@Hi OakD community, I’m working with an OAK-D setup using the OADW BNO086 IMU, and I noticed something odd in the calibration/intrinsics data. For the BNO086, all the q / b intrinsic values appear to be either perfect identity values or zeroes. For example, the matrix-like values look like exact identity matrices, and the bias/offset-style fields are all 0. I wanted to ask: * Is this expected behavior for the BNO086 on OAK-D/OADW? * Are these values placeholders because the BNO086 handles its own internal calibration/fusion? * Should I expect real calibrated IMU intrinsics/bias values to be exposed anywhere? * If I need accurate IMU-camera alignment or motion estimation, should I manually calibrate these parameters separately? I’m trying to understand whether the identity/zero values mean “already handled internally,” “not calibrated,” or “not exposed through the current API.” Any clarification would be really helpful. Thanks!

Why are OADW BNO086 q/b intrinsics identity matrices and zero values?

utkarshgupta846

@Hi OakD community,

I’m working with an OAK-D setup using the OADW BNO086 IMU, and I noticed something odd in the calibration/intrinsics data.

For the BNO086, all the q / b intrinsic values appear to be either perfect identity values or zeroes. For example, the matrix-like values look like exact identity matrices, and the bias/offset-style fields are all 0.

I wanted to ask:

Is this expected behavior for the BNO086 on OAK-D/OADW?
Are these values placeholders because the BNO086 handles its own internal calibration/fusion?
Should I expect real calibrated IMU intrinsics/bias values to be exposed anywhere?
If I need accurate IMU-camera alignment or motion estimation, should I manually calibrate these parameters separately?

I’m trying to understand whether the identity/zero values mean “already handled internally,” “not calibrated,” or “not exposed through the current API.”

Any clarification would be really helpful. Thanks!

Oskar-Luxonis

Hey utkarshgupta846 - the identity Q matrices and zero b values are valid neutral defaults. DepthAI applies them as calibrated = Q × uncalibrated + b, so these values apply no additional affine correction for scale, bias, or axis errors.

This calibration is separate from imuExtrinsics, which defines the IMU-to-camera transform. It is also separate from the BNO086 fused outputs, such as gravity and rotation vectors, which come from the IMU firmware.

For an older RVC2 device, these values most likely mean that no per-device affine IMU calibration was stored. calibration.

Thanks,
Oskar

utkarshgupta846

Hi, Thanks for the reply.
We have got the Extrinsics but we are searching for Factory Calibration Intrinsic values like bias, cross axis error, axis misalignment, scale factor error.
As written on: https://docs.luxonis.com/hardware/platform/sensors/imu/
Luxonis performs factory calibration of the IMU stack to compensate deterministic sensor errors

But when we are trying to retrive Factory IMU calibration data through DepthAI API we are getting Identity matrix (Q) and 0 values of vector (b). Is this actually correct there is no axis misalignment, bias or cross axis error in the IMU?
We got this device a month ago. we want to know If the factory calibration was missed for this device.

Oskar-Luxonis

Hey can you send me MXID of device, so I can check?

utkarshgupta846

Device MX_ID = "19443010D11F8E8B00", Batch = 2617TE

utkarshgupta846

Any Update?? If it is not factory calibrated, do you have any resource which we can use to calibrate it?

Oskar-Luxonis

Hey utkarshgupta846 - per-device IMU intrinsic calibration is a feature that we added recently, and it is still being tested in factory. Your device was produced before these calibration values were written to the device, which is why you see identity Q matrices and zero b values.

The IMU should still work normally. We do not currently have a reliable, supported resource for users to calculate and flash these values manually.

Thanks,
Oskar

Oskar-Luxonis

Hey Oskar-Luxonis - sorry for the confusion. The IMU itself is fully functional. The recent feature is the additional per-device Q|b correction, not IMU support.

You can use all supported BNO086 outputs:

Raw, uncalibrated, and calibrated accelerometer and gyroscope data
Magnetometer data
Gravity and linear acceleration
Rotation vectors and orientation quaternions

RAW uses the sensor-native axes. UNCALIBRATED applies imuExtrinsics to align the data with the camera coordinate frame. CALIBRATED also applies Q|b. Because your device has identity Q and zero b, it does not apply this additional affine correction.

Examples:

Thanks,
Oskar

utkarshgupta846

Then whats the point of IMU in this camera. Can you please let me know what is the use of IMU in this underdeveloped product?