Zigang Yang

Digital predistortion using direct learning with reduced bandwidth feedback

Digital predistortion (DPD) is a popular approach for linearizing power amplifiers (PAs) in wireless base stations and improving system efficiency. The feedback path in a DPD system typically requires 5x signal bandwidth to accommodate the bandwidth expansion caused by the DPD and nonlinear PA. In this paper, we propose a new approach for adapting DPD parameters using direct learning and reduced bandwidth feedback. The new approach is capable of achieving near full-rate DPD performance and linearization bandwidth with significantly reduced feedback bandwidth. Measurement results on a Doherty PA achieved more than 20 dB corrections over 200 MHz bandwidth for a 2-carrier WCDMA signal spanning 40 MHz with only 81.92 MHz feedback bandwidth.

Concurrent dual-band digital predistortion

With a power amplifier (PA) being able to support widely separated frequency bands, dual-band digital predistortion techniques have been proposed as an efficient alternative to traditional wideband digital predistortion by only compensating nonlinear distortions around the signal bands of interest. In this paper, we provide theoretic analysis of the connection between wideband and dual-band memory polynomial models and propose two efficient implementation schemes based on lookup tables. Performance was evaluated with an experimental test setup using a symmetric Doherty PA and a dual-band signal (10 MHz LTE and 2-carrier WCDMA) with 97 MHz center separation. Drain efficiency of 31% was achieved with 41.2 dBm output power and both bands meeting −45 dBc ACLR requirement.

Crest factor reduction for dual-band systems

As wideband power amplifiers (PAs) become more power and cost efficient, supporting widely separated frequency bands with a common PA and transmit chain is becoming more practical. Dual band digital pre-distortion (DPD) techniques have been proposed as efficient alternatives to traditional wideband solutions for such applications. In this paper, we propose a novel dual band crest factor reduction technique that works together with a dual band DPD solution. Theoretical analysis is performed to compare the dual band solution with the traditional wideband solution. Performance of the proposed technique was demonstrated with a dual-band signal that has two 10 MHz LTE bands with edge to edge separation of 200MHz.

Digital predistorter characterization using extended orthogonal Chebyshev polynomials

As communication signals become more dynamic and have wider bandwidth, it is desirable to have a tool that helps to understand digital predistortion (DPD) solution space for a given power amplifier (PA). Two-tone measurements have been used extensively for PA characterization. However, direct application of two-tone measurements to DPD characterization is difficult because of the correlation in conventional polynomial basis functions. In this paper, we establish a technique for characterizing the DPD solution space using 2-carrier narrowband signals by adopting the memory polynomial model and extending the orthogonal Chebyshev polynomials. This approach allows a direct view of different intermodulation products in the DPD solutions for various 2-carrier signals. We performed characterization on a Class AB PA and found interesting characteristics of the solution space. Based on the findings, we propose a new DPD adaptation scheme for frequency hopped 2-carrier GSM signals. The proposed scheme achieves excellent linearization performance for the frequency-hopped GSM signals.