VHDL implementation of a polyphase filterbank and Huffman encoder for an MP3 encoder chip

Abstract

Audio coding has increased in popularity since the advent of portable devices and the increased demand in digital audio broadcasting. One of the most popular audio coding technique developed is the MPEG-1 layer 3 (MP3) created by the Moving Picture Expert Group. It produces an audio file which has near CD quality but is smaller in size, thus reducing transmission time and storage space. Such efficient compression ratios are achieved with the fundamental roles played by the: psychoacoustic model, modified discrete cosine transform (MDCT) and Huffman coding. This final year project focuses on the hardware implementation of two MP3 encoder blocks, namely the Huffman coder and the Polyphase Filterbank. These implementations are achieved via VHSIC Hardware Description Language (VHDL) and are configured on a field programmable gate-array (FPGA). The aim of this project is to eliminate dependency on software algorithms hence alleviating the system microprocessor from processing the complex and CPU intensive tasks involved in audio coding. This final year project also highlights the importance in finding a compromise between the hardware used, speed and complexity. In fact, the implementations are designed such that they would use the minimum amount of FPGA resources possible. This comes at the cost of achieving an output in more clock cycles. However, the output must still be obtained in a specified timing constraint such that the different encoder blocks can work in parallel and thus reducing the use of storage memory.