Joint time delay and carrier frequency offset estimation for near equal power co-channel OFDM signals using low rank adaptive filters

Abstract

Carrier frequency offset (CFO) estimation in Orthogonal Frequency Division Multiplexing (OFDM) systems is difficult in co-channel interference (CCI), i.e. when a second user overlaps the signal-of-interest (SOI). Previous work presented a rank selection method applying MUltiple SIgnal Classification (MUSIC) to the time-domain signals, for jointly estimating the CFOs of two co-channel OFDM users, based on power level differences. In this paper, we apply a joint reduced rank implementation of this algorithm based on a more efficient method, the Reduced Order Correlation Kernel Estimation Technique (ROCKET). We compare ROCKET to the original MUSIC algorithm and show that ROCKET greatly improves CFO estimation accuracy when the CFOs of the two signals are nearly the same and time delays are random. We also compare ROCKET's CFO estimation accuracy to the Cramer-Rao Lower Bound (CRLB) and show it achieves better accuracy thanMUSIC, especially at low carrier-to-interference ratio (CIR), as well as improved bit error rate (BER) performance. Since synchronization must be obtained to estimate CFO, and given the ease with which OFDM signals may be transformed between time and frequency, we first apply the ROCKET algorithm in the frequency domain to jointly estimate the time delays of both users simultaneously. CFO (or delay) estimation accuracy is sufficient with a power difference between the SOI and the interferer, i.e. CIR, of just 0.5 dB, whereas the MUSIC algorithm only operates well down to CIR of 1–2 dB. The approach is to apply ROCKET in the time (or frequency) domain at rank D = k so that the frequency (or time) at which the spectrum peak occurs is the normalized CFO (or delay) estimate for the kth strongest user for all K users of the system. Hence, ranks D = 1 and D = 2 are applied when K = 2 users overlap.