Swift Quasi-Peak Detector Implementation using Neural Network

This paper presents an Python implementation of a CISPR 16-1-1 time-domain EMI receiver for post-processing of transient data originating from circuit simulation or oscilloscope measurement. The scientific novelty is the implementation of the Quasi-Peak (QP) detector with the help of machine learning (ML). The QP ML-model is trained with an automated lab-oratory setup using an actual EMI receiver and parameterized impulse burst train stimuli. The advantage of the proposed model is that it can very quickly and accurately predict QP readings from only one period of transient input data. In contrast, classic QP detector implementations require a waveform input of at least 2 seconds and a model of the critically damped meter to evaluate the settled QP reading.