is beyond the scope of this paper.
To summarize, compared to OFDM receiver with existing channel estimation techniques, the OFDM receiver with iterative turbo channel estimation can ap- proach the MSE lower bounds. It approaches performance of perfect CSI even with a small number of preambles and pilots in a rapid dispersive fading channel, which makes it an efficient solution in terms of both SNR and throughput. Fur- thermore, the iterative receiver is robust w.r.t residual CFO in practical OFDM systems.
3.7
Summary and Contributions
This chapter has investigated the problem of OFDM transmission in a rapid dis- persive fading channel. In such a highly mobile environment, the wireless channel undergoes fast variations both in time and frequency. In order to track the fast varying channel, large number of pilot tones are usually inserted to the OFDM symbol for existing receivers, which incurs huge SNR and throughput loss. An it- erative turbo channel estimation technique, which makes use of preamble, pilot and decoded soft data information for channel estimation is developed to improve the frame and bit error rate performance for a given signal to noise ratio while maximiz- ing the system throughput at the same time. The channel estimation is conducted by three estimation stages, where a frequency-domain and time-domain combining strategy is developed to combine the channel estimates from above signals in an efficient and low complexity manner. Numerical results and MSE analysis have shown that, compared to the OFDM receiver with existing conventional channel estimation, the OFDM receiver with iterative turbo channel estimation can ap- proach the performance with nearly perfect CSI at various mobility scenarios. In addition to the improvement in both SNR and throughput benefits, the receiver is robust to frequency error and has low computational complexity which means it is possible to implement in hardware.
Some specific contributions made in this chapter are as follows. First of all, the system model for the SISO-OFDM system model is investigated, the degradation from ICI due to realistic mobile radio channel is fully analyzed from theoretical perspective and validated through simulations. The effect of ICI in the channel estimation problem is modeled as Gaussian random process so that the system model for the channel estimation of the OFDM system is reformulated and simpli-
fied by combining the power of ICI with the white noise as effect noise. And the ML channel estimator and MMSE channel estimator are investigated and the MSE for MLE and MMSEE are analyzed through first and second order statistics of the channel estimator. The literature review on the conventional and iterative channel estimation techniques are presented.
Secondly, an iterative turbo channel estimator for the OFDM system is devel- oped. The iterative turbo channel estimator consists of three estimation stages, namely the initial coarse estimation stage, iterative estimation stage, and the final estimation stage. The initial coarse estimation is performed in the first iteration, the pilot estimation is performed. More specifically, in the downlink transmission, the preamble estimation is performed, and channel tracking through linear filter- ing is performed after the preamble estimation to obtain the channel estimation at the data subcarriers. On the other hand, in the uplink transmission, pilot-aided channel estimation is performed at the pilot subcarriers, and the linear interpo- lation is performed among the channel estimates from pilot subcarriers to obtain the channel estimates at the data subcarriers. After the initial coarse estimation, the data detection and decoding are performed. The soft coded information is fed back from channel decoder for the next iteration.
The iterative estimation is performed from the second iteration onwards. The channel estimates can be obtained through pilots and soft coded data symbols. The frequency-domain combining and time-domain combining are developed to explore the frequency-domain and time-domain correlations. The moving average windows along frequency and time directions are introduced to perform the combining. In the frequency-domain combining, the channel estimates from the pilot and data subcarriers are weighted in such a way the ratio of the pilot weights and the data weights are evolved adaptively over iterations. More specifically, at the beginning, the channel estimates from the soft coded data symbols are less weighted than the pilot symbols because the soft coded data symbols are less reliable than the pilot symbols due to the initial coarse channel estimation. Over iterations, the reliability or the energy of the soft coded data symbols improves, and eventually the soft coded data symbols can act as the pilot symbols. Hence, the wights between the pilot and data symbols are adjusted adaptively. The time-domain combining is similar to the frequency-domain combining, where the adjacent data symbols rather than the adjacent subcarriers are combined.
The final estimation is performed in the last iteration, where the soft coded data symbols can act as the pilot symbols eventually. If the channel statistics are
3.7 Summary and Contributions 71 not available at the receiver, the MLE is applied to perform the linear filtering to improve the channel estimates. On the other hand, if the channel statistics are known at the receiver, the MMSEE is employed.
Thirdly, the analysis of the lower MSE bounds that the iterative turbo channel estimation can achieve is presented. Theoretically, the iterative MMSEE has lower MSE than the iterative MLE where the additional gain is from the exploration of the channel statistics, which is usually difficult to obtain in the practical system. Furthermore, the complexity in terms of complex multiplications that the iterative turbo channel estimation requires are analyzed. Compared to the conventional MLE or MMSE estimation, the additional complexity from the iterative channel estimation remains low, which is very feasible for practical implementation.
Finally, the performances of the receivers with the iterative turbo channel es- timation and the conventional channel estimation techniques are compared. The numerical and analytical results show that the developed technique can approach the performance of systems with perfect CSI with much fewer preamble and pi- lots symbols compared to existing channel estimation methods. Therefore, under same system configuration, the iterative receiver improve the system performance over the time and frequency selective fading channel while maintaining the system throughput. Furthermore, the iterative receiver outperforms the conventional re- ceivers under pedestrian, low, intermediate, and high mobilities. And with marginal performance degradation, the iterative receiver is robust to within 4% carrier fre- quency offset after frequency acquisition.
Chapter 4
Iterative Receiver for
MIMO-OFDM system
4.1
Introduction
In Chapter 3, a receiver with novel iterative turbo channel estimation technique is presented, which shows how the near-optimal channel estimation and data detec- tion performance can be achieved in the realistic mobile radio channel environment. In this Chapter, the previous work is extended to MIMO-OFDM systems. A novel low complexity channel estimator with time-domain and frequency-domain combin- ing of channel estimates from preamble, pilots and soft decoded data information is proposed to track the dynamics of channel frequency response. This channel estimator is integrated with a MRC receiver for the Alamouti STC system and an interference canceler for the system with spatial multiplexing.