Chebyshev Polynomial-LSTM Model for 5G Millimeter-Wave Power Amplifier Linearization
Gaoming Xu, Huihui Yu, Changzhou Hua, Taijun Liu
Abstract
In this letter, a behavior model, namely CP- LSTM, composed of Chebyshev polynomials (CP) and a long short-term memory (LSTM) network is built to linearize the wideband millimeter-wave power amplifier (mmW PA) in the fifth-generation (5G) mobile communication system. In order to verify the linearization performance of the CP- LSTM predistorter, a 100-MHz bandwidth 5G new radio (5G NR) signal is employed to test the 28-GHz mmW PA under the text. Experimental results show that the adjacent channel power ratio (ACPR) of the mmW PA with the CP- LSTM can be improved by 20 dB which is 5-dB better than with the LSTM and 3-dB better than with the generalized memory polynomial (GMP). Therefore, the proposed CP- LSTM model is very effective to linearize 5G mmW PAs.