— With the increasing popularity of MP3 audio, there is a need to develop cost and power efficient architectures for the MP3 encoder and decoder. This paper describes dedicated architectures for computing the modified discrete cosine transform (MDCT) and its inverse (IMDCT). Recent profiling studies have shown that these operations represent about 30% of the total MP3 computations. MP3 format defines two frame sizes that can occur in the same data stream. We have developed the most efficient algorithms for MDCT and IMDCT suitable for both sizes. Unlike previous algorithms, our computations can be unified in a single ASIC architecture. This unified architecture implemented in 90 nm TSMC library is still 25% smaller and 25% faster than any previous single frame size architectures designed in the same technology. In addition, at 128 Kbits/sec data rates, our algorithms save nearly 1800 multiplications per second (18%) which can help reduce the power consumption.
Xingdong Dai, Meghanad D. Wagh