Research On Digital Predistortion Method Of Direct Learning Structure With Low Feedback Sampling Rate

With the commercialization of 5G wireless mobile communication technology,the signal bandwidth of wireless transmitters becomes wider to meet demand of highrate data transmission.Thus more distortion of power amplifiers occurs.The power amplifier is the major nonlinear device in wireless communication system.It is difficult to realize the balance between efficiency and linearity due to its inherent nonlinearity.The contradiction of the balance is more prominent in the application of broadband signal.In the meantime,the multi-band power amplifier is applied in many scenarios and the bandwidth is getting wider and wider.The nonlinear distortion of in-band and cross-band modulation in the multi-band power amplifier have serious impacts on the quality of communication.At present,predistortion becomes one of the most promising linearization techniques for PA,because of its high flexibility,robust stability and low cost.It can adapt to various scenarios.However,in the scenarios of both broadband and multi-band transmission,the expansion of bandwidth lead to a great increase of the feedback sampling rate,witch challenges the analogdigital converter.Therefore,it is necessary to improve the digital predistortion method to meet the new application scenarios and requirements.As the sampling rate of analog-digital converter increases,the cost and power consumption of system increase tremendously.So a digital predistortion method,which under samples the signal and broadens the spectrum,is proposed.Before analog-digital conversion,the band-limited filter is used to control the bandwidth of signal.Then,direct inverse algorithm is used to find an accurate predistortion model.In the experiment,the Volterra amplifier is applied in the simulation.As the simulation results shows,the predistortion method reduces the bandwidth of sampled signal and presents a good linearization.The ACPR and NMSE performance are improved by 16.80dB and 15.17dB respectively.In this paper,a two-dimensional digital predistortion method with a direct learning structure is proposed for concurrent dual-band transmitters.The PA modeling module match the actual power amplifier with the memory polynomial model.The predistortion coefficients training module inverses the PA model and obtain the accurate predistortion model through iterative operation.In the simulation,two kind of dual-band signals are applied.The predistortion results demonstrate that ACPR performance of lower band and upper band is reduced by 18.93dB and 20.59dB,NMSE performance is reduced by 21.38dB and 23.76dB for dual-band OFDM signal.And ACPR performance of lower band and upper band is reduced by 31.85dB and 31.51dB,NMSE performance is reduced by 20.86dB and 21.99dB for dual-band WCDMA signal.In conclusion,the predistortion presents a good linearization effect.