A Study of Pilot-Assisted OFDM Channel Estimation Methods With Improvements for DVB-T2

Abstract

We provide a comparative study of channel estimation methods for pilot-assisted orthogonal frequency-division multiplexing (OFDM) communication systems in time-varying frequency-selective fading channels. We also propose an improved low-complexity least-squares (LS) estimation method based on the domain-transform process, which estimates time-domain channel impulse response (CIR) using the LS estimate of pilot subchannels. Our method is particularly suited to commercial OFDM systems such as the second generation of digital terrestrial television broadcasting systems (DVB-T2) because it has the following properties: 1) It uses only one OFDM symbol; 2) it does not require knowledge of channel statistics; 3) it works for any pilot pattern; and 4) the size of the estimation matrix for obtaining CIR only depends on the number of pilot subcarriers, but not on the size of OFDM symbol. Because of the first three properties, our method is robust to high Doppler frequency shifts, timing synchronization errors, and pilot patterns in commercial systems. Moreover, through computer simulations, we verify that our method provides competitive bit error rate and channel estimation error performance, compared with some well-known methods, which suffer from either a much higher computational load or less robustness to Doppler shifts or timing errors. For irregular pilot patterns, efficient yet simple regularization methods are suggested to solve the ill-conditioned problem of the estimation matrix. An upper bound on the mean squared error of the estimated CIR is also provided.