Statistical modelling of a non-linear high-power amplifier with memory in multi-input-multi-output orthogonal frequency division multiplexing systems

A simple and accurate statistical model (SASM) for a high-power amplifier (HPA) with memory in multi-input multi-output (MIMO) orthogonal frequency division multiplexing (OFDM) systems is proposed. This model is statistically equivalent to the more complex generalised memory polynomial (GMP) model, which is one of the most accurate models. The HPA distortion is modelled using stationary white Gaussian noise sources that pass through finite impulse response (FIR) filters. The filter coefficients and the variance of noise sources are derived analytically from the GMP model parameters by spectral decomposition of the GMP noise power spectrum. The number of FIR filters is equal to the memory length of the GMP model. The HPA effect on the OFDM signal, considered as frequency-dependent attenuation coefficients, is modelled using another FIR filter. These FIR filters with small lengths make the model computationally inexpensive and mathematically tractable. The main characteristics of this model are identical bit error rate (BER) and error vector magnitude of subcarriers with those in the GMP model. The proposed model is the simplest representation of a HPA with memory effects. An excellent agreement between the simulation results of the SASM and GMP model is observed. This verifies the accuracy and efficiency of the proposed method.