Analysis of RF-Power Amplifier Modeling Performance using a 16-QAM Modulation over AWGN Channels

This paper presents the black-box modeling of a microwave power amplifier (PA) and its performance in the nonlinear operation region - low input back off (IBO). As input signal a rectangle 16-QAM corrupted by different levels of additive white Gaussian noise (AWGN) was used. The nonlinearities caused by the amplifier in low IBO operating conditions can change the output constellation and lead to difficulties in finding the correct model parameters, as the black-box identification procedure is based on measured input/output signals. These signals can be altered by different types of noise. The AWGN degrades the signal to noise ratio (SNR) and can lower the model accuracy that should ideally not be influenced. In this paper the models were first estimated using noise-free data (NFD) and secondly with different levels of noisy corrupted data (NCD). NFD was also applied to the models calculated with NCD and vice versa. An investigation of the loss of accuracy caused by the noise in the identified models is presented. Three situations were considered: 6dB IBO, 4dB IBO, and 2dB IBO. It is shown that the pseudo-inverse based models present reasonable results for PA modeling when AWGN is incident