Digital Communication Theory

COEC 324

Chapter 2
The Matched Filter
 Transmit Pulses in the Noisy Received Signal
𝒈(𝒕) Sampling
& “0100110101”
Data
Threshold

𝒘(𝒕)
AWGN

Assuming, we want to send data in baseband with the sequence "0101100100" coded in non polar Non-return-to-
zero (NRZ) through a certain channel.
Mathematically, a sequence in NRZ code can be described as a sequence of unit pulses or shifted rect functions,
each pulse being weighted by +1 if the bit is "1" and by -1 if the bit is "0".

Where T is the time length of one bit.

T
There is a high power of noise relative to the power of the desired signal (i.e., there is a low signal-to-noise ratio). If
the receiver were to sample this signal at the correct moments, the resulting binary message would possibly belie the
original transmitted one.
 Additive White Noise
▪ The additive noise channel

o White: meaning occupies all frequencies, so PSD is independent on the

▪ White noise operating frequency, White noise is often used to model the thermal noise in
electronic systems

o 𝑁𝑜 𝑖𝑠 watt/Hertz , 𝑁𝑜 = 𝐾𝑇 (𝐾: Bolman const., T: equivalent noise temperature

of the receiver)
𝑆𝑛 (𝑓)
o Its power spectrum density (PSD) is constant over all frequencies, i.e.,

Factor 1/2 is included to indicate that half

the power is associated with positive
𝑓 frequencies and half with negative.
 Additive White Noise

𝑋~𝒩(𝜇, 𝜎 2 )
𝔼 𝑋 = 𝜇 𝑚𝑒𝑎𝑛
2
𝑣𝑎𝑟 𝑋 = 𝜎 𝑣𝑎𝑟𝑎𝑖𝑛𝑐𝑒
2
1 𝑥−𝜇
𝑃𝑥 = 𝑒−
2𝜋𝜎 2 2𝜎 2
 Transmit Pulses in the Noisy Received Signal
𝒈(𝒕) Sampling
& “0100110101”
Data
Threshold

AWGN 𝒘(𝒕)

𝒈(𝒕) Sampling “0101100100”

Data Matched Filter &
Threshold
𝒘(𝒕)
AWGN

The matched filter is the optimal linear filter for maximizing the signal-to-noise ratio (SNR) in the presence of
additive AWGN.
 Matched Filter
▪ Matched Filter: is linear filter for the optimal detection of the digital pulse, the impulse response of the matched
filter matches of the pulse shape.
▪ Matched Filter: a LTI filter chosen to maximize the SNR at its output for a given signal at certain time

𝒕=𝑻
𝒉(𝒕)
𝒈(𝒕) 𝒙(𝒕) 𝒚(𝒕) Sampling 𝒚(𝑻)
Data Matched Filter &
Threshold

𝒘(𝒕)
AWGN

𝑥 𝑡 = 𝑔 𝑡 + 𝑤(𝑡)

𝑦 𝑡 = 𝑥 𝑡 ∗ ℎ(𝑡) ⟹ 𝑦 𝑡 = 𝑔 𝑡 + 𝑤(𝑡) ∗ ℎ(𝑡) ⟹ 𝑦 𝑡 = 𝑔 𝑡 ∗ ℎ 𝑡 + 𝑤(𝑡) ∗ ℎ 𝑡

⟹ 𝑦 𝑡 = 𝑔𝑜 𝑡 + 𝑛(𝑡)
 Matched Filter
𝑦 𝑡 = 𝑠𝑜 𝑡 + 𝑛(𝑡)

We now define the signal-to-noise ratio, which is our objective function, to be the ratio of the power of
the output due to the desired signal to the power of the output due to the noise:

𝑃𝑜𝑤𝑒𝑟𝑠𝑖𝑔𝑛𝑎𝑙 𝑃𝑔𝑜 𝑡
𝑆𝑁𝑅 = =
𝑃𝑜𝑤𝑒𝑟𝑛𝑜𝑖𝑠𝑒 𝑃𝑛 𝑡

2
▪ The power of the signal at 𝑡 = 𝑇: 𝑔𝑜 𝑇
▪ The power of noise:𝐸[𝑛(𝑡)2 ]

𝑔𝑜 𝑇 2
▪ The efficiency of the matched filter is 𝜂 can be expressed as: 𝜂=
𝐸[𝑛(𝑡)2 ]
 Matched Filter
𝑔𝑜 𝑇 2
𝜂=
𝐸[𝑛(𝑡)2 ]

∞
𝑔𝑜 𝑡 = 𝑔 𝑡 ∗ ℎ 𝑡 ⟹ 𝑔𝑜 𝑡 = න 𝐻 𝑓 𝐺 𝑓 𝑒 𝑖2𝜋𝑓𝑡 𝑑𝑓
−∞

▪ At t = T decision time at sample interval

∞ 2

2
𝑔𝑜 𝑇 = න 𝐻 𝑓 𝐺 𝑓 𝑒 𝑖2𝜋𝑓𝑡 𝑑𝑓
−∞

𝑛 𝑡 = 𝑤(𝑡) ∗ ℎ 𝑡
∞ ∞ 2
𝑁𝑜 𝑔𝑜 𝑇 2 ‫׬‬−∞ 𝐻 𝑓 𝐺 𝑓 𝑒 𝑖2𝜋𝑓𝑡 𝑑𝑓
𝐸𝑛 𝑡 2
=
2
න 𝐻 𝑓 2 𝑑𝑓
⟹ 𝜂=
𝐸[𝑛(𝑡)2 ]
=
𝑁𝑜 ∞ 2
−∞ 2 ‫׬‬−∞ 𝐻 𝑓 𝑑𝑓
 Matched Filter
∞ 2
‫׬‬−∞ 𝐻 𝑓 𝐺 𝑓 𝑒 𝑖2𝜋𝑓𝑡 𝑑𝑓
𝜂=
𝑁𝑜 ∞ 2 𝑑𝑓
‫׬‬ 𝐻 𝑓
2 −∞

▪ Schwarz inequality theory:

∞ ∞ ∞ 2 ∞ ∞

න ∅1 (𝑥) 2 𝑑𝑥 < ∞ & න ∅2 (𝑥) 2 𝑑𝑥 < ∞ ⟹ න ∅1 𝑥 . ∅2 (𝑥)𝑑𝑥 ≤ න ∅1 𝑥 2 𝑑𝑥

. න ∅2 (𝑥) 2 𝑑𝑥
−∞ −∞ −∞ −∞ −∞

𝑖𝑓 ∅1 𝑥 = 𝑘. ∅∗2 (𝑥)

∞ 2 𝑑𝑓 ∞
‫׬‬−∞ 𝐺 𝑓 ‫׬‬−∞ 𝐻(𝑓) 2 𝑑𝑓
𝑎𝑠𝑠𝑢𝑚𝑖𝑛𝑔 𝑡ℎ𝑎𝑡: ∅1 𝑥 = 𝐻(𝑓) & ∅2 𝑥 = 𝐺 𝑓 𝑒 𝑖2𝜋𝑓𝑡 ⟹ 𝜂≤
𝑁𝑜 ∞ 2 𝑑𝑓
𝐻 𝑓
∴𝐻 𝑓 = 𝑘. ∅∗2 𝑥 = 𝑘. 𝐺 ∗ 𝑓 𝑒 −𝑖2𝜋𝑓𝑡 2 ‫׬‬−∞
 Matched Filter
∞ ∞
‫׬‬−∞ 𝐺 𝑓 2 𝑑𝑓 ‫׬‬−∞ 𝐻(𝑓) 2 𝑑𝑓
𝜂≤
𝑁𝑜 ∞ 2 𝑑𝑓
‫׬‬ 𝐻 𝑓
2 −∞
∞
2 2 𝑑𝑓
𝜂≤ න 𝐺 𝑓
𝑁𝑜
−∞
2 ∞ 2 𝑑𝑓 2𝐸
𝜂𝑚𝑎𝑥 = ‫׬‬ 𝐺 𝑓 = ℎ𝑜𝑝𝑡 𝑓 = 𝑘𝑔(𝑇 − 𝑡)
𝑁𝑜 −∞ 𝑁𝑜
𝒕=𝑻
𝒉(𝒕)
𝒈(𝒕) 𝒙(𝒕) 𝒚(𝒕) Sampling 𝒚(𝑻)
Matched
𝐻 𝑓 = 𝑘. ∅∗2 𝑥 = 𝑘. 𝐺 ∗ 𝑓 𝑒 −𝑖2𝜋𝑓𝑡 &
Filter
Threshold
⟹ ℎ𝑜𝑝𝑡 𝑓 = 𝑘𝑔(𝑇 − 𝑡) 𝒘(𝒕)
 Matched Filter
ℎ𝑜𝑝𝑡 𝑓 = 𝑘𝑔(𝑇 − 𝑡)

𝒈(𝒕) 𝒉(𝒕)
ℎ𝑜𝑝𝑡 𝑓 = 𝑘𝑔(𝑇 − 𝑡)
𝒕=𝑻
∗ 𝒈(𝒕) 𝒙(𝒕)
𝒉(𝒕)
𝒚(𝒕) Sampling 𝒚(𝑻)
Matched
&
Filter
Threshold
𝟎 𝒕 𝟎 𝒕
𝒘(𝒕)
𝒚(𝒕)

𝟎 𝒕 𝟐𝒕
Intersymbol Interference
Line Coding
 Power Spectral Density
The function which describes how the power of a signal got distributed at various frequencies, in the frequency
domain is called as Power Spectral Density PSD. PSD is the Fourier Transform of Auto-Correlation
Bandlimited Channel

Time limited pulses are always band-unlimited

Bandlimited pulses are always time-unlimited

 Inter-symbol Interference (ISI)
This is a form of distortion of a signal, in which one or more symbols interfere with subsequent signals, causing
noise or delivering a poor output.
•Multi-path Propagation
•Non-linear frequency in channels
The ISI is unwanted and should be completely eliminated to get a clean output. The causes of ISI should also be
resolved in order to lessen its effect.
Nyquist Pulse Shaping
Nyquist Pulse Shaping
Raised Cosine Pulses
Raised Cosine Pulses
 Baseband Communications System
T
Bit Stream Bits to
Samples Filter 𝑝(𝑡) Channel ℎ(𝑡) Filter 𝑟(𝑡)

𝐵𝑒𝑐𝑎𝑢𝑠𝑒 𝑃(𝑡), ℎ(𝑡) 𝑎𝑛𝑑 𝑟(𝑡) 𝑎𝑟𝑒 𝐿𝑇𝐼: 𝑤𝑒 𝑐𝑎𝑛 𝑟𝑒𝑝𝑙𝑎𝑐𝑒 𝑡ℎ𝑒𝑠𝑒 𝑠𝑦𝑠𝑡𝑒𝑚 𝑤𝑖𝑡ℎ 𝑒𝑞𝑢𝑖𝑣𝑎𝑙𝑒𝑛𝑡 𝑓𝑖𝑙𝑡𝑒𝑟 𝑔(𝑡)
T
Bit Stream 𝑎(𝑡) 𝑞(𝑡)
Bits to Filter 𝑔(𝑡)
Samples 𝐺(𝑡) = 𝑝(𝑡) ∗ ℎ(𝑡) ∗ 𝑟(𝑡) 𝑞(𝑖𝑇)

𝑎 𝑡 = ෍ 𝑎𝑘 𝛿(𝑡 − 𝑘𝑇) 𝑞 𝑡 = ෍ 𝑎𝑘 𝛿 𝑡 − 𝑘𝑇 ∗ 𝑔 𝑡 + 𝑛(𝑡) = ෍ 𝑎𝑘 𝑔 𝑡 − 𝑘𝑇 + 𝑛(𝑡)

𝑘 𝑘 𝑘
 Baseband Communications System
T
Bit Stream 𝑎(𝑡) 𝑞(𝑡)
Bits to Filter 𝑔(𝑡)
𝑞 𝑖𝑇 = 𝑎𝑖
Samples 𝐺(𝑡) = 𝑝(𝑡) ∗ ℎ(𝑡) ∗ 𝑟(𝑡)

For what 𝑔(𝑡)?, 𝑞 𝑖𝑇 = 𝑎𝑖

𝑞 𝑡 = ෍ 𝑎𝑘 𝑔 𝑡 − 𝑘𝑇 + 𝑛(𝑡) ⟹ 𝑞 𝑖𝑇 = ෍ 𝑎𝑘 𝑔 𝑖𝑇 − 𝑘𝑇 + 𝑛(𝑖𝑇)
𝑘 𝑘

𝑞 𝑖𝑇 = 𝑎𝑖 𝑔 0 + ෍ 𝑎𝑘 𝑔 𝑖𝑇 − 𝑘𝑇 + 𝑛(𝑖𝑇)
𝑘=−∞
𝑘≠𝑖

our target is 𝑞 𝑖𝑇 = 𝑎𝑖
Noise
ISI
 Baseband Communications System
T
Bit Stream 𝑎(𝑡) 𝑞(𝑡)
Bits to Filter 𝑔(𝑡)
𝑞 𝑖𝑇 = 𝑎𝑖
Samples 𝐺(𝑡) = 𝑝(𝑡) ∗ ℎ(𝑡) ∗ 𝑟(𝑡)

Assume high SNR to ∞

ignore the noise effect 𝑞 𝑖𝑇 = 𝑎𝑖 𝑔 0 + ෍ 𝑎𝑘 𝑔 𝑖𝑇 − 𝑘𝑇
𝑘=−∞
𝑘≠𝑖

𝑤𝑒 𝑤𝑎𝑛𝑡 𝑡𝑜 𝑑𝑒𝑠𝑖𝑔𝑛 𝑖𝑚𝑢𝑙𝑠𝑒 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑜𝑓 𝑔 𝑡 𝑡𝑜 𝑠𝑎𝑡𝑖𝑠𝑓𝑦 𝑡ℎ𝑎𝑡 𝑞 𝑖𝑇 = 𝑎𝑖 Sinc function Nyquist Pulse
1, 𝑡=0
⟹ 𝑔 𝑡 =ቊ
0, 𝑡 = 𝑘𝑇
1, 𝑡=0
𝑔 𝑡 = ቐ𝑠𝑖𝑛 𝑡
, 𝑡≠0
𝑡
 equalizer
▪ Objective: design a receiver that
compensate for the channel (distorting
+AWGN) which is not known a priori to
reduce ISI. This compensation is known as
Equalizer. Equalizers are filters with
adjustable parameters to compensate for
the chancel distortion. Reducing the
complexity is an objective
▪ When the channel impulse
response is unknown but time-
invariant over the time of
transmission. The channel
characteristics may be
measured and used to adjust
the parameters of the equalizer.
(The equalizer coefficients are
fixed over the transmission
period).
Zero Forcing Equalizer
Zero Forcing Equalizer