EE141 Spring 2003 Discussion 7 - PowerPoint PPT Presentation

1 / 14
About This Presentation
Title:

EE141 Spring 2003 Discussion 7

Description:

Lab 6 reports due next week (same time and place as the project) ... methodology based on design considerations and analysis, you are in good shape. ... – PowerPoint PPT presentation

Number of Views:41
Avg rating:3.0/5.0
Slides: 15
Provided by: bwrcEecs
Category:

less

Transcript and Presenter's Notes

Title: EE141 Spring 2003 Discussion 7


1
EE141 Spring 2003Discussion 7
  • CMOS Gate Design and Circuit Optimization
  • Related Material Homework 6, Project 1

2
Class Dynamics
  • Project 1 due next week
  • No new lab or new homework next week
  • Lab 6 reports due next week (same time and place
    as the project)
  • It serves as a warm up for the project.
  • The work load there is reasonable, with part f-h
    assigned for extra credit. The extra credit can
    be turned in at the week after spring break, as
    requested by some of you. But evaluate the gain
    before you paid all the effort in.

3
Digital Circuit Design Considerations
  • Functionality
  • Design logic correctly
  • Increase noise margin and reduce noise
  • Area
  • Minimize gate / device count
  • Do compact layout
  • Speed
  • Optimizations on circuit structure, sizing and
    Vdd
  • Power
  • Minimize the area
  • Optimizations on sizing and Vdd

4
Logic Graph Euler Path
  • Logic Graph and Euler path are effective tools in
    CMOS logic design and layout implementation. (Pg.
    319)

Euler path A B C
Example of logic graph
5
Stick Diagram
Euler path A B C
6
A Previous CMOS Design HW Problem
  • Problem
  • Design the NMOS pull down network from the PMOS
    pull up network
  • There are actually multiple approaches to solve
    this problem

7
Solution 1 Logic Computation
  • Observed for PMOS PUN

8
Solution 2 Reverse the PMOS
  • Find those input patterns which block the PMOS
    paths from Vdd to F
  • AB, CD, AED, BEC

9
Solution 3 Logic Graph
10
Project 1 Clock Driver Design
  • There do exist a optimum sizing point for each
    group as the reference result.
  • But there is no standard method or equation
    leading to this result. Those are for homework
    or exam, not for project.
  • As long as your optimization result is close to
    the optimum point, and you support it with clear
    methodology based on design considerations and
    analysis, you are in good shape.
  • Actually you should be able to tell in the end
    whether you get it or not, since you should have
    understood all the trade offs in the design well
    if you are doing the right analysis.

11
Project 1 Grading Criteria
  • The correctness of your result (is it close to
    the optimum point?)
  • Your methodology behind the result (make it
    logical and CLEAR)
  • The quality of your report (present your work
    nicely)
  • Start working on the report EARLY. Dont wrap it
    up 2 hours before the project due. Thatll hurt
    your effort.

12
An Effective Approach
  • No matter how complicated the problem is, start
    from something simple, and then increase your
    understanding on the whole system, step by step.
  • Any engineering work is based on a simplified
    model of the reality. But building the model
    sometimes requires more expertise than solving
    it.
  • You cant get everything done in one strike.
    Dont try to model the whole system with one
    equation or equation group.
  • Identify those dominant factors and attack them,
    then check those others.

13
Project Delay Minimization
  • Why we want to increase the size of buffers
    geometrically?
  • tp (y) tp0 (1y/x?) tp0 (1z/y?)
  • To minimized tp by sizing the middle inverter
  • Conclusion geometric sizing minimizes buffer
    chain delay!

14
Project Energy optimization
  • Why happens with energy optimization?

0.8x2
  • Obviously the gate with the largest size deserves
    an attack
  • If its size is reduced by 20, the total energy
    reduction 20
  • Delay of this final stage increase by less than
    25. Delay of the
  • middle stage decreases. Thus the overall delay
    should increase by
  • less than 8.
  • Then how about the sizing of stages before the
    last one?
  • And what if we optimize the Vdd at the same time?
Write a Comment
User Comments (0)
About PowerShow.com