Performance Modeling and Noise Reduction in VLSI Packaging

Similar documents
Asynchronous IC Interconnect Network Design and Implementation Using a Standard ASIC Flow

Peak Dynamic Power Estimation of FPGA-mapped Digital Designs

EECS150 - Digital Design Lecture 2 - CMOS

Innovative Fast Timing Design

TKK S ASIC-PIIRIEN SUUNNITTELU

Testing Digital Systems II

Sharif University of Technology. SoC: Introduction

Future of Analog Design and Upcoming Challenges in Nanometer CMOS

EL302 DIGITAL INTEGRATED CIRCUITS LAB #3 CMOS EDGE TRIGGERED D FLIP-FLOP. Due İLKER KALYONCU, 10043

MR Interface Analysis including Chord Signaling Options

FDTD_SPICE Analysis of EMI and SSO of LSI ICs Using a Full Chip Macro Model

Performance Driven Reliable Link Design for Network on Chips

Low Power VLSI Circuits and Systems Prof. Ajit Pal Department of Computer Science and Engineering Indian Institute of Technology, Kharagpur

Optimizing BNC PCB Footprint Designs for Digital Video Equipment

Practical De-embedding for Gigabit fixture. Ben Chia Senior Signal Integrity Consultant 5/17/2011

Prototyping an ASIC with FPGAs. By Rafey Mahmud, FAE at Synplicity.

Timing with Virtual Signal Synchronization for Circuit Performance and Netlist Security

Power-Optimal Pipelining in Deep Submicron Technology

ECEN689: Special Topics in High-Speed Links Circuits and Systems Spring 2011

Scan. This is a sample of the first 15 pages of the Scan chapter.

Data Converters and DSPs Getting Closer to Sensors

SCSI Cable Characterization Methodology and Systems from GigaTest Labs

Design of Fault Coverage Test Pattern Generator Using LFSR

VLSI Chip Design Project TSEK06

Why FPGAs? FPGA Overview. Why FPGAs?

A video signal processor for motioncompensated field-rate upconversion in consumer television

Cascadable 4-Bit Comparator

Project 6: Latches and flip-flops

L12: Reconfigurable Logic Architectures

High Performance Microprocessor Design and Automation: Overview, Challenges and Opportunities IBM Corporation

L11/12: Reconfigurable Logic Architectures

COPY RIGHT. To Secure Your Paper As Per UGC Guidelines We Are Providing A Electronic Bar Code

data and is used in digital networks and storage devices. CRC s are easy to implement in binary

SoC IC Basics. COE838: Systems on Chip Design

Static Timing Analysis for Nanometer Designs

DEDICATED TO EMBEDDED SOLUTIONS

International Journal of Scientific & Engineering Research, Volume 5, Issue 9, September ISSN

High Speed Digital Design Seminar

nmos transistor Basics of VLSI Design and Test Solution: CMOS pmos transistor CMOS Inverter First-Order DC Analysis CMOS Inverter: Transient Response

Sequencing. Lan-Da Van ( 范倫達 ), Ph. D. Department of Computer Science National Chiao Tung University Taiwan, R.O.C. Fall,

Report on 4-bit Counter design Report- 1, 2. Report on D- Flipflop. Course project for ECE533

Digital Integrated Circuits EECS 312. Review. Remember the ENIAC? IC ENIAC. Trend for one company. First microprocessor

Digital Integrated Circuits EECS 312

GaAs MMIC Double Balanced Mixer

Design Project: Designing a Viterbi Decoder (PART I)

Laboratory 1 - Introduction to Digital Electronics and Lab Equipment (Logic Analyzers, Digital Oscilloscope, and FPGA-based Labkit)

Efficient Architecture for Flexible Prescaler Using Multimodulo Prescaler

Combining Dual-Supply, Dual-Threshold and Transistor Sizing for Power Reduction

VLSI Design Digital Systems and VLSI

Latch-Based Performance Optimization for FPGAs. Xiao Teng

SYNCHRONOUS DERIVED CLOCK AND SYNTHESIS OF LOW POWER SEQUENTIAL CIRCUITS *

Metastability Analysis of Synchronizer

VHDL Design and Implementation of FPGA Based Logic Analyzer: Work in Progress

Retiming Sequential Circuits for Low Power

Optimization of Multi-Channel BCH Error Decoding for Common Cases. Russell Dill Master's Thesis Defense April 20, 2015

A low-power portable H.264/AVC decoder using elastic pipeline

CS8803: Advanced Digital Design for Embedded Hardware

UNIT III COMBINATIONAL AND SEQUENTIAL CIRCUIT DESIGN

Power Distribution and Clock Design

Designing High Performance Interposers with 3-port and 6-port S-parameters

Interframe Bus Encoding Technique for Low Power Video Compression

Low Power Digital Design using Asynchronous Logic


GaAs MMIC Double Balanced Mixer

FPGA Design with VHDL

Ultra ATA Implementation Guide

EITF35: Introduction to Structured VLSI Design

GaAs DOUBLE-BALANCED MIXER

REDUCING DYNAMIC POWER BY PULSED LATCH AND MULTIPLE PULSE GENERATOR IN CLOCKTREE

POWER OPTIMIZED CLOCK GATED ALU FOR LOW POWER PROCESSOR DESIGN

Good afternoon! My name is Swetha Mettala Gilla you can call me Swetha.

GaAs MMIC Triple Balanced Mixer

AI-1204Z-PCI. Features. 10MSPS, 12-bit Analog Input Board for PCI AI-1204Z-PCI 1. Ver.1.04

Brian Holden Kandou Bus, S.A. IEEE GE Study Group September 2, 2013 York, United Kingdom

2.6 Reset Design Strategy

GaAs MMIC Double Balanced Mixer

Design for Testability

VLSI Technology used in Auto-Scan Delay Testing Design For Bench Mark Circuits

3-Channel 8-Bit D/A Converter

FPGA Implementation of Convolutional Encoder And Hard Decision Viterbi Decoder

ECEN620: Network Theory Broadband Circuit Design Fall 2014

GaAs DOUBLE-BALANCED MIXER

Random Access Scan. Veeraraghavan Ramamurthy Dept. of Electrical and Computer Engineering Auburn University, Auburn, AL

EE178 Spring 2018 Lecture Module 5. Eric Crabill

EECS150 - Digital Design Lecture 10 - Interfacing. Recap and Topics

Digital Systems Design

A Low Power Delay Buffer Using Gated Driver Tree

EECS150 - Digital Design Lecture 18 - Circuit Timing (2) In General...

Design and Simulation of a Digital CMOS Synchronous 4-bit Up-Counter with Set and Reset

VLSI Design: 3) Explain the various MOSFET Capacitances & their significance. 4) Draw a CMOS Inverter. Explain its transfer characteristics

AbhijeetKhandale. H R Bhagyalakshmi

OF AN ADVANCED LUT METHODOLOGY BASED FIR FILTER DESIGN PROCESS

EE273 Lecture 11 Pipelined Timing Closed-Loop Timing November 2, Today s Assignment

GaAs DOUBLE-BALANCED MIXER

Clock Gating Aware Low Power ALU Design and Implementation on FPGA

Parameter Input Output Min Typ Max Diode Option (GHz) (GHz) Input drive level (dbm)

GaAs DOUBLE-BALANCED MIXER

Design of a Low Power Four-Bit Binary Counter Using Enhancement Type Mosfet

Figure.1 Clock signal II. SYSTEM ANALYSIS

Clock Domain Crossing. Presented by Abramov B. 1

Transcription:

Performance Modeling and Noise Reduction in VLSI Packaging Ph.D. Defense Brock J. LaMeres University of Colorado October 7, 2005 October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 1

Problem Statement VLSI Packaging Limits System Performance 1) Supply Bounce 2) Signal Coupling 3) Bandwidth Limitation 4) Impedance Discontinuities 5) Cost & Scaling October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 2

Agenda 1) Problem Motivation 2) Research Overview 3) Advantages Over Prior Techniques 4) Broader Impact of this Work October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 3

1) Problem Motivation October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 4

Why is packaging limiting performance? IC Design/Fabrication is Outpacing Package Technology - We re seeing exponential increase in IC transistor performance - >1.3 Billion transistors on 1 die [Fall IDF-05] October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 5

Why is packaging limiting performance? Packages Have Been Designed for Mechanical Performance - Electrical performance was not primary consideration - IC s limited electrical performance - Package performance was not the bottleneck October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 6

Why is packaging limiting performance? VLSI Performance Exceeds Package Performance - Packages optimized for mechanical reliability, but still used due to cost - IC performance far exceeds package performance On-Chip -fic > 4GHz - large signal counts - exponential scaling Package -fpkg < 2GHz - limited signal counts - linear scaling October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 7

Why is packaging limiting performance? Package Interconnect Contains Parasitic Inductance and Capacitance - Long interconnect paths Wire Bond Inductance (~2.8nH) - Large return loops - L = Φ I - Q Aε C = = V t BGA Capacitance (~300fF) October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 8

Why is packaging limiting performance? Package Parasitics Limit Performance - Excess L and C causes package noise - Noise limits how fast the package can transmit date 1) Supply Bounce 2) Signal Coupling 3) Bandwidth Limiting 4) Impedance Discontinuities October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 9

Why is packaging limiting performance? Aggressive Package Design Helps, but is expensive - 95% of VLSI design-starts are wire bonded - Goal: Extend the life of current packages QFP Wire Bond : 4.5nH $0.22 / pin BGA Wire Bond : 3.7nH $0.34 / pin *** BGA Flip-Chip : 1.2nH $0.63 / pin October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 10

2) Research Overview October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 11

Research Overview Performance Modeling & Bus Sizing - algebraic model to predict performance and cost-effectiveness Bus Expansion CODEC - encoding data to avoid patterns on bus which cause excessive noise Bus Stuttering CODEC - encoding data to avoid patterns on bus which cause excessive noise Impedance Compensation - adding C or L near package to match impedance to system October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 12

Publications: Performance Modeling and Bus Sizing - FPGA I/O When to go serial, IEE Electronic Systems and Software, 2004 - Performance Model for Inter-Chip Busses Considering Bandwidth and Cost DesignCon, 2005 Best Paper Award - Performance Model for Inter-chip Com Considering Inductive Cross-talk and Cost, ISCAS, 2005 - Performance Model for Inter-Chip Busses Considering Bandwidth and Cost, DesignConEast, 2005 - Package Performance Model for Off-chip Busses Considering Bandwidth and Cost, IEE Journal on Computers and Digital Techniques (accepted for publication) October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 13

Publications: Bus CODECs to Avoid Package Noise - Encoding-based Minimization of Inductive Cross-talk for Off-chip Data Transmission, DATE, 2005 - Controlling Inductive Cross-talk and Power in Off-chip Buses using CODECS, ASP-DAC 2006 (accepted for publication) - Bus Stuttering: An Encoding Technique to Reduce Inductive Noise in Off-Chip Data, DATE 2006 (submitted) October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 14

Publications: Impedance Compensation - Time Domain Analysis of a Printed Circuit Board Via, Microwave Journal, 2000 - The Effect of Ground Vias on Changing Signal Layers in Multi-Layered PCBs, Microwave and Optical Technology Letters, 2001 - Broadband Impedance Matching for Inductive Interconnect in VLSI Packages, ICCD, 2005 Best Paper Award - Impedance Matching Techniques for VLSI Packaging, DesignCon, 2006 (accepted for publication) October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 15

New Performance Modeling Analytical Model To Predict Bus Performance VLSI/CAD integration Quick hand calculations We Can Use Package Noise As the Failure Parameter Any noise source can be used as limit Max ( di / dt ) or ( dv / dt ) is extracted and converted to bus throughput v(t) p VDD VDD NOISE t October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 16

New Performance Modeling Bus Notation - Analysis performed on repetitive segment, reducing computation time - A scalable framework is used to represent the bus configuration C 1(pC) C 13 C 13 C 12 C 12 C 1(pC) Mutual Capacitance M 1(pL) M 13 M 13 M 12 M 12 L 11, C 0 M 1(pL) Mutual Inductance Self Inductance, Capacitance P S S G S S P S S G S S P S S G S S segment (j-1) segment (j) segment (j+1) W bus (N g, N p ) October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 17

New Performance Modeling Use Ground Bounce as Failure Mechanism P S S G S S P S S G S S P S S G S S P C Z V M p V L C p 2 L pc L11 Wbus di di 1 ( k + 1) 0 di gnd bnc = + ( 1 ( k + 1) ) + = DD N g dt k k= p dt k k= p 0.8 dt k Self Contribution Mutual Inductive Contribution Mutual Capacitive Contribution Noise Limit October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 18

New Performance Modeling Slewrate v(t) dt dv t slewrate dv di = = Z dt dt load Risetime v(t) 10% 90% (0.8) VDD VDD t t rise = (0.8) VDD slewrate October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 19

New Performance Modeling Datarate DR max = pl pc 11 bus + M1( k + 1) + N g k= p k= p ( 1.5) ( 0.8) p Z 0 L W C Z (0.8) L C 2 1( k + 1) 0 Throughput UI = (1.5)(trise) = 1/DR Tx WBUS TP = W DR max BUS max October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 20

New Performance Modeling BGA Wire-Bond Package Simulations Per-Pin Data-Rate Bus Throughput Model Error Simulation Model - Model Matches Simulations to 11% for segments greater than 1 bit - Throughput does not increase linearly as channels are added October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 21

Covered in Comps Bus Expansion CODEC Encode the Data To Avoid Noise Causing Vector Sequences - Reducing noise allows faster per-pin datarate - Throughput is increased even after considering Overhead - Bus Expansion CODEC maps on-chip bus size (m) into off-chip bus size (n) No Encoding Core Package n-bits (un-encoded) n-bits w/ Encoding Core Encoder Package n-bits (encoded) m-bits October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 22

Covered in Comps Bus Expansion CODEC - Constraints For Each Possible Noise Source on the Bus, a Constraint is written 1) v j 0 = VDD - P bnc > (L/2) (# of v ij pins =1) < Pbnc 2) v j 1 = 1 k1 (v 2j ) + k2 (v 3j ) > P 1 3) v j 1 = -1 k1 (v 2j ) + k2 (v 3j ) < P -1 4) v j 1 = 0 - P 0 < k1 (v 2j ) + k2 (v 3j ) < P 0 Each Constraint is Evaluated to Find Illegal Transitions: v 1 j = 1 = rising v 1 j v 2 j v 3 j v 1 j = 0 = static v 1 j = -1 = falling 1 0 1 1-1 0 1-1 -1 1-1 1 1 1 0 1 1-1 1 1 1 violates user-defined glitch parameter violates user-defined supply bounce parameter October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 23

Covered in Comps Bus Expansion CODEC - Algorithm The Remaining Legal Transitions Construct a Directed Graph G The Directed Graph is evaluated to see if an m-bit bus can be encoded - A closed set of nodes S must exist such that: S > 2 m each vertex s in S has at least 2 m outgoing edges to vertices s in S October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 24

New Bus Expansion CODEC Physical Results TSMC 0.13um Synthesis Results - RTL design, synthesized and mapped - Segment sizes 2 8 implemented - Logic, delay, and area evaluated October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 25

New Bus Expansion CODEC Physical Results Xilinx FPGA, 0.35um Implementation Results - RTL design implemented - Xilinx, VirtexIIPro, FPGA October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 26

New Bus Expansion CODEC Physical Results Xilinx FPGA, 0.35um Implementation Results - RTL design, implemented - Segment sizes 2 8 measured - Logic operation verified - Noise Reduced from 16% to 4% (3 bit, SPG=4:1:1) October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 27

New Bus Stuttering CODEC Intermediate States are Inserted Between Noise Causing Transitions - Stutter states limit the number of simultaneously switching signals - The source synchronous clock is gated during stutter state No Encoding Un-encoded: Core A A A B B B C C C A A A B B B C C C Package B C Vector Sequence Causes Noise Limit Violation w/ Encoding A B C A B stutter C Encoded: Core A A B B C C Encoder A B stutter C A B stutter C Package B C Vector Sequence is eliminated using Stutter October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 28

New Bus Stuttering CODEC - Algorithm Constraints are Evaluated and a Legal Directed Graph is Created G Directed Graph is Used to Map Transitions Between any Two Vectors - A transition path (which may include stutters) exists between any two vectors if: There exists at least two outgoing edges for each vector v s G (including self-edge) There exists at least two incoming edges for each vector v d G (including self-edge) October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 29

New Bus Stuttering CODEC - Construction Multiple Stutter States can be used - between 0 and 2 (Wbus-1) stutters can be inserted between any two vectors - experimental results show that for segments up to 8 bits, more than 3 stutters is rare Overhead - Overhead increases as segments sizes increase - Still useful since segments greater than 8 bits are rarely used (SPG=8:1:1) Overhead = 2 ( W bus 1) k = 1 (#_ Trans _ Re quiring _ k _ stutters) k 2 (2 W ) bus October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 30

New Bus Stuttering CODEC Physical Results Circuit Implementation - 32 pipeline stages used - pipeline reset after 32 idle states (similar to SRIO, HT, and PCI Express) - protocol inherently handles pipeline overflow October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 31

New Bus Stuttering CODEC Physical Results TSMC 0.13um Synthesis Results - RTL design, synthesized and mapped - Segment sizes 2 8 implemented - Logic, delay, and area evaluated October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 32

New Bus Stuttering CODEC Physical Results Xilinx FPGA, 0.35um Implementation Results - RTL design, implemented - Segment sizes 2 8 measured - Logic operation verified - Noise Reduced from 16% to 4% (3 bit, SPG=4:1:1) October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 33

Covered in Comps Impedance Compensation Add Capacitance Near Bond Wire to Reduce Impedance - Adding additional capacitance lowers the wire bond impedance - Impedance can be matched to system, reducing reflections Z WireBond = L C WireBond WireBond Add Capacitance to lower Z Γ= Z Z L L + Z Z 0 0 Better Impedance Match results in less reflections October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 34

Covered in Comps Impedance Compensation If the capacitance is close to the wire bond, it will alter its impedance - Electrical lengths less than 20% of risetime are treated as lumped elements - For modern dielectrics, anything within 0.15 of wire bond is lumped Treated as Lumped Element Treated as Distributed Element October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 35

Covered in Comps Static Impedance Compensator Capacitor values chosen prior to fabrication - Equal amounts of capacitance are used on-chip and on-package On-Package Capacitor On-Chip Capacitance Ccomp1 Ccomp2 Z WireBond LWB = = 50 Ω' s C + C + C WB pkg MIM October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 36

Covered in Comps Static Impedance Compensator Time Domain Analysis (TDR) Worst Case : 5mm No Static Capacitance = 19.8% w/ Static Capacitance = 4.8% 1mm 2mm 3mm 4mm 5mm October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 37

Covered in Comps Dynamic Impedance Compensator Pass Gates are used to switch in on-chip capacitors - Pass gates connect on-chip capacitance to the wire bond inductance - Pass gates have control signals which can be programmed after fabrication Z WireBond = LWB 50 ' s C + C = Ω WB Comp October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 38

Covered in Comps Dynamic Impedance Compensator Time Domain Analysis (TDR) Worst Case : 5mm No Dynamic Capacitance = 19.8% w/ Dynamic Capacitance = 6.0% 1mm 2mm 3mm 4mm 5mm October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 39

Comp Question 3) Advantages Over Prior Techniques October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 40

Performance Modeling and Bus Sizing Currently Packages are Modeled Using SPICE - Analog simulators are computationally expensive [BSIM, BPTM} - Time of simulation reduces the number of configurations to be evaluated [Agilent Ft. Collins] Model is Linear in the size of the bus - Fast computation is enabled using key assumptions - More configurations can be evaluated, which expands usefulness - Narrows hundreds of configurations into 2 or 3 for SPICE evaluation Cost is Considered - Analog simulators do not account for cost - This adds even more time to analysis October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 41

Bus CODECs to Avoid Package Noise Current Approaches Have Physical Limitations - Operate by reducing ( di / dt ) or skewing transitions [pipeline_damping, Multi-Level] - Reducing ( di / dt ) will ultimately limit performance - Skewing data increases data invalid window, will ultimately limit performance Our CODECs operate above the physical layer - Only data vectors are altered - Off-chip drivers are left unchanged, no skewing is necessary - This allows usefulness up to higher frequencies - This also allows implementation in various process and package technologies October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 42

Impedance Compensation Currently, Package Interconnect is Not Addressed - Only primary impedance is terminated (i.e., the PCB T-line) [HS_Design, MGT] - No broadband solution exists Our Techniques Target Package Directly - Impedance of wire bond or bumping can be addressed - Broadband operation suited well for digital VLSI Static Compensator - Developed using embedded construction, no cost - Simple and requires no active circuitry Dynamic Compensator - Accounts for process variation by allowing programmability after fabrication October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 43

Comp Question 4) Broader Applications of this Work October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 44

The Move Toward FPGAs 80% of Design Starts Have FPGAs Design Starts per Year FPGA Xilinx ASIC 4x 2004 1992 1996 2001 2002 97 98 Source: 99 00 Agilent 01 02 03 04 October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 45

FPGA Business Model The Move Toward FPGAs - Single design is packaged in multiple technologies - This enables multiple performance price-points - Designer cannot optimize for particular package RAM LUTs Cores Wire-Bond Flip-Chip - Performance Modeling - Noise Reduction CODECs - Impedance Compensators Cost / Performance October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 46

Power Minimization Power is Predicted to Limit Moore s Law - Large amounts of power are consumed in the off-chip drivers - CODECs can remove patterns which result in noise violations - CODECs can also remove patterns with high power consumption > 100W [ITRS] October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 47

Internet Fabric Network Congestion Slows Internet Performance - CODECs can remove patterns which result in noise violations - Can extend CODECs to remove redundant patterns in streaming A/V October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 48

Backplanes and Connectors All Interconnect Has Parasitic Inductance and Capacitance - Backplanes are popular to provide design segmentation and scalability - Connectors are present in all digital designs - Modeling, CODECs, and Compensation can be applied to backplanes/connectors October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 49

Questions? October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging 50