Since its release, the MP3 is the most efficient way to store sound data. The primary application areas of this format are portable audio players and mobile phones, but it is used nearly everywhere where it is needed to play digital sound (On-board computers, public address systems, etc.).
By portable devices the power consumption is a very important matter. That is why most of the data processing is done by hardware components made especially for this purpose. In other realizations it is also an alternative path to use s DSP-s (Digital Signal Processor) or general purpose processors.
In the case of on an FPGA based “System on a Programmable Chip” realization, it is an expected thing to solve the audio decoding task inside one chip, without using any dedicated external hardware units. The aim of my thesis is to analyze a microprocessor system implemented in a FPGA and decide whether it is able to accomplish real-time decoding or not. If not I also have to determine the degree of hardware support, which is needed to fulfil the real-time condition.
In the first part of my thesis, I will examine a general purpose Xilinx MicroBlaze processor running a decoder software in respect of runtime and operating speed. The open source decoder I’m using is called MAD and operates with fix point variables. After that I will search for the bottle-neck of the software and make a proposal for an FPGA implemented hardware design that will substitute the specified part. Finally I will present the achieved speed growth and will make a proposal for further developments that may be done in the future.