# imu noise parameters

**URL:** <https://community.stereolabs.com/t/imu-noise-parameters/11644>\
**Category:** ZED2\
**Created:** [August 21, 2026, 2:45pm UTC](https://community.stereolabs.com/t/imu-noise-parameters/11644 "2026-08-21T14:45:37Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Francesco](https://community.stereolabs.com/letter_avatar_proxy/v4/letter/f/e47c2d/32.png) [@Francesco](https://community.stereolabs.com/u/Francesco)\
**Post date:** [August 21, 2026, 2:45pm UTC](https://community.stereolabs.com/t/imu-noise-parameters/11644/1 "2026-08-21T14:45:37Z")

</div>

Hi [Stereolabs](https://www.stereolabs.com/) team,  
i am using a ZED i2 and i need the following parameters:  
-accelleration bias random walk  
-gyroscope bias random walk.  
If it is possible i need the reference frame of imu with respect to zed\_camrera\_link or zed\_camera\_center.

Thank you for your help

---

<div class="post-metadata">

**Author:** ![TristanLed](https://community.stereolabs.com/letter_avatar_proxy/v4/letter/t/ee59a6/32.png) [@TristanLed](https://community.stereolabs.com/u/TristanLed)\
**Post date:** [August 24, 2026, 4:11pm UTC](https://community.stereolabs.com/t/imu-noise-parameters/11644/2 "2026-08-24T16:11:29Z")

</div>

Hi Francesco,

Both values are available through the SDK, so the reliable way to get them is to read them from your own camera rather than from a datasheet. They sit in `getCameraInformation().sensors_configuration`, as `accelerometer_parameters.random_walk` and `gyroscope_parameters.random_walk`, with the unit in `sensor_unit`.

These two values are defined per IMU model, not per unit, so every ZED 2i returns the same figure for instance. The field is NAN when the camera does not provide it, and the gyroscope is in degrees, so convert to radians before feeding Kalibr or an EKF.

The IMU to camera transform is calibrated per unit and comes from the same structure, as `camera_imu_transform`. Its reference is the left camera, not the camera center. In ROS 2 the wrapper publishes `left_camera_frame` to `imu_link`, and the same transform on the `left_cam_imu_transform` topic. A tf2 lookup from `zed_camera_center` to `zed_imu_link` resolves, but only if you launch with `publish_imu_tf:=true`, which is false by default.

For better accuracy you can recompute the parameters with an Allan variance over a long static sequence, more than six hours, using ZED\_Sensor\_Viewer to grab the raw data.

Best regards,

Tristan Leduc
