Degrees of Freedom (3, 6 or 9)

3-DOF is accelerometer only (tilt, shock, vibration). 6-DOF adds a gyro for smooth pitch and roll. 9-DOF adds a magnetometer so you can also estimate heading (yaw) relative to magnetic north.

Measurement Range

Most parts offer selectable ±2/4/8/16 g and ±250 to ±2000 °/s. Use the smallest range that won’t saturate: ±2 g for tilt, ±16 g for impacts, ±2000 °/s or more for fast-spinning drones and rockets.

Raw Data vs Onboard Fusion

Raw IMUs (MPU6050, LSM6DS3) give you numbers and leave the filtering to your MCU. Fusion modules (BNO055, BNO085) run the algorithm internally and output stable quaternions or Euler angles.

Noise, Drift & Temperature Stability

Gyro bias drift and noise density decide how stable your angles are. Newer chips (ISM330DHCX, ICM-20948, BMI270) beat the 2010-era MPU6050, and industrial parts stay stable across wide temperature swings.

Interface, Voltage & Connectors

I2C is easiest; SPI is faster for high sample rates. Check for an onboard 3.3V regulator before wiring to a 5V Arduino, and look for Qwiic/STEMMA QT connectors if you hate soldering.

The Adafruit BNO055 was the first affordable “smart” IMU and is still the gentlest introduction to absolute orientation. Bosch’s BNO055 runs its own fusion on an internal Cortex-M0 and hands you Euler angles, quaternions, angular velocity, linear acceleration, gravity vector and magnetic field over I2C, with a calibration status register so you know when the heading is trustworthy. Adafruit’s breakout adds a regulator and level shifting so it works with 3.3V and 5V boards, and the Adafruit_BNO055 library has you reading heading, roll and pitch within minutes. It shares hardware with the BNO085 but runs Bosch’s simpler firmware, which is why we rank it just below.

👉 Check Price on Amazon →

The ISM330DHCX is ST’s industrial-grade 6-axis IMU, and SparkFun’s 1″×1″ Qwiic breakout makes it plug-and-play. The gyro reaches an unusually wide ±4000 °/s, the sensor is specified from -40 to +105°C with embedded temperature compensation, and ST’s Machine Learning Core and finite-state machine can classify motion on-chip without waking your MCU. That combination makes it the right choice for vibration and condition monitoring, fast-spinning robots, rockets and anything that lives outdoors or in an enclosure that gets hot. The SparkFun library supports Arduino and ESP32, and the Qwiic connector means zero soldering.

👉 Check Price on Amazon →

When TDK discontinued the MPU9250, the ICM-20948 became its official successor, and it is the 9-DOF chip to design around today. It pairs a 6-axis accel/gyro with the AK09916 magnetometer, uses noticeably less power than the MPU9250, and keeps an onboard Digital Motion Processor that can output fused quaternions. This GY-ICM20948V2 two-pack is the budget way in, and both the SparkFun ICM-20948 library and several MicroPython drivers support it. The listing is light on detail, so treat it as a 3.3V part unless you confirm a regulator on the board.

👉 Check Price on Amazon →

The LSM6DS3 is ST’s “always-on” 6-axis IMU, the same family used on boards like the Arduino Nano 33 IoT and Seeed XIAO nRF52840 Sense. Its standout feature is a giant 8 KB FIFO: the sensor can buffer thousands of samples while your MCU sleeps, then hand them over in one burst, which is a big win for battery-powered loggers. It also has hardware pedometer, tap, double-tap, free-fall and tilt interrupts, and samples the accelerometer up to 6.7 kS/s for vibration work. The NOYITO breakout exposes I2C and SPI, and the SparkFun LSM6DS3 library works with it.

👉 Check Price on Amazon →

The Bosch BMI270 was designed for smartwatches and fitness bands, and SparkFun’s Micro Qwiic breakout makes it easy to prototype with. It draws very little current, and its on-chip features (step counter, wrist-gesture and activity recognition, any-motion and no-motion interrupts) let your MCU stay in deep sleep until something actually happens. It is the same IMU found on the Arduino Nano 33 BLE Sense Rev2, so there’s good Arduino support, and SparkFun’s library covers ESP32 and other cores. The tiny footprint fits wearables and handheld controllers where every millimetre counts.

👉 Check Price on Amazon →

The MPU9250 is the MPU6050’s big brother: an MPU6500 accel/gyro and an AK8963 magnetometer in one package, giving a full 9-DOF for heading estimation on a tight budget. The HiLetgo GY-9250 board accepts 3V to 5V, exposes both I2C and SPI, and works with the popular hideakitai MPU9250 and Bolder Flight libraries (see our MPU9250 Arduino tutorial). The honest catch: TDK has discontinued the chip, quality varies from batch to batch, and some cheap “MPU9250” boards turn out to be MPU6500s with no magnetometer. Check the WHO_AM_I register (0x71 for a genuine MPU9250) as soon as it arrives.

👉 Check Price on Amazon →

Not every project needs a gyro. If you only care about tilt, shock, vibration, tap or free-fall, the ADXL345 on a GY-291 board is simpler, cheaper to run and easier to use than any IMU. Analog Devices’ chip offers ±2 to ±16 g ranges with a fine 4 mg/LSB resolution, and its built-in single-tap, double-tap, activity, inactivity and free-fall detection can fire interrupts so your MCU doesn’t have to poll. It talks I2C or SPI, and the Adafruit ADXL345 and SparkFun libraries make it a five-minute job on Arduino, ESP32 or Pico. The three-pack covers a tilt alarm, a knock sensor and a spare.

👉 Check Price on Amazon →

Need reliable orientation fast?

Get the BNO085: drift-corrected quaternions from onboard fusion, no filter code needed.

Learning IMUs on a budget?

The HiLetgo MPU6050 three-pack has the most tutorials of any IMU and costs almost nothing.

Want Euler angles and great docs?

The Adafruit BNO055 outputs heading, roll and pitch directly with an excellent library.

Harsh environment or vibration monitoring?

The SparkFun ISM330DHCX is industrial-grade with ±4000 °/s and -40 to 105°C range.

Need raw 9-DOF for your own fusion?

Choose the ICM-20948 for new designs, or the MPU9250 if you’re following an older tutorial.

Running on a coin cell or LiPo?

Use the BMI270 or LSM6DS3: low current, smart interrupts and big FIFOs.

What’s the difference between the MPU6050, MPU9250 and ICM-20948?

The MPU6050 is a 6-axis IMU (accelerometer and gyroscope). The MPU9250 adds an AK8963 magnetometer for 9 axes, and the ICM-20948 is TDK’s newer, lower-power 9-axis replacement with an AK09916 magnetometer. The MPU6050 and MPU9250 are both discontinued, though modules are still easy to find. For new 9-DOF designs, use the ICM-20948 (raw data) or BNO085 (fused orientation).

Do I need a 6-DOF or a 9-DOF IMU?

A 6-DOF IMU gives accurate pitch and roll because gravity provides a stable reference, which is enough for self-balancing robots, tilt sensing and most gimbals. Yaw (heading) has no such reference, so it slowly drifts. If you need a heading that stays put, such as for a rover, a compass or head tracking, you need a 9-DOF IMU so the magnetometer can correct yaw. Keep the magnetometer away from motors and batteries and calibrate it in place.

Why does my gyroscope angle drift, and how do I fix it?

A gyro measures rotation rate, so you get an angle by integrating it, and any tiny bias accumulates over time into drift. The fix is sensor fusion: combine the gyro’s smooth short-term response with the accelerometer’s (and magnetometer’s) long-term reference. A complementary filter is a few lines of code and works well for pitch and roll; Madgwick, Mahony or Kalman filters handle full 3D orientation. Also calibrate the gyro bias at startup by averaging a few hundred samples while the board is still. Fusion modules like the BNO085 and BNO055 do all of this internally.

Can I connect these modules to a 5V Arduino?

It depends on the breakout, not the chip. The GY-521 MPU6050, GY-9250 MPU9250 and Adafruit BNO055 boards include a 3.3V regulator and are commonly used on 5V Arduinos (the Adafruit board also level-shifts I2C). Many bare LSM6DS3, ICM-20948 and generic BNO085 boards are 3.3V-only, so power them from 3.3V and use an I2C level shifter with a 5V MCU. With ESP32, STM32, Raspberry Pi Pico and other 3.3V boards, everything here works directly.

How do I know if my MPU9250 is genuine?

Read the WHO_AM_I register (0x75) over I2C. A genuine MPU9250 returns 0x71; an MPU6500 returns 0x70, and some MPU9255 parts return 0x73. Then try to reach the AK8963 magnetometer at I2C address 0x0C (after enabling bypass mode): if it doesn’t respond, the board has no magnetometer. Since the MPU9250 is discontinued, buying from a seller with easy Amazon returns, as with the HiLetgo listing here, is worth it.