NARMA based novel closed loop digital predistortion using Moore-Penrose inverse technique

Abstract

A novel digital predistortion (DPD) technique based on single loop feedback, nonlinear auto-regressive moving average (NARMA) structure is presented. The distinctive characteristics of this DPD is its straightforward deduction from the NARMA model of the power amplifier (PA) characteristics to be linearized, and the direct learning approach used to identify the DPD function. Moore-Penrose pseudo inverse technique is used to calculate the least mean square solutions, and it's performance is also reported. The proposed DPD technique is tested and validated on a 12 Watt Doherty power amplifier (DPA), with Long Term Evolution (LTE), and multicarrier Wideband Code Division Multiplexing Access (WCDMA) signals at 2.4 GHz. Simulation and experimental results show that such DPD is able to assure the linearity requirements of LTE and WCDMA standards by reducing the Adjacent Channel Power Ratio (ACPR) level at DPA output, from −35dBc to −50dBc relative to the carrier, and the Error Vector Magnitude (EVM) reduced from 24.6% to 1.6%. Moreover, a comparative study between the proposed DPD technique and Memory Polynomial DPD (MP-DPD) has also been carried out. The experimental comparison demonstrates the superior performance of the novel DPD formulation with respect to MP-DPD achieving an improvement of 6 dBc and 12 percentage points in ACPR and EVM, respectively.