Analog VLSI Circuits for the Perception of Visual Motion

Foreword

Although we are now able to integrate many millions of transistors on a single chip, our ideas of how to use these transistors have changed very little from the time when John von Neumann first proposed the global memory access, single processor architecture for the programmable serial digital computer. That concept has dominated the last half century, and its success has been propelled by the exponential improvement of hardware fabrication methods reflected in Moore's Law. However, this progress is now reaching a barrier in which the cost and technical problems of constructing CMOS circuits at ever smaller feature sizes is becoming prohibitive. In future, instead of taking gains from transistor count, the hardware industry will explore how to use the existing counts more effectively by the interaction of multiple general and specialist processors. In this way, the computer industry is likely to move toward understanding and implementing more brain-like architectures.

Carver Mead, of Caltech, was one of the pioneers who recognized the inevitability of this trend. In the 1980s he and his collaborators began to explore how integrated hybrid analog–digital CMOS circuits could be used to emulate brain-style processing. It has been a hard journey. Analog computing is difficult because the physics of the material used to construct the machine plays an important role in the solution of the problem. For example, it is difficult to control the physical properties of sub-micron-sized ...