Implementation of Communication Receivers as Multi-Processor Software

Over the years, we have seen changes in the mobile communication systems starting from Advanced Mobile Phone System (AMPS) to 3G Universal Mobile Telecommunications System (UMTS) and now to 4G Long Term Evolution (LTE) advanced. Also the mobile terminals have more features to offer comparatively when it comes to supported applications for example Wireless Local Area Network (WLAN), Global-Positioning System (GPS) and high speed multimedia applications. As the mobile terminals are now evolving towards multistandard systems, the traditional approach of designing radio platforms has now been replaced by more flexible and cost-effective solutions. The challenge imposed by this multistandard approach in the implementation of mobile terminals is to integrate several radio technologies into a single device. Sharing components and processing resources between different radio technologies is the key in the implementation of multistandard terminals. Software implementation of the components is preferred because of shorter lead-time of software development and it also costs less to carry out necessary redesigns with software. In an effort to take up this challenge, the designers proposed Software Defined Radio (SDR) that allows multiple protocols to work on a System-on-Chip (SoC). The SDR implementations can follow either the Multi-Processor System-on-Chip (MPSoC) or the Coarse-Grain Reconfigurable Array (CGRA) paradigm. For this thesis work, a homogeneous MPSoC platform is used to accelerate the signal processing baseband algorithms of WCDMA and OFDM IEEE 802.11a WLAN standards. The performance comparison between single core and multi-core platforms has been made based on the number of clock cycles consumed. The idea is to exploit the inherent parallelism offered by homogeneous MPSoC platform and improve the execution times of computationally intensive algorithms like correlation operation and Fast Fourier Transform (FFT). The baseband signal processing components have been implemented in software and executed on an MPSoC platform to evaluate their performance. The multiprocessor platform has been used in an asymmetric manner in which each processing node has its own copy of application software and uses shared memory space for multiprocessor communication. Each of the processing nodes fetches and executes instructions from its own local instruction memory and is therefore independent from each other. Data Level Parallelism (DLP) has been exploited in the software implementation of the algorithms by performing identical operations simultaneously on different processors