A Band-Divided Memory Polynomial for Wideband Digital Predistortion With Limited Bandwidth Feedback

With a real analog bandpass filter (BPF) in the feedback loop, band-limited digital predistortion (DPD) normally uses a narrower bandwidth low-pass filter at baseband to further constrain the bandwidth of the power amplifier (PA) output signal. This discards the nonlinear information at the BPF'...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on circuits and systems. II, Express briefs Express briefs, 2015-10, Vol.62 (10), p.922-926
Hauptverfasser: Zhang, Qi, Liu, Youjiang, Zhou, Jie, Jin, Shubo, Chen, Wenhua, Zhang, Silong
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:With a real analog bandpass filter (BPF) in the feedback loop, band-limited digital predistortion (DPD) normally uses a narrower bandwidth low-pass filter at baseband to further constrain the bandwidth of the power amplifier (PA) output signal. This discards the nonlinear information at the BPF's roll-off regions. The PA's spectral regrowth at the edge of the BPF is thus unable to be suppressed with the conventional band-limited DPD. To make efficient use of digital-to-analog converter/analog-to-digital converter sampling rate without losing information at the BPF's roll-off regions, this brief proposes a band-divided DPD method, in which the restricted feedback bandwidth is divided into two parts in frequency domain-center and border regions of PA's band-limited output. These two parts are characterized separately using two different basis functions with a band-divided memory polynomial (BDMP) model. It provides higher flexibility in selecting the digital band-limiting filter's bandwidth and also reduces the filter order. The measurement results for a class-AB GaN PA with orthogonal frequency division multiplexing signal of 60-MHz bandwidth have shown the superiority of BDMP DPD over the previous state-of-the-art.
ISSN:1549-7747
1558-3791
DOI:10.1109/TCSII.2015.2457793