Algorithms and Techniques for Conquering Extreme Physical Variation in Bottom-Up Nanoscale Systems

, 2010

Category: VLSI/CAD

Overall Rating

2.3/5 (16/35 pts)

This paper does not offer a unique, actionable path for direct modern research, as the specific NanoPLA technology and its assumed variation profile are largely superseded.
However, it presents a nuanced approach: explicitly counteracting physical device variation by leveraging logical/architectural variation (fanout).
While highly specific to the NanoPLA implementation, the core principle of identifying and matching different, predictable sources of variation against unpredictable ones could be a niche, actionable gem if applicable contexts in other highly variable, non-CMOS emerging technologies can be identified.

The paper's core premise is built upon a specific type and magnitude of variation inherent in a particular speculative technology: bottom-up assembled nanowire-based programmable logic (NanoPLA) using amorphous silicon switches, aiming for feature sizes around 5nm.
The paper likely faded because the core technology it addresses – this specific NanoPLA architecture with its assumed extreme variation profile – did not gain widespread traction or prove scalable enough to justify further detailed architectural and CAD research like VMATCH.
The modeling relies on simplified Elmore delay models and assumes independent Gaussian distributions for variations, which are likely insufficient for accurately capturing complex nanoscale phenomena, spatial correlations, and non-ideal device behavior under extreme variation.
Modern approaches to variability in conventional and emerging technologies often employ more sophisticated techniques integrated earlier in the design flow... None of these approaches require the fine-grained, post-fabrication characterization and specific fanout-to-RoffFET matching central to VMATCH.

Watch