Cross-subframe channel estimation for low-complexity devices in LTE

Abstract: One of the most critical issues of wireless communication systems is the timevaryingand frequency-selective channel. Knowledge about the channel greatlyimproves communication performance. It enables coherent demodulation andmeasurements of Signal to Noise Ratio (SNR) and Channel Quality Indicators(CQI). Theses measurements are used to optimize the transmission schemes dependingon the channel conditions. Therefore, the channel estimation is one ofthe most important feature of modern wireless communication devices. In the3rd Generation Partnership Project (3GPP) Long Term Evolution (LTE) system,estimation of the channel is achieved using pilots called Reference Signals (RSs)which are scattered in time and frequency. The full estimation of the channel isdone by filtering and interpolation of the estimated pilots.The 3GPP Release 13 issued in June 2016 defines a new category of UserEquipment (UE) named category M1 (Cat-M1) which should support low SNRscenarios. At such low SNR, legacy channel estimation techniques based on persubframeestimation are not efficient enough. The standard thus enables crosssubframechannel estimation by insuring persistence of the channel conditionsduring a group of subframes.This thesis presents techniques for cross-subframe channel estimation. It showshow algorithms can be devised to obtain improved estimation accuracy comparedto single-subframe channel estimates while being resistant to Doppler effect andclock frequency offset. Three types of algorithms are studied: linear averaging,first-order Infinite Impulse Response (IIR) filters and Finite Impulse Response(FIR) Wiener filters. An analytic study of these algorithms is performed to findoptimal parameters in terms of channel estimation Mean Square Error (MSE).Algorithm validation is done with computer simulations to show that the BitError Rate (BER) performance of low-complexity algorithms (linear averagingand first-order IIR filtering) are very close to optimal Wiener filtering ones andthat they provide significant improvement over single-subframe techniques.