Developments in `smart´ custom VLSI sensing for industrial automata

Visual guidance of robots for assembly tasks is the `Cinderella´ of industrial machine vision applications, lagging far behind automated visual inspection, gauging and part recognition/orientation in achieving its potential within the industrial scene. A prototype binary edge-extracting (`smart´) custom image sensor array, `Edge1´, was developed previously at the University of Brighton. It featured a 32×32 array with a unique unit cell architecture, featuring a diode photo-site, and associated processing circuitry capable of performing binary edge detection within the focal plane of the sensor. It was implemented in a 2 micron ES2 CMOS process and featured 72 I/O pins because of its total lack of on-chip support circuitry. Thus, although the total die area was only 6.65 square mm, it required an 84 pin grid array package of around 26 mm square. A number of detail changes were suggested by testing this prototype and many of these have been tested in a 1.5 micron ES2 CMOS reimplementation of this design, `Edge 2´. Implemented in 1 micron ES2 CMOS it features a resolution of 64×64, as this is considered to be the optimum compromise between useful spatial resolution, speed and simplicity of data transfer and postsensor processing