It shouldn't be a surprise, then, if the skeleton is occasionally visible through the skin. Today, according to our data, about half of chip designers report having some involvement in analog design.

More often than not, that analog content will have something to do with clocks: either driver circuitry, PLLs or DLLs. Again, they're examples of analog in service of the digital abstraction.

But it's a good bet that tomorrow's chips will include more explicit analog content, and not all of it in support of digital circuits.

One of the drivers will certainly be the growing complexity of clocking. With multiple clock regions, the growing need for increasingly complex clock multiplication and synchronization techniques, and the growing interest in spread-spectrum techniques for clock noise control, the use of analog circuitry in clock generation and distribution can only grow. Add to that the exquisite sensitivity of deep-submicron cells to noise and jitter in their clocks, and you have not just an analog design problem, but a precision analog mess. Already, clock craft has moved from yet another annoying side issue with which digital designers must cope to an area of specialization and the basis of some very lucrative careers.

High-speed I/O is another example. The use of fast differential serial techniques such as LVDS in pad rings is growing rapidly, necessitating yet another area of analog expertise. Don't be surprised to see those techniques spread to on-chip, interblock interconnect in the future as SoC designers struggle with timing convergence and routing complexity.

There is an equal emergence of analog for analog's sake. One interesting example is mentioned in this month's Behind the News, where an RF design team has worked out a means of digitally manipulating an analog downconverter to provide a feedback loop all the way from baseband processing back into the RF stage. By willfully violating the rule to isolate the analog circuitry from digital signals, the team appears to have broken ground in RF performance.

Another example is found in our cover story. It has been widely predicted that the move to extreme-submicron processes would be the death of precision analog design. Quite the contrary: Innovative design teams are learning to exploit the abundance of transistors, even if they are mediocre transistors, to implement entirely new analog architectures.

We are coming upon a flowering of analog design not seen since the monolithic op amp first appeared. In an age when the PC no longer rules the industry, the Web is discredited as a source of boundless wealth and investors are beginning to turn greedy eyes back toward consumer applications, this reemergence of precision analog as an art form is timely. And for the lucky few with the skills to practice analog design as art rather than a survival skill, it must feel welcome indeed.

http://www.isdmag.com

© 2001 CMP Media LLC.
10/1/01, Issue # 13148, page 6.