Adaptive averaging channel estimation for DVB-T2 systems

Abstract

In modern communication systems, the rate of transmitted data is growing rapidly. This leads to the need for more sophisticated methods and techniques of implementation in every block of the transmitter-receiver chain. The weakest link in radio communications is the transmission channel. The signal, which is passed through it, suffers from many degrading factors like noise, attenuation, diffraction, scattering etc. In the receiver side, the modulated signal has to be restored to its initial state in order to extract the useful information. Assuming that the channel acts like a filter with finite impulse, one has to know its coefficients in order to apply the inverse function, which will restore the signal back to its initial state. The techniques which deal with this problem are called channel estimation. Noise is one of the causes that degrade the quality of the received signal. If it could be discarded, then the process of channel estimation would be easier. Transmitting special symbols, called pilots with known amplitude, phase and position to the receiver and assuming that the noise has zero mean, an averaging process could reduce the noise impact to the pilot amplitudes and thus simplify the channel estimation process. In this thesis, a novel channel estimation method based on noise rejection is introduced. The estimator takes into account the time variations of the channel and adapts its buffer size in order to achieve the best performance. Many configurations of the estimator were tested and at the beginning of the research fixed size estimators were tested. The fixed estimator has a very good performance for channels which could be considered as stationary in the time domain, like Additive White Gaussian Noise (AWGN) channels or slowly time-varying channels. AWGN channel is a channel model where the only distorting factor is the noise, where noise is every unwanted signal interfering with the useful signal. The properties of the noise are that it is additive, which means that the noise is superimposed on the transmitted signal, it is white so the power density is constant for all frequencies, and it has a Gaussian distribution in the time domain with zero mean and variance σ2=N. A slowly time varying channel refers to channel with coherence time larger than the transmitted symbol duration. The performance of a fixed size averaging estimator in case of fast time-varying channels is subject to the buffering time. When the buffering time is smaller or equal to a portion of the coherence time the averaging process offers better performance than the conventional estimation, but when the buffering time exceeds this portion of the coherence time the performance of the averaging process degrades fast. So, an extension has been made to the averaging estimator that estimates the Doppler shift and thus the coherence time, where the channel could be assumed as stationary. The improved estimator called Adaptive Averaging Channel Estimator (AACE) is capable to adjust its buffer size and thus to average only successive Orthogonal Frequency Division Multiplexing (OFDM) symbols that have the same channel distortions. The OFDM is a transmission method where instead of transmitting the data stream using only on carrier, the stream is divided into parallel sub-streams where the subcarriers conveying the sub-streams are orthogonal to each other. The use of the OFDM increases the symbol duration making it more robust against Inter-Symbol Interference (ISI), which the interference among successive transmitted symbols, and also divides the channel bandwidth into small sub-bandwidths preventing frequency selectivity because of the multipath nature of the radio channel. Simulations using the Rayleigh channel model were performed and the results clearly demonstrate the benefits of the AACE in the channel estimation process. The performance of the combination of AACE with Least Square estimation (AACE-LS) is superior to the conventional Least Square estimation especially for low Doppler shifts and it is close to the Linear Minimum Mean Square Error (LMMSE) estimation performance. Consequently, if the receiver has low computational resources and/or the channel statistics are unknown, then the AACE-LS estimator is a valid choice for modern radio receivers. Moreover, the proposed adaptive averaging process could be used in any OFDM system based on pilot aided channel estimation. In order to verify the superiority of the AACE algorithm, quantitative results are provided in terms of BER vs SNR. It is demonstrated that AACE-LS is 7dB more sensitive than the LS estimator.