Robust Training Symbol Assisted Least Squares Estimation for Direct Conversion Receivers with Oscillator IQ-Imbalance

Abstract

The direct conversion receiver is attractive for being able to provide low-cost and compact receiver design for wireless communications. However, the oscillator imbalance in the in-phase and quadrature components of its mixing sinusoid introduces additional self-image interference, which is a non-linear effect to the received signal. Relying on just one OFDM training symbol, we formulate the joint estimation of oscillator imbalance, DC offset, and channel response as a linear model. A low-complexity least squares (LS) estimation method is then proposed. The transformation matrix in our model is shown to be of full rank with probability 1, which allows our LS approach to be realizable for a variety of training OFDM symbols. Justified by simulation results, our method outperforms the other three LS methods in the literature.