DESIGN OF OPTIMIZED FIXED-POINT WCDMA RECEIVER Hai-Nam Nguyen, Daniel Menard, and Olivier Sentieys IRISA/INRIA, University of Rennes 1, 6 rue de Kerampont F-22300 Lannion Email: hanguyen@irisa.fr ABSTRACT To satisfy energy and complexity constraints, embedded wireless systems require fixed-point arithmetic implementation. To opti- mize the fixed-point specification, existing approaches are based on fixed-point simulations to evaluate the performances. In this paper, the approach used to optimize the fixed-point specification for a WCDMA receiver is presented. The dynamic range and the fixed-point accuracy are evaluated analytically in our approach. The analytical accuracy constraint expression according to the bit error rate (BER) is proposed. The results show that the optimized fixed- point specification depends on the input receiver signal-to-noise ra- tio (SNR). 1. INTRODUCTION Wireless communication is one of the most important sectors for digital signal processing (DSP) applications [12]. The low cost and low power terminal design is one of the key challenges in this do- main. New services, such as image, video and Internet access re- quire higher data rate. Consequently, the complexity of the base- band digital part is growing. Different aspects have to be considered to optimize the implementation cost and the power consumption. Especially, the arithmetic aspects offer opportunities to reduce the cost and the power consumption. Efficient embedded wireless system implementation requires the use of fixed-point arithmetic. Therefore, the vast majority of embedded DSP applications is implemented in fixed-point architec- tures [2, 3, 11]. Indeed, fixed-point architectures are cheaper and more energy efficient than floating-point architectures because of their lower data word-lengths. The fixed-point conversion process is made up of two main steps corresponding to the dynamic range estimation and the fixed- point data word-length optimization. The aim of this optimization process is to minimize the implementation cost as long as the appli- cation performances are fulfilled. To optimize the fixed-point spec- ification, existing approaches [14, 4] are based on fixed-point sim- ulations to evaluate the performances. In [10], the fixed-point error is analyzed for a CDMA receiver, but the performances in terms of bit error rate is also measured with fixed-point simulations. To evaluate accurately the bit error rate, a great number of samples are needed. Each modification of the fixed-point data requires a new fixed-point simulation. Thus these approaches suffer from a ma- jor drawback which is the long optimization time. Consequently, the fixed-point design space cannot be explored and multiple word- length approaches [5] cannot be used. In this paper the approach used to optimize the fixed-point spec- ification is presented for the case of a WCDMA receiver. The WCDMA technology is used for the physical layer of the Univer- sal Mobile Telecommunications System (UMTS), one of the third generation wireless communication systems. A new approach is proposed to estimate more accurately the data dynamic range. The properties of the application are taken into account to reduce the pessimistic effects of classical analytical approaches like interval arithmetic. Then, the accuracy constraint used in the fixed-point optimization problem is determined from the required application performances. For the bit error rate, the analytical expression of the accuracy constraint according to the bit error rate is proposed. The experiment results show the opportunity to code the fixed-point data according to the receiver signal-to-noise ratio. Our approach can be easily adapted to any communication system. The paper is organized as follows. In Section 2, the fixed-point conversion process is summarized and the WCDMA receiver is de- scribed in Section 3. The design of the symbol decoder module is detailed in Section 4. First, the dynamic range estimation is pre- sented and secondly, the fixed-point specification optimization is described. In Section 5, the design of the searcher module is pre- sented. 2. FIXED-POINT CONVERSION The fixed-point conversion can be divided into two main steps cor- responding to binary-point position determination and word-length optimization. The first step corresponds to the determination of the integer word-length of each datum. The number of bits iwl i for this integer part must allow the representation of all the values taken by the data, and is obtained from the data bound. Thus, firstly the dynamic range is evaluated for each datum. Then, these results are used to determine, for each datum, the binary-point position which minimizes the integer word-length and which avoids over- flow. Moreover, scaling operations are inserted in the application to adapt the fixed-point format of a datum to its dynamic range or to align the binary-point of the addition inputs. The second step corresponds to the determination of the frac- tional word-length. The number of bits fwl i for this fractional part defines the computational accuracy, which is usually measured by the output quantization noise power P e q . The implementation cost is minimized under the accuracy constraint P max e q . Let wl be an N-size vector representing the N application data word-lengths. Let C(wl) be the implementation cost and P e q (wl) be the computational accu- racy obtained for the word-length vector wl. The implementation cost C(wl) is minimized under the accuracy constraint P max e q : min ( C (wl)) such as P e q (wl) ≤ P max e q (1) To obtain reasonable optimization times, an analytical approach is used to evaluate the fixed-point accuracy. Moreover, in a wireless communication system, the performance is measured by the error rate. Therefore a relation between the application performance and the accuracy constraint must be done. An analytical expression of the maximal quantization noise power according to the bit error rate at the Rake receiver output is proposed. 3. PRESENTATION OF WCDMA WCDMA is the air interface of 3G mobile telecommunications net- works, which is based on DS-CDMA (Direct Spread Code Division Multiple Access) technology. In WCDMA, two layers of spreading codes are used [13]: channelization codes and scrambling codes. The channelization codes C ch are based on the Orthogonal Vari- able Spreading Factor (OVSF) technique to allow the change of spreading factor and to maintain the orthogonality between differ- ent spreading codes of different lengths. The scrambling codes used 17th European Signal Processing Conference (EUSIPCO 2009) Glasgow, Scotland, August 24-28, 2009 © EURASIP, 2009 993