VLSI Digital Signal Processing Systems: Design and Implementation

1.3    DSP APPLICATION DEMANDS AND SCALED CMOS TECHNOLOGIES

The field of VLSI DSP continues to be driven by the increasing demand of DSP applications, particularly in multimedia and wireless communications, and by advances in scaled CMOS technologies. This section presents the computation requirements needed in digital video processing applications and in scaled CMOS VLSI technologies. It is shown that video processing systems require computation rates in the range of 10–100 GOPs/sec (gigaoperations per second) and, therefore, design of these systems is quite challenging!

A color digital image consists of picture elements (pixels), which are represented using three primary color elements, red (R), green (G) and blue (B). The RGB representation is converted to YUV representation based on human visual systems, where Y stands for the luminance information and U and V are the color differences between Y and blue, Y and red, respectively, and are called chrominances. A full-sampling of YUV is represented as 4 : 4 : 4 sampling and the resulting pixel is represented using 24 bits, 8 bits for each variable. With 4:4:4 sampling, a CIF (common intermediate format) frame with a frame size of 288 × 352 pixels and a frame rate of 30 frame/sec requires storage of size 2.433 megabits (Mb) and the video source data rate is 72.99 Mb/sec for a single frame. For high-definition TV (HDTV) video with a frame size of 1920 × 1250 pixels and a frame rate of 50 frame/sec, one frame requires storage size ...