UNIVERSITY OF NAIROBI

Department of Electrical Information Engineering

FEE 422: Telecommunications and Electro-Acoustics B
Practice Questions
Q1. A signal to be quantized has a range normalized to 1 and a probability density
function p( x)  1  x with -1  x  1, as shown in the adjoining figure.
p  x

1  x 1  x  0
1
p ( x)  
1  x 0  x 1
1 0 1 x
Figure 1
(a) Find the quantizer step size and levels for a uniform quantizer with eight levels.
(b) For the given p(x) and the range of x, determine the ms signal power S = E{X2}.
(c) Use the form of p(x) to compute the mean square quantization error Nq.
(d) Express the signal to quantization error ratio S/Nq in dB.

Q2. A signal to be quantized has a range normalized to 1 and a probability density

function (pdf) p(x) shown in Figure 2.
p(x)
0.5

 0.5 for  1  x  1
p( x)  
 0 for x  1 -1 0 1 x
Figure 2
(a) Find the quantizer step size and levels for a uniform quantizer with four levels.
(b) Plot the quantizer characteristic for part (a).
(c) Calculate the mean square quantization error Nq.
(d) Find the mean square signal power S., and express the signal-to-quantization noise
(S/Nq) ratio in dB.

Q3. The S/Nq ratio characteristic in PCM exhibits three distinct regions.
(a) With the aid of a suitable sketch, illustrate the regions, and state which one is
preferred for a well-designed PCM system.
(b) A 16-bit linear PCM system has a sampling frequency of 44 kHz .
(i). What is the resulting data rate?
(ii). What is the maximum S/Nq ratio in dB?

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.1
Q4.
(a) A sinusoidal signal m(t )  Am sin(2 f mt ) is applied to a delta modulation (DM)
system with a fixed step-size, . The sampling period is TS. In terms of the
given parameters, what is the maximum value of the amplitude Am to just avoid
slope overload?
(b) In an 8-bit linear PCM system, the bit rate is 64106 bits/sec. What is the
maximum message bandwidth for which the system operates satisfactorily?
(c) What is the maximum output signal-to-noise ratio for the PCM system in (b) ?
Give your answer in dB.
Q5.
(a) In designing a linear predictor using the Levinson-Durbin algorithm, explain how
you would determine the order of the predictor.
(b) Why is the predictor called a linear predictor?
(c) Given a signal with the autocorrelation sequence,
r = [ 2.56, 1.36, 0.94, 0.67, 0.48, 0.32 ]
use the Levinson-Durbin algorithm [provided] to construct a linear predictor of a
suitable order, and justify your answer.

Q6. A pulse of shape p(t) = s1(t) is non-zero from t = a/4 to t = a as shown in Fig.Q4.

s1(t) Noise, n(t)

Gn(f)
Sample at
1.0 t =T

0.5 r (t) Matched

v0(T)
+
Filter
0.0
0 t
(b)
a/4 a/2 a/4 a

(a)
Figure Q6
(a) Sketch the impulse response hMF(t) of a physically realizable filter matched to p(t).
(b) If the matched filter output is to be sampled at t=T, determine and state the
required value of T.
(c) What is required of the noise power spectral density G n(f) for the result in (b)
above to be valid?
(d) If the noise power spectral density Gn(f) does not conform to the requirement in (c),
and you must incorporate a filter of transfer function HW(f) [impulse response
hW(t)] prior to the matched filter, so that at the input to matched filter, the signal
component is g(t):
i. What is the relationship between Gn(f) and HW(f)?
ii. Express g(t) in terms of hW(t) and p(t) [do not evaluate]
iii. What is the impulse response hMF(t) of the matched filter?
[i.e. to what is it matched ? ]

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.2
Q7.
(a) In the transmission of a pulse waveform of known shape [in time], what does a
matched filter guarantee? State the necessary conditions for its use (if any).
(b) What is the role of system consisting of the RC network and the op-amp?
(c) A pulse p(t) is transmitted through a system with an additive noise whose power
spectral density Gn(f) has the shape shown in Figure Q7. Determine the value of
the RC time constant.

(d) What is the impulse response of the matched filter?

Noise, n(t)
Gn(f)

p(t)
+
q(t ) Matched
v0(T)
R
  + Filter
N0
Gn ( f )  1 f 2 Sample at
2 C t =T

impulse response
-W 0 W h1(t)

Figure Q7.

Q8.
A body having a mass of 15 kg is suspended from a spring which deflects 12 mm under
the weight of the mass.
(a) Determine the frequency of the free vibrations.
(b) What is the viscous damping force needed to make the motion aperiodic at a speed
of 1 mm/s?
(c) If, when damped to this extent, a disturbing force having a maximum value of 100 N
and vibrating at 6 Hz is made to act on the body, determine the amplitude of the
ultimate motion.
Q9. A simple oscillator whose natural frequency is 5 rad/s is displaced a distance 0.03
m from its equilibrium position and released.
Find
(a) the initial acceleration,
(b) the amplitude of the resulting motion
(c) the maximum speed attained.
Q10. A mass of 0.5 kg hangs on a spring. When an additional mass of 0.2 kg is attached
to the spring, the spring stretches an additional 0.04 m. When the 0.2 kg mass is
abruptly removed, the amplitude of the ensuing oscillations of the 0.5 kg mass is
observed to decrease to 1/e of its initial value in 1.0 s.
Compute values for Rm, d, A, and ϕ.

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.3
Q11.
(a) Describe what is meant by the statement claiming that “acoustic processes are fast
and small”
(b) In deriving the acoustic wave equation in an elastic medium, describe three
principles that are used, and name the corresponding equations obtained by the
application of each principle.
(c) A string of a cello has a linear density = 1.70 g m−1 and a length L = 0.70 m. The
tension in the string is adjusted so that the fundamental frequency is 220 Hz.
Determine:
(i) the value of the tension in the string
(ii) the mass m which, when suspended from the string, produces the same
tension in the string.
(iii) the wavelength of the sound from the string. (Take the velocity of sound in
air to be 340 m s−1 and the acceleration due to gravity to be 9.81 m s−2).

++++++++++++++++++++++++++++++++++++++++++++++++++++

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.4
 FEE 422: Telecommunications and Electro-Acoustics B
Solutions to Practice Questions: April 2014

Q.1: Solution
(a) Range = 1-(-1)=2
Number of levels = Q = 8
Step Size =  = Range/Q = 2/8 = 1/4 = 0.25

Decision thresholds are dk, k=0,1,2,..,8 with

d0 = -1, and d8 = +1
and dk = dk-1 +  for k = 1,2,3…,8

The quantizer (reconstruction) levels rk are half-way between the decision

thresholds: rk = 0.5(dk+1 + dk), for k = 0,1,2,3…,7.
Therefore:
r0 = 0.5(d1 + d0) = 0.5(2d0+ ) = d0+ /2 = -7/8
r1 = 0.5(d2 + d1) = 0.5(2d0+ 3) = d0+ 3/2 = -5/8

etc
In Summary:
Thresholds,
d0 = –1, d1 = –3/4, d2 = –1/2, d3 = –1/4, d4 = 0, d5 =1/4, d6 =1/2, d7 = 3/4, d8 = 1
dk
Levels, r0 = –7/8, r1 = –5/8, r2 = –3/8, r3 = –1/8, r4 = 1/8, r5 = 3/8, r6 = 5/8, r7 = 7/8
rk

(b) The quantizer characteristic is plotted with the results in(a).

Q(x)
r7
7/8

5/8 r6

3/8 r5

1/8 r4
-1 -3/4 -1/2 -1/4 0 1/4 1/2 3/4 1 x
d0 d1 d2 d3 d4 d5 d6 d7 d8
r3
-1/8

-3/8 r2

-5/8 r1

r0
-7/8

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.5
 (c) Since the signal has uniform pdf, (and the quantizer is uniform), the
quantization error Nq is the same for each level. Furthermore, the levels are
equally likely.
/2
2  x3 
3
/2 /2 2    2
 
1 2
Nq  x 2 dx  x 2 dx       
  / 2  0   3  3  2  12
0
11 1
With  = 1/4, we have N q   
12  16  192

(d) The mean square signal power S., is found as follows:

1
1 1 2  1 2  x3 
S  E  X    x p  x  dx   x dx  2   x dx    x dx   
2 2
1 1 1 2
1 2 1 2 0  0 30
which when we evaluate, we obtain S = 1/3
1 / 3 
192
 64
Then the signal-to-quantization noise (S/Nq) ratio is (S/Nq) =
1 / 192 3
which in dB, becomes, [S/Nq]dB = 10 log10 [64] = 18 dB

Q.2: Solution
(a) Quantizer step size and levels for a uniform quantizer with four levels.
Q = Number of levels = 4, Range = 2, S = step size = Range/Q  S = 0.5
Let the decision thresholds be d0, d1, d2, d3, d4,
We must have d0 = –1, d4 = +1 and d1 = d0+S, d2 = d1+S, d3 = d2+S
Thus, d0 = –1, d1= – 0.5, d2=0, d3= +0.5, d4 = +1
The quantizer levels are r0, r1, r2, r3 and are half-way between adjacent threshold

levels: rk 
1
d k  d k 1  for k=0, 1, 2, and 3
2
Substituting, we obtain: r0 = –0.75, r1 = –0.25, r2,= 0.25, r3 = 0.75

-1 -0.5 0 0.5 1

thresholds, d k d0 d1 d2 d3 d4

levels, rk r3=-0.75 r1=-0.25 r2=0.25 r3=0.75

Relation between decision thresholds and the quantizer levels

r3
0.75

0.25 r2
-1 -0.5 0.5 1
0
d0 d1 d2 d3 d4
r1
-0.25

r0
-0.75

Designed quantizer characteristic

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.6
 (b) The given p(x) and the range of x, are interpreted in following figure:
p(x)
p( x)  1  x 1 1-x
1+x

1  x x0
 p ( x)  
1  x x0
-1 0 1
x

The mean square(MS) signal power S = E{X2} is computed as

  
S E X2 
1

1
x 2 p( x)dx  
0

1
x 2 (1  x)dx 
1
 x (1  x)dx
0
2

If we replace x by –x in the first integral, we find that the two are equal, and so

  x
S  E X2 2
1

0
2
(1  x)dx which gives S = 1/6

(c) By symmetry,

x  0.252 (1  x)dx  2 x  0.752 (1  x)dx

0.5 1
Nq  2  0 0.5
 We obtain Nq = 1/48

 1 / 6 
(d) [S/Nq ]dB = 10 log10   = 10 log10[8] = 9 dB.
 1 / 48 

Q.3: Solution
(a) The regions of the SNR characteristic for PCM are illustrated below.

S 
  dB
 N out Transition
Region
50

48

46

44 Quantization Noise Limited

Region
42
36
40

38
Thermal Noise Limited
36 Region

34

32

30

S 
20 21 22 23 24 25 26 27 28   dB
 N  in

Illustration is for N = 8 bits.

The preferred region is the quantization noise limited region, since in that region the output
SNR is independent of the input SNR.

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.7
 (c) Number of bits per sample = N =16.
(i). Data rate = (samples per second) × (bits per sample) = 44,000×16 = 704 kbit/s
S 22N S
(ii). For PCM,  , and this has a maximum of  22N ,
Nq 1  4 Pe 2 2 N
Nq
which in dB, becomes S / N q dB = 2N log 10[2] = 6N
For N=16, maximum [S/Nq] = 96 dB

Q.4: Solution
dm (t )
(a) The maximum of the input signal is
 2 Am f m
dt max
The slope of the delta modulator (DM) output is /TS.

To just avoid slope overload, the maximum slope of the input signal
should be just less than /TS. Thus 2 Am f m   / TS

Thus the maximum value of the amplitude is A 
2 f mTS
m

(b) The number of bits per sample is N = 8.

Sampling rate = FS (samples per second)
Date rate = R (bits per second)

Now Data Rate = (Bits per sample)  (samples per second)

That is, R = NFS

Therefore, FS = R/N = 64106 /8 = 8106 samples per second

The sampling rate must be at least twice the message bandwidth.
Thus the maximum message bandwidth is half the sampling rate

Maximum message bandwidth = 4106 Hz = 4 MHz

(c) In PCM the output signal-to-noise ratio is given by

S 22N

N 1  4 Pe 2 2 N
As the input signal-to-noise ratio is increased, the probability of bit error
Pe in the denominator gets smaller and smaller.

A point is soon reached when the expression achieves a maximum value

of S/N = 22N

which in dB is [S/N]dB = 2N log10[2] = 6N

For N = 8, we have a maximum value of [S/N]dB = 48 dB

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.8
Q.5: Solution
(a) The Levinson-Durbin algorithm designs a linear predictor recursively,
proceeding from lower to higher orders, computing the prediction error
variance m for each order. The order m of the predictor is determined when
there is no significant difference between the values for m+1 and m
(b) The predictor is called a linear predictor because the prediction is a linear
combination of past values of the signal. The prediction is a sum of linearly
weighted set of past values of the signal, the weights being referred to as
the predictor coefficients.
(c) The given autocorrelation sequence indicates that
r0 = 2.56, r1 = 1.36, r2 = 0.94, r3 = 0.67, r4 = 0.48, r5 = 0.32
For m=1: k1 = r1/r0 = 1.36/2.56 = 0.53125
1 = r0[1– (k1)2] = 1.8375
a11 = k1 = 0.53125
m=2: k2 = [r2 – a11r1] / 1 = [0.94 – 0.53125*1.36] / 1.8375 = 0.118367
2 = 1 [1– (k2)2] = 1.8118
a21 = a11 – k2*a11 = a11[1–k2] = 0.53125*[1–0.118367] = 0.468367
a22 = k2 = 0.118367

m=3: k3 = [r3–a21r2– a22r1] / 2

= [0.67–0.468367*0.94–0.118367*1.36]/ 1.8118 = 0.037949
3 = 2 [1– (k3)2] = 1.8091
Since the difference between 3 and 2 is less than 0.01, we consider that a
second order predictor will be sufficient, with the predictor coefficients as
already found:
a1 = 0.468367 and a2 = 0.118367

Q.6: Solution
(a) The impulse response hMF(t) is a time-reversed shifted copy of p(t). For it to be
physically realizable, it must be zero for negative time.
s1(t) s1(-t)

1.0
time 1.0
reversed
0.5 0.5

0.0
0 t -a
a/4 a/2 3a/4 a -3a/4 -a/2 -a/4 0

shited forward
hMF(t) = s1(a-t)
(delayed) by "a"
1.0

0.5

0.0 t
0 a/4 a/2 3a/4

(b) The best sampling time is determined by the amount of the time shift: T=a.
FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.9
 (c) The matched filter maximizes the output SNR when the additive noise is white. That
is, the noise power spectral density G n(f) must be a constant for all frequencies of
interest (e.g. Gn(f) = N0/2).
(d) If the noise power spectral density G n(f) does not conform to the requirement in (c),
we must incorporate whitening filter transfer function HW(f) [impulse response hW(t)]
prior to the matched filter, so that at the input to matched filter, the signal component
is g(t):

i. Since the whitening filter must “flatten” the noise, we must have
K
HW ( f )  , where K is some constant (which could be set to unity).
Gn ( f )
ii. The signal component (and not the noise component) is g(t) = hW(t)p(t). That
is, g(t) is the convolution of the original pulse with the impulse response of the
whitening filter.
iii. At the input of the matched we now have white noise [not necessarily Gaussian
noise], and a signal component g(t). The impulse response hMF(t) of the matched
filter is now matched to g(t): hMF(t) = g(T-t).
All that remains at this point is to dteremine the value of T from the shape of g(t),
so as to make hMF(t) physically realizable.

Q.7 Solution
(a) A matched filter guarantees that at the sampling instant, the signal-to-noise
ratio will be maximum:
+ for the pulse to which the filter is matched, and
+ Provided that the additive noise is white

(b) The rôle of system consisting of the RC network and the op-amp is to
“flatten” the noise power spectral density (PSD) so that the matched filter
may be used. It’s plays the role of a noise “whitening” filter.

(c) The transfer function of the RC network is

1
H2 ( f ) 
1  j 2 fRC
Its magnitude is
1
H2( f ) 
1  2 fRC 2
At the input to the matched filter, the noise power spectral density is required to
be constant. Therefore, the transfer function of the RC network must be such
that its squared magnitude will be the reciprocal of the noise PSD. For us to
have
2 1
H2 ( f ) 
1 f 2
we must set RC = 1/2

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.10
 (d) At the matched filter input the signal q(t) has a noise component and a
message component. Let the message component be g(t). Then g(t) is the
convolution of the pulse p(t) and the impulse response h2(t) of the RC
network. That is: g(t) = p(t)h2(t)

The filter is matched to g(t). If the filter impulse response is h(t), then it is given
by:

h(t) = g(T – t)

All that remains is to find the value of the parameter T to make the filter
physically realisable.

Q.8. Solution
Given:
(a) The frequency is found as

Or equivalently, the angular

frequency
(b) The damping compliance Rm  2mn  2 15 28.59   857 Ns/m  0.857 Ns/mm
(c) The amplitude A is then F0
A
 k  m    R 
2 2
m
2

, , ,
k  12, 260 N / m

Q.9. Solution
The equation of the motion is x  t   A cos t   
dx
The system starts at rest:  0    A sin   0  sin  0    0
dt t 0
The equation of motion becomes x  t   A cos t 

(a) The initial acceleration is x  0   0   2 x  t 
t 0
 52  0.03  0.75 ms-2

(b) The amplitude of the motion: x  t  t 0  A  0.03 m

 
(c) The speed of the motion is x  t    A sin t  ( x) max   A  5  0.03  0.15 ms
-1

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.11
Q.10. Solution
The equation of motion:
Rm
where n  k / m 
2 km
x  t   Aent cos d t   
Initial displacement: d  n 1   2  n2   2   n
x  0   0.04 m  A cos   0.04
 
Initial velocity = 0:  Aen n cos   d sin    0  tan    n    

d  1  2 
 
 sin    cos   1   2
W 0.2  g 0.2  9.81
k    49.05 Nm-1
 0.04 0.04
k 49.05
n    9.9045 rad/s
m 0.5
After 1 sec, magnitude decreases to e 1
That is, en  e1  n  1
1 0.5
    0.100964 cos   1   2  0.99489
n 49.05
sin     0.100964
d  n 1   2  9.854 rad/s

  4.099  0.0715 rads
0.04
A cos   0.04  A   0.04021 m
1  2
Rm  2 km  2  0.100964  49.05 0.5  1.000 N-s/m
Rm = 1.000 N-s/m
d = 9.854 rad/s
A= 0.04021 m
ϕ = 4.099  0.0715 rads

The solution is therefore x  t   0.04021et cos  9.854 t  0.0715

Q 11. Solution
(a) Acoustic processes are fast and small:

 They are considered to be fast because pressure waves propagate through

air without appreciable heat transfer taking place
– that is, acoustic processes are adiabatic.

 They are considered small because pressure fluctuations that create audible
sounds are extremely tiny when compared to atmospheric pressure.

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.12
 In fact, a relatively loud 100 dB tone is equivalent to only 2 Pa of fluctuating
air pressure (compare this to the atmospheric pressure, which is 101,325 Pa).
Particle velocities are similarly small.

(b) The acoustic wave equation is derived by combining three principles:

1. The conservation of mass, which gives the continuity equation

2. The conservation of momentum, which gives the momentum equation (also

known as the forces equation)

3. Equation of State (from The Ideal Gas Law), which gives the constitutive
relation

(c)
(i) f   c and /2L for the fundamental frequency, giving c  2 Lf .
 1.70 
T   c 2 =   2Lf      2  0.70  220   161 N
2 2

 1000 
T 161
(ii) T  mg  m   = 16.4 kg .
g 9.81
This result illustrates the fact that stringed instruments are subject to large forces.
(iii) The frequency of the sound wave is the same as the frequency of the vibrating
340
string. Hence, the wavelength of the sound wave is equal to = 1.54 m . This is
220
different from the wavelength of the fundamental of the string (= 2L= 1.40 m) because
of the different wave velocities in the string and in air.

FEE 422: Telecomm and Electro-Acoustics B: Practice Questions V. K. Oduol 2014 – P.13