Analysis of digital predistortion architectures for direct digital-to-RF transmitter systems

This work presents an analysis of digital predistortion (DPD) architectures for radio-frequency digital-to-analog convertor (RFDAC) based direct digital to RF transmitter (DRF). The nonlinearity of the DRF frontend is determined by circuit simulation. Direct learning and indirect learning structures using least mean square (LMS) and normalized LMS (NLMS) estimation algorithm are evaluated with respect to the lowest normalized mean square error (NMSE) and the error vector magnitude (EVM) and spectral emissions when processing broadband OFDM based IEEE 802.11a (WLAN) signals. MATLAB simulations show the possibility to reduce out-of-band emissions by 8 dB and to improve the EVM from 4.27% without predistortion to 2.50%using DPD. All DPDs show similar results, hence the architecture with the lowest complexity is evaluated for digital implementation. The impact of quantization and delay is evaluated, increasing the EVM to 3.71% and showing a strong dependence of the resulting EVM from digital circuit delay.