Adaptive linearization of transmitter in the presence of I/Q Imbalance using distributed spatio-temporal neural network

For distortion-free data transmission, digital predistortion (DPD) is now a widely accepted method to linearize the power amplifier (PA) in a transmitter. DPD requires inverse modeling of PA and processing input through inverse model before feeding it to PA. However, when modulator in transmitter also have I/Q imbalance and local oscillator (LO) leakage, they cause extra intermodulation distortion (IMD) to appear at the PA output and conventional models such as Volterra series, memory polynomial, Weiner-Hammerstein and look up table methods are not able to model the inverse modeling properly and even worsen the spectral regrowth. This paper focuses on an adaptive distributed spatiotemporal neural network (DSTNN) for adaptive predistortion which has been shown to be robust to all the linear imperfections (i.e., gain/phase errors) as well as nonlinearity (i.e., IMD) to finally mitigate all the imperfections in the transmitter system in one step, due to its unique nonlinear mapping which does not depend on quadrature relation between I and Q components. DSTNN provides one-step solution for online adaptive application, which low cost in terms of floating point operations and does not require complex matrix operations such as matrix inversion.