Professional Documents
Culture Documents
Communication Engineering
Chapter 1-3, Baseband Transmission
1.6 Matched Filter
AWGN Step 1 Step 2
Receiver
Threshold
Sample at t=T
comparison
𝑠1 ( 𝑡 ) Linear
𝑠 𝑖 ( 𝑡 )= 𝑜𝑟 Filter 𝑠^ 𝑖 ( 𝑡 )
𝐻1
𝑠2 ( 𝑡 ) h(t) 𝑧 (𝑇 ) ≷ 𝛾
Binary waveform 𝐻2
• We wish to find the filter transfer function, , that maximizes the above equation.
Apply inverse Fourier transform to T(f) and the filter transfer function H(f) to find a(t):
∞
𝑎 ( 𝑡 )= ∫ 𝐻 ( 𝑓 ) 𝑇 ( 𝑓 ) exp ( 𝑗2 𝜋 𝑓𝑡 ) 𝑑𝑓
−∞
If the two sided power spectral density of the input noise is N0/2, then the output noise
power:
∞
2 𝑁0 2
𝜎 0= ∫ |𝐻 (𝑓 )| 𝑑𝑓
2 −∞
∞ 2
𝑆
2
𝑎𝑖 |∫
−∞
𝐻 ( 𝑓 ) 𝑇 ( 𝑓 ) exp ( 𝑗 2 𝜋 𝑓𝑡 ) 𝑑𝑓
|
( ) 𝑁 𝑇
=
𝜎2
0
=
𝑁0 ∞
2
2
∫ |𝐻 (𝑓 )| 𝑑𝑓
−∞
This
equality holds if where k is arbitrary constant
and * represents complex conjugate. If we assume this to be true, then,
∞ 2 ∞ ∞
| | 2
∫ 𝐻 ( 𝑓 ) 𝑇 ( 𝑓 ) 𝑒𝑥𝑝 ( 𝑗2 𝜋 𝑓𝑇 ) 𝑑𝑓 ≤ ∫ |𝐻 ( 𝑓 )| 𝑑𝑥 ∫ |𝑇 ( 𝑓 )| 𝑑𝑥
−∞ −∞ −∞
2
∞ ∞
2 2
∫ |𝐻 ( 𝑓 )| 𝑑𝑥 ∫ |𝑇 ( 𝑓 )| 𝑑𝑥
𝑆
𝑇h𝑎𝑡 𝑖𝑠 : ( ) 𝑁 𝑇
≤ −∞
𝑁0
∞
2
−∞
2 ∫ |𝐻 ( 𝑓 )| 𝑑𝑓
−∞
∞
𝑆 2 2
𝑖𝑒: 𝑚𝑎𝑥 ( )
= ∫ |𝑇 ( 𝑓 )| 𝑑𝑓
𝑁 𝑇 𝑁 0 −∞
∞
𝑆 𝐸 2
𝑜𝑟 :𝑚𝑎𝑥 ( )
=
𝑁 𝑇 𝑁 0 /2
wherethe energy E of s ( 𝑡 ) 𝑖𝑠 E= ∫ |𝑇 ( 𝑓 )| 𝑑𝑓
−∞
Complex
conjugation
Matched Filter: ℎ(𝑡)=𝑘[𝑠(𝑇−𝑡)]
• A filter that produces maximum output SNR.
• Impulse response of this filter h(t) is the mirror image of the message signal s(t).
• Impulse response of this filter is delayed by a symbol duration T.
t t t
T -T T
Correlator Realization of the Matched Filter
Another way of implementing a Matched filter is too use a correlator
𝑡
𝑧 ( 𝑡 )=𝑟 ( 𝑡 ) ⊗ h (𝑡 ) =∫ 𝑟 ( 𝜏 ) h ( 𝑡 − 𝜏 ) 𝑑 𝜏
0
𝑇
The matched filter output equals to the product of the received signal, r(t), with a replica
of the transmitted waveform, s(t), over a one symbol interval.
Thus a matched filter is often used synonymously with product integrator or correlator.
𝑟 ( 𝑡 ) =𝑠 𝑖 ( 𝑡 ) +𝑛 ( 𝑡 ) Integrator
𝑧 𝑐 ( 𝑡 )
(known signal)
Reference input
When t=T, the correlator output equals the matched filter output
𝑇
This is easier to implement practically and is
𝑧 𝑐 ( 𝑇 )=𝑍 (𝑇 )=∫ 𝑟 ( 𝜏 ) 𝑠 ( 𝜏 ) 𝑑 𝜏 used to replace the matched filter h(t)
0
z(T)
Correlator output zc(t)
t
T
AWGN Step 1 Step 2
Receiver
Threshold
Sample at t=T
comparison
𝑠1 ( 𝑡 )
𝑡
𝑠 𝑖 ( 𝑡 )= 𝑜𝑟 𝑟 (𝜏 ) 𝑠 (𝜏 )𝑑 𝜏 𝑠^ 𝑖 ( 𝑡 )
∫ 𝐻1
𝑠2 ( 𝑡 ) 0
𝑧 (𝑇 ) ≷ 𝛾
Binary waveform S(t) 𝐻2
(known signal)
Reference input
Example by L W Couch
Example: L. W. Couch,
Matched Filter for Detection of a BPSK Signal
1.7 Error Probability Performance of Binary
Signaling
1.7.1 Unipolar NRZ Signal
si(t) 𝑇 𝐻1
❑ 𝑧 (𝑇 ) ≷ 𝛾 0
𝑟 ( 𝑡 )
∫ 𝑧 ( 𝑇 ) 𝑠^ 𝑖 ( 𝑡 )
0 𝐻2
A
Reference signal
0 t s1(t) – s2(t)=A
0 T 2T 3T 4T 5T
Unipolar NRZ Example Correlator Detector
The difference of the reference signal is
When
𝑇
{ 0
} Note: average noise
2
𝑎1 ( 𝑇 )= 𝐸 [ 𝑧 ( 𝑇 ) ] = 𝐸 𝐴 ∫ [ 𝐴 +𝑛 ( 𝑡 ) ] 𝑑𝑡 = 𝐴 𝑇
component is zero
Similarly, when ,
𝑇 2
2
𝑁0 2 𝑁0 𝐴 𝑇
With 𝜎 0= ∫ 𝐴 𝑑𝑡 =
2 0 2
𝑎1 − 𝑎2 𝐴2 𝑇 −0
And the threshold distance 𝑑= =
2 2
1 1
Recall 𝑃𝐵 =𝑄 ( 𝑑 / 𝜎 0 ) +𝑄 ( 𝑑 / 𝜎 0 ) =𝑄 ( 𝑑 / 𝜎 0 )
2 2
𝐴2 𝑇 𝐸𝑏
𝑃𝐵 =𝑄 ( √ ) (√ )
2𝑁0
=𝑄
𝑁0
𝐴2𝑇
Where Eb is the normalized average energy per bit 𝐸𝑏 =
2
Note: average bit energy Eb = average power T
2
∗
𝑂𝑟 𝐸𝑏 =∑ [ 𝑠 𝑖 ( 𝑡 ) 𝑠 ( 𝑡 ) 𝑑𝑡 ] 𝑃𝑟𝑜𝑏 [ 𝑠 𝑖 ( 𝑡 ) ] 𝐴𝑠𝑠𝑢𝑚𝑖𝑛𝑔 1 Ω𝑙𝑜𝑎𝑑𝑖𝑛𝑔
𝑖
𝑖=1
1.7.2 Bipolar NRZ Signal (i.e. NRZ-L)
❑ 𝑧 1 ( 𝑇 )
𝑇
si(t) ∫
0
+ 𝐻1
Σ
A 𝑧 (𝑇 ) ≷ 𝛾 0
Reference signal
𝑟 ( 𝑡 ) z(T) 𝐻2 𝑠^ 𝑖 ( 𝑡 )
0 t s1(t)=A
0 2T 4T -
𝑇
-A ❑
T 3T 5T ∫ 𝑧 2 ( 𝑇 )
0
{
𝑎1 ( 𝑇 )= 𝐸 [ 𝑧 ( 𝑇 ) ] = 𝐸 𝐴 ∫ [ 𝐴 +𝑛 ( 𝑡 ) ] 𝑑𝑡 − ( − 𝐴 )∫ [ 𝐴 + 𝑛 (𝑡 ) ] 𝑑𝑡
0 0
}
𝑖 .𝑒 . 𝑎1 ( 𝑇 )=2 𝐴 2 𝑇 , 𝑠𝑖𝑚𝑖𝑙𝑎𝑟𝑙𝑦 , 𝑎 2 ( 𝑇 )=− 2 𝐴 2 𝑇
The threshold distance
( 𝐴 )2 𝑇 ( − 𝐴 )2 𝑇
Average energy per bit is; 𝐸𝑏 = [
2
+
2 ] = 𝐴2𝑇
The noise amplitude component at the upper and lower branch respectively;
𝑇 𝑇
∫ 𝐴𝑛 (𝑡 ) 𝑑𝑡 𝑎𝑛𝑑 −∫ 𝐴𝑛 ( 𝑡 ) 𝑑𝑡
0 0 𝑇
Noise component after subtraction; 2∫ 𝐴𝑛 ( 𝑡 ) 𝑑𝑡
0
2
24𝑁0 𝐴 𝑇
𝜎 = 0 ¿ 2 𝑁 0 𝐴2𝑇
2
1 1
The error probability is; Recall 𝑃𝐵 =𝑄 ( 𝑑 / 𝜎 0 ) +𝑄 ( 𝑑 / 𝜎 0 ) =𝑄 ( 𝑑 / 𝜎 0 )
2 2
2
2
1
2
1
𝑃𝐵 =𝑄 ( 𝑑 /𝜎 0 ) +𝑄 ( 𝑑 / 𝜎 0 ) =𝑄 ( 𝑑 /𝜎 0 )=Q
2 (√ (2 𝐴 𝑇 )
2 𝑁0 𝐴 𝑇2 ) (√ )
=𝑄
2 𝐴2𝑇
𝑁0
2𝐴 𝑇2
2 𝐸𝑏
𝑃𝐵 =𝑄 (√ ) (√ )
𝑁0
=𝑄
𝑁0
PB is usually expressed in terms on Eb/N0
(or C/N) for comparative purposes.
In digital/analogue communications, dB
is normally used to describe quantity.
For example,
2 𝐸𝑏
The probability of bit error for bipolar signaling is:
𝑃𝐵 =𝑄 (√ )
𝑁0
𝑏
2
2𝐴 𝑇
The probability of bit error for bipolar signaling is:
𝑃𝐵 =𝑄 (√ ) 𝑁0
𝑏