作者: M.-T. Sun , T.-C. Chen , A.M. Gottlieb
DOI: 10.1109/31.92893
关键词:
摘要: The implementation of a 16*16 discrete cosine transform (DCT) chip using concurrent architecture is presented. contains 32 processing elements working in parallel and random-access memory (RAM) which performs matrix transposition. structure highly regular modular, thus very efficient for VLSI implementation. was designed real-time 14.3-MHz sample video data. It an equivalent half billion multiplications accumulations per second. Fabricated 2- mu m double-metal CMOS technology, the approximately 73000 transistors occupy 7.2*7.0-mm/sup 2/ area. 68-pad die size 8.3*8.1 mm/sup 2/. fully functional first DCT chip. accuracy studies finite-wordlength are circuit design layout symbolic tool MULGA described detail. Possible variations also discussed multipurpose (variable sizes, forward-inverse transform) applications. >