The gain of the CMSIS pid are not the same as the one calculated in Matlab
While testing the PID from zephyr we realized that the PID library used required different gains that the matlab block used as input. This difference in gain lead to a basic workaround of multiplying the error per 0.01. Without really understanding why it was required. Although it kind of worked so far, the pid was unstable when pushed beyond is zone of controllability and not able to converge back after such an event leading to high suspicion on the value of KI.