise of the gyro. Its angle error grows with the square root of time, so it dominates over seconds to minutes and decides how calm a stabilized camera or antenna is.
Bias instability, in °/h, is the slow wander of the gyro offset. Its angle error grows roughly with time, so it dominates over minutes to hours and decides how long a vehicle can hold its heading without GNSS or other aiding.
Both values come from the Allan deviation of the gyro and are given as typical values in the Epson data sheets. Compare them at the same conditions, and remember that temperature changes and vibration add errors in a real system.
DRIFT ESTIMATOR
How far does heading drift without aiding?
First-order estimate for one gyro axis, free integration, constant temperature and a calibrated start bias. Real drift also depends on temperature changes, vibration, the navigation filter and any aiding such as GNSS, odometry or magnetometers. Use it to compare models, not as a performance guarantee.
STEP 3
Choose interface and housing
The interface follows the installation. Compact modules are built into your electronics, while IP67 units are mounted directly on a vehicle, mast or machine.
UART / SPI
Short connections to a controller board, drones, robots and embedded navigation. Smallest size and weight.
RS-422
Long cables and electrically noisy environments, for example antennas, masts and vessels.
CANopen
Machines and vehicles with several nodes on one CAN bus.
SAE J1939
Agricultural and construction machinery that already speak J1939.
STEP 4
Check the system requirements
Output rate and latency
Fast control loops need high output rates. The M-G366PDG and M-G370PDG deliver up to 2000 samples per second, the M-G355QDG up to 400.
Temperature range
The PDG series covers −40 to +85 °C. Epson IMUs are factory-calibrated over temperature, which reduces your own calibration effort.
Functional safety
For safety-related functions, the M-G355QDG complies with IEC 61508 SIL 1.
Size and weight
The PDG series shares one 24 × 24 × 10 mm footprint of about 10 g, so you can move between performance levels without a new board design.
COMPARISON
Epson IMUs at a glance
All models measure six axes with digital output. Start with the row that matches your task and interface, then compare the gyro numbers.
| Model | Interface | Bias instability | ARW | Housing / feature | Typical use |
|---|---|---|---|---|---|
| M-G330PDG | UART / SPI | 3 | 0.1 | 24 × 24 × 10 mm, approx. 10 g | Cost-sensitive navigation, service and mobile robots |
| M-G366PDG | UART / SPI | 1.2 | 0.08 | 24 × 24 × 10 mm, approx. 10 g | Small UAVs, camera gimbals, robots with tight size and power budgets |
| M-G370PDG | UART / SPI | 0.8 | 0.06 | 24 × 24 × 10 mm, approx. 10 g | Precise heading in compact systems, EO/IR gimbals, INS boards |
| M-G370PDT | UART / SPI | 0.8 | 0.03 | Compact, lowest angle noise | High-end gimbals and EO/IR payloads |
| M-G355QDG | UART / SPI | 1.2 | 0.08 | IEC 61508 SIL 1 | Safety-related functions in robotics and machinery |
| M-G552Px1 | RS-422 / CANopen / J1939 | 1.2 | 0.08 | IP67 housing | Vehicles and mobile machines |
| M-G552Px7 | RS-422 / CANopen / J1939 | 0.8 | 0.06 | IP67 housing | Mining, agricultural and construction machines, vessels |
| M-G570PR | RS-422 | 0.5 | 0.04 | IP67 housing | Survey-grade INS, mapping and stabilization |
QWhat is the most important IMU specification?
It depends on the job. For stabilization, angular random walk (ARW) decides how smooth the result is. For navigation without GNSS, gyro bias instability decides how fast heading drifts. For motion control, output rate and latency matter as much as noise.
QHow much heading drift should I expect without GNSS?
As a first-order estimate, multiply the gyro bias instability by the time without aiding. With 0.8 °/h, ten minutes give about 0.13°. Real systems drift more because temperature, vibration and the filter add errors, so aiding from odometry, LiDAR or vision is used for longer outages.
QWhat is the difference between ARW and bias instability?
Angular random walk describes the short-term noise of the gyro and grows with the square root of time. Bias instability describes the slow wander of the gyro offset and limits how well an angle or heading is held over minutes and hours.
QWhen do I need an IP67 IMU?
When the sensor sits outdoors, on vehicles or vessels, or at the end of a long cable. The Epson M-G552 series and M-G570PR combine an IP67 housing with RS-422, CANopen or J1939. Compact UART/SPI modules are meant to be built into a protected housing.
QIs there an Epson IMU for functional safety?
Yes. The Epson M-G355QDG complies with IEC 61508 SIL 1 and uses the same compact footprint as the M-G366PDG.
QCan I test an IMU before I decide?
Yes. AMCORIS provides Epson evaluation kits as a free loan for qualified projects, so you can compare models with your own data.
EVALUATION KIT
Compare two candidates in your own system
Tell us your task and interface. We suggest two Epson IMUs and lend you an evaluation kit for qualified projects.