ZED X One S Stereo Calibration Fails at 0.5 m Baseline

I’m trying to calibrate a custom stereo setup using two ZED X One S (narrow FOV), but calibration fails due to high reprojection errors even at a relatively short baseline of 0.5m.

My Setup:
Cameras: 2 × ZED X One S (narrow)
Mounting: Rigidly fixed on a static bar (no movement)
Current baseline tested: 1 m and 0.5 m
Future goal: ~2 m baseline

Calibration pattern:
Chessboard: 11 × 8 inner corners
Square size: 65.3 mm
Printed and mounted on a rigid flat surface

Followed the official guide:
https://www.stereolabs.com/docs/cameras/zed-x-one/zed-x-one-stereo
Achieved 100% coverage in all metrics: X,Y,Size,Skew.

Calibration fails due to reprojection error:

Even when coverage is perfect, I cannot get below the 0.5 px threshold, although i am very close to success, there is always something that above 0.5 and fails me.

This happens at 0.5 m baseline, not long baseline

Observations

  • This happens at 0.5 m baseline, not long baseline

  • Cameras are completely rigid

  • Chessboard is flat and stable

  • Images look sharp and well exposed

  • Repeating calibration gives slightly different results, but always above threshold

Eventually I want to calibrate a ~2 m baseline stereo rig, so I’m trying to validate the process first at 0.5 m. what size of square i will need at this case if i cannot handle the 0.5 m ?

You could try to use bigger squares (80 mm) to improve the detection when the target is moved far from the cameras.
Also, check that there are no flares and shadows on the target.

You are not too far from the threshold (0.5 px), it’s only a matter of a few steps: