COMP ENG 4TL4:

Digital Signal Processing

Notes for Lecture #17

Wednesday, October 15, 2003
Frequency shift (modulation):

Proof: Similar to that for the circular shift property.

2
Parseval theorem:

Proof: Using the matrix formulation of the DFT, we obtain:

3
Conjugation:

Proof:

4
Circular convolution:

Here ~ stands for circular convolution, defined by:

5
Illustration of circular convolution for N = 8:

n=0 0 n=1 0 n=2 0 n=3 0

7 1 7 1 7 1 7 1
1 0 7 2 1 0 3 2 1 4 3 2
6 2 6 2 6 3 7 2 6 4 2 6 5 1 2
0
3 4 5 4 5 6 5 6 7 6 7 0
5 3 5 3 5 3 3 5
4 4 4 4

n= 4 0 n=5 0 n=6 0 n=7 0

7 1 7 1 7 1 7 1
5 4 3 5 4
6 7 6 5 0 7 6
6 6 2 2 6 7 3 2 6 0 4 2 6 1 5 2
7 0 1 0 2 1 2 3 2 3 4
1
5 3 5 3 5 3 5 3
4 4 4 4

x[n] spread clockwise

y[n] spread counterclockwise 6
Example #1: Consider the circularly convolved sequences:

giving z[n] = x[n] ~ y[n]. Then:

7
Example #1 (cont.):

8
Example #1 (cont.): Illustration of the circular convolution
process:
7
0 4 6
5

12 3
x[m]
0 0 0 0
2 2 2
3 3 3
5 5 5
6 6
6 y[(–m) mod N]
7 4 7 4 7 4
(i.e., n= 0)
1 1 1

0 0 0
2 2 2
3 3 3
5 5 5
6 6 6

7 4 7 4 7 4
y[(1–m) mod N]
(i.e., n=1) 9
1 1 1 1
Circular convolution and linear convolution:
– A consequence of the circular convolution property is that circular
convolution in the time domain can be computed efficiently via
multiplication in the Fourier domain.
– If two discrete-time sequences of length L and P, respectively, are
zero-padded to length N, such that
N ≥ L + P−1, then the circular convolution of the sequences is equal
to the linear convolution of the sequences.
– If N < L + P−1, then the circular convolution of the sequences is a
time-aliased version of the linear convolution of the sequences.
⇒ Sampling in the time-domain produces aliasing in the
frequency domain, and sampling in the frequency-domain
produces aliasing in the time domain!
– The upper two properties above allow efficient implementation of FIR
filters in the Fourier domain on DSPs that have specialized hardware
and/or software for computing DFTs.

10
Example #2: Consider the two discrete-time sequences, (i) x[n]: a
rectangular pulse of length 4, and (ii) y[n]: the sequence x[n] zero-padded
to length 8. The circular convolution of y[n] with itself is identical to the
linear convolution of x[n] with itself, while the circular convolution of x[n]
with itself is a time-aliased version of the linear convolution of x[n] with
itself. Discrete–time sequence

x[n] 4
1
3

|DFT|
Y[k]
0.5 2

y[n]
1
X[k]
0 0
0 2 4 6 8 0 2 4 6 8
n k
Discrete–time sequence

Discrete–time sequence
4 4 x[n] ~ x[n]

3 x[n] ∗ x[n] 3 y[n] ~ y[n]

2 2

1 1

0 0 11
0 2 4 6 8 0 2 4 6 8
n n
Proof of the circular convolution property:

12
Multiplication:

Proof: Similar to that for the circular convolution property.

13