Automated Electrical Behaviour Modelling of Analogue and Mixed Signal Circuits using Data-based Methods

P. Senger, R. Doelling, and W. Rosenstiel (Germany)


Automatic Data-based Modelling, Electrical Behaviour, Machine Learning, Simulation Time Acceleration


This paper presents an automated approach for electrical black box behaviour modelling. The introduced strategy is based on the transient simulation data of in- and out put pins of an analogue or mixed signal circuit. The pre sented method extracts the magnitude and the phase in formation by measuring in- and output currents and volt ages and creates a new approximated electrical pin compat ible black-box model. The behaviour model is made up of cross connected intelligent admittances blocks and their re spective functional interconnections. These complex con nections between the admittance blocks are approximated using Support Vector Machines. The electrical behaviour models have a speed-up factor up to 37x and an accuracy greater then 95% compared to the original circuit simula tion. A special filter structure enables the modeling of fre quency, amplitude and load dependent circuits. The electri cal functioning of the methodology is shown by modeling a linear and passive analogue Butterworth filter and an active time variant mixed signal circuit.

Important Links:

Go Back