Estimation in CDMA ~~~~~~n~~~~io ia Maximram Likelihood Techniques
Raghu Madyastha, Behnaam Aazhang · 1997
We present a maximum likelihood method for delay estimation in a CDMA wireless com- munication system. An antenna array is used at the receiver, which facilitates the concomitant estimation of the direction of arrival (DQA) of each user. In ad- dition, the amplitude of each user is also estimated. A single path is considered for each user thereby re- sulting in uncorrelated signals at the receiver. The delay estimation reduces to the solution of a set of quadratic equations while the DQA estimation prob- lem is equivalent to an eigenvalue problem. I. INTRODUCTION We assume a I(-user direct sequence CDMA system with BPSK (Binary Phase Shift Keying) modulation with each transmitted signal limited to (O,T). Each user transmits a zero mean bit sequence with i.i.d components and different users are independent of each other. The bitstream of each user is further modulated by a spreading sequence of length Ai, which is assumed to be periodic with the bit interval. In this development we will assume a single path channel; each of the users transmits through a different time varying channel whose parameters, we will assume, are constant in the time taken to estimate them. The first stage in the demodulation is the so-called acquisi- tion stage wherein the receiver attempts to lock onto the phase of the desired user's code (I). The acquisition algorithm in this paper is based on a computationally elegant ML algorithm for direction of arrival (DOA) estimation presented in (2). The re- ceiver at the base station incorporates an antenna array of M sensors in a specified geometry. It can be shown that increas- ing the number of sensors in the receiver improves detection performance especially with linear multiuser detectors. We extend the ML algorithm discussed in (a) to the simultane. ous estimat.ion of DOA as well as code delay. We assume the transmission of training sequences by all the users that are being acquired. By incorporating the single path assumption the received bit streams are rendered uncorrelated. The ad- ditive noise is assumed to be a circularly complex zero mean Gaussian random vector; however no a priori knowledge is assumed of its covariance structure. 11. ML ALGORITHM