Channel Estimation and Prediction in LTE

Studenteropgave: Kandidatspeciale og HD afgangsprojekt

  • Lathaharan Somasegaran
The 3rd Generation Partnership Project (3GPP) currently works on developing the third generation (3G) mobile telecommunication system towards a future 4th generation system. The evolution of the current 3G UMTS system was given the name Long Term Evolution (LTE). This project focuses on the downlink of the UMTS LTE, where orthogonal frequency division multiplexing (OFDM) is utilized as multiple access scheme. Based on working assumptions of 3GPP the project probes into different low-complexity methods in order to estimate and predict a time-varying channel for the UMTS LTE. The estimation is performed in two dimensions, i.e. the channel frequency response (frequency-domain) needs to be estimated at different time indices (time-domain). Common for all methods is the utilization of discrete prolate spheroidal (DPS) sequences for interpolation in the time-domain. The investigated methods are evaluated by simulations in Matlab. The channel model is chosen as a typical urban scenario modeled by Spatial Channel Model Extended (SCME), which is implemented with the LTE downlink structure in this project. The performance is measured using the mean square error (MSE) between the actual and the estimated frequency response. The performance is compared to the performance of a 2x1 dimensional Wiener interpolator, which consistently yields the lowest MSE but also the highest complexity. In general the investigated estimators have the same performance for channel prediction and which is close to the one of the 2x1D Wiener interpolation at a speed of 120 km/h. One of the investigated estimators, the linear minimum mean square error channel impulse response (LMMSE CIR) is a good compromise between complexity and performance. It is shown that the performance of this estimator for channel estimation purpose is close to the 2x1D Wiener filter at 120 km/h for different signal-to-noise ratios.
Antal sider89
Udgivende institutionAalborg Universitet
ID: 10608549