New Concepts of Asynchronous Circuits Worst-case Delay and Yield Estimation

Although the benefits of asynchronous design style are undeniable, this style is still a road that designers rather avoid. There are, however, serious advantages of this digital design concept that are making it favourable for many applications. Asynchronous circuits need no clock generation and distribution (Sparso, 2006; Martin & Nystrom 2006), which leaves the problems related to clock skew behind and saves a lot of chip area. Asynchronous circuits are characterized with much easier technology migration and good modularity. Very low EMI occurs during operation, while achieving high noise immunity (Lewis & Brackenbury 2001). Power is consumed only when useful work is done. The absence of the clock itself reduces the power consumption. These issues are very important while designing portable systems where battery size and lifetime are important. Synchronous circuit design styles have enormous commercial practice and very significant pedigree, and those are the major reasons for the lack of motivation to apply asynchronous circuit techniques (Davis & Nowick, 1997). Nevertheless, the motivation to pursue the study of asynchronous circuits is based on the simple fact that all high-performance ‘‘synchronous” design styles are ‘‘asynchronous in the small” (Cortadella et al. 1999) Because of that, some techniques for desynchronization of synchronous circuits have appeared lately (Cortadella et al. 2006; Andrikos 2007). Beside their benefits, some problems related to asynchronous circuit design are still waiting to be solved. One of the most important is the estimation of asynchronous circuit performances. That is, determining the delays of the paths in a particular asynchronous circuit. Early evaluation of the path delays in the circuit helps avoiding early timing problems as well as circuit performance characterization (Sokolovic, Litovski & Zwolinski 2009). Precise paths delays, of course, can be estimated only in the final steps of the design process. That is because the delay is extracted from the circuit after layout synthesis. If the delays do not satisfy the required speed of the circuit, the circuit has to be redesigned. The same conclusion stands when timing problems occur. This strongly suggests that new methods are to be offered enabling delay estimation to be performed during the early phases of digital system design. Our aim here is to establish the application of a standard logic simulator in asynchronous circuit path delay analysis and parametric yield estimation. The simplest way to determine the circuit delay is simulation. At the transistor level complex circuits’ simulation becomes inefficient. To verify the logic function and the timing