A PROGRAMMABLE BASEBAND PLATFORM FOR SOFTWARE-DEFINED RADIO

Future wireless terminals will have to be multi-band, multistandard and able to execute multiple standards concurrently. In this paper we describe a flexible and programmable baseband platform for a large variety of mobile and WLAN standards. For the SDR platform architecture our primary design goal was to find the most flexible and easyto-program solution within a specified power budget. The result is an architecture consisting of a cluster of four singleinstruction multiple-data (SIMD) DSP cores each containing four processing elements and operating at 300 MHz. The cluster of SIMD cores is accompanied by dedicated processors for filtering operations, and channel encoding and decoding. The programming environment of this platform consists of an application programming interface (API), compiler and debugger, and a virtual prototype of the hardware. Profiling results for the digital signal processing software performing the PHY layer of IEEE 802.11b on the virtual prototype underline the feasibility of our approach. Figure 1: Trend of flexibility Figure 1 In the recent past, the design criteria were chosen with regard to what could be realized economically with 0.5 – 0.13μm CMOS technologies. This lead to architectures with minimal area and power consumption. Macros absorbed the compute-intensive signal-processing parts of the physical layer whereas layer-1 control processing was executed on a DSP ( ). With the advent of new standards and the shift to ubiquitous communication, continuation of this style of design would have meant to increase the number of macros to an intolerable height. Instead, the idea emerged of emulating the macros by a small number of reconfigurable data path units with adjacent small control units. This way, the firmware increased significantly, the programs could be written by the designers of the chip, only, and the customer had to be involved into the partitioning of the system into hardware and software. Furthermore, the size as well as the number of these data paths would grow with the number of standards and applications, which, in turn, would increase control-overhead and area, let alone the complexity of the programming model.