Study On Channel Estimation In MIMO-OFDM System Based On Pilots

1. IntroductionNew Generation Communication (beyond 3G/4G) will supply high data rate transmission up to 100Mbit/s or higher, and it will provide full range of services from voice to multimedia and even including stream media service. Data rates can be arranged dynamically according to the services. Very high spectral efficiency communication technology is required to realize high speed rate and large capability in limited spectrum resource. MIMO technology can take advantage of the spatial diversity obtained by spatially separated antennas, so it can improve channel capacity gain without increasing additional spectrum resource and antenna transmission power. OFDM is a technology of muticarriers. The time domain waveforms of the subcarriers are orthogonal. Hence, the available bandwidth is used very efficiently. OFDM converts a frequency-selective channel into parallel collection of frequency flat subchannels and it can restraint frequency selective fading efficiently. The combination of the two powerful techniques, MIMO and OFDM, is very attractive, and has become a most promising scheme for future wireless communication.Channel estimation is the key technology before signal detection, so study on channel estimation has very important significance. There are three distinct technologies in channel estimation. The first one is based on pilots or training sequence. The second class is blind channel estimation method and the third is semi-blind channel estimation method. Channel estimation in MIMO-OFDM system based on pilots is the technology studied on in this paper. This kind algorithm has good performance and is easy to realize, so it is applied broadly. The basic principle for channel estimation in MIMO-OFDM system based on pilots is make use of some known pilots in transmitter and estimate channel parameter. In order to make the channel estimation more accurately and have strong tracking capability, Doppler shift and maximum time delay is worst condition need to be considered. Under this situation we should insert sufficient pilots to keep up with the varying of time and frequency.This thesis mainly focused on channel estimation in MIMO-OFDM system based on pilot and training sequence. We analyzed LS and RLS channel estimation technology in MIMO-OFDM system, and introduced 2D-RLS algorithm and applied it in channel estimation in MIMO-OFDM system. This algorithm makes use of the LS estimation result of past symbols for basic estimation and makes a further accurate estimation with RLS. This method exploits channel structure in the time domain and the frequency domain, and it improves performance in channel estimation. In order to reduce computational complexity, a new algorithm 2D-NALMS based on 2D-RLS is proposed in this paper. Two dimensional estimation is also exploited in this method, but RLS is replaced by less complexity NALMS algorithm in time domain estimation. The pilot used in this paper also makes algorithm less complex, because Estimation of the full matrix is done with successive applications of the scalar channel estimation instead of vector.2. Pilot DesignIn case of channel estimation for a MIMO-OFDM system with M_Ttransmitters and M_R receivers, there are M_T×M_R channels to be estimated. As all transmitters send their signals simultaneously, the received signal at each receiver is a superposition of the transmitted signals that distorted by the channel. Hence in the estimation process of channel between i'th transmitters and j'th receiver the signal transmitted by other transmitters are interference. So in the MIMO channel estimation, whenever a pilot tone is inserted in a subcarrier, all other transmitters don't send anything in that subcarrier.3. 2D-RLS Channel Estimation for MIMO-OFDM System2D-RLS based on MIMO-OFDM system is a kind of two dimensional adaptive channel estimation algorithms based on LS and RLS. This method doesn't require accurate channel statistics, it use LS get the general result of channel parameter and make more accurate estimation by RLS. With properly chosen parameters, 2D-RLS can converge in steady state in several OFDM symbols time and has a better performance than LS and RLS. It is very adaptive and efficient for a time-varying channel. Computational complexity can be decreased by designing pilots. Estimation of the channel matrix if done with successive applications of the scalar channel algorithm by pilot method detailed above. LS estimation is use the pilot in the frequency. For the interpolation, a low-pass interpolation is performed by taking the IDFT of the estimated subcarriers, zero-padding, and then taking of DFT. The simulation results proves that 2D-RLS has fast convergence, steady state performance and good tracking ability. It achieves channel estimation fairly good when channel varying quickly.4. Proposed 2D-NALMS algorithmThe advantage of LMS algorithm compared to RLS is low computational complexity and can be realized easily, and the disadvantage is much slowly convergence rate. The 2D-NALMS algorithm is proposed based on 2D-RLS. This method makes use of the LS estimation result of frequency domain for basic estimation and makes a further accurate estimation with NALMS algorithm in time domain, so it can reduce computational complexity. For the time varying channel, the change of step-size can not match the instantaneous channel condition, so an algorithm 2D-NALMS is proposed in this paper. Normalized power method and adaptive trimming factor is introduced in this algorithm, and it can improves the estimation performance. The simulation results proves that 2D-NALMS has faster convergence rate and better performance than LS and RLS. It can achieve channel estimation good under Rayleigh time-varying channel.