Estimation of the Number of Multipath Components in a Delay-Dispersive Environment for LTE OFDM Downlink

Malta, Vincenzo

10. term

Elektronik og IT, Kandidatuddannelsen (Spec. Signal and Information Processing for Communication)

2010

89

LTE’s downlink bruger OFDM, som deler data op over mange små frekvenskanaler. Modtageren skal estimere, hvordan radiosignalet påvirkes af kanalen (kanalestimering), typisk ved hjælp af pilotsignaler. Hvis man antager en parametermodel for kanalen, kan nøjagtigheden forbedres. En sådan model beskriver, at signalet når frem via nogle få veje med hver sin forsinkelse (tap-delay-model). For at disse estimationsalgoritmer kan fungere, skal man kende, hvor mange forsinkelser der er (modelorden). Specialet undersøger, hvordan man bestemmer dette antal, med fokus på estimators som ESPRIT. Vi afprøver MDL (Minimum Description Length) til at detektere modelordenen på en foreslået LTE-kanalmodel og på en dynamisk, tidsvarierende kanal, og sammenligner med en enkel tærskelregel. Begge metoder anvendes sammen med en forbehandling kaldet spatial smoothing. Resultaterne viser, at på en statisk kanal er MDL robust, selv når observationstiden reduceres, mens tærskelmetoden får markant større fejl. På en dynamisk kanal falder ydeevnen generelt, men MDL’s gennemsnitsfejl forbliver omtrent konstant over et bredt interval af vinduesstørrelser og forværres ikke drastisk som tærskelmetoden. Konklusionen er, at state-of-the-art ikke overgås, men de observerede udfordringer gør området fortsat åbent for videre arbejde.

LTE’s downlink uses OFDM, which spreads data across many small frequency channels. The receiver must estimate how the radio signal is altered by the channel (channel estimation), typically using pilot signals. Assuming a parametric channel model can improve accuracy. Such a model describes the signal arriving over a few paths, each with its own delay (tap-delay model). To run these estimators, one must know how many delays there are (the model order). This thesis studies how to determine that number, focusing on estimators like ESPRIT. We evaluate MDL (Minimum Description Length) for model-order detection on a proposed LTE channel model and on a dynamic, time-varying channel, and compare it with a simple threshold rule. Both methods are used with a preprocessing step called spatial smoothing. Results show that on a static channel, MDL is robust even when observation time is reduced, while the threshold method suffers a significant increase in error. On a dynamic channel, performance drops overall, but MDL’s average error stays roughly constant across a wide range of window sizes and degrades less than the threshold method. In conclusion, state-of-the-art performance is not exceeded, and the observed issues leave this topic open for further work.

[This abstract was generated with the help of AI]

OFDM ; Estimation ; Multipath ; Delays ; LTE

Download
View record in AAU Student Projects