Confronting and exploiting operating environment uncertainty in predictive analysis of signal integrity

With device feature size continuing to decrease and 3D IC integration continuing to mature, high-density, high-speed/high-frequency IC functionality integration is making significant strides forward. The opportunities made available by such integration cannot be fully exploited without addressing effectively the reliability challenges resulting from such integration. More specifically, in addition to the reduced device reliability and increased device parameter uncertainty associated with the sub-45 nm technology nodes, the electromagnetic interference (EMI) between densely integrated components and circuits stands out as a major hurdle in robust mixed-signal IC design. While integration complexity is one of the major obstacles in the use of electromagnetic computer-aided design (EM-CAD) tools for comprehending EMI in the development of rules and guidelines for noise-aware, functional integration of such systems, the uncertainty in layout and operating conditions after integration looms as another hindrance of reliable functionality.

Alternatively, the presence of post-integration uncertainty in both topography and operating conditions of a portion of the signal/power distribution network or a specific functional block, suggests the opportunity of relaxing the accuracy with which electrical performance needs to be predicted early in the design phase. This, in turn, can be exploited to expedite predictive electrical performance assessment through the use of simplified electromagnetic models for signal integrity analysis, where the pertinent simplification is introduced in a stochastic manner.

It is the objective of this talk to elaborate on these issues through the discussion of specific, representative examples pertinent to signal integrity analysis at different stages of design. An essential element in all examples is the modeling of the uncertainty introduced in the simplification of the model in a manner such that the resulting model represents in physically-consistent manner the impact of the anticipated uncertainty in the operating environment of the structure.

Several examples will be discussed addressing predictive assessment of signal integrity at different stages of the design of the signal/power distribution network. Depending on the specific stage of the design, different types of modeling and simulation methodologies will be presented, and their validity and merit discussed and examined.

The ultimate objective of such studies is to motivate new ways of thinking in the context of noise-aware design of high-density, multifunctional ICs. For example, in the context of tools for design for manufacturability, even though process design kits provide information about process-induced variations, current generations of EM-CAD tools do not take advantage of the availability of such information. Thus, for interconnect electrical performance analysis, the ideal output of an EM-CAD tool should be error-bar plots for scattering parameters that comprehend the impact of manufacturing inaccuracies (surface roughness, topology distortion, etc.) on the electromagnetic behavior of the interconnect. Such modeling capability is not available today.

As another example, with digital, baseband and RF modules tightly integrated in SoC and SiP designs, a variety of EMI sources and coupling mechanisms need to be anticipated, quantified and accounted for to ensure reliable functionality. The complexity of EMI modeling in such a mixed-signal environment forces, more often than note, approximations in the model used that could be described in terms of geometric/material/layout variability. This variability, which is also relevant to uncertainties in floor-planning, layout and operating conditions after integration, calls for efficient methodologies and tools for predictive component and system electromagnetic performance and functionality assessment in the presence of uncertainty. While such modeling capability is not available today, it is hoped that the ideas and examples presented will motivate its consideration for further investigation and potential adoption in future generations of EM-CAD tools.