Digital Communications – Final Exam.

1. (20 %) Let the received signal of a binary communication system be given by

y=u+ v
where u is the n -dimensional transmitted signal vector which is equally likely to
N0
v ~ N (0 , I )
be u A or u B and 2 n .
(A) (10 %) Determine the ML detection rule.
(B) (10 %) Determine the detection error probability by the ML detection rule.

SOL
(A) The ML detection rule gives
⃗
u
f ( y|u A ) ¿ f ( y|uB )
A

¿ ⃗uB

1
( 0)
πN n/2 (
exp −
N 0
)
‖ y−u A‖2 ¿⃗u A 1
¿ u
⃗ B (
πN 0)
n/ 2
exp −
N0(
‖ y−u B‖2
)
(答道這裏算對)
⃗u A
2¿
‖y −u A‖ ‖y−u B‖2
¿ ⃗u
B

(B) The error probability is given by

Pb =Q ( 2dσ ) where d is the Euclidean
distance between u A and u B , and σ is the noise variance. Therefore, we have

(√ )
‖u A−u B‖
Pb =Q
N0
2
(C)
2

2. (20 %) Simplex Signals: An orthogonal signal set for M=3 is given below
 (A) (10 %) Determine the vector representation of the signal set given the basis
, , and as indicated on the figure.

(B) (10 %) Construct the simplex signal set (using vector representation) based
on the result from (A).
SOL
1 The signals s1 (t ) , s2 (t ) , and s3 (t ) can be represented as

s1 (t )=1⋅φ 1 (t )+0⋅φ2 (t )+0⋅φ3 (t ) -> s1 =[ 1 ,0 ,0 ]

s2 (t )=0⋅φ1 (t )+1⋅φ2 (t )+ 0⋅φ 3 (t ) -> s2 =[0 , 1 ,0 ]

s3 (t )=0⋅φ1 (t )+ 0⋅φ 2 (t )+1⋅φ 3 (t ) -> s1 =[ 0 , 0 ,1 ]

M
1
s̄=
M
∑ si
2 Define i =1 , we have

1 1 1
[1 1 1
s̄= [1 , 0 , 0 ]+ [0 , 1 , 0 ]+ [ 0 , 0 ,1 ]= , ,
3 3 3 3 3 3 ] '
By simplex signal set can be generated by the formula si =s i− s̄ . Therefore

s'1 =[ 1 ,0 ,0 ]−[ 1/3 ,1 /3 , 1/3 ]=[ 2/3 ,−1/3 ,−1 /3 ]

s'2 =[0 , 1 ,0 ]−[1/3 ,1 /3 ,1/3 ]=[−1 /3 , 2/3 ,−1 /3 ]
'
s3 =[0 , 0 , 1 ]−[1/3 , 1/3 ,1/3 ]=[−1 /3 ,−1/3 ,2 /3 ]

3 (10%) Assume M signals are employed for digital

 communication over an AWGN channel with noise power-spectral density

. Let
Pm be the a priori probability of transmitting sm and φ 1(t ),φ2 (t ),...,φ N (t ) be

the orthonormal basis for this signal set. Draw the block diagram of the
correlation receiver with N correlators.

SOL:

4 (15 %) Consider a binary PSK signal represented by the equivalent lowpass signal
∞
u(t )= ∑ I n g(t−nT )
n=−∞

where
I n takes on one of the two possible values −1 and +1 with equal

probability. The sequence of information symbol

{I n }is statistically

independent.
(A) (10 %)Determine the power spectrum density of u(t ) when

g(t )= sinc T )
( t

(B) (5 %) Draw the power spectrum density.

 SOL
1
Su (f )= |G(f )|2 S I ( f ) {I }
(A) We have that T . Since n is statistically independent,
we have

R I (m )= {10 m=0
m≠0
1
Su (f )= |G(f )|2
Therefore, S I ( f )=1 and T . For the sinc pulse
G( f )=F [ g (t ) ] =T ×rect (fT )
1
Su ( f )= |G(f )|2 S I ( f )=rect ( fT )
T
(B)

5. (20 %) Assume M equal-power orthogonal signals are

employed for digital communication over an AWGN channel with noise power-
spectral density . Assume that all signals are equally likely. The symbol

period is and for .

(A) (15%) Derive the symbol error probability using union bound.
(B) (5 %) By using the result from (A), determine the bit error probability.
 SOL:
5. The union bound for symbol error probability can be represented as
M
1
PeM ≤
M
∑
¿
¿¿¿
M
T
d kj =∫0 [ si (t )−s j (t ) ] dt=2 E
2 2
where . Therefore,

PeM ≤( M −1)Q
(√ )
E
N0

6.
Pb =
M
2( M −1)
M
P eM ≤ Q
2 (√ ) E
N0

以下答案也算對
1
Pb ≈ P eM ≤
2
M −1
2
Q
(√ )E
N0

6. (15 %) Comparison of M-FSK, M-PSK, and M-QAM: Consider three digital

communication systems, the M-FSK, M-PSK, and M-QAM, where M is the number
of constellation points.
(A) (5 %) As M is large, which modulation has the best power efficiency.
(B) (5 %) As M is large, which modulation has the worst bandwidth efficiency.
(C) (5 %) For large
Eb / N 0 , which modulation should be employed to achieve the

−5
symbol error rate of 10 as M is large.

SOL
(A) M-FSK has the best power efficiency, because under the same symbol error
rate, the M_FSK requires the lowest
Eb / N 0 .

(B) M-FSK has the worst bandwidth efficiency as M is large.

2 log 2 M
r=
Bandwidth efficiency for M-FSK is M
Bandwidth efficiency for M-PSK and M-QAM are r =log 2 M
 2 log 2 M
< log 2 M
As M is large M
(C) M-QAM should be selected as
Eb / N 0 is large.