Principles of Top-Down Mixed-Signal Design
tional circuitry increases the area and, if used to monitor high speed nets, the extra load- ing can increase the power of a design and reduce its performance.
An important piece of the strategy to increase the productivity of mixed-signal designers is reuse . There are two types of reuse, IP (intellectual property) and derivatives. With IP, individual blocks are used in more than one system. With derivatives, an existing sys- tem is modified to fit the needs of a new application.
While IP reuse is promising, as yet the promise is largely unfulfilled. Part of the prob- lem is the relentless pace of process technology evolution. For various reasons, it is cur- rently difficult to migrate a mixed-signal designs from one process node to another, which limits the lifetime of IP. A limited lifetime also limits the degree one can leverage a particular piece of IP, which limits the marginal investment one can make to support reuse. Without an investment in documentation, modeling, and publishing, it is difficult for designers that are interested in incorporating existing blocks in their design to find potential IP and know whether it is suitable. The problem is heightened by the need share IP between companies, where competitive pressures force suppliers to make it easy to evaluate and integrate their IP.
Derivatives are more common, but currently require that the design team remains rela- tively intact, because much of the design knowledge that went into the system is never written down and it is difficult to pass on to a new team. This presents companies with an unpleasant choice. Generally the top design teams are the ones to take on the large new systems that will lead to several derivatives. Once the initial design is complete, should the design team continue to work on derivatives, in which case they would not be available to take on the design of new systems. Or do they move on to the new systems, in which case the derivatives will at best take considerably longer and may flounder.
Increasingly Fluid Design and Process Requirements
Between time-to-market pressures and the need to support reuse, designers are increas- ingly being faced with having to design systems with either the design requirements or the process shifting during the design process, or with the requirement that the design be easily migrated to a new application or process after completion of the initial version.
As mentioned before, mixed-signal design is increasingly focussed on communications, and communications design is dominated by industry standards. Time-to-market is par- ticularly challenging when trying to design to standards, because the design process usually must begin before the standard is finalized, meaning that the design require- ments often change late in the design process. The design team must be able to quickly react to those changes, and still produce a high quality design without the errors that would cause a respin. Thus, a design methodology is needed that is tolerant to late changes, and that allows a design to be quickly and completely reverified after such changes.
As CMOS processes shrink, secondary effects (short channel effects, mobility reduc- tion, coupling, etc.) become increasingly important. It is difficult to anticipate the effect of these on the design at the beginning of the design process. The design components need to be repeatedly simulated with estimated or extracted layout parasitics, and itera- tively refined. As a consequence, the system design often needs to be revisited. While
8 of 31
The Designer’s Guide Community