Abstract
The impact of feedback noise on digital predistortion with intuitive direct learning is investigated. The noise leads to an extra NMSE, which is NMSE η = 1/Gp(1-|1-θ| 2 ). σ n 2 /P y . Processing gain G p is the ratio of block size to the number of coefficients, θ is the approximation of the first order derivative for PA functions and p y /σ n 2 is the SNR of feedback path. For processing gain 22.5 dB, our experiments show noise of 35 dB SNR almost has no impact on linearization. If we consider 12 dB PAPR (8 dB PAPR + 4 dB margin), an 8-ENOB ADC could support DPD systems with intuitive direct learning.