Differential mode attenuation of RF suppression filters can be measured directly. Alternatively, this attenuation can be derived as modal S-Parameters from the S-Parameters measured for the single multiport paths which in fact the filter is. This paper outlines the differences between the two approaches. The first consists in direct measurement with separation transformers and the second with calculation from single ended measurement. In addition, the paper presents the method of determining the differential mode attenuation for the case of input and output impedance different from those used in direct measurements, i.e. by 50Ω/50Ω. At the same time, the presented method of determining the differential mode attenuation for different input and output impedances will require a change in the assumptions of the CISPR 17 document entitled ”Methods for measuring the attenuation characteristics of passive EMC filtering devices”, since the modal S-Parameter formulas presented in this document are only valid for the case when the RF suppression filter is in perfect balance. Otherwise, for different impedances it should be done taking into account the modification of the modal S-Parameters, as shown in this paper.The inclusion of this innovative change in the CISPR 17 document will allow the approach presented in this document to be used for various cases of the RF suppression filter, which is undoubtedly a significant progress in determining its attenuation.Frequency span by measurement of attenuation of RF Suppression filter is spanned from 10 kHz to 30 MHz.The comparison presented in this paper confirms that the S parameters are identical only if the filter is in equilibrium (ZL=ZN).From the results, three frequency areas can be distinguished. In the first interval between 10 kHz and 150 kHz the filter is in balance, which was demonstrated with identical measurement results. Between 150 kHz and 1.5 MHz, the attenuation is superimposed on the dynamics of the measurement system and therefore on the random frequency dependence of both measurements. Above 1.5 MHz there is a clear filter asymmetry, so the curves move away from each other.
Read full abstract