Analysis on RLCG Parameter Matrix Extraction for Multi-Core Twisted Cable Based on Back Propagation Neural Network Algorithm

Abstract

A new method based on back propagation (BP) neural network for extracting RLCG parameter matrix of multi-core twisted cable is presented. With the properly selected parameter matrix sample, the variation characteristics of the parameter matrix of the multi-core twisted cable can be learned by the Levenberg-Marquardt (L-M) algorithm based on BP neural network. The proposed method is combined with the finite-difference time-domain (FDTD) method to calculate the near end crosstalk (NEXT) and the far end crosstalk (FEXT) of the multi-core twisted cable. To verify the new method, a three-core twisted cable is measured and analyzed in the frequency band of 100 kHz - 1 GHz. The results show that the verification error of the extraction network of the RLCG parameter matrix has good accuracy, which does not exceed 0.8%. Compared with the full wave method, the maximum deviations of FEXT and NEXT solved by the proposed methods are -2.71 dB and 10.56 dB, respectively, which are better than 29.32 dB and 32.45 dB solved by the conventional method.

Keywords

• Multi-core twisted cable
• crosstalk
• finite-difference time-domain (FDTD) method
• neural network