Files in this item



application/pdfMilosevic_Pavle.pdf (6MB)
(no description provided)PDF


Title:Crosstalk mitigation of high-speed interconnects using modal signaling
Author(s):Milosevic, Pavle
Director of Research:Schutt-Ainé, José E.
Doctoral Committee Member(s):Bernhard, Jennifer T.; Rosenbaum, Elyse; Shanbhag, Naresh R.
Department / Program:Electrical & Computer Eng
Discipline:Electrical & Computer Engr
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):crosstalk mitigation
modal decomposition
multiconductor transmission lines
high-speed signaling
Abstract:Today's high-speed I/O signaling links are faced with difficult challenges: due to manufacturing and technology limitations, the resources (pins and interconnects) available for off-chip signaling remain almost constant, while the throughput needed is increasing and the required aggregate bandwidths are moving into the Tb/s range. The solutions need to simultaneously satisfy the requirements for low power (typically 1-2 mW/Gb/s for short chip-to-chip links) and extremely reliable signaling (with target bit error rates as low as 1E-21 for memory links). The routing density is being increased, and faster edge rates of signals are being used, thus causing increased levels of electromagnetic coupling between interconnects. This results in various signal integrity impairments, which in turn limit the system performance and signaling rate. The focus of this research is to explore the application of modal decomposition of coupled transmission lines to crosstalk mitigation of high-speed interconnects, in particular far-end crosstalk (FEXT), which is the dominant noise source for modern single-ended memory links. Special attention is devoted to addressing the issues that arise over realistic tightly coupled cascaded channels with discontinuities in the signal path. First, we propose the application of generalized modal decomposition theory to the class of tightly coupled, nonhomogeneous and nonuniform channels with discontinuities. The proposed approach offers a robust method of extracting modal properties of the channel starting from the presumed channel geometry and structure, or from actual measured channel data, or a combination of both. Based on the results of generalized modal decomposition, optimal encoder, decoder and termination blocks for the modal signaling system are extracted from channel geometry or measurements. Due to the nonuniform structure of the communication channels, we propose the use of a frequency-dependent termination network for optimal signaling performance. We demonstrate the performance benefits over the suboptimal resistive grid termination network in terms of decision margin improvement and crosstalk-induced jitter reduction. In order to facilitate transceiver design, we explore a MIMO system perspective of modal signaling. In this context, some of the important system performance metrics are analyzed. We demonstrate the method of obtaining the modal decoder coefficients for near-optimum SNR for a given channel. We outline a methodology for determining the required number of bits of modal encoder and decoder precision given the target bit-error rate. Finally, we propose two approaches for practical system implementation of modal signaling, using (a) digital cores present in ADC/DAC based transceivers, and (b) analog frontend transceiver structure. For the analog transceiver, the design flow is demonstrated using the low-power digital CMOS process, using a case study of a typical controller-memory microstrip bus.
Issue Date:2012-02-06
Rights Information:Copyright 2011 Pavle Milosevic
Date Available in IDEALS:2012-02-06
Date Deposited:2011-12

This item appears in the following Collection(s)

Item Statistics