J4 ›› 2010, Vol. 37 ›› Issue (3): 565-569.doi: 10.3969/j.issn.1001-2400.2010.03.032

• Original Articles • Previous Articles     Next Articles

Separate BPNN-predistortion method for nonlinear HPA with memory

CUI Hua1,2;ZHAO Xiang-mo1;AI Bo3   

  1. (1. School of Information Eng., Chang'an Univ., Xi'an  710064, China;
    2. Shaanxi Eng. and Tech. Research Center for Road and Traffic Detection, Xi'an  710064, China;
    3. National Key Lab. of Control and Safety of Railway Traflic, Beijing Jiaotong Univ., Beijing  100044, China)
  • Received:2009-04-20 Online:2010-06-20 Published:2010-07-23
  • Contact: CUI Hua E-mail:cuihua7276@yahoo.com.cn

Abstract:

To circumvent the predistortion limitation for nonlinear high power amplifiers with memory (NHPAM) in OFDM communication systems, a separate BPNN (back propagation neural network) predistortion method with its structure and algorithm and their principles is proposed by considering the distortion characteristics and structure features of NHPAM. This method separates the whole predistortion for NHPAM into memory, AM/AM and AM/PM predistortion subsystems. Then the three NN predistorters are identified by the Levenberg-Marquardt BP algorithm based on the indirect training method. Simulation results show this proposed method can reduce ACEPR (Adjacent Channel Error Power Ratio) 30dB.

Key words: neural network, nonlinearity with memory, high power amplifiers, predistortion