PAPR Control in Digital Predistortion Systems Using PC-CFR and RLS Forgetting Factor Adjustment

Abstract The high peak-to-average power ratio (PAPR) of orthogonal frequency-division multiplexing (OFDM) signals presents a challenge for power amplifier (PA) linearization. Cascaded crest factor reduction (CFR) and digital predistortion (DPD) suppress signal peaks and compensate for PA nonlinearity, but unsuitable parameter settings can cause excessive waveform distortion or peak regrowth after predistortion. This paper investigates the joint adjustment of peak-cancellation CFR (PC-CFR) and recursive least-squares (RLS) identification to control DPD-output PAPR while preserving linearization performance. The PC-CFR threshold determines the distortion introduced before predistortion, whereas the RLS forgetting factor influences coefficient adaptation and subsequent peak regrowth. A sequential procedure first determines the target PAPR immediately before normalized mean-square error (NMSE) first worsens and then selects the lowest-NMSE forgetting factor satisfying the PAPR constraint. Experimental results show that joint adjustment reduces DPD-output PAPR from 12.985 to 11.593 dB and improves PA feedback NMSE from -34.877 to -38.537 dB. The adjacent-channel power ratio (ACPR) in the upper and lower channels improves from -49.370/-50.332 to -51.923/-52.747 dBc. These results demonstrate coordinated peak suppression and improved linearization for the measured operating condition.