A Multistep Extrapolated S-Parameter Model for Arbitrary On-Chip Interconnect Structures - VLSI-SoC: Technologies for Systems Integration Access content directly
Conference Papers Year : 2011

A Multistep Extrapolated S-Parameter Model for Arbitrary On-Chip Interconnect Structures

Abstract

Accurate high-frequency interconnect models are needed for the precise estimation of signal delays, crosstalk, and energy losses in complex on-chip communication structures, such as hierarchical bus architectures and networks-on-chip. In this chapter we introduce a computationally-efficient wide-bandwidth characterization method based on an incremental extrapolation of S-parameters for arbitrary interconnect structures. Our method defines a systematic set of a priori parameter extractions and performs on-demand multistep extrapolations for interconnect segments with specified wire length, widths, spacings, metal layer, and neighboring routing information. Experimental evaluations show a maximum absolute error of less than 2·10− 2 (magnitude) and 7 degrees (angle) between our model and an industry-standard full-wave field simulator for a 90-nm CMOS process. We consistently enforce the passivity of the admittance matrices for each set of measured or generated parameters to eliminate the possible errors introduced during parameter measurements and extrapolation. Circuit-level simulations with the extrapolated model show a maximum signal delay error of less than 12.5% across multiple metal layers and wire configurations.
Fichier principal
Vignette du fichier
978-3-642-23120-9_9_Chapter.pdf (500.79 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-01569366 , version 1 (26-07-2017)

Licence

Attribution

Identifiers

Cite

Petru B. Bacinschi, Manfred Glesner. A Multistep Extrapolated S-Parameter Model for Arbitrary On-Chip Interconnect Structures. 17th International Conference on Very Large Scale Integration (VLSISOC), Oct 2009, Florianópolis, Brazil. pp.156-180, ⟨10.1007/978-3-642-23120-9_9⟩. ⟨hal-01569366⟩
94 View
240 Download

Altmetric

Share

Gmail Facebook X LinkedIn More