<title>Implementation of a MAP decoder for use in UMTS receivers</title>

Abstract

Inter-user interference in mobile radio systems based on CDMA can be reduced with the aid of multi-user detection. As a consequence of the reduced interference the system capacity is increased for a given quality of service. The forthcoming UMTS system based on CDMA and FDD will apply long scrambling codes, thus, from a receiver implementation point of view random spreading codes are used. Random spreading codes, however, increase implementation complexity dramatically in a multi-user detection receiver. A receiver which shows both an acceptable implementation complexity and a bit error rate performance which is close to the single user bound, is the coded parallel interference cancellation (PIC) receiver. In principle, the coded PIC algorithm performs the following operations: after despreading and deinterleaving the receiver calculates soft values of the coded data with the aid of the maximum a posteriori (MAP) algorithm. The soft values are then respread and rechanneled. If all users are considered in the receiver, the multiple access interference can, ideally, be cancelled out. Through further stages in the receiver, i.e., doing the same procedure again, the performance of the single user receiver can approximately e obtained, as simulations have shown. In order to implement the coded PIC algorithm, its most complex component, which is the MAP decoder, has to be investigated in more detail. MAP decoding requires the calculation of probabilities and the storage over the entire data block because of a forward and backward recursion. These probabilities are obtained through multiplications and additions. To simplify the MAP algorithm on a large scale, conventionally the LOGMAP algorithm is used, which substitutes multiplications into additions with the aid of the logarithmic function and an additional correction function. To come up with a VHDL implementation of the LOGMAP algorithm, firstly a study into the bit resolution of a fixed point implementation is required. The paper investigates this problem for all variables and gives results of the fixed point implementation. With the aid of a professional simulation tool the VHDL implementation of the MAP decoder is investigated within a UMTS floating point simulation environment.© (2001) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.