EECS 427 Lecture 5: Logical Effort. Reminders

Similar documents
EE241 - Spring 2003 Advanced Digital Integrated Circuits

Lecture 12: HEMT AC Properties

2/20/2013. EE 101 Midterm 2 Review

PHYS 1443 Section 001 Lecture #4

Chapter 6: AC Circuits

Advanced time-series analysis (University of Lund, Economic History Department)

Outline. Energy-Efficient Target Coverage in Wireless Sensor Networks. Sensor Node. Introduction. Characteristics of WSN

Normal Random Variable and its discriminant functions

Chapter 6 MOSFET in the On-state

EECS 141: FALL 00 MIDTERM 2

Lecture 7: Multistage Logic Networks. Best Number of Stages

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4

EP2200 Queuing theory and teletraffic systems. 3rd lecture Markov chains Birth-death process - Poisson process. Viktoria Fodor KTH EES

Estimating Delays. Gate Delay Model. Gate Delay. Effort Delay. Computing Logical Effort. Logical Effort

Logic effort and gate sizing

First-order piecewise-linear dynamic circuits

Lecture 11 SVM cont

FTCS Solution to the Heat Equation

Bandlimited channel. Intersymbol interference (ISI) This non-ideal communication channel is also called dispersive channel

Mechanics Physics 151

グラフィカルモデルによる推論 確率伝搬法 (2) Kenji Fukumizu The Institute of Statistical Mathematics 計算推論科学概論 II (2010 年度, 後期 )

Chapters 2 Kinematics. Position, Distance, Displacement

Mechanics Physics 151

More Digital Logic. t p output. Low-to-high and high-to-low transitions could have different t p. V in (t)

Displacement, Velocity, and Acceleration. (WHERE and WHEN?)

ELECTRONIC DEVICES MOSFET METAL-OXIDE-SEMICONDUCTOR FIELD EFFECT TRANSISTOR. Piotr Dziurdzia, Ph.D

12d Model. Civil and Surveying Software. Drainage Analysis Module Detention/Retention Basins. Owen Thornton BE (Mech), 12d Model Programmer

CHAPTER 16 ADVANCED MOS AND BIPOLAR LOGIC CIRCUITS

E c. i f. (c) The built-in potential between the collector and emitter is. 18 ae bicb

EE 560 MOS INVERTERS: DYNAMIC CHARACTERISTICS. Kenneth R. Laker, University of Pennsylvania

Lecture 18: The Laplace Transform (See Sections and 14.7 in Boas)

Physical Limitations of Logic Gates Week 10a

10. A.C CIRCUITS. Theoretically current grows to maximum value after infinite time. But practically it grows to maximum after 5τ. Decay of current :

EE 330 Lecture 41. Digital Circuits. Propagation Delay With Multiple Levels of Logic Overdrive

Different kind of oscillation

(,,, ) (,,, ). In addition, there are three other consumers, -2, -1, and 0. Consumer -2 has the utility function

Density Matrix Description of NMR BCMB/CHEM 8190

Density Matrix Description of NMR BCMB/CHEM 8190

Lecture 2 M/G/1 queues. M/G/1-queue

Scattering at an Interface: Oblique Incidence

Notes on the stability of dynamic systems and the use of Eigen Values.

DEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL

EE 466/586 VLSI Design. Partha Pande School of EECS Washington State University

Ordinary Differential Equations in Neuroscience with Matlab examples. Aim 1- Gain understanding of how to set up and solve ODE s

Endogeneity. Is the term given to the situation when one or more of the regressors in the model are correlated with the error term such that

EE141. Administrative Stuff

CHAPTER 10: LINEAR DISCRIMINATION

Chapter Lagrangian Interpolation

Diode rectifier with capacitive DC link

CS286.2 Lecture 14: Quantum de Finetti Theorems II

P-MOS Device and CMOS Inverters

EE 330 Lecture 40. Digital Circuits. Propagation Delay With Multiple Levels of Logic Overdrive

Energy Storage Devices

Lecture 9: Dynamic Properties

EE 247B/ME 218: Introduction to MEMS Design Lecture 27m2: Gyros, Noise & MDS CTN 5/1/14. Copyright 2014 Regents of the University of California

NPTEL Project. Econometric Modelling. Module23: Granger Causality Test. Lecture35: Granger Causality Test. Vinod Gupta School of Management

The Finite Element Method for the Analysis of Non-Linear and Dynamic Systems

Solution in semi infinite diffusion couples (error function analysis)

Response of MDOF systems

Foundations of State Estimation Part II

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany

FI 3103 Quantum Physics

VLSI Design I; A. Milenkovic 1

J i-1 i. J i i+1. Numerical integration of the diffusion equation (I) Finite difference method. Spatial Discretization. Internal nodes.

Including the ordinary differential of distance with time as velocity makes a system of ordinary differential equations.

EE 330 Lecture 41. Digital Circuits. Propagation Delay With Multiple Levels of Logic Optimally driving large capacitive loads

II. Light is a Ray (Geometrical Optics)

Application Notes for AP3772 System Solution

On One Analytic Method of. Constructing Program Controls

2015 Sectional Physics Exam Solution Set

Lecture 11 VTCs and Delay. No lab today, Mon., Tues. Labs restart next week. Midterm #1 Tues. Oct. 7 th, 6:30-8:00pm in 105 Northgate

WiH Wei He

Hidden Markov Models Following a lecture by Andrew W. Moore Carnegie Mellon University

Digital Integrated Circuits

Chapter 3: Maximum-Likelihood & Bayesian Parameter Estimation (part 1)

SHOALING OF NONLINEAR INTERNAL WAVES ON A UNIFORMLY SLOPING BEACH

Outline. Chapter 2: DC & Transient Response. Introduction to CMOS VLSI. DC Response. Transient Response Delay Estimation

Motion in Two Dimensions

10/10/2011. Signals and Systems EE235. Today s menu. Chicken

Objectives. Image R 1. Segmentation. Objects. Pixels R N. i 1 i Fall LIST 2

Area Minimization of Power Distribution Network Using Efficient Nonlinear. Programming Techniques *

Variants of Pegasos. December 11, 2009

Vector autoregression VAR. Case 1

NATIONAL UNIVERSITY OF SINGAPORE PC5202 ADVANCED STATISTICAL MECHANICS. (Semester II: AY ) Time Allowed: 2 Hours

Quantum Mechanics II Lecture 11 Time-dependent perturbation theory. Time-dependent perturbation theory (degenerate or non-degenerate starting state)

CptS 570 Machine Learning School of EECS Washington State University. CptS Machine Learning 1

VLSI Design Issues. ECE 410, Prof. F. Salem/Prof. A. Mason notes update

Cairo University Faculty of Engineering Chemical Engineering Department

Slides: CMOS Basics.

COMPUTER SCIENCE 349A SAMPLE EXAM QUESTIONS WITH SOLUTIONS PARTS 1, 2

R th is the Thevenin equivalent at the capacitor terminals.

Logical Effort: Designing for Speed on the Back of an Envelope David Harris Harvey Mudd College Claremont, CA

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!") i+1,q - [(!

( ) [ ] MAP Decision Rule

Digital Integrated Circuits

Wp/Lmin. Wn/Lmin 2.5V

( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model

Chapter 5. Circuit Theorems

A Deterministic Algorithm for Summarizing Asynchronous Streams over a Sliding Window

Embedded Systems 5. Midterm, Thursday December 18, 2008, Final, Thursday February 12, 2009, 16-19

Transcription:

9//009 EES 47 Lecure 5: Local Eor Readn: handou Remnders Semnar announcemen: Dr. Mchael Mcorquodale, TO and ounder, Mobus Mcrosysems Toc: Srah Down he rooked Pah The Dynamc Process o ommercalzn Research Frday 9/5, :0-:0 m, 00 EES AD wll be done n eams. AD s due nex Wednesday Sar early Try no o sk lecure Lookn ahead: HW Projec roosal due Wednesday 0/7, weeks away Quz Wednesday 0/4, weeks away

9//009 Inro o local eor : Local eor : Elecrcal eor : Parasc delay Las Tme Delay comonens revew Normalzed Delay 5 4 Inverer: = ; = -nu NAND: = 4/; = Eor Delay 0( ) Inrnsc Delay 4 5 Fanou I. Suherland, e al, Local Eor, Academc Press, 999

9//009 More comlex crcu Every ae s szed accordn o a : mn-szed nverer nand 4 nand 4 ae, nor 4 4 0 ( nand 4 ae, nv ae, nand 4 nand 4 ae, nand 4 ) 0 0 Mulsae eor Pah eecve an-ou F F F

9//009 4 Mnmum delay Toal delay Subsue F F Subsue F Mnmze he delay Solve or he mnmum delay 0 mn F 0 Solve or he mnmum delay Delay s omal when sae eors are equal h h 0 Gae eor Branchn h h H ) ( ) ( Dene ah eor H GF G ) ( ) ( G: ah local eor F: ah eecve an-ou ah on ah o ah on b b ah o ah on Branchn eor

9//009 Equvalen Pah Eors F B FB G ou n b H Pah Eor GFB Pah Eor: Does no chane wh added nverers Does no deend on szes, bu on ooloy Mnmum Delay Sae eor o sae : Omal sae eor: For N saes: Mnmum Delay: H GFB h hˆ h H hˆ N 0 0 ( hˆ ( H N NH ) / N ) 5

9//009 Examle o comue mn. delay Un szed nand ae G a F L n 4 0 B H GFB 47.4 hˆ H.6 0 b 6.9 c ( 0 nand hˆ).7 Sae szn Aer comun ĥ hˆ n hˆ ou ou n Work backwards o sze each ae n, c nand ˆ h L 6

9//009 Sae szn examle nand L n, c hˆ 4 0 un _ nand n, c. 68.6 un _ nand Sae szn Aer comun ĥ hˆ ou h ˆ n ou n an also work orward o sze each ae h ˆ.6 n, a un _ nand n, b n, b. 6 un _ nand nand 4 / 7

9//009 alculan omal # o saes Pah eor H can be used o deermne he omal number o saes Assumn we add n nverers New number o saes N=n +n G, F, B don chane H s xed Bu nrnsc delay ncreases ( n n ) ( n n) F n n Omum s echnoloy deenden nv Summary o he ermnoloy Gae level Parasc delay Local eor Elecrcal eor Sae eor Sae delay h n ou h Pah elecrcal eor Pah local eor Branch eor Branchn acor Pah F G B ou ah n ah b b BF on ah on ah o ah 8

9//009 Mehod Pah eor Omal sae eor Omal ah delay Sae szn H GFB N hˆ H ou, NH n, N hˆ n,. omue ah eor. omue omal sae eor. Add buers (deermne omal number o saes) 4. omue an-ou o each sae 5. Sze ndvdual aes (workn backward or orward) Mehod examle / b = because s he same ye o ae 5 5 5 G 9 B F 60 6 H GFB. Nˆ 5 5 hˆ.. 0 (5. 6 ) 4 0 hoose # o saes such ha sae eor s close o 4. Add nverers o he exsn saes 9

9//009 Mehod examle: Szn / ˆ. h ˆ.9 h.. 9 4. 5. 5/ / n, e n, d n, c n, b n, a ou, e 5 60 8.75. 8.75 5.86. 5.86.05.9.05.9..9 0.989.9 Branchn Always check hs Mehod examle: Gae szn / e) (/)x(8.75) (/)x(8.75)(8 c) a) 4/5x.05 4/5x.05 /5x /5x /5x /5x /5x /5x.05 /5x.05 /5x 0

9//009 Wron number o saes 5 4 Hher number o saes han omal s less bad D^(N) D^(N^ ).5.5.6.6 0 0.5 0.5 4 8 N N^ Wron ae sze.6.5 D(s) D().4....0...044.044 Penaly s he same n any szn drecon Snle ae s szed wron 0 0.5 0.5 0.67.5 4 s

9//009 0 or our 0nm echnoloy Inverer Delays Tau's 70 60 0 Delay (s) 50 40 0 0 Nom nal Fasas Slow -slow Tau (s) 8 6 4 Nom nal Fasas Slow -slow 0 0 0 4 5 Fanou 0 D()-D(0) D()-D() D()-D() D(4)-D() D(5)-D(4) Tau and Inrnsc Delay (Pnv) or 0nm Tech P_nv P_nv 4.5 4.5.5.5 05 0.5 0 0/60 560/80 0/560 Nom nal Fasas Slowslow Averae Tau value Averae Pnv W/Wn No Fas slow m 0/60 6.6 5.74 6.98 560/80 6. 5.66 7.4 0/560 6.7 5.55 7.4 0/60.46.64 4.4 560/80 7.7 86.86 75.75 0/560.4.77.9 W/Wn

9//009 Bes number o saes nv.8 Pah eor H Bes number o Mn. delay saes Sae eor h 0.8 0-9.57 957 9.57.9.09-7.8 54.4.5.79-6.65 94 0. 4 4.4-6.9 56 8.7 5 4.5-6.07 846 47. 6 4.49-5.9 47 55.8 7 IBM 0.um haracerscs har / Un dsance NMOS j PMOS j NMOS PMOS NMOS R eq PMOS R eq Parameer.06F/um (W).0F/um (W).57F/um (W).44F/um (W) 4.4 k/um (W) 86.95 k/um (W) * All values obaned rom devce smulaon

9//009 Devce Smulaons I DB I G Juncon aacance Gae aacance I d Q jv I d Q V Devce Smulaons R ull eq R eq V I DD ull D ull hal Req Req R hal eq V I DD hal D 4

9//009 IBM 0.um haracerscs har / Un dsance Poly Ressance M Ressance Parameer 58./um (l) 0.44/um (l) M-M6 Ressance 0./um (l) Uer layer Meals 0.09/um (l) Plae Above, Below Isolaed Poly aacance 0.69 F/um 0.9 F/um M aacance 0.69 F/um 0.565 F/um M-M6 aacance 0.4 F/um 0.59 F/um Uer layer Meals 0.970 F/um 0.50 F/um * All values obaned rom desn manual IBM 0nm haracerscs MOS Tye Parameer alculaed Smulaed % Error NMOS.0F/um.57F/um.7% PMOS.94F/um.44F/um 5.8% NMOS j.09f/um.06f/um.% PMOS j.44f/um.0f/um 7.% NMOS (FO4) 4.8s 5.5s % PMOS (FO4) 86.s 68.s 6% * alculaed = 0.69 * R eq * L L = 4,P + 4,N + j,n + j,p * Smulaed = 50% - 50% ranson 5

9//009 Why use P/N =? P/N rao Nose marns are balanced Equal sloes How abou P/N =.5? Lmaons Inernal caacance aacance n nernal nodes Body eec 6

9//009 Lmaons - Taern Transsor szes n sack are deren The laes arrvn nu should be closes o ouu node Wha sze s he bes choce? ca n nernal node Lmaons Branchn Assume ha he sze o he o-ah ae racks he sze o he ae on-ah o-ah h o h on on-ah Szn one crcal ah o a branch may make he oher ahs worse 7

9//009 Lmaons Seres Devces Models seres devces o sze o be equvalen n drve srenh o a snle devce o sze No very accurae Soluon smulae aes o drecly nd local eors Lmaons Sloe & Inerconnec Inores mac o nu sloe on sae delay Inerconnec caacances are sde loads. They do no necessary scale wh aes and canno be lumed n he local eor ormulaon 8

9//009 Lmaons - Scaln Scaln s no lnear wh wdh k S R 0. 69 c k S R R R k S k S R c k v k Permeer Narrow wdh eecs Process varaons Parasc delay o mul-nu aes usually much less han smle model redcs Duson sharn, nu deendences Szn ool Tool: TILOS [Dunlo 89] Sar wh all ranssors o mn. sze Fnd crcal ah (Omze ah) omue delays Increase sze o ranssors n crcal ah Sze ah wh bes sensvy Reea Goal o ah dsrbuon All ahs equal n lenh 9

9//009 Area - Delay Bu he mac o rocess varaons can be worse or he omzed ahs more on hs laer n he course Summary Local eor s useul or hnkn o delay n crcus NANDs are aser han NORs n MOS Pahs are ases when eor delays are ~4 Pah delay s weakly sensve o saes, szes Bu usn ewer saes doesn mean aser ahs Delay o ah s abou lo 4 F FO4 nverer delays Inverers and NAND bes or drvn lare cas Provdes lanuae or dscussn as crcus Bu requres racce o maser 0