Investigations On Digital Predistortion Technology For 5G Broadband Power Amplifiers

With the development of wireless communication technology,the fifth-generation(5G)and future communication systems are developing in the direction of higher user density,throughput and data capacity.Digital predistortion(DPD)has been widely used in linearization of transmitters due to its flexibility and stability.However,the new 5G application scenarios have brought some problems to the technology,such as large bandwidth and high peak-to-average power ratio(PAPR).The thesis focuses on the behavioral model and digital predistortion of broadband power amplifiers in 5G systems.Firstly,in the aspect of the high PAPR of broadband signals,the application of crest factor reduction(CFR)technology in DPD systems is studied.Based on the the joint CFR/DPD structure,the hard clipping(HC),the clip-and-filter(CAF),the clipping-and-bank-filtering(CABF)and other peak clipping techniques are evaluated.In order to reduce the sampling rate of the feedback loop in the DPD system,based on the band-limited structure,the band-limited magnitude selective affine(BL-MSA)model is proposed and it can reduce the complexity and the number of multipliers in conventional band-limited models.Furthermore,the modified random demodulation-based reduced sampling(MRDRS)model,which is based on the under-sampling structure,is applied for reducing the requirements for hardware resources in the feedback loop.In addition,the 2D simplified memery polynomial(2D-SMP)model is proposed to linearize dual-band power amplifiers and the work has been published in IEEE Microwave and Wireless Components Letters.The above models are all verified on the instrument platform.Theoretical analysis and experimental results show that,compared with the HC and the CAF,the CABF has better linearization performance while reducing crest factor in the joint CFR/DPD structure.When the memory depth and thresholds are the same,the number of multiplications of the BL-MSA model is only 1/5 of that of the BL-CPWL model,but the BL-MSA has about 1.5d B adjacent channel power ratio(ACPR)improvements.The MDRRS model still has good predistortion performance when the sampling rate is 1/5 of the signal bandwidth.Under the same model complexity,the 2D-SMP_(odd)model provides slightly higher model accuracy than the 2D modified memery polynomial(2D-MMP)model and its algorithm for parameter estimation is simpler.Secondly,a millimeter-wave transmitter front-end operating from 27.5GHz to28.35GHz is designed and implemented.The experimental results show that the transmission channel has good performance.The gain flatness,mirror suppression,local oscillator suppression and EVM meet the requirements.The C-band frequency synthesizer in the local oscillator module has good phase noise performance.At 6.5GHz,the phase noise is-106d Bc/Hz@100k Hz and-130d Bc/Hz@1MHz,respectively.Finally,the DPD test is carried out on the self-developed baseband hardware platform.The experimental results show that for the 100MHz 5G NR signal at the sub 6 GHz frequency band,the ACPR of a Skyworks power amplifier can be improved to below-45d Bc after DPD when using the BL-MSA model.For the 400MHz 5G NR signal at the millimeter wave frequency band,the ACPR of the millimeter-wave transmitter front-end can be reduced to about-41 d Bc if the BL-MSA model is adopted.