By now technologists have become familiar with a variety of semiconductors, each of which can boast some advantage over silicon-the electronics industry standby. But silicon's ascendancy over its competitors is due more to its ability to compromise rather than outperform-it is the ideal medium for hosting a variety of dopants, metals and epitaxially deposited films. In short, silicon's prime advantage has been its ability to support integrated components.
Now, as designers try to crowd more and more on a single chip, the system-on-chip generation is casting a shadow on silicon's seemingly endless ability to integrate diverse functions and materials. Part of the problem is that silicon chips have diverged somewhat into distinct families. So integration means something slightly different in each case. Digital logic circuits are manufactured in a distinct process line from DRAM chips, for example. Analog circuits are easily perturbed by rapid ly switching digital signals, making it difficult for the two types to coexist.
There are many possible limits to consider. Economically, it may be cheaper to use separate designs on different chips. The sheer complexity of large-scale, single-chip systems may make the design phase too long. A comprehensive fabrication process may not be available for mixing different technologies. Companies are increasingly asking the question of whether very large-scale integration is worth the effort.
In this week's Focus on Mixed Technologies, experts grappling with leading-edge integration problems offer some insights into the problem. One lesson is that each design project has unique aspects and it is difficult to set down general rules for diverse integration.
Engineering ingenuity is an important factor. The lesson for now is that highly integrated systems have to be approached as unique design problems. Eventually, general rules and processes for diverse integration may emerge.