Sei sulla pagina 1di 12

Signals and Systems

Problem Set:
From Continuous-Time to Discrete-Time

Updated: October 22, 2017


Problem Set

Problem 1 - Properties of unit delay


In the third lecture, you introduced the z-transform. Consider the LTI system G with transfer
function G(z) = z 1 . Determine whether G is a) memoryless, b) causal, c) stable.

u y
G
Input Output

Problem 2 - LCCDE from system diagram


Consider the DT LTI system in Fig. 1. Write a difference equation that relates the output y to
the input u.

u q y
+ +

z 1
2 3

Figure 1

Problem 3 - Periodicity of DT sequences


In the lecture we saw that the complex exponential sequence

{x[n]} = {ej0 n }

is periodic only if 0 /2 is a rational number. Prove this.

Problem 4 - Periodicity of sampled sinusoids


Consider the sinusoidal continuous-time signal

x(t) = cos(15t)

a) Find all sampling times Ts for which the sampled signal {x[n]} = {x(nTs )} is periodic.

b) Find the fundamental period of {x[n]} if Ts = 0.1 seconds.

Problem 5 - The sum of periodic signals


Let the sequences x1 and x2 be periodic with fundamental periods N1 and N2 , respectively.
Under what conditions is the sum x = x1 + x2 periodic, and what is the fundamental period of
x if it is periodic?

2
Problem 6 - System properties using MATLAB
Consider the system described by the following difference equation:

y[n] = u[n] 2u[n 1]

We assume initial rest, that is y[n] = 0 for all n < 0.

a) Using analytical methods, determine whether the system is:

i) Linear
ii) Time-Invariant

b) Determine the impulse response h of the system analytically.

c) Based on the impulse response, determine whether the system is:

i) Causal
ii) Memoryless
iii) Stable

d) Create a MATLAB function that simulates the system. Your function should accept a causal,
DT input sequence u of finite length. The first element of this vector is assumed to be at n = 0
and the signal is assumed to be zero at all points not defined by the vector. Furthermore,
the system is assumed to be at initial rest.

e) Now use your function to test your answers to a), b) and c)

Problem 7 - Periodic inputs and outputs


Show that if the input u to a discrete-time LTI system is periodic with period N , then the
output y is also periodic with period N .

Problem 8 - FIR vs IIR systems


Find the impulse response h for each of the causal LTI discrete-time systems satisfying the
following difference equations and indicate whether each system is an FIR or an IIR system.

a) y[n] = u[n] 2u[n 2] + u[n 3]

b) y[n] + 2y[n 1] = u[n] + u[n 1]

c) y[n] 0.5y[n 2] = 2u[n] u[n 2]

Problem 9 - Stability from LCCDE


Consider an LTI system described by the difference equation

y[n] = ay[n 1] + bu[n],

where u[n] is the input, y[n] is the output, and a and b are real numbers. For which values of
a 6= 0 and b 6= 0 is the system stable?

3
Problem 10 - Cascaded systems
Consider the following cascade of causal LTI systems:

u q y
G1 G2

where

q[n] = 0.5q[n 1] + 2u[n]


y[n] = 0.75y[n 1] + q[n]

a) What is the transfer function of the cascaded system, that is, the transfer function between
input u and output y?

b) Is the cascaded system stable?

Problem 11 - Stable interpretation for a system


In lecture 3, we introduced the following theorem:
Given the transfer function H(z), there exists a stable, but not necessarily causal interpretation
for the underlying system iff H(z) has no poles on the unit circle. That is, if H(p) = , then
|p| =
6 1.
We proved it for one direction. You should now prove the other direction: if H(z) has no pole
p with |p| = 1 there exists a stable interpretation for the system. Prove this fact for the special
case that H(z) is a rational transfer function, as would arise from LCCDE and SS descriptions
of a system. Also assume that the finite poles of H(z) are distinct.
Hint: Consider separately poles pk with |pk | > 1 and poles pk with |pk | < 1. Show that, with the
appropriate interpretation (causal/non causal), it is possible to obtain an absolutely summable
time-domain sequence for both cases. Then do the same for the poles at infinity.

Problem 12 - DT Transfer function


Let G be a causal LTI system with transfer function
1
H(z) = .
1 + a1 z 1
If the system outputs the sequence {y[n]} when given input {u[n]}, what is a1 ?
1
a) a1 = 2

b) a1 = 1
1
c) a1 = 3

1 1
0.5 0.5
u[n]

y[n]

0 0
0.5 0.5
1 1
1 0 1 2 3 4 5 6 1 0 1 2 3 4 5 6
n n

4
Problem 13 - DT Transfer function
Let G be a causal LTI system with input u, output y and transfer function

1 + 14 z 1 + z 2
H(z) = .
1 + 31 z 1

Which of the following sequences shows the first values of the impulse response {h[n]} of G?

1 1
0.5 0.5
h1 [n]

h2 [n]
0 0
0.5 0.5
1 1
2 0 2 4 2 0 2 4
n n

5
Sample Solutions
Problem 1 - Solution
The transfer function G(z) = z 1 corresponds to a time delay of 1 between input and output,
thus the systems input-output relation is given by
{y[n]} = G{u[n]} = {u[n 1]}
a) Since the output value at time n depends on the input value at time n 1, the system is not
memoryless.
b) Since the output does not depend on future input values, the system is causal.
c) We calculate the impulse response of the system as
{h[n]} = {[n 1]}.
We see that the system has an FIR and thus

X
|h[k]| < .
k=

Therefore the system is stable.

Problem 2 - Solution
From the system diagram and by using the z-transform, we can immediately write
Y (z) = Q(z) + 3z 1 Q(z) = (1 + 3z 1 )Q(z) (1)
Q(z) = 2z 1 Q(z) + U (z). (2)
Rearranging Equation (2) for Q(z) yields
Q(z) = (1 2z 1 )1 U (z),
which when substituted into Equation (1) gives
1 + 3z 1
Y (z) = U (z),
1 2z 1
(1 2z 1 )Y (z) = (1 + 3z 1 )U (z).
From which we can directly write the difference equation as
y[n] 2y[n 1] = u[n] + 3u[n 1].

Problem 3 - Solution
{x[n]} is periodic with period N if {x[n]} = {x[n + N ]}. In our case we have
{ej0 n } = {ej0 (n+N ) } = {ej0 n }ej0 N .
This is only the case if ej0 N = 1. Therefore 0 N must be a multiple of 2:
0 N = m2, m Z.
Rearranging we get
0 m
= = rational number. (3)
2 N

6
Problem 4 - Solution
a) Using 0 = 0 Ts , where 0 is the frequency of the CT sinusoid, and the periodicity crite-
rion (3) we get
0 0 Ts 15Ts m
= = =
2 2 2 N
2 m
Ts = with m, N N+ (4)
15 N

b) Sampling x(t) with sampling time Ts = 0.1 seconds, leads to the DT sequence x[n] = cos 32 n ,


which has frequency 0 = 32 . The fundamental period N of a sequence is given by the small-
2
est N that satisfies N = m 0
, with m, N N+ . We therefore have:

4
N = m = 4.
3

Problem 5 - Solution
Sequences x1 and x2 have fundamental periods N1 and N2 , respectively. We therefore know that
these sequences will be periodic with any integer multiple 1 , 2 Z of the fundamental period:

{x1 [n]} = {x1 [n + 1 N1 ]}


{x2 [n]} = {x2 [n + 2 N2 ]}.

Thus,

{x[n]} = {x1 [n]} + {x2 [n]} (5)


= {x1 [n + 1 N1 ]} + {x2 [n + 2 N2 ]}

In order for {x[n]} to be periodic with period N it is required that

{x[n]} = {x[n + N ]}
{x1 [n + 1 N1 ]} + {x2 [n + 2 N2 ]} = {x[n + N ]} (6)

Therefore, comparing Equation 5 with Equation 6 we see that

1 N1 = 2 N2 = N

Since we can always find integers 1 , 2 and N to satisfy the above equation, it follows that
the sum of two periodic sequences is also periodic and its fundamental period N is the least
common multiple of N1 and N2 .

Problem 6 - Solution
Sample solutions in the form of MATLAB files can be found in the corresponding zip file on
the website.

a) Using analytical methods, determine whether the system is:

i) Linear: By definition, a system described by an LCCDE is linear (assuming initial rest).


This can be shown using the methods from Problem Set 2.
ii) Time-Invariant: By definition, a system described by an LCCDE is time-invariant. This
can be shown using the methods from Problem Set 2.

7
b) The system is described by the LCCDE
y[n] = u[n] 2u[n 1] (7)
This system is written non-recursively. Thus applying an impulse input u = yields the
required result directly:
h[n] = [n] 2[n 1]
{h[n]} = {. . . , 0, 1, 2, 0, . . . }

c) Based on the impulse response, determine whether the system is:


i) Causal: Causal as the impulse response satisfies h[n] = 0 for all n < 0.
ii) Memoryless: Not memoryless as the impulse response does not satisfy h[n] = 0 for all n 6= 0.
iii) Stable: The impulse response is absolutely summable, and the LTI system is therefore
stable.
d) See attached file.
e) Now use your function to test your answers to the above. Note in particular that one cannot
prove the linearity or time-invariance of a system by analyzing its input/output behavior, as
these properties need to be shown for all possible combinations of inputs (an infinitely large
set). It is, however, possible to show that a system is not linear or not time-invariant, by
finding a single counter example.

Problem 7 - Solution
Let h be the impulse response of the system. Writing the output y as the convolution of u and
h we have

X
X
y= u[k]{h[n k]} = h[k]{u[n k]}
k= k=

Let n = m + N . Then

X
X
{y[n]} = {y[m + N ]} = h[k]{u[m + N k]} = h[k]{u[(m k) + N ]}
k= k=

As u is periodic with period N , we have {u[(m k) + N ]} = {u[m k]}. Thus,



X
{y[m + N ]} = h[k]{u[m k]} = {y[m]}
k=

which indicates that the output y is periodic with period N .

Problem 8 - Solution
a) If the input to the system is the unit step, that is u = , it follows that the output is, by
definition, the impulse response of the system: y = h. We subsitute and rewrite the difference
equation as:
h[n] = [n] 2[n 2] + [n 3]
= {. . . , 0, 1, 0, 2, 1, 0, . . . }

Since h is finitely summable, the system is an FIR system.

8
b) The difference equation can be rewritten as:

h[n] = 2h[n 1] + [n] + [n 1]

Since the system is causal, h[1] = 0. Thus

h[0] = 2h[1] + [0] + [1] = [0] = 1


h[1] = 2h[0] + [1] + [0] = 2 + 1 = 1
h[2] = 2h[1] + [2] + [1] = (2)(1) = 2
h[3] = 2h[2] + [3] + [2] = (2)(2) = 22
..
.
h[n] = 2h[n 1] + [n] + [n 1] = (1)n 2n1 for n > 0.

Therefore

h[n] = [n] + (1)n 2n1 s[n 1]

Since h has infinitely many non-zero terms, the system is an IIR system.

c) The difference equation can be rewritten as:

h[n] = 0.5h[n 2] + 2[n] [n 2]

Since the system is causal, h[2] = h[1] = 0. Then

h[0] = 0.5h[2] + 2[0] [2] = 2[0] = 2


h[1] = 0.5h[1] + 2[1] [1] = 0
h[2] = 0.5h[0] + 2[2] [0] = 0.5(2) 1 = 0
h[3] = 0.5h[1] + 2[3] [1] = 0
..
.

Since h is finitely summable, the system is an FIR system.

Problem 9 - Solution
First, we compute the transfer function of the system

b zb
H(z) = = ,
1 az 1 za
which has a pole at p = a. We have seen in the lecture that there exists a stable interpretation
for the system iff |p| =
6 1. Specifically, the above system can be interpreted in the following ways:

Causal interpretation: The system is stable if |a| < 1.

Non causal interpretation: The system is stable if |a| > 1.

The number b can take any value < .

9
Problem 10 - Solution
a) Consider the two systems, given by the LCCDEs:

q[n] = 0.5q[n 1] + 2u[n]


y[n] = 0.75y[n 1] + q[n]

We form the transfer function for each system:


Q(z) 2
H1 (z) = =
U (z) 1 0.5z 1
Y (z) 1
H2 (z) = = .
Q(z) 1 + 0.75z 1
We saw in the lecture that the impulse response of the combined, cascaded system is given
by:

hCascade = h1 h2 .

Convolution in the time domain corresponds to multiplication in the frequency domain. It


follows that
Y (z)
HCascade (z) = = H1 (z) H2 (z)
U (z)
2
= .
(1 0.5z )(1 + 0.75z 1 )
1

b) The systems poles can be directly read from the transfer function as: z = 0.5 and z = 0.75.
As both poles have magnitude |z| < 1, and as we are dealing with a causal system, the
combined system is stable.

Problem 11 - Solution
We will prove this result for the special case that H(z) is a rational transfer function, as would
arise from LCCDE and SS descriptions of a system. We also assume that the finite poles of
H(z) are distinct. In that case, H(z) can be factored as:
NA NC N
X z X z X
H(z) = k + k + ck z k ,
z pAk z pCk
k=1 k=1 k=0

where the finite NA poles with magnitude larger than one are denoted with pAk , the NC poles
with magnitude smaller than one are denoted with pCk . Also note that the third term creates
poles of H(z) at z = .
We can now look at the three terms above, and come up with an absolutely summable sequence
in the time-domain for each of them.

For the first term, we know from the lecture the corresponding time-domain sequence
for |z| < |pAk | (anti-causal interpretation). Note that we call a sequence x anti-causal if
x[n] = 0 for n 0.

NA
X
k pnAk

n<0
hA [n] =
k=1

0 otherwise

10
Since |pAk | > 1, it can be readily shown that

X
|hA [n]| < .
n=

For the second term, we know from the lecture the corresponding time-domain sequence
for |z| > |pCk | (causal interpretation).

NC
X
k pnCk n0
hC [n] =
k=1

0 otherwise

Since |pCk | < 1, it can be readily shown that



X
|hC [n]| < .
n=

For the third term, we use the z-transform pairs [n + k] z k and we get
(
cn N n 0
h [n] =
0 otherwise
Because it is finite, h is absolutely summable.

We can then combine the three sequences (by linearity) and obtain the following result:

X
X
|h[n]| |hA [n]| + |hC [n]| + |h [n]| < .
n= n=

Thus, there always exists a stable intepretation for the underlying system if H(z) has no pole p
with |p| = 1. Note that the resulting h[n] does not have to be causal nor anti-causal. 

Problem 12 - Solution
Rewrite the system in LCCDE form using the definitions
b0 + b1 z 1 + + bM z M
H(z) =
a0 + a1 z 1 + + aN z N
N
X M
X
ak y[n k] = bk u[n k],
k=0 k=0

where a0 = b0 = 1, a1 is to be determined and all other coefficients ak , bk are zero. Therefore,


the output sequence is given by

y[n] + a1 y[n 1] = u[n].


Using this LCCDE and the plot of {y[n]}, we can infer that
1
y[3] = u[3] a1 y[2] = a1 = 0.
2
Thus, the correct answer is
1
a1 = .
2

11
Problem 13 - Solution
This system can readily be rewritten in LCCDE form as

1 1
y[n] + y[n 1] = u[n] + u[n 1] + u[n 2].
3 4
Using the LCCDE and {u[n]} = {[n]}, we get the output sequence

1 37 37 37
{h[n]} = {. . . , 0, 1, , , , , . . . },
12 36 108 324

which is the sequence {h1 [n]}.

12

Potrebbero piacerti anche