Td-scdma System, Joint Detection Algorithm And Its Dsp Implementation

Independantly developed and submitted by China, TD-SCDMA is known as the low chip-rate time division duplex (LCR TDD) mode of the third generation partnership program (3GPP). The TD-SCDMA system adopted joint-detection techniques to eliminate Multiple Address Interference (MAI) and Inter Symbol Interference (ISI). A joint detector combines the knowledge about all users that share one burst into one large system of equations. This knowledge consists of the channel impulse responses that have been estimated from training sequences, the spreading codes, and the received antenna samples from all antenna elements.After giving an introduction of the physical channel and the uplink baseband signal model in TD-SCDMA system, not only do we present three commonly used joint detectors which include DMF(Decorrelating Matched Filter), ZF-BLE(Zero Forcing-Block Linear Equalizer) and MMSE-BLE(Minimum Mean Square Error Block Linear Equalizer), but also analyze and compare their algorithm efficiency and complexity.Mentioned in TDD standard 25.945, ZF-BLE algorithm is widely used in practical applications. It is performed by solving a least squares problem, where the system matrix has a block-Sylvester structure. However, ZF-BLE still involves the inverse computation of large matrices requiring immense computing power. Thus in chapter 4, we focuses on two algorithms with the advantage of fast matrix inverse, which are Approximate Cholesky Decomposition algorithm and Block Fourier algorithm.In chapter 5, first we build a simulation platform of the TD-SCDMA uplink physical layer with the help of MATLAB tools. Then based on this platform, we give a comprehensive analysis of the performance of joint detection techniques.Cholesky decomposition is fundamental for implementing Block-Fourier algorithm which is based on ZF-BLE linear joint-detection. In chapter 6, after giving a deep insight of Cholesky decomposition, we analyze some key issues that might baffle us when it is implemented in fixed point DSP devices for real-time requiring applications. Taking advantage of the unique characteristics of the structure of core processing unit of the Starcore 8144 DSP provided by Freescale, we present a fixed-point routine satisfying real-time requirements based on Newton-Raphson method. It significantly reduces the computing complexity, and is very likely to have a good application prospect in the base station receiver end in real system's uplink transmission.