Equalization of LTE-OFDM Systems over Doubly Selective Channels

Abstract

OFDM is employed in numerous existing standards for local and metropolitan area networks like IEEE 802.11 a, g, n (WLAN), IEEE 802.15.3 a (PAN and UWB), IEEE 802.16 e (WMAN) and IEEE 802.20 (MBWA) families because it is an effective method to overcome frequency-selective fading without complex equalizers at the receiver. OFDM along with MIMO have been proposed as the fourth generation technologies for LTE and WiMax. However, both require data rates from mobile devices that are 15 to 100 times greater than 3G technologies. In such scenarios where the transmitter and/or receiver are moving with respect to each other the channel becomes doubly selective (time selective and frequency selective). Data transfer rates of these mobile devices are limited by the effects of fading which introduces ICI between orthogonal carriers in OFDM. Hence, channel estimation and equalization becomes important in high mobility scenarios for achieving high data rates. In this thesis work, the effects of fading on the OFDM system are studied and low complexity equalization with LSRQ and GMRES algorithm is compared with traditional equalization methods. As the next generation technologies like LTE and WiMax have a large number of subcarriers the traditional equalization methods like MF, ZF, MMSE, VBLAST, and SAGE which requires matrix inversion to become highly complex in time and consume a lot of time to equalize the received signal. An alternative approach that involves solving the system of linear equations using Krylov subspace based methods like LSQR and GMRES is hence employed to equalize the signal in Doubly Selective channels and their performance is compared to the traditional methods. It is found that LSQR algorithm gives performance comparable to MMSE equalization, but not as well as VBLAST and SAGE which give a performance gain of about 4.0477 dB and 4.5593 dB at 30dB SNR; trading accuracy for speed. Extending the endeavor to reduce complexity a new channel model based on Piecewise Time Invariant Approximation (PITIA) of the channel is used to estimate the OFDM system in a Doubly Selective channel and LSQR algorithm is used for equalization. This method is compared with the traditional LS and MMSE estimators and it found that it outperforms them while offering low complexity. PITIA achieves this by decoupling the frequency selective and time varying effects of the channel and thereby reducing the number of parameters to be estimated. The simulation is carried out on channel specifications of ITU-models for LTE systems in vehicles and urban ETU environments and it is observed that PITIA equalizer outperforms the LS and MMSE equalizers. PITIA achieves a BER of 0.0019 at 18 dB SNR in vehicular environment and a BER of 10-2 at 27 dB in the Urban ETU environment. Simulations in these scenarios reveal that PITIA is suitable for small to moderate delay spreads.