MILA - A noise mitigation technique for RF power amplifier linearization

Abstract

Standard indirect learning architecture (ILA) based methods for estimating the parameters in digital pre-distortion produce biased estimates of the parameters due to noise. This paper proposes using a direct model of the power amplifier to generate a noise-free estimate of the output signal and subsequently use this estimated signal to extract the parameters of the pre-distorter using the ILA. Measurement results shows that this technique improves the performance of the pre-distorter for all signal-to-noise ratios, but especially in the range of 20-30 dB signal-to-noise ratio.