Frequency-Domain C/S of LTI Systems

x(n) y(n)
LTI

I LTI: Linear Time-Invariant system

I h(n), the impulse response of an LTI systems describes the
time domain c/s.
I H(), the frequency response describes the frequency-domain
c/s.
F
I h(n)
H()
I study: system response to excitation signals that are a
weighted linear combination of sinusoids or complex
exponentials.

1/ 25
Frequency-Domain C/S of LTI Systems

I Recall the response of an LTI system to input signal x(n)

X
y (n) = h(n)x(n k) (1)
k=

2/ 25
Frequency-Domain C/S of LTI Systems

I Excite the system with a complex exponential, i.e. let

x(n) = Ae jn , < n < , < <

X h i
y (n) = h(k) Ae j(nk)
k=
" #
X
jk
(2)
=A h(k)e e jn
k=
= Ae jn H()

where,

X
H() = h(k)e jk (3)
k=

3/ 25
Frequency-Domain C/S of LTI Systems

Observations
I y (n) is in the form of a complex exponential with same
frequency as input, multiplied by a factor.
I The complex exponential signal x(n) is called an eigenfunction
of the system.
I H() evaluated at the frequency of the input is the
corresponding eigenvalue of the system.

4/ 25
Frequency-Domain C/S of LTI Systems

Observations, cont.
I As H() is a Fourier transform, it is periodic with period 2
I H() is a complex-valued function, can be expressed in polar
form
H() = |H()|e
I h(k) is related to H() through
Z
1
h(k) = H()e jk d
2

5/ 25
Frequency-Domain C/S of LTI Systems

Real-valued impulse response

An LTI system with a real-valued impulse response, exhibits
symmetry properties as derived in section 4.4.1
I |H()| even function of
I () odd function of .
I Consequently, it is enough to know |H()| and () for
,

6/ 25
Example 5.1.1

I Determine the output sequence of the system with impulse

response
1
h(n) = ( )n u(n)
2
when the input signal is

x(n) = Ae jn/2 , < n <

7/ 25
Example, cont.
I Solution: evaluate H()

X 1
H() = h(n)e jn = 1 j
n=
1 2e
1 2 o
H( = )= 1
= e j26.6
2 1+j2 5
I Thus, output is
 
2 o
y (n) = A e j26.6 e jn/2
5
2 o
= Ae j(n/226.6 ) , < n <
5
I system effect on the input: amplitude scale and phase shift
I change frequency, we change amount of scale change and
phase shift.
8/ 25
Example: Moving Average Filter 5.1.2

I Determine the magnitude and phase of H() for the

three-point moving average (MA) system.

1
y (n) = [x(n + 1) + x(n) + x(n 1)]
3
and plot these functions for 0

9/ 25
Example: Moving Average Filter, cont.

I Solution:
1 1 1
h(n) = { , , }
3 3 3
consequently,
X 1 1
H() = h(k)e jk = (e j + 1 + e j ) = (1 + 2 cos )
n
3 3
Hence
1
|H()| = |1 + cos |
3
2

0, 0

() = 3
2
, <

3

10/ 25
11/ 25
Example: system response to sinusoids 5.1.3

I Determine the response of the system in Example 5.1.1, for

the input signal

x(n) = 10 5 sin n + 20 cos n, < n <
2
I Recall, the frequency response of the system is
1
H() = 1 j
1 2e

12/ 25
Example, cont.
I Solution
I Idea Recognise the frequency of each part of the input signal,
and find corresponding system response.
I First term 10, fixed signal = 0
1
H(0) = 1
=2
1 2

I 5 sin 2 n has a frequency = /2, thus

2 o
H( ) = e j26.6
2 5
I 20 cos n has a frequency =
2
H() =
3
13/ 25
Frequency Domain c/s of LTI systems
The General Case
I Most general case: input to the system is an arbitrary linear
combination of sinusoids of the form
L
X
x(n) = Ai cos(i n + i ), < n <
i=1

Where {Ai } and {i } amplitude and phase of corresponding

sinusoidal component i .
I System response will be of the form:
L
X
y (n) = Ai |H(i )| cos [i n + i + (i )]
i=1

Where |H(i )| and (i ) are the magnitude and phase

imparted by the system to the individual frequency
components of the input signal. 14/ 25
Steady-State and Transient Response to Sinusoidal Input
Signals
I If excitation signal (exponential or sinusoidal) applied at some
finite time instant,e.g. n = 0

response = steady-state + transient

I Example: let
y (n) = ay (n 1) + x(n)
system response to any x(n) applied at n = 0

Xn
y (n) = an+1 y (1)+ ak x(nk), n 0, y (1) initial condition
k=0

I Let x(n) be a complex exponential

x(n) = Ae jn , n 0
15/ 25
Steady-State and Transient Response to Sinusoidal Input
Signals, cont.

I We get,
n
X
y (n) = an+1 y (1) + A ak e j(nk)
k=0
" n #
X
n+1 j k
=a y (1) + A (ae ) e jn
k=0
1 an+1 e j(n+1) jn
= an+1 y (1) + A e ,n 0
1 ae j
Aan+1 e j(n+1) jn A
= an+1 y (1) j
e + e jn
1 ae 1 ae j

16/ 25
Steady-State and Transient Response to Sinusoidal Input
Signals, cont.

I BIBO stable, if |a| < 1

I Since |a| < 1, the terms containing an+1 0, as n .
I Steady-state response:
A
yss (n) = lim y (n) = e jn
n 1 ae j
= AH()e jn

I Transient response of the system:

Aan+1 e j(n+1) jn
ytr (n) = an+1 y (1) e ,n 0
1 ae j

17/ 25
Steady-State Response to Periodic Input Signals

I Let the input signal x(n) to a stable LTI be periodic with

fundamental period N.
I Periodic < n < total response of the system is
the steady-state response.
I Using the Fourier series representation of a periodic signal
N1
X
x(n) = ck e j2kn/N , k = 0, 0, ..., N 1
k=0

18/ 25
Steady-State Response to Periodic Input Signals, cont.

I Evaluating the system response for each complex exponential

xk (n) = ck e j2kn/N , k = 0, 1, ..., N 1

 
2k
yk (n) = ck H e j2kn/N , k = 0, 1, ..., N 1
N
where
2k
H( ) = H()|=2kn/N , k = 0, 1, ..., N 1
N

19/ 25
Steady-State Response to Periodic Input Signals, cont.

I Superposition principle for linear systems:

N1  
X 2k
y (n) = ck H e j2kn/N , < n <
N
k=0

I LTI system response to a periodic input signal is also periodic

with the same period N, with coefficients related by
 
2k
dk = ck H , k = 0, 1, ..., N 1
N

20/ 25
Response to Aperiodic Input Signals

I Let {x(n)} ne the aperiodic input sequence, {y (n)} output

sequence, and {h(n)} unit sample response.
I By Convolution theorem

Y () = H()X ()
In polar form, magnitude and phase of the output signal:
|Y ()| = |H()||X ()|
Y () = H() + X ()

I Output of an LTI system can NOT contain frequency

components that are not contained in the input signal.

21/ 25
Response to Aperiodic Input Signals, cont.

I Energy density spectra of input and output

|Y ()|2 = |H()|2 |X ()|2

Syy () = |H()|2 Sxx ()

I Energy of the output signal

Z
1
Ey = Syy ()d
2

Z
1
= |H()|2 Sxx ()d
2

22/ 25
Example 5.1.5

I A linear time-invariant system is characterized by its impulse

response
 2
1
h(n) = u(n)
2
I Determine the spectrum and energy density spectrum of the
output signal when the system is excited by the signal
 2
1
x(n) = u(n)
4

23/ 25
Example: Solution
I Frequency response of the system
 2
X 1
H() = e jn
2
n=0
1
= 1 j
1 2e

I Similiarly, Fourier transform of the input sequence

1
X () = 1 j
1 4e

I Spectrum of the output signal

Y () = H()X ()
1
= 1 j
(1 2e )(1 14 e j )
24/ 25
Example, cont.

I energy density spectrum

Sxx () = |Y ()|2 = |H()|2 |X ()|2

1
= 5
( 4 cos )( 17 1
16 2 cos )

25/ 25