You are on page 1of 17

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

30. Signal Space Analysis of BASK, BFSK, BPSK, and QAM

The vector-space representation of signals and the optimum detection process which chooses the signal closest to the received signal is particularly useful in signal design and in probability-of-error calculations. The material presented herein includes the error performance of binary amplitude-shift keying (BASK), binary frequency-shift keying (BFSK), binary phase-shift keying (BPSK), and M-ary quadrature amplitude modulation (QAM) signals.

A binary amplitude-shift keying (BASK) signal can be defined by

s(t) =

A cos(

0

,

2

f

c

t

),

0

elsewhere

tT

(30.1)

where A is a constant, f c is the carrier frequency, and T is the bit duration. It has a

power P = A 2 /2, so that A = 2P . Thus equation (30.1) can be written as s(t)

=

=

= 22
P cos(
t
),
f c
0 ,
2
PT
T cos(
2
t
f c
0 ,
2
E
cos(
2
t
),
f c
T

0

,

),

0

elsewhere

tT

0

tT

elsewhere

0

tT

elsewhere

(30.2)

where E = PT is the energy contained in a bit duration. Figure 30.1 shows the signal constellation diagram of BASK signals and the conditional probability density functions associated with the signals.

Figure

30.1

BASK signal constellation diagram and the conditional probability density functions associated with the signals.

The energy contained in a bit duration is

E = d 2

(30.3)

and so

30.1

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

d = E

(30.4)

It is assumed that the noise is additive white Gaussian noise (AWGN) with a two-sided

σ 2 = n 0 /2. In the

presence of AWGN, the conditional probability density function of φ 1 assuming that s 0 is transmitted is

power spectral density of n 0 /2, zero mean and fixed variance

f(φ 1 / 0)

=

1 2 1
2
e
2
2

2

and the probability of error given that s 0 is transmitted is

P e0 =

 • • f(φ 1 /0)dφ 1 =

1 =

d

/

2

1 =

d /

2

2

1 1
2 2
e
2
2

d

1

(30.5)

Similarly, the probability of error given that s 1 is transmitted is

where

P e1 =

1

= d / 2

f(φ 1 /1)dφ 1 =

1

= d / 2

 • •

f(φ 1 /1) = 2
(
d )
1
1
2 2
e
2
2

1 2
2

e

(

1

d

)

2

2

2

d

1

=

P e1

(30.6)

Let p 0 be the probability of sending s 0 and p 1 be the probability of sending s 1 . For equally likely transmission of binary signals, we have p 0 = p 1 = 0.5. The average probability of error is given by

P e

= p 0 P e0 +

=

P e0

p 1 P e1

=

1 =

d / 2

2

1 1
2 2
e
2
2

d

1

30.2

(30.7)

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

Let

y

=

1 .
2

Then y 2

equation (30.7), we get

=

2 d

1

2

2

and d y

1

= 2

P e

=

=

=

=

= 1
e -y 2 dy
d
y =
2
2
1 2
e -y 2 dy ]
2 [
π

y 1
2 )
d
erfc(
2
2
d
1 erfc(
2
2 n 0 )
1
E
2 erfc(
4n
0 )

where the complementary error function is

erfc (x) =

2 π

x

e -α 2 dα

σ 2 = n 0 /2 for AWGN, and d = E .

Error Performance of Binary PSK

. Substituting y and dy

(30.8)

(30.9)

into

A binary phase-shift keying (BPSK) signal can be defined by

s(t) = +A cos 2πf c t,

0 < t < T

(30.10)

where A is a constant, f c is the carrier frequency, and T is the bit duration. It has a

power P = A 2 /2, so that A = 2P . Thus equation (30.10) can be written as s(t)

= + 2P cos 2πf c t 2
= cos 2πf c t
+
PT
T
2
= cos 2πf c t
+
E
T

(30.11)

30.3

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

where E = PT is the energy contained in a bit duration. Figure 30.2 shows the signal constellation diagram of BPSK signals and the conditional probability density functions associated with the signals.

Figure

30.2

BPSK signal constellation diagram and the conditional probability density functions associated with the signals.

The energy contained in a bit duration is

and so d
E = (
) 2
2
d =
4E

(30.12)

(30.13)

In the presence of AWGN, the conditional probability density function of φ 1 assuming that s 0 is sent is

f(φ 1 /0) =

1 2
2

e d
2
(
+
)
1
2
2
2

(30.14)

and the probability of error given that s 0 is transmitted is

P e0 =

f(φ 1 /0)dφ 1 =

0 0

1 2
2

e d
2
(
+
)
1
2
2
2

d

1

(30.15)

Similarly, the probability of error given that s 1 is sent is

where

P e1 =

0

0

f(φ 1 /1) dφ 1 =

1 2
2

30.4

e d
2
(
)
1
2
2
2

d

1

=

P e0

(30.16)

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

f(φ 1 /2) =

1 2
2

e d
2
(
)
1
2
2
2

(30.17)

Let p 0 be the probability of sending s 0 and p 1 be the probability of sending s 1 . For equally likely transmission of binary signals, we have p 0 = p 1 = 0.5. The average probability of error is P e
= p 0 P e0 + p 1 P e1
= P e0
d
2
(
+
)
1
2
1
2 2
=
e
d
2
2
0
2
+ d
(
+ d
)
1
1
2
2
Let
y =
.
Then y 2 =
2
2
2
equation (30.18), we get
1
=
e -y 2 dy
P e
π
d

1

and dy 1 +
2
y =
2
1
2
=
e -y 2 dy
2
π
d
1 +
2
y =
2
+ d
1
1
2
=
erfc (
)
2
2
1
d
=
erfc (
)
2
2
2
1
d
=
erfc(
2
2
n 0 )
1
=
erfc(
2
n
E 0 )

= d
1
2

(30.18)

. Substituting y and dy into

(30.19)

where φ 1 = 0 is the threshold value, the complementary error function is

30.5

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

erfc (x) =

2 π

x

e -α 2 dα

σ 2 = n 0 /2 for AWGN, and d = 4E .

(30.20)

Error Performance of Orthogonal Binary FSK

A binary frequency-shift keying (BFSK) signal can be defined by

s(t) =

A

A

cos(

cos(

2

2

f

c

f

c

1

2

t

),

t

),

0

elsewhere

tT

(30.21)

where A is a constant, f c 1 and f c 2 are the carrier frequencies, and T is the bit duration.

It has a power P = A 2 /2, so that A = 2P . Thus equation (30.21) can be written as

s(t)

=

=

= 2
P
cos(
2
f
t
),
c
1
2
P
cos(
2
f
t
),
c 2
2
PT
T cos(
2
f
t
),
c
1
2
PT
T cos(
2
f
t
),
c
2
E
cos(
2
f
t
),
c
1
E
cos(
2
f
t
),
c 2

0

elsewhere

tT

0

tT

elsewhere

0

tT

elsewhere

(30.22)

where E = PT is the energy contained in a bit duration. Figure 30.3 shows the signal constellation diagram of orthogonal BFSK signals and the conditional probability density functions associated with the signals.

Figure

30.3

Orthogonal BFSK signal constellation diagram and the conditional probability density functions associated with the signals.

The energy in a bit duration is

and

E = (

d 2

) 2

(30.23)

30.6

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

d = 2E

(30.24)

If we rotate the two orthonormal axes φ 1 and φ 2 to u 1 and u 2 , we end up with a signal constellation diagram for BPSK signals with a separation distance of d. Thus, the average probability of error is d
=
1 erfc(
P e
2 )
2
2
d
=
1 erfc(
2
2 n 0 )
E
=
1
2 0 )
erfc(
2n
σ 2 = n 0 /2 for AWGN, and d =
2E .

(30.25)

Table 30.1 summarises the performance of BPSK, orthogonal BFSK, and BASK signals in the presenec of AWGN.

Modulation

BPSK

Orthogonal BFSK d
1 erfc(
2 )
2
2
d
1 erfc(
2
2 n 0 ) n
E 0 )
)
E
)
4n
0

d

P e

1

2

1

2

1

2

=

=

erfc (

erfc ( E

erfc(

Relative (dB)

0

-3

-6

Table 30.1 Performance of various modulation systems.

Also, we can express the performance of these modulation systems in terms of the average

signal energy E , where

E = E/2

30.7

(30.26)

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

E = E

for orthogonal BFSK signals, and

E = E

(30.27)

(30.28)

for BPSK signals. This is summarised in Table 30.2.

Modulation

BPSK

Orthogonal BFSK

E

E

E

E/2

P e

1

2

1 n E
2n
0
E
4n
0

1

E 0 ) = 2 n

E 0 )

erfc(

erfc(

erfc(

1

2 erfc(

erfc(

) ) 2 ) =

1 2 erfc(

1

) = 2

Relative (dB)

0

-3

-3

Table

30.2

Performance of various binary modulation systems in terms of the average signal energy.

Error Performance of M-ary QAM

Consider the (M=16)-ary quadrature amplitude modulation (QAM) signal constellation diagram of Figure 30.4 as an example.

Figure 30.4 16-ary QAM signal constellation diagram.

The average signal energy of an M-ary QAM signal is

E

=

1

M

1

M

i = 0

E i

(30.29)

where E i is the energy of s i . For M = 16, we have

E

=

1

16

[ 4 2
2
d
d
+
+
8
4
4

9 d

2

d

2

+

4

4

+ 4

9 d

2

9

d

2

+

4

4

1442443

s

5 ,,, 6 9 10

1442443 144424443

s

1 ,,, 2 4 111314 , , ,, 7 8

s

0 ,, 312 , 15

30.8

]

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

and so

d =

5d 2

=

2 2 E
5

(30.30)

(30.31)

To find the probability of error, it is simpler to first find the probability P c of correct detection. It is apparent from Figure 30.4 that we can group signal points into 3 classes and compute P c based on the 3 decision regions as shown in Figure 30.5.

Figure 30.5

Decision regions.

Let n 1 and n 2 be additive white Gaussian noise samples with two-sided power spectral

density of n 0 /2, axis, respectively.

zero mean and fixed variance σ 2

= n 0 /2

along the φ 1 -axis and φ 2 -

Case 1 - four inner signal points (s 5,6,9,10 ) :

In the presence of AWGN, the probability of correct detection given that s i , i = 5, 6, 9, 10, is transmitted is

where

P(C/s i )

= P(-d/2 < n 1 < d/2) x P(-d/2 < n 2 < d/2)

= P(-d/2 < n 1 < d/2) x P(-d/2 < n 1 < d/2)

p = p x
p
= P(-d/2 < n 1 < d/2)
2
1
d /
2
2 2
= 1
e
2
2
d
/ 2
2
1
d
/ 2
1
2 2
= 2
e
2
2
0

dφ 1

dφ 1

30.9

(30.32)

(30.33)

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

Let

y

=

1 .
2

Then y 2

equation (30.33), we get

=

2 d

1

2

2

and d y

1

= 2

d 2
2
1
p =
2
π
0
d
= erf(
)
2
2
d
= erf(
n 0 )

e -y 2 dy

where the error function is

erf (x) =

2 π

x

e -α 2 dα

0

and σ 2 = n 0 /2 for AWGN.

.

Case 2 - four corner signal points (s 0,3,12,15 ) :

Substituting y and dy

(30.34)

(30.35)

into

In the presence of AWGN, the probability of correct detection given that s i , i = 0, 3, 12, 15, is transmitted is

 P(C/s i ) = P( • < n 1 < d/2) x P(-d/2 < n 2 < • ) (30.36) = P(-d/2 < n 1 < • ) x P(-d/2 < n 2 < • ) = P(-d/2 < n 1 < • ) x P(-d/2 < n 1 < • ) = r x r

where

r = P(-d/2 < n 1 < )

=

d / 2

1 2
2

e

2

1

2

2

dφ 1

30.10

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

=

=

=

0

d / 2

1 2
2

e

d / 2

0

1 2
2

e

0.5p + 0.5

2

1

2

2

1

2

2

2

dφ 1 +

dφ 1 +

0

0 1
e
2
2
1
e
2
2

2

1

2

2

1

2

2

2

dφ 1

dφ 1

(30.37)

Case 3 - eight edge signal points (s 1,2,4,7,8,11,13,14 ) :

In the presence of AWGN, the probability of correct detection given that s i , i = 1, 2, 4, 7, 8, 11, 13, 14, is transmitted is

 P(C/s i ) = P(-d/2 < n 1 < d/2) x P(-d/2 < n 2 < • ) = P(-d/2 < n 1 < d/2) x P(-d/2 < n 1 < • ) = p x r (30.38) Let P(s i ) be the probability of sending s i , for i = 0, 1, correct detection is , 15. Thus, the total probability of

P c =

M

1 =

i = 0

15

(

P s

i

)

123

a priori probability

P(C/s i )

With equally likely transmission of s i , we have

P c

p x

1

p) + 4 ( 16

r x

[ 4 p 2 + 4( p 2

+

1

2

1

1

= 4 ( 16

1

=

r) + 8( 16

+

) 2 + 8p ( p 2

16

( 3p + 1) 2

=

16

The probability of error is

P e

= (1 - P c )

= 1 -

(

3p + 1) 2

=

4

2 ( 3p + 1) 2

16 16

3(1 p)(5 + 3p)

=

16

30.11

 (30.39) p x r) 1 2 )] (30.40)

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

For normal operation, p approaches 1 and P e << 1, and then

P e

3(1 p)8

=

3

2

16

(1 - p)

Substituting p = erf (

d ) and d = P e

)] Example

30.1 2 E
5

(30.41)

into equation (30.41), we have

(30.42)

Consider the 4-ary QAM signal constellation diagram as shown in Figure 30.6.

Figure 30.6 4-ary QAM signal constellation diagram.

The average signal energy is

and

E

=

1 4 [ 4 d =
2 E
.
5

d

2 d

2

+

4

4

]

= d 2

2

From our previous analysis of 16-ary QAM signals, the probability of correct detection in 4-ary QAM signals given that s i , i = 0, 1, 2, 3, is transmitted is

where

P(C/s i )

= P(

= P(-d/2 < n 1 < ) x P(-d/2 < n 2 < )

= P(-d/2 < n 1 < ) x P(-d/2 < n 1 < )

= r x

< n 1 < d/2) x P(-d/2 < n 2 < )

r

30.12

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

 r = 0.5p + 0.5 = erf( d ) = erf( d ) p 2 2 2 n 0

and σ 2 = n 0 /2 for AWGN.

The total probability of correct detection is

P c =

M

1 =

i = 0

3

)

123

a priori probability

(

P s

i

P(C/s i )

With equally likely transmission of s i , we have

P c

= 4 ( 1

4 r x

= ( p 2

(1

1

+ 2

) 2

+ p

=

4

r)

) 2

The probability of error is

P e

= (1 - P c ) = 1 -

2 (1

=

(1

+ p

)

p

)(3

+

p

)

4

4

For normal operation, p approaches 1 and P e << 1, and then

P e

4 1

(

p) = (1 - p)

4

Substituting p = erf (

d 2
n 0

) and d = 2E into equation (30.43), we have

P e

1

erf ( E
2 n
0

) = erfc ( E
2 n
0

) .

30.13

(30.43)

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

References

  M. Schwartz, Information Transmission, Modulation, and Noise, 4/e, McGraw- Hill, 1990.  H. Taub and D. L. Schilling, Principles of Communication Systems, 2/e, MaGraw-Hill, 1996.

30.14

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac

1 -

2 π σ 2

e

Figure

30.1

1 -

2 π σ 2

e

Probability density function 2
=
sin
2 π f c t
φ 2
T
2
φ
1
1
=
f ( φ
/0)
f
( φ
/1) =
2σ 2
1
1
2π σ 2
P e1
P e0
s 0
s 1
φ
=
1
0
d/2
d =
E

e

-

( φ

d

1 2

) 2

2σ 2 T cos 2 π f c t

BASK signal constellation diagram and the conditional probability density functions associated with the signals. 2
d
φ
+
=
) 2
sin 2 π f c t
1
2
T
2
1
f
( φ
/1) =
2σ 2
= f
( φ
/0)
1
1
2π σ 2
P e1
P e0
s
s
0
1
φ
0
1
-d
d
=
E
2
2

e

=

-

( φ

1

-

d

2

) 2

2σ 2

cos 2 π f c t Figure
30.2
BPSK signal constellation diagram and the conditional probability density
functions associated with the signals.
d
(u
+
) 2
1
2
1
-
2
f
(u
/0) =
e
T cos 2 π f c
t
=
1
2σ 2
φ 2
2
2
σ
u 2
d
(u
-
) 2
1
2
s 0
d
1
-
f
=
( u
/1) =
E
e
1
2σ 2
2
d
σ 2
0
s 1
2
=
cos 2 π f c
t
φ 1
0
T
1
d
=
E
2
u 1
Figure
30.3
Orthogonal BFSK signal constellation diagram and the conditional
probability density functions associated with the signals.

30.15

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac 2
=
sin
2 π f c t
φ 2
T
Decision boundary
s 0
s 1
s 2
s 3
d/2
d/2
d/2
d/2
d/2
s 7
s 6
s 5
s 4
2
=
cos 2 π f c t
φ 1
0
T
s 8
s 9
s 10
s 11
d
s 15
s 14
s 13
s 12
d

Figure 30.4 16-ary QAM signal constellation diagram. φ 2
d/2
s 10
-d/2
0 d/2
-d/2

(a) φ 2
s 0
φ 1
φ 1
0 d/2
-d/2

(b)

Figure 30.5

Decision regions.

30.16

φ 2 s 2
-d/2
0
d/2
-d/2

(c)

φ 1

Signal Space Analysis of BASK, BFSK, BPSK, and QAM on Mac 2
=
sin
φ 2
T

2 π f c t Decision boundary
s 0
s 1
d/2
2
=
cos 2 π f c t
φ 1
0
T
d/2
s 3
s 2
d/2
d/2

Figure 30.6

4-ary QAM signal constellation diagram.

30.17