20 Pages

Lect_17

Course: ECE 497, Fall 2008
School: University of Illinois,...
Rating:
 
 
 
 
 

Word Count: 1284

Document Preview

497 ECE JS Lecture - 17 On-Chip Power Distribution Spring 2004 Jose E. Schutt-Aine Electrical & Computer Engineering University of Illinois jose@emlab.uiuc.edu ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 1 Power-Supply-Level Fluctuations Total capacitive load associated with an IC increases as minimum feature size shrinks Average current needed to charge...

Register Now

Unformatted Document Excerpt

Coursehero >> Illinois >> University of Illinois, Urbana Champaign >> ECE 497

Course Hero has millions of student submitted documents similar to the one
below including study guides, practice problems, reference materials, practice exams, textbook help and tutor support.

Course Hero has millions of student submitted documents similar to the one below including study guides, practice problems, reference materials, practice exams, textbook help and tutor support.
497 ECE JS Lecture - 17 On-Chip Power Distribution Spring 2004 Jose E. Schutt-Aine Electrical & Computer Engineering University of Illinois jose@emlab.uiuc.edu ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 1 Power-Supply-Level Fluctuations Total capacitive load associated with an IC increases as minimum feature size shrinks Average current needed to charge capacitance increases Rate of change of current (dI/dt) also increases Total chip current may change by large amounts within short periods of time Fluctuation at the power supply level due to self inductance in distribution lines ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 2 Reducing Power-Supply-Level Fluctuations Minimize dI/dt noise Decoupling capacitors Multiple power & ground pins Taylored driver turn-on characteristics Decoupling capacitors Large capacitor charges up during steady state Assumes role of power supply during current switching Leads should be small to minimize parasitic inductance Must be placed as close as possible to the chip ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 3 Effects of SSN SSN can affect circuits in 3 ways 1) SSN may increase chip-to-chip delays 2) Affects the operation of the receiving chips 3) May affect gates on the sending chip Current driven off-chip has only one return path: power and ground pins of the chip carrier to minimize effective inductance of the return path and noise, many power/ground pins must be supplied for off-chip drivers On-chip circuitry can close the loop by small inductance on chip lines ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 4 Model for On-Chip Power Distribution VDD PINS L VDD RPU C + VDD L RPD C RPD GROUND C C RPD RPU RPU GROUND PINS 1) 2) 3) 4) 5) Portion of circuits switch Vdd-VSS is reduced Non-switching devices come to rescue (through low inductance) Share charge with switching capacitors Power-level collapse is prevented ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 5 Design Criteria for SSN Inductive time constant must be much smaller than its capacitive time constant L R RC Valid for external power distribution lines that carry the current to the chip and for external lines that distribute it on the chip - Presently satisfied by on-chip lines - Board and package power lines are too high to satisfy criteria ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 6 Design Criteria for SSN Cannot distribute power on the board simply by using the printed circuit wires and connecting them to the power/ground pins of the chip directly. Off chip power distribution must employ methods that reduce the effects of line and pin inductance. To insure reliable circuit operation: dI L dt VDD Equivalence of 2 conditions is obtained by setting: dI = VDD / R dt = RC ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 7 Resonance Condition at Power-Supply Lines - Periodic nature of digital circuits can cause resonance - Large fluctuation can build up and cause circuit to fail Rchip Lmod Cchip Z(jw) f chip 1 = 2 Lmod Cchip Lmod Z ( j chip ) Rchip Cchip - fchip should be much larger than the clock frequency - Impedance should be kept small ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 8 Delta-I Noise in CMOS Circuits In a CMOS chip the portion of the circuit that is not switching (R1, C1) at a given system cycle helps the switching portion of the chip (R2, C2). L ext + VDD - R1 R2 C1 C2 C1 VDD + V = VDD C1 + C2 ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 9 Delta-I Noise in ECL Bipolar Circuits Because of diode structure of BJTs, current can only flow in one direction L Vx ext + VDD R1 R2 DC current of gates (proportional to Vx) help reduce powersupply-level fluctuations R1 R1 V = VDD For turn on V = VDD for turn off + R1 R2 R2 ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 10 Model for CMOS Power Distribution Network POWER SUPPLY Lcon 2 1 L n cap 1 R n cap n Ccap Lcon 2 Lmod 2 BOARD MODULE Lmod 2 + VDD - Cboard - n decoupling capacitors - Lcon is due to power connectors at edge of board - Cboard is intrinsic power and ground capacitance ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 11 Impedance of Power Distribution Network Influenced by Package and Bypass Cap Increase in low-frequency due to resonance frequency of board connector Increase in the high-frequency impedance due to resonance frequency of decoupling capacitor Keep both resonance frequencies away from operating frequency Zo Frequency ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 12 Off-Chip Driver SSN Calculations - Worst case on-chip delta-I noise generated at beginning of clock cycle - Main problem for on-chip drivers is lack of low-inductance return path - Off-chip drivers are the major source of SSN Problem: 32 low-impedance CMOS buffers (RS << Zo) are switched simultaneously. In addition, the line impedance is 50 , rise time is 2 nsec, output swing is 5 V, and the allowed power-supply-level fluctuation is 0.25V Solution: First, calculate the rate of change of the output voltage from the voltage swing and rise time dV 80% Vswing 80% 5V = = = 2V / n sec dt tr 2n sec ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 13 Off-Chip Driver SSN Calculations The current driven into the transmission line is I=V/Zo and its rate of change is: dI 1 dV 2V / n sec = = = 0.04 A / n sec 50 dt Z o dt Total current transient for 32 drivers: dITOT dI = N drv = 1.28 A / n sec dt dt dI Through a 1nH inductance the voltage drop is V = L = 1.28V dt To guarantee a maximum of 0.25V voltage fluctuation, the effective inductance is L= ECE 497-JS, Spring 2004 V = 0.2nH dI / dt Copyright by Jose E. Schutt-Aine , All Rights Reserved On-Chip Interconnection Resistance R Zin C l R : series resistance per unit length C : shunt capacitance per unit length Rl C Rl coth (1 + j ) 2 R Zin = Rl C (1 + j ) 2 R For very high , arg(Zin ) 45 ECE 497-JS, Spring 2004 Copyright by Jose E. Schutt-Aine , All Rights Reserved 15 RC Transmission Line R Zin C l First order approximation: 2 If << 2 then RCl Zi...

Find millions of documents on Course Hero - Study Guides, Lecture Notes, Reference Materials, Practice Exams and more. Course Hero has millions of course specific materials providing students with the best way to expand their education.

Below is a small sample set of documents:

University of Illinois, Urbana Champaign - ECE - 497
ECE 497 JS Lecture -10 Packaging TechnologiesSpring 2004 Jose E. Schutt-Aine Electrical &amp; Computer Engineering University of Illinois jose@emlab.uiuc.eduECE 497-JS, Spring 2004Copyright by Jose E. Schutt-Aine , All Rights Reserved1Radio Spe
University of Illinois, Urbana Champaign - ECE - 497
ECE 497 JS Lecture - 14 Projects: FDTD &amp; LVDSSpring 2004 Jose E. Schutt-Aine Electrical &amp; Computer Engineering University of Illinois jose@emlab.uiuc.eduECE 497-JS, Spring 2004Copyright by Jose E. Schutt-Aine , All Rights Reserved1ECE 497 J
University of Illinois, Urbana Champaign - ECE - 497
1 TIME DOMAIN REFLECTOMETRY1 Introduction The determination of the propagation parameters is essential in the design of microstrip interconnections, and reliable measurement techniques must be made available in order to verify the validity and the
University of Illinois, Urbana Champaign - ECE - 497
ECE 497 JS Lecture - 07 MulticonductorsSpring 2004 Jose E. Schutt-Aine Electrical &amp; Computer Engineering University of Illinois jose@emlab.uiuc.eduECE 497-JS, Spring 2004Copyright by Jose E. Schutt-Aine , All Rights Reserved1TELGRAPHER'S EQ
University of Illinois, Urbana Champaign - ECE - 497
ECE 497 JS Lecture - 06 Coupled LinesSpring 2004 Jose E. Schutt-Aine Electrical &amp; Computer Engineering University of Illinois jose@emlab.uiuc.eduECE 497-JS, Spring 2004Copyright by Jose E. Schutt-Aine , All Rights Reserved1Crosstalk NoiseS
University of Illinois, Urbana Champaign - ECE - 497
ECE 497 JS Lecture 02 Resistance, Capacitance, InductanceSpring 2004 Jose E. Schutt-Aine Electrical &amp; Computer Engineering University of Illinois jose@emlab.uiuc.eduECE 497-JS, Spring 2004Copyright by Jose E. Schutt-Aine , All Rights Reserved
University of Illinois, Urbana Champaign - ECE - 497
ECE 497 JS Lecture 16 Power DistributionSpring 2004 Jose E. Schutt-Aine Electrical &amp; Computer Engineering University of Illinois jose@emlab.uiuc.eduECE 497-JS, Spring 2004Copyright by Jose E. Schutt-Aine , All Rights Reserved1OverviewMotiv
University of Illinois, Urbana Champaign - ECE - 497
ECE 497 JS Lecture - 13 ProjectsSpring 2004 Jose E. Schutt-Aine Electrical &amp; Computer Engineering University of Illinois jose@emlab.uiuc.eduECE 497-JS, Spring 2004Copyright by Jose E. Schutt-Aine , All Rights Reserved1ECE 497 JS - Projects
University of Illinois, Urbana Champaign - ECE - 497
ECE 497 JS Lecture - 23 EMISpring 2004 Jose E. Schutt-Aine Electrical &amp; Computer Engineering University of Illinois jose@emlab.uiuc.eduECE 497-JS, Spring 2004Copyright by Jose E. Schutt-Aine , All Rights Reserved1Announcements- Read Chapt
University of Illinois, Urbana Champaign - ECE - 497
HOMEWORK No 3Solutionsr s t w dDrive Line 1 Sense Line Drive Line 3Terminations for both cases: near ends: 50 , far ends: 100 kA. Chip Level- the lines are 1.3 microns above the ground plane, (d=0.0013 mm) - line width=0.25 micron, (w=0.00
University of Illinois, Urbana Champaign - ECE - 497
ECE 497-JS1.HOMEWORK No 2 SolutionsUsing your knowledge of multiconductor transmission lines, determine the load network that will provide an exact match to a coupled line system in terms of the even and odd-mode impedances. Compare your solutio
Allan Hancock College - COMP - 5028
COMP5028 Object Oriented Analysis and DesignSemester 2, 2008Laboratory session OneSIT Lab 114/115 Monday August 4, 2008School of Information Technologies The University of Sydney OBJECTIVE The lab will practice the use case modeling on a web b
Allan Hancock College - COMP - 5028
COMP5028 Object Oriented Analysis and DesignSemester 2, 2008Laboratory session TWOSIT 114 Monday August 11, 2008School of Information Technologies The University of Sydney OBJECTIVES Practice the domain modeling based on problem description.
Allan Hancock College - COMP - 5028
COMP5028 Object Oriented Analysis and DesignSemester 2, 2008Laboratory session THREESIT 114 Monday August 18, 2008School of Information Technologies The University of Sydney OBJECTIVES Practice the domain modeling based on problem description.
Allan Hancock College - COMP - 5028
COMP5028 Object Oriented Analysis and DesignSemester 2, 2008Laboratory session FourSIT Lab 114 Monday August 25, 2008School of Information Technologies The University of SydneyObjective Learn to read communication diagram and map it to code
Allan Hancock College - COMP - 5028
COMP5028 Object Oriented Analysis and DesignSemester 2, 2008Laboratory Session SixSIT Lab 114 Wednesday September 15, 2008School of Information Technologies The University of SydneyDesign Pattern Practice IObjective Recognize particular d
Allan Hancock College - COMP - 5028
COMP5028 Object Oriented Analysis and DesignSemester 2, 2008Laboratory session SevenSIT 114 Monday September 22, 2008School of Information Technologies The University of SydneyObjective Understand the Open-Closed Principle Improve an initia
Allan Hancock College - COMP - 5028
COMP5028 Object Oriented Analysis and DesignSemester 2, 2008Laboratory session EightSIT Lab 114 Monday October 13, 2008School of Information Technologies The University of Sydney Objective Improve an initial design to reorganize and reduce dep
Allan Hancock College - COMP - 5028
Iteration 1: Domain Model and SSDLecture 4Wednesday Apr 6, 2005 6:00-8:00COMP5028: Object-Oriented Analysis and Design (2005 S1) Ying ZHOU, School of IT, University of Sydney1AgendaCase Requirements Elaboration process Domain modelIntrodu
University of Illinois, Urbana Champaign - MATH - 220
Math 220 AD9 Spring 2009 Worksheet 211. How could you justify using the following (often used) approximation for small x? 1 1+x1+ x 2How do the approximation and the functions actual value compare for x = 0.01, x = 0.1 and x = 1? 2. Consider the
Allan Hancock College - COMP - 5028
Granularity: the principles of package cohesionPackage design &amp; persistenceWeek 12 LectureFunctional cohesive types (classes and interfaces) are grouped togetherQuantifiable measure of cohesion of a package relational cohesion RC = number of
Allan Hancock College - COMP - 5028
Introduction OO BasicsLecture 1 Wednesday July 26, 20061AgendaAdministrative Course objective and outline OO BasicsWhat's OOAD? Functional decomposition and its problem What's UML? Software Development ProcessCOMP5028 Object-Oriented Analysi
UMass (Amherst) - BE - 640
Assignment 2. Linear Regression2.1. For this problem, please use the data on cholesterol, height, and age in the LA Heart Study in 1950. See Resources for the data. a. Construct a scatterplot for cholesterol versus age. After the scatter plot, writ
UMass (Amherst) - BE - 640
Assignment 3. Linear Regression (Please attach a copy of your program to the end of your solutions.) 3.1. a. Enter the data for weight and age into SAS and create a SAS data set. Child 1 2 3 4 5 wgt 64 71 53 67 55 age 8 10 6 11 8 b. Fit a simple line
UMass (Amherst) - BE - 640
Assignment 4. Multiple Linear Regression ApplicationsProblems (Note: Any SAS output should include program documentation.) The SEASONS study collected data on volunteer subjects over a one year time period. Information was collected at the study st
UMass (Amherst) - BE - 640
Assignment 4. Multiple Linear Regression Applications Solutions Problems (Note: Any SAS output should include program documentation.) The SEASONS study collected data on volunteer subjects over a one year time period. Information was collected at the
UMass (Amherst) - BE - 640
Assignment 5. Multiple Linear Regression Applications using Nutrition and Activity Data collected via repeated 24 hour Recalls.Problems (Note: Any SAS output should include program documentation.) This assignment investigates the importance of the
UMass (Amherst) - BE - 640
Assignment 5. Multiple Linear Regression Applications using Nutrition and Activity Data collected via repeated 24 hour Recalls. Problems (Note: Any SAS output should include program documentation.) This assignment investigates the importance of the v
UMass (Amherst) - BE - 640
Assignment 8. Mixed Models Problems (Note: Any SAS output should include program documentation.)8.1. Data were collected in the Seasons study on the number of glasses of water consumed per day (GLASS) using a 24-hour questionnaire. The data are con
UMass (Amherst) - BE - 640
Assignment 8. Mixed Models Solutions Problems (Note: Any SAS output should include program documentation.)8.1. Data were collected in the Seasons study on the number of glasses of water consumed per day (GLASS) using a 24-hour questionnaire. The da
UMass (Amherst) - BE - 640
Assignment 9 Review Programs in esb05p27.sas; esb05p28.sasProblems (Note: Any SAS output should include program documentation.)Exercise: 9.1. A sample of 500 college students participated in a study designed to evaluate the level of college studen