Leading IC design companies have identified that better analog design tools are desperately needed to improve the predictability of the design cycle and to reduce the true cost of this cycle. Today, these companies are alleviating their design and re-spin problems by turning to analog circuit optimization, which automatically sizes and biases circuit devices for optimal performance. But what role should analog circuit optimizers play in the AMS design process? Should they complement what designers do, or replace them entirely?

Historically, research teams have tried to build optimization tools to replace designers. While these "push button" replacement optimizers achieved interesting academic results, their usefulness in real production environments was questionable. Companies complained that these early optimizers removed the designer from the design process, did nothing to increase the designer's knowledge, and simply did not work on reasonably complicated commercial designs.

Early optimizers failed because they tried to do too much and didn't solve the right problem. They didn't allow the user to input any design knowledge, so only the simplest designs could be optimized. Trying to optimize a reasonably complicated design was like trying to boil the ocean.

Analog design is based as much on what is possible as what is needed. Optimizers aimed at replacing designers did not highlight all the possible trade-offs, selecting instead the designs that came closest to meeting all specifications. By failing to allow designers to choose, these optimizers settled for mediocrity and failed to deliver the best circuit solutions for the applications.

For analog circuit optimization to be useful in real production environments, it needs to complement designers, increase their knowledge, and show them the possible circuit solutions, rather than try to replace designers.

At the start of a new design or migration to a new manufacturing process, an analog circuit's full performance potential across all the speed, power, noise, and area tradeoff options is largely unknown. For instance, the circuit solution with the lowest power may not be the fastest nor have the lowest area and noise. A design decision in this case should be based on understanding how much speed may be lost to achieve a reasonable reduction in power.

Replacement optimizers generate one "optimum" circuit. But which optimum circuit will the designer get? The circuit with less area and more noise, or the larger circuit with less noise? Since the replacement optimizer cannot read the designer's mind and make tradeoff decisions, it cannot find the best circuit for the application. In other words, replacement optimizers frequently return adequate but less than "optimum" circuits.

Optimizers that are complementary to the design process and allow designers to control automation are key to creating the best circuits for an application. Complementary optimizers highlight the tradeoffs among various circuits and let designers direct the optimization.

If an optimizer doesn't replace designers, why go to the expense of using optimization at all? Complementary optimizers facilitate a rigorous and re-usable design methodology. Optimized circuits are tested over all manufacturing and environmental corner conditions. The increased design robustness created by this thorough testing reduces the true cost of IC design, often eliminating costly re-spins and decreasing time to market.

Complementary optimization reduces design time, especially when a design specification or the manufacturing process changes. The design setup is reused and the circuit is simply and interactively re-optimized to meet the modified requirements. Complementary optimization removes the iterative, time-consuming task of manually re-sizing devices and re-simulating for every corner condition.

Complementary optimization makes intelligent use of the simulation farm when the designer cannot be present and often creates circuit solutions that exceed performance specifications. Complementary optimization allows IC design companies to better deploy their scarce analog design resources.

With the right complementary optimization tools, designers can graphically explore the tradeoffs between design alternatives. These tradeoff exploration tools allow design teams to clearly communicate design alternatives and provide design managers with greater visibility into the design process. Design managers can better predict the time and cost of the AMS design cycle since robust designs can be created, without a re-spin, using a rigorous and visible design methodology. Using optimization to complement design effort gives IC design companies the benefits of optimization and gives designers the control and knowledge they need to produce quality AMS ICs.

Christopher Labrecque is product marketing manager for the Circuit Explorer product at Synopsys Inc.