Chapter 8 - Error Probability Calculation

Digital Communications
Instructor: Dr. Phan Van Ca

Lecture 07 - Probability of Error
Calculations for Optimal Receivers
Kyung Hee
University
Modulation
Modulation is used to transmit digital data over a channel
Gramm-Schmidt procedure allows vector representation

of any signal constellation
Optimal receiver consists of a correlator, weighted to

adjust for signal energy and a priori probabilities
Receiver implementation can be simplified with reduced

correlations and matched filter
We now turn to calculating performance

Young S. Kim Wireless Communication Systems Lab. - 2 -
Kyung Hee
University
Digital Communications
Spring 2006
Instructor: Young Soo Kim

Lecture #7 - Probability of Error
Calculations for Optimal Receivers
Modulation
Modulation is used to transmit digital data over a channel
Gramm-Schmidt procedure allows vector representation

of any signal constellation
Optimal receiver consists of a correlator, weighted to

adjust for signal energy and a priori probabilities
Receiver implementation can be simplified with reduced

correlations and matched filter
We now turn to calculating performance

Problem Statement
We transmit a signal s( t ) ∈{s1( t ), s2 ( t ),… , s M ( t )} , where

s(t ) is nonzero only on t ∈[0, T ].
Let the various signals be transmitted with probability:

p1 = Pr[ s1( t )],… , p M = Pr[ s M ( t )]
The received signal is corrupted by noise:

r ( t ) = s( t ) + n( t )
Given r ( t ) , the receiver forms an estimate s ( t ) of the signal

s(t ) with the goal of minimizing symbol error probability
Ps = Pr[ s ( t ) ≠ s(t )]
Kyung Hee
University
Final Form of MAP Receiver
Multiplying through by the constant N 0: 2
N0 K 1 K
s = arg max ln[ pm ] + ∑ rk sm,k − ∑ sm,k 2
{s1,…,s M } 2 k =1 2 k =1

Kyung Hee
University
Problem Statement
We transmit a signal s( t ) ∈{s1( t ), s2 ( t ),… , s M ( t )} , where

s(t ) is nonzero only on t ∈[0, T ].
Let the various signals be transmitted with probability:

p1 = Pr[ s1( t )],… , p M = Pr[ s M ( t )]
The received signal is corrupted by noise:

r ( t ) = s( t ) + n( t )
Given r ( t ) , the receiver forms an estimate s ( t ) of the signal

s(t ) with the goal of minimizing symbol error probability
Ps = Pr[ s ( t ) ≠ s(t )]
Final Form of MAP Receiver
Multiplying through by the constant N 0: 2
N0 K 1 K
s = arg max ln[ pm ] + ∑ rk sm,k − ∑ sm,k 2
{s1,…,s M } 2 k =1 2 k =1
Matched Filter Implementation of
Correlation
t=T
r (t ) h1(t ) r1
r = [ r1 rK ]
t=T
r (t ) hK (t ) rK
Reduced Complexity Implementation -

Kyung Hee
University Processing Stage
K
∑ s1,k rk Σ Σ
k =1
r ×
− E1 2 N0
ln( p1 ) Choose
2
⎡ s1,1 sM ,1 ⎤ Largest
⎢ ⎥
⎢ ⎥
⎢⎣s1, K sM , K ⎥⎦ K Σ Σ
∑s M , k rk
k =1
− EM 2 N0 ln p
( M)
2
Matched Filter Implementation of
Kyung Hee
University Correlation
t=T
r (t ) h1(t ) r1
r = [ r1 rK ]
t=T
r (t ) hK (t ) rK
Reduced Complexity Implementation -

Processing Stage
K
∑ s1,k rk Σ Σ
k =1
r ×
− E1 2 N0
ln( p1 ) Choose
2
⎡ s1,1 sM ,1 ⎤ Largest
⎢ ⎥
⎢ ⎥
⎢⎣s1, K sM , K ⎥⎦ K Σ Σ
∑s M , k rk
k =1
− EM 2 N0 ln p
( M)
2
Symbol Error Probability
Ps ( e) = Pr[s ≠ s]is the average probability of symbol error

M
Ps ( e) = ∑ Pr[si ] Pr[ s ≠ si s = si ]
i =1
z where Pr[s ≠ si s = si ] is the conditional probability of the

receiver not deciding on si given si was transmitted
Pr[ s ≠ si s = si ] = 1 − ∫ p(r si )dr
Ri
z This may be a multidimensional integral over the decision region Ri
⎛ K ⎞
p(r si ) = ( πN 0 ) −K 2
(
exp⎜ − ∑ rk − si ,k
⎝ k =1
2
N0⎟
⎠
)
Symbol Error Probability Calculation for

Kyung Hee
University BPSK
Two antipodal signals ( M = 2) :
s1( t ) = 2 P cos(2 πf ct ) s2 ( t ) = − 2 P cos(2 πf ct )
T T
0 0
T
2
f
One basis function : 1 ( t ) = cos(2 πf ct )
T 0
Signal Space Representation:
s1( t ) = P ⋅ T f1 ( t ) = Eb f1( t ), s2 ( t ) = − Eb f1( t )
s1 = Eb , s2 = − Eb
1
Pr[s1] = Pr[s2 ] =
2
Kyung Hee
University
Symbol Error Probability
Ps ( e) = Pr[s ≠ s]is the average probability of symbol error

M
Ps ( e) = ∑ Pr[si ] Pr[ s ≠ si s = si ]
i =1
z where Pr[s ≠ si s = si ] is the conditional probability of the

receiver not deciding on si given si was transmitted
Pr[ s ≠ si s = si ] = 1 − ∫ p(r si )dr
Ri
z This may be a multidimensional integral over the decision region Ri
⎛ K ⎞
p(r si ) = ( πN 0 ) −K 2
(
exp⎜ − ∑ rk − si ,k
⎝ k =1
2
N0⎟
⎠
)

BPSK
Two antipodal signals ( M = 2) :
s1( t ) = 2 P cos(2 πf ct ) s2 ( t ) = − 2 P cos(2 πf ct )
T T
0 0
T
2
f
One basis function : 1 ( t ) = cos(2 πf ct )
T 0
Signal Space Representation:
s1( t ) = P ⋅ T f1 ( t ) = Eb f1( t ), s2 ( t ) = − Eb f1( t )
s1 = Eb , s2 = − Eb
1
Pr[s1] = Pr[s2 ] =
2
Boundaries of Decision Regions for BPSK
Decision rule says choose s1 if:

p(r s2 ) Pr (s2 ) ≤ p(r s1 ) Pr (s1 )
1 1 ⎛ ( r − E )2 ⎞ 1 1 ⎛ ( r + E )2 ⎞
b ⎟
⇔ ⋅ exp⎜ − b ⎟
≥ ⋅ exp⎜ −
2 πN 0 ⎜ N0 ⎟ 2 πN 0 ⎜ N0 ⎟
⎝ ⎠ ⎝ ⎠
⎛ ( r − E )2 ⎞ ⎛ ( r + E )2 ⎞
⎟ ≥ exp⎜ − b ⎟
⇔ exp⎜ − b
⎜ N0 ⎟ ⎜ N0 ⎟
⎝ ⎠ ⎝ ⎠
⇔ (r − Eb ) ≤ (r + Eb )
2 2
⇔r≥0
Kyung Hee
University
Decision Region for BPSK
R2 R1
s2 = − Eb 0 s1 = Eb

Kyung Hee
University
Boundaries of Decision Regions for BPSK
Decision rule says choose s1 if:

p(r s2 ) Pr (s2 ) ≤ p(r s1 ) Pr (s1 )
1 1 ⎛ ( r − E )2 ⎞ 1 1 ⎛ ( r + E )2 ⎞
b ⎟
⇔ ⋅ exp⎜ − b ⎟
≥ ⋅ exp⎜ −
2 πN 0 ⎜ N0 ⎟ 2 πN 0 ⎜ N0 ⎟
⎝ ⎠ ⎝ ⎠
⎛ ( r − E )2 ⎞ ⎛ ( r + E )2 ⎞
⎟ ≥ exp⎜ − b ⎟
⇔ exp⎜ − b
⎜ N0 ⎟ ⎜ N0 ⎟
⎝ ⎠ ⎝ ⎠
⇔ (r − Eb ) ≤ (r + Eb )
2 2
⇔r≥0
Decision Region for BPSK
R2 R1
s2 = − Eb 0 s1 = Eb
Calculation of Error Probability for BPSK
0⎛ ( r + E )2 ⎞
Pr[ s ≠ s2 s = s2 ] = 1 − ∫
1 b ⎟
exp⎜ − dr
⎜ ⎟
−∞ πN 0 ⎝ N0 ⎠
∞1 ⎛ ( r + E )2 ⎞
b ⎟
= ∫ exp⎜ − dr
πN ⎜ N ⎟
0 0 ⎝ 0 ⎠
∞
= ∫
Eb
1
πN 0 ( )
exp − y 2 N 0 dy , y = r + Eb
=
∞
∫
1
2π
(
exp − x 2 2 dx, ) x= y
N0
2
2 Eb N 0
Kyung Hee
University
Error Probability for BPSK
Pr[ s ≠ s2 s = s2 ] = Q( 2 Eb N o ),
∞ 1
where Q( u) = ∫
u 2π
( )
exp − x 2 2 dx
By symmetry:
⎛ 2 Eb ⎞
Pr[ s ≠ s1 s = s1] = Pr[s ≠ s2 s = s2 ] = Ps ( e) = Q⎜ ⎟
⎝ No ⎠
Although we have derived the result with BPSK in mind,

the result holds for any signal set with this constellation.

Kyung HeeCalculation of Error Probability for BPSK
University
0⎛ ( r + E )2 ⎞
Pr[ s ≠ s2 s = s2 ] = 1 − ∫
1 b ⎟
exp⎜ − dr
⎜ ⎟
−∞ πN 0 ⎝ N0 ⎠
∞1 ⎛ ( r + E )2 ⎞
b ⎟
= ∫ exp⎜ − dr
πN ⎜ N ⎟
0 0 ⎝ 0 ⎠
∞
= ∫
Eb
1
πN 0 ( )
exp − y 2 N 0 dy , y = r + Eb
=
∞
∫
1
2π
(
exp − x 2 2 dx, ) x= y
N0
2
2 Eb N 0
Error Probability for BPSK
Pr[ s ≠ s2 s = s2 ] = Q( 2 Eb N o ),
∞ 1
where Q( u) = ∫
u 2π
( )
exp − x 2 2 dx
By symmetry:
⎛ 2 Eb ⎞
Pr[ s ≠ s1 s = s1] = Pr[s ≠ s2 s = s2 ] = Ps ( e) = Q⎜ ⎟
⎝ No ⎠
Although we have derived the result with BPSK in mind,

the result holds for any signal set with this constellation.
Error Probability Curve for BPSK
-1
Error Rate vs. Eb/No for BPSK
10
-2
10
-3
10
Ps(e)
-4
10
-5
10
-6
10
0 2 4 6 8
Eb/No (dB)

Kyung Hee
University Binary Coherent FSK
Two orthogonal signals ( M = 2) :
s1( t ) = 2 P cos(2 πf1t ) s2 ( t ) = 2 P cos(2 πf 2 t )

T T
0 0
Two basis functions:
T T
2 2
f 1( t ) = cos(2 πf1t ) f 2 (t ) = cos(2 πf 2 t )
T 0 T 0
Signal space representation:

s1( t ) = Eb f1( t ), s2 ( t ) = Eb f 2 ( t )
s1 = [ Eb 0], s2 = [0 Eb ]

Kyung Hee
University
Error Probability Curve for BPSK
-1
Error Rate vs. Eb/No for BPSK
10
-2
10
-3
10
Ps(e)
-4
10
-5
10
-6
10
0 2 4 6 8
Eb/No (dB)

Binary Coherent FSK
Two orthogonal signals ( M = 2) :
s1( t ) = 2 P cos(2 πf1t ) s2 ( t ) = 2 P cos(2 πf 2 t )

T T
0 0
Two basis functions:
T T
2 2
f 1( t ) = cos(2 πf1t ) f 2 (t ) = cos(2 πf 2 t )
T 0 T 0

s1( t ) = Eb f1( t ), s2 ( t ) = Eb f 2 ( t )
s1 = [ Eb 0], s2 = [0 Eb ]
Decision Regions for Binary Coherent FSK
f2 (t )
R2
Eb s2 R1
s1
f1(t )
Eb
Rotating and translating the coordinates gives:
R2 R1
0
s2 ′ = − s1′ =
Eb Eb
2 2
Kyung HeeError Calculation for Binary Coherent FSK

University
Any translation, rotation, or reflection operation on the

coordinates which does not change the distance between
signals will not effect the error probability.
By repeating the calculation for BPSK (with substituted Eb

for Eb 2 ), we find that :
⎛ Eb ⎞
Ps ( e) = Q⎜ ⎟
⎝ No ⎠

Kyung Hee
University
Decision Regions for Binary Coherent FSK
f2 (t )
R2
Eb s2 R1
s1
f1(t )
Eb
Rotating and translating the coordinates gives:
R2 R1
0
s2 ′ = − s1′ =
Eb Eb
2 2
Error Calculation for Binary Coherent FSK
Any translation, rotation, or reflection operation on the

coordinates which does not change the distance between
signals will not effect the error probability.
By repeating the calculation for BPSK (with substituted Eb

for Eb 2 ), we find that :
⎛ Eb ⎞
Ps ( e) = Q⎜ ⎟
⎝ No ⎠
BER Curves for BPSK and FSK
FSK is approximately 3dB worse than BPSK

0
Error Probabilities for Binary Coherent PSK and FSK
10
BPSK
FSK
-1
10
-2
10
Ps(e)
-3
10
-4
10
-5
10
-6
10
0 2 4 6 8 10 12
Eb/No (dB)

Kyung Hee
University Binary Coherent ASK
Two signals ( M = 2) :
s1( t ) = 2 P cos(2 πf ct )
T
s2 ( t ) = 0
0
One basis function:
T
2
f1( t ) = cos(2 πf ct )
T 0

s1 ( t ) = 2 Eb f1( t ), s2 ( t ) = 0
s1 = 2 Eb , s2 = 0

Kyung Hee
University
BER Curves for BPSK and FSK
FSK is approximately 3dB worse than BPSK

0
Error Probabilities for Binary Coherent PSK and FSK
10
BPSK
FSK
-1
10
-2
10
Ps(e)
-3
10
-4
10
-5
10
-6
10
0 2 4 6 8 10 12
Eb/No (dB)

Binary Coherent ASK
Two signals ( M = 2) :
s1( t ) = 2 P cos(2 πf ct )
T
s2 ( t ) = 0
0
One basis function:
T
2
f1( t ) = cos(2 πf ct )
T 0

s1 ( t ) = 2 Eb f1( t ), s2 ( t ) = 0
s1 = 2 Eb , s2 = 0
Decision Regions for Binary Coherent ASK
R2 R1
s2 = 0 Eb s1 = 2 Eb
2
Translating the coordinates gives:
R2 R1
0
s2 ′ = − s1′ =
Eb Eb
2 2
Kyung HeeError Calculation for Binary Coherent ASK

University
Translated signal constellation is idential to that for FSK
Error calculation is identical to that of FSK, so:
⎛ Eb ⎞
Ps ( e) = Q⎜ ⎟
⎝ No ⎠

Kyung Hee
University
Decision Regions for Binary Coherent ASK
R2 R1
s2 = 0 Eb s1 = 2 Eb
2
Translating the coordinates gives:
R2 R1
0
s2 ′ = − s1′ =
Eb Eb
2 2
Error Calculation for Binary Coherent ASK
Translated signal constellation is idential to that for FSK
Error calculation is identical to that of FSK, so:
⎛ Eb ⎞
Ps ( e) = Q⎜ ⎟
⎝ No ⎠
2-D Symbol Error Probability Example:
QPSK
Consider a QPSK signal set:
s1(t ) = 2 P cos(2 πf ct ) , s2 (t ) = 2 P sin(2 πf ct ) ,
T T
0 0
s3 (t ) = − 2 P cos(2 πf ct ) , s4 ( t ) = − 2 P sin(2 πf ct )
T T
0 0
These can be represented with the basis functions:

T T
2 2
f1 ( t ) = cos(2 πf ct ) , f 2 (t ) = sin(2 πf ct )
T 0 T 0
This representation results in the signal vectors:
s1 = [ P ⋅ T = E s 0], s2 = [0 E s ],
s3 = [ − E s 0], s4 = [0 − E s ]
QPSK Signal Constellation and Decision

Kyung Hee
University Regions
0.8
0.6
0.4
0.2
-0.2
-0.4
-0.6
-0.8
-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

2-D Symbol Error Probability Example:
Kyung Hee
University QPSK
Consider a QPSK signal set:
s1(t ) = 2 P cos(2 πf ct ) , s2 (t ) = 2 P sin(2 πf ct ) ,
T T
0 0
s3 (t ) = − 2 P cos(2 πf ct ) , s4 ( t ) = − 2 P sin(2 πf ct )
T T
0 0
These can be represented with the basis functions:

T T
2 2
f1 ( t ) = cos(2 πf ct ) , f 2 (t ) = sin(2 πf ct )
T 0 T 0
This representation results in the signal vectors:
s1 = [ P ⋅ T = E s 0], s2 = [0 E s ],
s3 = [ − E s 0], s4 = [0 − E s ]
QPSK Signal Constellation and Decision

Regions
0.8
0.6
0.4
0.2
-0.2
-0.4
-0.6
-0.8
-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
QPSK Signal Constellation after 45 Degree
Rotation
s1 = [ E s 2 Es 2 ] 0.8
s2 = [ − E s 2 Es 2 ]
0.6
0.4
s3 = [ − E s 2 − Es 2 ]
0.2
s4 = [ E s 2 − Es 2 ]
-0.2
-0.4
-0.6
-0.8
-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

Kyung Hee
University QPSK
Pr[ s ≠ s1 s = s1] = 1 − ∫ p(r s1)dr

R1
2 2
⎛ Es ⎞ ⎛ E ⎞
∞∞ 1 − ⎜ x − ⎟ N0 −⎜ y − s ⎟ N0
⎝ ⎠
= 1− ∫ ∫ e 2 ⋅e ⎝ 2⎠ dxdy
0 0 πN 0
x2 y2
∞ 1 −2 ∞ 1 −2
= 1− ∫ e dx ∫ e dy
Es 2π Es 2π
− −
N0 N0

QPSK Signal Constellation after 45 Degree
Kyung Hee
University Rotation
s1 = [ E s 2 Es 2 ] 0.8
s2 = [ − E s 2 Es 2 ]
0.6
0.4
s3 = [ − E s 2 − Es 2 ]
0.2
s4 = [ E s 2 − Es 2 ]
-0.2
-0.4
-0.6
-0.8
-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

QPSK
Pr[ s ≠ s1 s = s1] = 1 − ∫ p(r s1)dr

R1
2 2
⎛ Es ⎞ ⎛ E ⎞
∞∞ 1 − ⎜ x − ⎟ N0 −⎜ y − s ⎟ N0
⎝ ⎠
= 1− ∫ ∫ e 2 ⋅e ⎝ 2⎠ dxdy
0 0 πN 0
x2 y2
∞ 1 −2 ∞ 1 −2
= 1− ∫ e dx ∫ e dy
Es 2π Es 2π
− −
N0 N0
Symbol Error Probability Calculation
for QPSK (continued)
⎡ ⎛ Es ⎞ ⎤ ⎡ ⎛ Es ⎞ ⎤
= 1 − ⎢1 − Q⎜ ⎟ ⎥ ⋅ ⎢1 − Q⎜ ⎟⎥
⎣ ⎝ N 0⎠⎦ ⎣ ⎝ N 0⎠⎦
2
⎛ Es ⎞ ⎡ ⎛ Es ⎞ ⎤
= 2 Q⎜ ⎟ − ⎢Q⎜ ⎟⎥
⎝ N0 ⎠ ⎣ ⎝ N0 ⎠ ⎦
2
⎛ 2 Eb ⎞ ⎡ ⎛ 2 Eb ⎞ ⎤
= 2 Q⎜ ⎟ − ⎢ Q⎜ ⎟⎥ (Because E s = 2 Eb )
⎝ N0 ⎠ ⎣ ⎝ N0 ⎠ ⎦
⎛ 2 Eb ⎞
≈ 2 Q⎜ ⎟ (Because Q2 (⋅) << Q(⋅) )
⎝ N0 ⎠
Kyung Hee
University
Symbol Error Probability of QPSK
The conditional error probability given any of the four

signals is identical:
Ps ( e) = Pr[ s ≠ s1 s = s1] = Pr[ s ≠ s2 s = s2 ] =
The symbol error probability of QPSK is approximately

twice that of BPSK:
⎛ 2 Eb ⎞
Ps ( e) ≈ 2Q⎜ ⎟
⎝ N0 ⎠
When we discuss Bit Error Rate, we will find that the BER
of QPSK and BPSK are identical
Symbol Error Probability Calculation
Kyung Hee
University for QPSK (continued)
⎡ ⎛ Es ⎞ ⎤ ⎡ ⎛ Es ⎞ ⎤
= 1 − ⎢1 − Q⎜ ⎟ ⎥ ⋅ ⎢1 − Q⎜ ⎟⎥
⎣ ⎝ N 0⎠⎦ ⎣ ⎝ N 0⎠⎦
2
⎛ Es ⎞ ⎡ ⎛ Es ⎞ ⎤
= 2 Q⎜ ⎟ − ⎢Q⎜ ⎟⎥
⎝ N0 ⎠ ⎣ ⎝ N0 ⎠ ⎦
2
⎛ 2 Eb ⎞ ⎡ ⎛ 2 Eb ⎞ ⎤
= 2 Q⎜ ⎟ − ⎢ Q⎜ ⎟⎥ (Because E s = 2 Eb )
⎝ N0 ⎠ ⎣ ⎝ N0 ⎠ ⎦
⎛ 2 Eb ⎞
≈ 2 Q⎜ ⎟ (Because Q2 (⋅) << Q(⋅) )
⎝ N0 ⎠
Symbol Error Probability of QPSK
The conditional error probability given any of the four

signals is identical:
Ps ( e) = Pr[ s ≠ s1 s = s1] = Pr[ s ≠ s2 s = s2 ] =
The symbol error probability of QPSK is approximately

twice that of BPSK:
⎛ 2 Eb ⎞
Ps ( e) ≈ 2Q⎜ ⎟
⎝ N0 ⎠
When we discuss Bit Error Rate, we will find that the BER
of QPSK and BPSK are identical
Symbol Error Probabilities
for BPSK and QPSK
0
Symbol Error Probabilities for BPSK and QPSK
10
BPSK
QPSK - approximate
10
-1 QPSK - exact
-2
10
Ps(e)
-3
10
-4
10
-5
10
-6
10
0 2 4 6 8 10
Eb/No (dB)
Kyung Hee
University
Notes on Error Probability Calculations
Error probability is found by integrating conditional

probability of error over the decision region
z This becomes difficult for a large number of dimensions
z Difficult multidimensional integrations can be simplified by
appropriate rotation, translation or reflection of coordinates
Error performance depends only on the distance properties

of the signal constellation.
Calculations for nonbinary signal constellations may be

reduced to a set of binary calculations via the “Union
Bound”

Symbol Error Probabilities
Kyung Hee
University for BPSK and QPSK
0
Symbol Error Probabilities for BPSK and QPSK
10
BPSK
QPSK - approximate
10
-1 QPSK - exact
-2
10
Ps(e)
-3
10
-4
10
-5
10
-6
10
0 2 4 6 8 10
Eb/No (dB)
Notes on Error Probability Calculations
Error probability is found by integrating conditional

probability of error over the decision region
z This becomes difficult for a large number of dimensions
z Difficult multidimensional integrations can be simplified by
appropriate rotation, translation or reflection of coordinates
Error performance depends only on the distance properties

of the signal constellation.
Calculations for nonbinary signal constellations may be

reduced to a set of binary calculations via the “Union
Bound”

Chapter 8 - Error Probability Calculation

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 8 - Error Probability Calculation

Uploaded by

Copyright:

Available Formats

Digital Communications

Instructor: Dr. Phan Van Ca

 Modulation is used to transmit digital data over a channel

 Gramm-Schmidt procedure allows vector representation

 Optimal receiver consists of a correlator, weighted to

 Receiver implementation can be simplified with reduced

 We now turn to calculating performance

Instructor: Young Soo Kim

 Modulation is used to transmit digital data over a channel

 Gramm-Schmidt procedure allows vector representation

 Optimal receiver consists of a correlator, weighted to

 Receiver implementation can be simplified with reduced

 We now turn to calculating performance

 We transmit a signal s( t ) ∈{s1( t ), s2 ( t ),… , s M ( t )} , where

 Let the various signals be transmitted with probability:

 The received signal is corrupted by noise:

 Given r ( t ) , the receiver forms an estimate s ( t ) of the signal

 Multiplying through by the constant N 0: 2

Young S. Kim Wireless Communication Systems Lab. - 4 -

 We transmit a signal s( t ) ∈{s1( t ), s2 ( t ),… , s M ( t )} , where

 Let the various signals be transmitted with probability:

 The received signal is corrupted by noise:

 Given r ( t ) , the receiver forms an estimate s ( t ) of the signal

Final Form of MAP Receiver

 Multiplying through by the constant N 0: 2

Reduced Complexity Implementation -

Young S. Kim Wireless Communication Systems Lab. - 5 -

Reduced Complexity Implementation -

 Ps ( e) = Pr[s ≠ s]is the average probability of symbol error

z where Pr[s ≠ si s = si ] is the conditional probability of the

Symbol Error Probability Calculation for

 Ps ( e) = Pr[s ≠ s]is the average probability of symbol error

z where Pr[s ≠ si s = si ] is the conditional probability of the

Symbol Error Probability Calculation for

 Decision rule says choose s1 if:

Young S. Kim Wireless Communication Systems Lab. - 10 -

 Decision rule says choose s1 if:

Young S. Kim Wireless Communication Systems Lab. - 9 -

Decision Region for BPSK

 Although we have derived the result with BPSK in mind,

Young S. Kim Wireless Communication Systems Lab. - 12 -

Young S. Kim Wireless Communication Systems Lab. - 11 -

Error Probability for BPSK

 Although we have derived the result with BPSK in mind,

Symbol Error Probability Calculation for

s1( t ) = 2 P cos(2 πf1t ) s2 ( t ) = 2 P cos(2 πf 2 t )

 Signal space representation:

Young S. Kim Wireless Communication Systems Lab. - 14 -

Young S. Kim Wireless Communication Systems Lab. - 13 -

Symbol Error Probability Calculation for

s1( t ) = 2 P cos(2 πf1t ) s2 ( t ) = 2 P cos(2 πf 2 t )

 Signal space representation:

Kyung HeeError Calculation for Binary Coherent FSK

 Any translation, rotation, or reflection operation on the

 By repeating the calculation for BPSK (with substituted Eb

Young S. Kim Wireless Communication Systems Lab. - 16 -

Error Calculation for Binary Coherent FSK

 Any translation, rotation, or reflection operation on the

 By repeating the calculation for BPSK (with substituted Eb

 FSK is approximately 3dB worse than BPSK

Symbol Error Probability Calculation for

 Signal space representation:

Young S. Kim Wireless Communication Systems Lab. - 18 -

 FSK is approximately 3dB worse than BPSK

Modulation is used to transmit digital data over a channel

Gramm-Schmidt procedure allows vector representation

Optimal receiver consists of a correlator, weighted to

Receiver implementation can be simplified with reduced

We now turn to calculating performance

Modulation is used to transmit digital data over a channel

Gramm-Schmidt procedure allows vector representation

Optimal receiver consists of a correlator, weighted to

Receiver implementation can be simplified with reduced

We now turn to calculating performance

We transmit a signal s( t ) ∈{s1( t ), s2 ( t ),… , s M ( t )} , where

Let the various signals be transmitted with probability:

The received signal is corrupted by noise:

Given r ( t ) , the receiver forms an estimate s ( t ) of the signal

Multiplying through by the constant N 0: 2

We transmit a signal s( t ) ∈{s1( t ), s2 ( t ),… , s M ( t )} , where

Let the various signals be transmitted with probability:

The received signal is corrupted by noise:

Given r ( t ) , the receiver forms an estimate s ( t ) of the signal

Multiplying through by the constant N 0: 2

Ps ( e) = Pr[s ≠ s]is the average probability of symbol error

Ps ( e) = Pr[s ≠ s]is the average probability of symbol error

Decision rule says choose s1 if:

Decision rule says choose s1 if:

Although we have derived the result with BPSK in mind,

Although we have derived the result with BPSK in mind,

Signal space representation:

Signal space representation:

Any translation, rotation, or reflection operation on the

By repeating the calculation for BPSK (with substituted Eb

Any translation, rotation, or reflection operation on the

By repeating the calculation for BPSK (with substituted Eb

FSK is approximately 3dB worse than BPSK

Signal space representation:

FSK is approximately 3dB worse than BPSK

Signal space representation:

Translating the coordinates gives:

Translated signal constellation is idential to that for FSK

Error calculation is identical to that of FSK, so:

Translating the coordinates gives:

Translated signal constellation is idential to that for FSK

Error calculation is identical to that of FSK, so:

These can be represented with the basis functions:

These can be represented with the basis functions:

The conditional error probability given any of the four

The symbol error probability of QPSK is approximately

The conditional error probability given any of the four

The symbol error probability of QPSK is approximately

Error probability is found by integrating conditional

Error performance depends only on the distance properties

Calculations for nonbinary signal constellations may be

Error probability is found by integrating conditional

Error performance depends only on the distance properties

Calculations for nonbinary signal constellations may be