Evaluation of a Safety Multi-Core Platform for the Sensorless Drive of a PMSM

To reduce hardware requirements and costs recent applications apply sensorless methods to drive electrical machines. However, such methods have a high demand on CPU resources, reducing the possibility to integrate further applications into the system. Multi-core architectures potentially provide the necessary performance for joint integration of sensorless driving methods and additional, typically outsourced, applications. However, such parallel execution can induce cross-core interference which can corrupt the time-critical execution of the drive controller and reduce the overall system performance. In this paper we evaluate a speed control application for a sensorless drive and quantify its sensitivity to cross-core interferences. This allows estimating the influence of genuine parallelism on the sensorless drive functions. The same applies to the performance when integrating high-level functions into the same system. The evaluation is carried out on a practically oriented multi-core platform with safety features to directly apply the results for practical use. It is shown that the sensorless algorithm suffers from a maximum slowdown of 6.8% while the frequency of the speed control can be increased by up to 75%.