EE123 - Digital Signal Processing

Fall 2003


[Course Details] [Course Handouts] [Homework] [Announcements] [Lecture Notes]


Lectures

Wednesdays and Fridays 9:30 - 11:00 am, 203 McLaughlin Hall


Lecturer

Professor A. Zakhor
avz@eecs.berkeley.edu
507 Cory Hall
ext. 3-6777

Prof.  office hours : Fridays 11:00 am - 12:00 pm.


Teaching Assistant

Dan Hazen
hazen@eecs.berkeley.edu

Discussion Section 1: Tuesday 2:00 - 3:00 pm, 293 Cory Hall
Discussion Section 2: Wednesday 4:00 - 5:00 pm, 293 Cory Hall
Office Hours: Tuesday 4:00 - 5:00 pm, Thursday 3:00 - 5:00 pm, 469 Cory Hall


Course Administrative Assistant

Rosita Alvarez
195M Cory Hall
ext. 3-6633


Course Handouts

Handouts and graded problems sets not picked up during lectures can be picked up in 195M Cory from Rosita Alvarez.


Prerequisites

EE 120, graduate standing, or consent of the instructor.


Text book

A. V. Oppenheim and R. W. Schafer, Discrete Time Signal Processing, Second Edition, Prentice Hall, 1999. (required)

 


Outline of Topics

   

1.

Fast review of LTI systems, DTFT, sampling

2.

Multirate signal processing, Bilateral Z Transform

3.

Discrete Fourier Transform, Fast Fourier Transform

4.

Quantization, finite word length effects

5.

FIR and IIR filter design techniques

6.

Filter banks, Wavelets

7.

Multidimensional signal processing: Z transform, filter design, reconstruction from phase, magnitude, DFT

8.

Applications: speech and video processing


Homework

1. Problem sets will be issued approximately once a week. They will contain Matlab problems. Problem sets are due in class, either on Wednesday or Friday, at the beginning of class.

 

2. Problem sets handed in late will not be accepted unless consent is obtained from the teaching staff prior to the due date.

 

HW #1: 2.42, 2.49, 2.56, 2.58, 2.60, 2.67 due Friday, September 5th at the beginning of class.

HW #2: 3.28, 3.34, 3.37, 3.39, 3.40, 3.43 due Friday, September 12th at the beginning of class.

HW #3: 4.29, 4.30, 4.34, 4.36, 4.37, 4.38 due Friday, September 26th at the beginning of class.
Note
: You have 2 weeks to complete this assignment.

HW #4: 6.23, 6.26, 6.27, 6.28, 6.29, 6.32 due Friday, October 3rd at the beginning of class.

Lab 1 handed out in class. Note: for anyone who did not receive a copy of this homework, you need to see Rosita Alvarez in 195M. NOTE: IF YOU DO NOT HAVE ACCESS TO MATLAB, YOU SHOULD MAKE SURE THAT YOU CAN ACCESS MATLAB USING THE LAB MACHINES IN 199 CORY ASAP!!!

FORMAT FOR WRITEUP: Please include all answers to any questions, all figures required, and all code used (preferably in an appendix).  Make sure plots in the frequency domain go from –π to π, not 0 to 2π.  Hand in your own lab report representing your work.  DO NOT COPY OTHER PEOPLE’S CODE AND/OR ANSWERS.

HW #5: See sheet. Note: You will need the file prob3.mat

HW #6: 7.34, 7.35, 7.36, 7.37, 7.38 due Friday, October 31st at the beginning of class. NOTE: Problem 7.37 has a bunch of typos.  Please do the question as follows: H(ejw)=A(ejw)e-jw(N-1)/2.  Then, for part b, change hd[n] to h[n].  Then, for part c, change hd[n] to h[n]. 

Lab 2: Handed out in class 10/31/03.  Notes for lab 2: For question #1, “ripples” means how many local maxima/minima are in the signal (do NOT count the values at the passband frequency wp and the stopband frequency ws because they do not contribute to what is defined as “ripple”).  For question #2, part b should read: What is the discrete-time frequency corresponding to the Ωc given (forget about the “prewarped analog band edge”, this makes no sense).  For part c, the prewarped band edge is the Ωc given. For question 3, assume fp = 0.25 and fs = 0.3.  For question 4, note that in part c, the definitions of fs and fp have been switched.

HW #7: 8.25, 8.26, 8.27, 8.29, 8.31, 8.32, 8.33

HW #8: 8.43, 8.44, 8.47, 8.50, 8.59, 8.60, 8.63

HW #9: 9.23, 9.24, 9.27, 9.30, 9.33 due Monday December 1st by 11:00 AM in Rosita Alvarez-Croft’s office (195M Cory Hall).  Note: no exceptions will be made to this submission time.  Also, you are of course welcome to hand the homework in early.

Note: There were also a couple of errata for previous homework solutions (Please see Rosita in 195M for the homework solutions/errata.)


Grading

Homeworks : 20%

 

In class midterms (Tentative dates in parenthesis):
Midterm 1 (10-22-03) : 40%
Midterm 2 (12-5-03) : 40%

 

Any questions about graded homework should go to the reader for the course, Thinh Nguyen. His email is thinhq@eecs.berkeley.edu.  Any questions about graded labs should go to the TA for the course, Dan Hazen (see above for email address).


Announcements (bold denotes new/valid announcements)

1. Discussion Sections for Tuesday 08-26-03 and Wednesday 08-27-03 will be cancelled.

2. Instructions for logging onto the UNIX machines for accessing Matlab can be found here.

3. TA Office Hours for the week of 09-01-03 through 09-05-03 will be held in 199 Cory. On Thursday,

09-04-03, office hours may move to 469 Cory, so check there if Dan cannot be found.

4. Makeup lecture scheduled for Monday (09-08-03) from 9:30-11 AM in the usual room.

5. Professor Zakhor’s office hours will move from Friday to Wednesday, 11 AM - 12 PM this week.

6. There will be no lecture Wednesday, September 17th or Friday, September 19th

7. Makeup lecture scheduled for Monday, 09-29-03 from 9:30-11 AM in the usual room.

8. Both Monday makeup lecture videos have been posted. See the link below.

9. Errata for previous homework assignments have been posted.

10. Midterm #1 on Wednesday, October 22nd in class

11. Dan’s Office Hours for the week of 10/6-10/10 will be held in 199 Cory to work on the lab assignment. 

12. The format for the lab has been posted.  It is (listed above also):

FORMAT FOR WRITEUP: Please include all answers to any questions, all figures required, and all code used (preferably in an appendix).  Make sure plots in the frequency domain go from –π to π, not 0 to 2π.  Hand in your own lab report representing your work. DO NOT COPY OTHER PEOPLE’S CODE AND/OR ANSWERS.

13. Midterm #1 will be held in Sibley Auditorium

14. For homework #5, use the command zplane for plotting pole-zero diagrams.  The Chebyshev error is defined as the average of the differences between the desired filter and the obtained filter squared IN THE FREQUENCY domain.

15. For the guy who let prof. Zakhor use his book today (Friday, October 17th), we have your book.  Email the TA if you want to get it back to study for the midterm.

16. The review session on Tuesday, October 21, 2003 will be held in the Hogan room, 5th floor Cory.

17. Discussion sections for the week of 10/20 - 10/24 are cancelled.  A review session will be held in place of the Tuesday discussion section.

18. Problem 7.37 has a bunch of typos.  Please do the question as follows: H(ejw)=A(ejw)e-jw(N-1)/2.  Then, for part b, change hd[n] to h[n].  Then, for part c, change hd[n] to h[n]. 

19. Lab #2 Notes: For question #1, “ripples” means how many local maxima/minima are in the signal (do NOT count the values at the passband frequency wp and the stopband frequency ws because they do not contribute to what is defined as “ripple”).  For question #2, part b should read: What is the discrete-time frequency corresponding to the Ωc given (forget about the “prewarped analog band edge”, this makes no sense).  For part c, the prewarped band edge is the Ωc given. For question 3, assume fp = 0.25 and fs = 0.3.  For question 4, note that in part c, the definitions of fs and fp have been switched.

20. As mentioned in class, we’ll have the extra review session for people who are struggling to understand some of the material on Monday (11-10-03).  Let’s meet at the second floor lounge in Cory at 12:30.  Bring your books and we’ll go through the material from the beginning, along with doing some questions and fixing any confusion.  NOTE: this session is for people struggling with the material.  I strongly encourage those of you who got below a sixty on the midterm to come to help you for the final.

21. Discussion section and Office hours are cancelled on Tuesday 11-11-03 due to the holiday.  Dan will hold extra office hours on Wed. 5-6 PM in 469 Cory to make up for the lost Tuesday office hours.

22. Extra Review Session #2 will be on Friday from 2-4.  Let’s meet outside of the second floor lounge and we’ll move to an open room from there.

23.  Extra Review Session #3 will be on Tuesday from 5-7 (NOTE: to students in review session #2, I’ve moved it to Tuesday because I’ll be out of town on Monday).

24. Extra Review Session #4 will be held on Friday.  The time is TBD (see me in class to figure out when, I’ve had conflicting suggestions for when is best).  The place, as always, will be 2nd floor Cory Hall by the lounge.

25. The last problem of the homework solutions for homework #8 was done assuming that N = 100.  It was given, however, that N = 128.  Go here to find a revised version of the homework solutions that makes these changes.  Second Note: the solution was re-updated on 11-24-03 to reflect an error in the new solution that was pointed out.  The current solution should be the definitive solution to the homework.  This is the same as the original solution.

26.  Extra Review Session #6 will be on Monday, December 1st from 3-5PM.

27.  Due to travels, I (Dan) will not be able to hold my normal discussion section on Wednesday.  Since the review for the exam is on Tuesday, I move the discussion to 11 AM – 12 PM on Wednesday in 469 Cory Hall.  The discussion will be more like an office hours where you can ask questions on the material.

 

28.  The solution to homework #9 can be found here.  Note: there was an error in problem 9.30 part c (it should have read … (x[n]-x[n+N/2])).  This was corrected as of Tuesday, December 02 at 4:30 PM.

 

29.  Extra Review Session #7, the final extra review session, will be held on Wednesday, December 3rd from 6:00 PM – 8:00 PM.  If I’m not there right away, meet together and I’ll try to get there as soon as possible.

 

30.  The Second Midterm is in 3 Leconte on Friday, December 5th, 2003 from 9:30-11 AM.  Good luck!


Lectures

Some lectures might move from Wednesday or Friday to Monday from 9:30 to 11:00 am. This will be done with advance notice. Please make sure that you either attend Monday 9:30 to 11:00 classes, or watch the video tapes of the lecture. The course is digitally recorded, webcasted, and stored to be viewed on line. The URL for the lectures is given below along with the URL for Fall 2000. This year’s class will be different from 2000. So, if you miss a lecture, make sure you watch the 2003 digitized video, rather than the 2000 one.

 

Pdf versions of the lecture notes will be posted shortly after the lecture was given.

 

For Fall 2003 Lectures, click here.

For Fall 2000 Lectures, click here.

Lecture Notes

          Lecture 1: What is DSP? Intro to LTI systems

            Lecture 2: The DTFT

            Lecture 3: Basics of Sampling

            Lecture 4: The Z-Transform

            Lecture 5: Upsampling and Interpolation

            Lecture 6: LTI

            No Class on Sept. 17 and 19

            Lecture 7: Different Realization of LCC DE

            Lecture 8: Flowgraphs/Realizations of IIR Filters

            Lecture 9: Transposition Theorem

            Lecture 10: Linear Phase Filtering

            Lecture 11: Generalized Linear Phase Filter

            Lecture 12: Filter Design

            Lecture 13: FIR Filter Design Using Windows

            Lecture 14: Optimal FIR Filter Design

            Lecture 15: Optimal FIR Filter Design, continued

            Lecture 16: Optimal FIR Filter Design, continued

            Lecture 17: IIR Filter Design

            Lecture 18: Bilinear Transformation

            Lecture 19: IIR Filter Design Transformation

            Lecture 20: Discrete Fourier Series

            Lecture 21: DFT = Discrete Fourier Transform

            Lecture 22: How to Use DFT to do Convolution

            Lecture 23: Using DFT for Filtering Infinitely Long Sequences

            Lecture 24: DCT and Its Relation to DFT

            Lecture 25: Fast Fourier Transform

            Lecture 26: Fouric Analysis of Signals using DFT