Baseband Digital Transmission and

Matched Filter

1 Modern Digital Communication Systems (Dr. Muqaibel)

 Outlines and Class Objectives
• Baseband pulse transmission

• Dispersive channels

• Matched filter
– Properties of matched filter

– Matched filter for rectangular pulses

2 Modern Digital Communication Systems (Dr. Muqaibel)

 Baseband Pulse Transmission & ISI
• In transmission of digital data over baseband channel, digital data have significant low
frequency content.
→ Requires: Low-pass channel (baseband) with bandwidth large enough to accommodate the
essential frequency content.
• Baseband channel is dispersive
→Frequency response of channel is not an ideal LPF
→Received pulses will be affected by adjacent pulses
→Inter-Symbol Interference (ISI)
• Dispersive baseband channel is a major source for bit error

• How can we correct or control ISI effect resulting from dispersive channels?
• It can be controlled by pulse shaping (or equalization)

3 Modern Digital Communication Systems (Dr. Muqaibel)

Sources of Error in Baseband Transmission
• Sources of bit error in baseband transmission:
1. ISI
2. Channel noise
• In practice, the effects of ISI and channel noise are simultaneous; however, we will study
their effects separately.
• Let us start with finding the best receiver in the presence of noise.

4 Modern Digital Communication Systems (Dr. Muqaibel)

 Matched Filter
• Optimum detection of pulse (known waveform) in additive white noise.
• Matched filter is linear time-invariant (LTI) where impulse response matches the
pulse signal.
• Here, we assume that major limitation in communication system is the noise.
• Examples of pulses in digital communication: Bit Value 0 1 0 1 1 1 0 0 1

5V
▪ Unipolar (ON-OFF)
• 𝑠1 𝑡 = 𝐴
0V
0 ≤ 𝑡 ≤ 𝑇 (binary 1)
• 𝑠2 𝑡 = 0 0 ≤ 𝑡 ≤ 𝑇 (binary 0)
Bit Value 0 1 0 1 1 1 0 0 1
▪ Polar 5V

• 𝑠1 𝑡 = +𝐴 0 ≤ 𝑡 ≤ 𝑇 (binary 1) 0V

• 𝑠2 𝑡 = −𝐴 0 ≤ 𝑡 ≤ 𝑇 (binary 0)
–5 V

5 Modern Digital Communication Systems (Dr. Muqaibel)

 Matched Filter
Block Diagram
• At the receiver, recovering the shape of
transmitted pulse is not important; but the 𝑥 𝑡
decision of which pulse was transmitted
𝑔(𝑡)
(detection problem). ℎ(𝑡)

• The filter input, 𝑥 𝑡 , is composed of

pulse 𝑔(𝑡) corrupted by additive noise 𝑤 𝑡
𝑤(𝑡).
• Channel LTI filter has impulse response,
ℎ(𝑡).
Objective:
• Additive white noise, 𝑤(𝑡), is zero mean
Design a receiver that detects the pulse 𝑔(𝑡) in an
with PSD of 𝑁0 /2
optimum way given 𝑥(𝑡)
𝑥 𝑡 =𝑔 𝑡 +𝑤 𝑡 , 0≤𝑡≤𝑇
→ Minimizing the effect of noise at the filter output
• 𝑇: arbitrary observation interval.

6 Modern Digital Communication Systems (Dr. Muqaibel)

 Matched Filter
• The filter ℎ 𝑡 is linear Block Diagram
𝑦 𝑡 = 𝑔𝑜 𝑡 + 𝑛(𝑡)
𝑔0 𝑡 = 𝑔(𝑡) ∗ ℎ(𝑡)
𝑥 𝑡
▪ 𝑛 𝑡 : filtered noise
▪ 𝑔𝑜 𝑡 : output signal component 𝑔(𝑡)
ℎ(𝑡)
• After sampling at time 𝑡 = 𝑇
𝑦 𝑇 = 𝑔𝑜 𝑇 + 𝑛 𝑇
𝑤 𝑡
• The objective is to find ℎ(𝑡) that maximizes
the instantaneous power in output signal
𝑔𝑜 (𝑡) (at time 𝑇) compared to the average
power of output noise 𝑛(𝑡)
𝑔𝑜 𝑇 2
𝑔𝑜 𝑇 2 Maximize 𝜂 =
Maximize 𝜂 = 𝐸[𝑛2 (𝑡)]
𝐸[𝑛2 (𝑡)]

7 Modern Digital Communication Systems (Dr. Muqaibel)

 Matched Filter
Block Diagram
• Design ℎ(𝑡) such that
𝑔𝑜 𝑇 2 𝑥 𝑡
Maximize 𝜂 = 𝐸[𝑛2 (𝑡)]
𝐺(𝑓)
• Here, 𝐺0 𝑓 = 𝐺 𝑓 𝐻 𝑓 and 𝐻(𝑓)
∞
𝑔𝑜 𝑡 = න 𝐺 𝑓 𝐻 𝑓 exp 𝑗2𝜋𝑓𝑡 𝑑𝑓
−∞ 𝑤 𝑡
∞ 2
𝑔𝑜 𝑇 2
= න 𝐺 𝑓 𝐻 𝑓 exp 𝑗2𝜋𝑓𝑇 𝑑𝑓 • Hence, substitute in 𝜂
−∞ ∞ 2
‫ 𝑓 𝐻 𝑓 𝐺 ׬‬exp 𝑗2𝜋𝑓𝑇 𝑑𝑓
• PSD of the output noise, 𝜂 = −∞
𝑁0 ∞ 2
2 ‫׬‬−∞ 𝐻 𝑓 𝑑𝑓
𝑁0 2
𝑆𝑁 (𝑓) = 𝐻 𝑓
2 • For a given 𝐺 𝑓 , find the transfer function 𝐻(𝑓) that maximizes 𝜂
∞
𝑁0 ∞
𝐸 𝑛2 𝑡 = න 𝑆𝑁 𝑓 𝑑𝑓 = න 𝐻 𝑓 2
𝑑𝑓 → The solution for this optimization problem involves the application
−∞ 2 −∞ of mathematical result known as Schwarz’s inequality

8 Modern Digital Communication Systems (Dr. Muqaibel)

 Schwarz’s inequality

• For two complex functions 𝜙1 𝑡 and 𝜙2 (𝑡) in the real variable 𝑡 of finite
energy, i.e.
∞ ∞
න 𝜙1 𝑡 2 𝑑𝑡 < ∞, න 𝜙2 𝑡 2 𝑑𝑡 <∞
−∞ −∞
• Then
∞ 2 ∞ ∞
2 𝑑𝑡 න 2 𝑑𝑡
න 𝜙1 𝑡 𝜙2 𝑡 𝑑𝑡 ≤න 𝜙1 𝑡 𝜙2 𝑡
−∞ −∞ −∞
• Equality holds if and only if
𝜙1 𝑡 = 𝑘𝜙2∗ 𝑡
• where 𝑘 is constant.

9 Modern Digital Communication Systems (Dr. Muqaibel)

 𝐻 𝑓 = 𝑘𝐺 ∗ 𝑓 exp(−𝑗2𝜋𝑓𝑇)
Matched Filter 2 ∞
2 𝑑𝑓
𝜂≤ න 𝐺 𝑓
𝑁0 −∞
• Now let 𝜙1 𝑥 = 𝐻 𝑓 , and 𝜙2 𝑥 = 𝐺 𝑓 exp 𝑗2𝜋𝑓𝑇
• Then
∞ 2 ∞ ∞
න 𝐻 𝑓 𝐺 𝑓 exp 𝑗2𝜋𝑓𝑇 𝑑𝑓 ≤න 𝐻 𝑓 2 𝑑𝑓 න 𝐺 𝑓 2 𝑑𝑓
−∞ −∞ −∞
• Here, |exp 𝑗2𝜋𝑓𝑇 | = 1. Equality holds if
𝐻 𝑓 = 𝑘𝐺 ∗ 𝑓 exp(−𝑗2𝜋𝑓𝑇)
• Peak pulse signal-to-noise ratio
∞ 2 ∞ ∞
‫׬‬−∞ 𝐺 𝑓 𝐻 𝑓 exp 𝑗2𝜋𝑓𝑇 𝑑𝑓 ‫׬‬−∞ 𝐻 𝑓 2 𝑑𝑓 ‫׬‬−∞ 𝐺 𝑓 2 𝑑𝑓 2 ∞ 2 𝑑𝑓
𝜂= 𝑁0 ∞ ≤ 𝑁0 ∞ = ‫׬‬ 𝐺 𝑓
‫׬‬ 𝐻 𝑓 2 𝑑𝑓 ‫׬‬ 𝐻 𝑓 2 𝑑𝑓 𝑁0 −∞
2 −∞ 2 −∞

• Note that it does not depend on 𝐻(𝑓), but signal energy and noise PSD.

10 Modern Digital Communication Systems (Dr. Muqaibel)

 Matched Filter
• The maximum of 𝜂 is
∞
2 2 𝑑𝑓
𝜂𝑚𝑎𝑥 = න 𝐺 𝑓
𝑁0 −∞
• when 𝐻𝑜𝑝𝑡 𝑓 = 𝑘𝐺 ∗ 𝑓 exp(−𝑗2𝜋𝑓𝑇). Therefore,
∞
ℎ𝑜𝑝𝑡 𝑡 = 𝑘 න 𝐺 ∗ 𝑓 exp −𝑗2𝜋𝑓𝑇 exp(𝑗2𝜋𝑓𝑡) 𝑑𝑓
−∞
∞
= 𝑘 න 𝐺 ∗ 𝑓 exp −𝑗2𝜋𝑓(𝑇 − 𝑡) 𝑑𝑓
−∞
• For real signals, 𝐺 ∗ 𝑓 = 𝐺 −𝑓
∞
ℎ𝑜𝑝𝑡 𝑡 = 𝑘 න 𝐺 −𝑓 exp −𝑗2𝜋𝑓(𝑇 − 𝑡) 𝑑𝑓
−∞

11 Modern Digital Communication Systems (Dr. Muqaibel)

 ℎ𝑜𝑝𝑡 𝑡 = 𝑘𝑔(𝑇 − 𝑡)
Matched Filter 2𝐸
𝜂𝑚𝑎𝑥 =
𝑁0
∞
ℎ𝑜𝑝𝑡 𝑡 = 𝑘 න 𝐺 −𝑓 exp −𝑗2𝜋𝑓(𝑇 − 𝑡) 𝑑𝑓
−∞
= 𝑘𝑔(𝑇 − 𝑡)
• which is the time-reversed delayed version of the input signal 𝑔(𝑡), that is matched
to the input signal.
• The assumption that is made to determine the matched filter is that noise is
𝑁
stationary, white, with zero mean and PSD 0 .
2
• The filter ℎ𝑜𝑝𝑡 𝑡 = 𝑘𝑔 𝑇 − 𝑡 will result in maximum signal to noise ratio, given by
∞
2 2
2𝐸
𝜂𝑚𝑎𝑥 = න 𝐺 𝑓 𝑑𝑓 =
𝑁0 −∞ 𝑁0
𝐸: the energy of pulse signal 𝑔(𝑡)

12 Modern Digital Communication Systems (Dr. Muqaibel)

 Matched Filter Output
• To show that, the spectrum of signal 𝑔𝑜 (𝑡)
𝐺𝑜 𝑓 = 𝐻𝑜𝑝𝑡 𝑓 𝐺(𝑓)
= 𝑘𝐺 ∗ 𝑓 exp(−𝑗2𝜋𝑓𝑇)𝐺(𝑓)
= 𝑘 𝐺 𝑓 2 exp(−𝑗2𝜋𝑓𝑇)
• Matched filter output at 𝑡 = 𝑇 (inverse Fourier transform),
∞
𝑔𝑜 𝑇 = න 𝐺𝑜 𝑓 exp 𝑗2𝜋𝑓𝑡 𝑑𝑓
−∞
∞ 2
= 𝑘 ‫׬‬−∞ 𝐺 𝑓 𝑑𝑓
∞ ∞
𝐸= ‫׬‬−∞ 𝑔2 𝑡 𝑑𝑡 = ‫׬‬−∞ |𝐺 𝑓 |2 𝑑𝑓 (Rayleigh energy theorem)

13 Modern Digital Communication Systems (Dr. Muqaibel)

 Matched Filter Maximum SNR
𝑔𝑜 𝑇 = 𝑘𝐸
• As for the average power of output noise
2
𝑁0 ∞ 2
𝐸𝑛 𝑡 = න 𝐻 𝑓 𝑑𝑓
2 −∞ 𝑜𝑝𝑡
𝑘 2 𝑁0 ∞ 2
= න 𝐺 𝑓 𝑑𝑓
2 −∞
𝑘 2 𝑁0 𝐸
=
2
• Therefore,
𝑘𝐸 2 2𝐸
𝜂𝑚𝑎𝑥 = 2 =
(𝑘 𝑁0 𝐸/2) 𝑁0
• Dependence on 𝒈(𝒕) has been removed by the matched filter. All signals with same energy
are equally effective.

14 Modern Digital Communication Systems (Dr. Muqaibel)

Example 1:Matched Filter
• Matched filter for a rectangular pulse, 𝑔(𝑡)
𝑔𝑜 (𝑡)
𝑔(𝑡)
Energy = 𝐴2 𝑇 𝑘𝐴2 𝑇
𝐴

𝑡 𝑡
𝑇 𝑇
Matched filter output, 𝑔𝑜 𝑡
ℎ 𝑡 = 𝑘𝑔(𝑇 − 𝑡) = 𝑘𝑔(−(𝑡 − 𝑇))

𝑘𝐴 Maximum value of output signal 𝑔𝑜 (𝑡) is 𝑘𝐴2 𝑇, which occurs at 𝑡 = 𝑇.

Maximum of output signal equals the energy of the input signal 𝑔(𝑡)
𝑡 scaled by 𝑘.
𝑇

15 Modern Digital Communication Systems (Dr. Muqaibel)

Example 2: Integrate-and-dump (Special Case)
For the special case of 𝑔 𝑡 is a rectangular pulse, matched filter is
implemented using integrate-and-dump circuit.
Rectangular
pulse Integrator
Sample at
𝑡=𝑇

Output of integrate-and-dump

𝐴𝑇

𝑡
𝑇
16 Modern Digital Communication Systems (Dr. Muqaibel)
 Example 3: Matched Filter
• Design a filter matched to the signal s(𝑡). 𝑠𝑜 (𝑡)

𝑠(𝑡) ℎ(𝑡) 𝐴2 𝑇/4

𝐴/2 𝐴/2

𝑡 𝑡 𝑡
𝑇 𝑇 𝑇 𝑇 𝑇 𝑇 3𝑇 2𝑇
2 2 2 2
−𝐴/2 −𝐴/2 −𝐴2 𝑇/8

Impulse response for the matched Output of the matched filter, 𝑘 = 1,𝑠𝑜 (𝑡)
Energy of 𝑠(𝑡) = ? filter for 𝑘 = 1, ℎ 𝑡

17 Modern Digital Communication Systems (Dr. Muqaibel)

Example 3: Matched Filter
• Design a filter matched to the signal s(𝑡). 𝑠𝑜 (𝑡)

ℎ(𝑡) 𝐴2 𝑇/4

𝐴/2

𝑡 𝑇 𝑡
𝑇 𝑇 𝑇 3𝑇 2𝑇
2 2 2
−𝐴/2 −𝐴2 𝑇/8

Impulse response for the matched Output of the matched filter, 𝑘 = 1,𝑠𝑜 (𝑡)
Energy of 𝑠(𝑡) = ? filter for 𝑘 = 1, ℎ 𝑡

18 Modern Digital Communication Systems (Dr. Muqaibel)

Tabatabaei, Nima. 2018. "Matched-Filter Thermography" Applied Sciences 8, no. 4: 581. https://doi.org/10.3390/app8040581
 Block Diagram
Summary
𝑥 𝑡
𝑔(𝑡)
ℎ(𝑡)

ℎ𝑜𝑝𝑡 𝑡 = 𝑘𝑔(𝑇 − 𝑡)
𝑤 𝑡
𝐻 𝑓 = 𝑘𝐺 ∗ 𝑓 exp(−𝑗2𝜋𝑓𝑇)
∞
2 2 𝑑𝑓
2𝐸
𝜂≤ න 𝐺 𝑓 𝜂𝑚𝑎𝑥 = For the special case of 𝑔 𝑡 is a rectangular pulse, matched filter is
𝑁0 −∞ 𝑁0 implemented using integrate-and-dump circuit.

Performance Does not depend on the exact shape of

the pulse but on the energy content Rectangular
pulse Integrator
Sample at
𝑡=𝑇

𝐴𝑇

𝑡
𝑇
Output of integrate-and-dump
Error Rate Analysis
(Baseband)

Modern Digital Communication Systems (Dr. Muqaibel) 1

 Outlines and Class Objectives
• Error rate due to noise

• Optimal threshold

• Leibniz’s rule

2 Modern Digital Communication Systems (Dr. Muqaibel)

 Error Rate Due to Noise
• For signaling interval 0 ≤ 𝑡 ≤ 𝑇𝑏 , the received signal is
+𝐴 + 𝑤 𝑡 , symbol 1 was sent
𝑥 𝑡 =ቊ
−𝐴 + 𝑤 𝑡 , symbol 0 was sent
𝑇𝑏 : bit duration
𝐴: transmitted pulse amplitude
𝑤 𝑡 : additive white Gaussian noise of zero mean and PSD 𝑁0 /2
• The pulse shape is assumed to be known at receiver, but not the polarity.
(Polar non-return-to-zero (NRZ) signaling)
• The start and end of each transmitted pulse are assumed to be known at
receiver (acquired).

3 Modern Digital Communication Systems (Dr. Muqaibel)

 Error Rate Due to Noise
• There are two possible kinds of errors:
1. Symbol 1 is selected when a 0 was transmitted (error of first kind)
2. Symbol 0 is selected when a 1 was transmitted (error of second kind)
• Average probability of symbol error, 𝑃𝑒 , in the receiver is
𝑃𝑒 = 𝑝0 𝑃𝑒|0 + 𝑝1 𝑃𝑒|1
𝑝0 , 𝑝1 : a priori probabilities of binary symbols 0 and 1, respectively.
𝑃𝑒|0 , 𝑃𝑒|1 : conditional error probabilities given that symbols 0 and 1 were transmitted, respectively.
𝐴𝑜 1 − 𝑃𝑒|0
𝐵0
𝑃𝑒|0

𝑃𝑒|1 1 − 𝑃𝑒|1
𝐴1 𝐵1
4 Modern Digital Communication Systems (Dr. Muqaibel)
 Receiver Model for Error Rate Analysis
• The receiver that performs decision making consists of matched filter, sampler, and decision device.
• Sample value 𝑦 is compared to a threshold 𝜆 in decision device.
• What if the sample 𝒚 = 𝝀? is it going to change average error probability?
• If 𝒚 = 𝝀, the receive makes a guess in selecting 0 or 1. This will not change the average error
probability.

𝑦 Say 1 if y > λ

Say 0 if y < λ

λ
Block Diagram

5 Modern Digital Communication Systems (Dr. Muqaibel)

 ℎ 𝑡 = 𝑘𝑔(𝑇𝑏 − 𝑡) = 𝑘𝑔(−(𝑡 − 𝑇𝑏 ))

Error Rate Due to Noise 𝑘𝐴

𝑡
𝑇𝑏

𝑇𝑏
• Suppose that symbol 0 was sent, then noisy received signal is 𝑦 = 𝑘𝐴 න 𝑥 𝑡 𝑑𝑡
𝑥 𝑡 = −𝐴 + 𝑤 𝑡 , 0 ≤ 𝑡 ≤ 𝑇𝑏 0

• Matched filter output after sampling at 𝑡 = 𝑇𝑏 , (assume 𝑘𝐴𝑇𝑏

= 1 for convenience) is 𝑓𝑌 𝑦 0
𝑘𝐴
−𝑘𝐴2 𝑇𝑏
1 𝑇𝑏 𝑃𝑒|0
𝑦 = −𝐴 + න 𝑤 𝑡 𝑑𝑡
𝑇𝑏 0
• Characteristics of the random variable 𝑦: −𝐴 λ
1. Random variable 𝑦 is Gaussian distributed with mean −𝐴.
2. Variance of the random variable 𝑦
2 2
𝜎𝑌 = 𝐸 𝑌 − −𝐴 = 𝐸[ 𝑌 + 𝐴 2 ]

6 Modern Digital Communication Systems (Dr. Muqaibel)

 𝑁0
Variance of the Output Noise 𝜎𝑌2 =
2𝑇𝑏

Recall that
2 1 𝑇𝑏 𝑇𝑏 1 𝑇𝑏 𝑇𝑏 𝑇𝑏
𝜎𝑌 = 2 න න 𝐸 𝑤 𝑡 𝑤 𝑢 𝑑𝑡 𝑑𝑢 = 2 න න 𝑅𝑤 𝑡, 𝑢 𝑑𝑡 𝑑𝑢 𝑦 = 𝑘𝐴 න 𝑥 𝑡 𝑑𝑡
𝑇𝑏 0 0 𝑇𝑏 0 0
0
𝑅𝑤 𝑡, 𝑢 : autocorrelation of the white noise And because
𝑘𝐴𝑇𝑏 = 1
• Since 𝑤(𝑡) is white with PSD 𝑁0 /2, then 1 𝑇𝑏
𝑦 = න 𝑥 𝑡 𝑑𝑡
𝑁0 𝑇𝑏 0
𝑅𝑤 𝑡, 𝑢 = 𝛿(𝑡 − 𝑢)
2
• Therefore, the variance of 𝑌 is

1 𝑇𝑏 𝑇𝑏 𝑁0
𝜎𝑌2 = 2න න 𝛿(𝑡 − 𝑢) 𝑑𝑡 𝑑𝑢
𝑇𝑏 0 0 2
𝑁0
𝜎𝑌2 =
2𝑇𝑏

7 Modern Digital Communication Systems (Dr. Muqaibel)

 Error Rate Due to Noise
• The probability density function of 𝑌 given 0 was sent is Gaussian with mean – 𝐴 and
𝑁
variance 0
2𝑇𝑏
1 𝑦−𝜇 2 1 𝑦+𝐴 2
𝑓𝑌 𝑦 0 = 𝑒𝑥𝑝 − = exp −
2𝜋𝜎 2 2𝜎 2 𝜋𝑁0 /𝑇𝑏 𝑁0 /𝑇𝑏

𝑓𝑌 𝑦 0

𝑃𝑒|0

−𝐴 λ

8 Modern Digital Communication Systems (Dr. Muqaibel)

Error Rate Due to Noise 𝑓𝑌 𝑦 0

• The conditional probability of error given that symbol 0 was

transmitted is given by
∞
𝑃𝑒|0
𝑃𝑒|0 = 𝑃 𝑦 > 0|symbol 0 was sent = න 𝑓𝑌 𝑦 0 𝑑𝑦
0

1 ∞ 𝑦+𝐴 2 λ
= ‫׬‬ exp − 𝑑𝑦 −𝐴
𝜋𝑁0 /𝑇𝑏 0 𝑁0 /𝑇𝑏

• In the analysis, the threshold 𝜆 was assigned the value 0

based on the equiprobable symbols 0 and 1, i.e. 𝑝0 = 𝑝1
= 0.5

• Above integral has no closed-form solution, we evaluate it by

using the so called complementary error function

2 ∞
erfc(𝑢)= න exp −𝑡 2 𝑑𝑡
𝜋 𝑢

9 Modern Digital Communication Systems (Dr. Muqaibel)

 Complementary Error & Q Functions
• The complementary error function is closely related to Gaussian distribution.
• For large positive values of 𝑢, we have the following upper bound on complementary error function
exp(−𝑢2 )
erfc(𝑢) <
𝜋𝑢
• Relation to 𝑄-function:
2 ∞
erfc(𝑢)= න exp −𝑡 2 𝑑𝑡
𝜋 𝑢
∞
1 𝑥2
𝑄 𝑣 = න exp − 𝑑𝑥
2𝜋 𝑣 2
1 𝑣
𝑄 𝑣 = erfc( )
2 2
erfc 𝑢 = 2𝑄( 2 𝑢)

10 Modern Digital Communication Systems (Dr. Muqaibel)

 2 ∞
Error Rate Due to Noise erfc(𝑢)= න exp −𝑡 2 𝑑𝑡
𝜋 𝑢

1 ∞ 𝑦+𝐴 2
• We need to rewrite 𝑃𝑒|0 = ‫׬‬ exp − 𝑑𝑦 in terms of the erfc(𝑢)
𝜋𝑁0 /𝑇𝑏 0 𝑁0 /𝑇𝑏

𝑦+𝐴
• Define a new variable 𝑧 =
𝑁0 /𝑇𝑏

• Therefore,
∞
1 𝑦+𝐴 2 1 ∞
𝑃𝑒|0 = න exp − 𝑑𝑦 = න exp −𝑧 2 𝑑𝑧
𝜋𝑁0 /𝑇𝑏 0 𝑁0 /𝑇𝑏 𝜋 𝐸𝑏 /𝑁0

• where 𝐸𝑏 is the transmitted signal energy per bit defined by 𝐸𝑏 = 𝐴2 𝑇𝑏

1
𝑃𝑒|0 = erfc 𝐸𝑏 /𝑁0
2
• Similar analysis can be carried if symbol 1 was transmitted and conditional probability of error is
1
𝑃𝑒|1 = erfc 𝐸𝑏 /𝑁0
2

11 Modern Digital Communication Systems (Dr. Muqaibel)

 Error Rate Due to Noise
• The conditional probability of error was defined as
𝑃𝑒|1 = 𝑃 𝑦 < 0|symbol 1 was sent Water Fall
𝑓𝑌 𝑦 1

𝑃𝑒|1

λ +𝐴
• Note that 𝜆 = 0. Since 𝑝0 = 𝑝1 = 0.5, and 𝑃𝑒|0 = 𝑃𝑒|1 , then the average
probability of symbol error 𝑃𝑒 at receiver is
1 The average probability of symbol error in a binary-
𝑃𝑒 = 𝑝0 𝑃𝑒|0 + 𝑝1 𝑃𝑒|1 = erfc( 𝐸𝑏 /𝑁0 )
2 encoded PCM receiver depends entirely on 𝐸𝑏 /𝑁0

Exercise: Use MATLAB to generate this plot

12 Modern Digital Communication Systems (Dr. Muqaibel)
 2 ∞
Finding the Optimal Threshold erfc(𝑢)= න exp −𝑡 2 𝑑𝑡
𝜋 𝑢

• In our previous analysis, 𝜆 was assigned the value 0 (why?)

• To show that 𝜆 = 0, we will evaluate the conditional probability of error
(assuming 0 was sent) as
∞
1 𝑦+𝐴 2
𝑃𝑒|0 = න exp − 𝑑𝑦
𝜋𝑁0 /𝑇𝑏 𝜆 𝑁0 /𝑇𝑏
𝑦+𝐴 𝑑𝑦
• Substituting 𝑧 = , then 𝑑𝑧 = , and
𝑁0 /𝑇𝑏 𝑁0 /𝑇𝑏

1 ∞ 1 𝐴+𝜆
𝑃𝑒|0 = න exp −𝑧 2 𝑑𝑧 = erfc
𝜋 (𝐴+𝜆)/ 𝑁0 /𝑇𝑏 2 𝑁0 /𝑇𝑏

13 Modern Digital Communication Systems (Dr. Muqaibel)

 Finding the Optimal Threshold
• Similarly, assuming 1 was transmitted

1 ∞ 1 2
𝐴−𝜆
𝑃𝑒|1 = න exp −𝑧 𝑑𝑧 = erfc
𝜋 (𝐴−𝜆)/ 𝑁0 /𝑇𝑏 2 𝑁0 /𝑇𝑏
• Average probability of symbol error is
𝑝0 𝐴+𝜆 𝑝1 𝐴−𝜆
𝑃𝑒 = erfc + erfc
2 𝑁0 /𝑇𝑏 2 𝑁0 /𝑇𝑏
• Clearly, 𝑃𝑒 is function of the threshold 𝜆, we need the optimal threshold to minimize 𝑃𝑒
• → Leibniz’s Rule

14 Modern Digital Communication Systems (Dr. Muqaibel)

 Leibniz’s Rule
• Consider the integral
𝑏(𝑢)
න 𝑓 𝑧, 𝑢 𝑑𝑧
𝑎(𝑢)
• Then, the derivative of this integral with respect to 𝑢 is
𝑑 𝑏(𝑢) 𝑑𝑏 𝑢 𝑑𝑎 𝑢 𝑏(𝑢) 𝜕𝑓 𝑧,𝑢
‫׬‬ 𝑓 𝑧, 𝑢 𝑑𝑧 = 𝑓 𝑏 𝑢 , 𝑢 −𝑓 𝑎 𝑢 ,𝑢 + ‫)𝑢(𝑎׬‬ 𝑑𝑧
𝑑𝑢 𝑎(𝑢) 𝑑𝑢 𝑑𝑢 𝜕𝑢
2 ∞
• From complementary error function , erfc(𝑢)= ‫ ׬‬exp −𝑡 2 𝑑𝑡
𝜋 𝑢
2
𝑓 𝑧, 𝑢 = exp(−𝑧 2 )
𝜋
𝑎 𝑢 = 𝑢, 𝑏 𝑢 =∞

15 Modern Digital Communication Systems (Dr. Muqaibel)

 𝑁0 𝑝0
Finding the Optimal Threshold 𝜆𝑜𝑝𝑡 =
4𝐴𝑇𝑏
𝑙𝑜𝑔
𝑝1

𝑝0 𝐴+𝜆 𝑝1 𝐴−𝜆
• Applying Leibniz’s rule to complementary error function 𝑃𝑒 = erfc
2
+ erfc
2
𝑁0 /𝑇𝑏 𝑁0 /𝑇𝑏
𝑑 2
erfc 𝑢 =− exp(−𝑢2 )
𝑑𝑢 𝜋
• Differentiate the expression for 𝑃𝑒 with respect to 𝜆 then set the
result to zero. After simplifications, the optimum threshold is
𝑁0 𝑝0
𝜆𝑜𝑝𝑡 = 𝑙𝑜𝑔
4𝐴𝑇𝑏 𝑝1
• For the case, 𝑝0 = 𝑝1 = 0.5
→ 𝝀𝒐𝒑𝒕 = 𝟎
• Average probability of symbol error for PCM receiver is upper
bounded as
exp(−𝐸𝑏 /𝑁0 )
𝑃𝑒 <
2 𝜋𝐸𝑏 /𝑁0

16 Modern Digital Communication Systems (Dr. Muqaibel)

 Matlab Simulation (Optimal Threshold)
𝐸𝑏 /𝑁0 = 5𝑑𝐵 𝑁0 𝑝0
𝑝0 = 0.1 𝜆𝑜𝑝𝑡 = 𝑙𝑜𝑔
4𝐴𝑇𝑏 𝑝1
𝜆𝑜𝑝𝑡 = −0.1737

𝜆𝑜𝑝𝑡

𝜆𝑜𝑝𝑡
Check proj3.m
17 Modern Digital Communication Systems (Dr. Muqaibel)
 Nyquist Criteria for Zero ISI

1 Modern Digital Communication Systems: Dr. Muqaibel

 Outlines
• Inter-symbol interference

• Nyquist criterion

• Pulse shaping

2 Modern Digital Communication Systems: Dr. Muqaibel

 Intersymbol Interference
• Given a pulse shape, how can it be used to transmit data in M-ary form?
• Using discrete pulse modulation, and varying amplitude, duration, or position
of pulses in discrete manner.
• Intersymbol interference (ISI) occurs when communication channel is dispersive,
which leads to bit error in baseband pulse transmission.
• An efficient scheme for baseband transmission is pulse amplitude modulation (PAM)
• Binary sequence {𝑏𝑘 } → sequence of pulses with amplitudes 𝑎𝑘

+1, → 𝑏𝑘 = 1
𝑎𝑘 = ቊ
−1, → 𝑏𝑘 = 0

3 Modern Digital Communication Systems: Dr. Muqaibel

 Intersymbol Interference
Input {𝑎𝑘 }
binary
data
{𝑏𝑘 }

Block Diagram
• Incoming binary sequence {𝑏𝑘 } of symbols 0 and 1
• Pulse amplitude modulator transforms this binary sequence into a new sequence of
short pulses with amplitude 𝑎𝑘 is represented in the polar form

+1, if the symbol 𝑏𝑘 is 1

𝑎𝑘 = ൞
−1, if the symbol 𝑏𝑘 is 0
4 Modern Digital Communication Systems: Dr. Muqaibel
 𝑛(𝑡) is the noise produced at the output of

Block Diagram for Analyzing ISI the receiver filter, 𝑐 𝑡 , due to additive
white Gaussian noise 𝑤(𝑡) of zero mean
and PSD of 𝑁0 /2 at receiver input.
Transmitted signal 𝑠(𝑡) in terms of the transmit filter 𝑔(𝑡) is ∞
∞
𝑦 𝑡 = ෍ 𝑎𝑘 𝑝 𝑡 − 𝑘𝑇𝑏 + 𝑛(𝑡)
𝑠 𝑡 = ෍ 𝑎𝑘 𝑔(𝑡 − 𝑘𝑇𝑏 ) 𝑘=−∞
𝑘=−∞

Input {𝑎𝑘 } 𝑠(𝑡) 𝑦(𝑡) 𝑦(𝑡𝑖 )

binary
data
{𝑏𝑘 }

∞
where 𝑝 𝑡 = 𝑔 𝑡 ⊛ ℎ(𝑡) ⊛ 𝑐(𝑡).
𝑦 𝑖𝑇𝑏 = ෍ 𝑎𝑘 𝑝 𝑖𝑇𝑏 − 𝑘𝑇𝑏 + 𝑛[𝑖𝑇𝑏 ]
ℎ 𝑡 : impulse response of the channel
𝑘=−∞
𝑐 𝑡 : impulse response of the receive filter
In frequency domain: 𝑃 𝑓 = 𝐺 𝑓 𝐻 𝑓 𝐶(𝑓)
5 Modern Digital Communication Systems: Dr. Muqaibel
 Intersymbol Interference
• Receiver filter output 𝑦(𝑡) is sampled at time, 𝑡𝑖 = 𝑖𝑇𝑏 resulting in
∞ ∞

𝑦 𝑖𝑇𝑏 = ෍ 𝑎𝑘 𝑝 𝑖𝑇𝑏 − 𝑘𝑇𝑏 + 𝑛[𝑖𝑇𝑏 ] = 𝑎𝑖 + ෍ 𝑎𝑘 𝑝 𝑖𝑇𝑏 − 𝑘𝑇𝑏 + 𝑛[𝑖𝑇𝑏 ]

𝑘=−∞ 𝑘=−∞
𝑘≠𝑖
ISI
𝑎𝑖 is contribution of 𝑖𝑡ℎ transmitted bit
• The unavoidable presence of ISI and noise in the system introduces
errors in the decision device at the receiver output.

6 Modern Digital Communication Systems: Dr. Muqaibel

 Intersymbol Interference
• In the design of transmit and receive filters, the objective is to minimize the effects
of noise and ISI

→ Deliver digital data to the destination with least error rate (as possible).

• If the SNR is high the operation of the communication system is limited by ISI rather
than noise, e.g. in telephone systems.

→ Noise term, 𝑛(𝑖𝑇𝑏 ), can be neglected for high SNR case.

• Techniques to control ISI in high SNR include determining the pulse waveform, 𝑝(𝑡),
for which ISI is eliminated.

7 Modern Digital Communication Systems: Dr. Muqaibel

 Nyquist’s Criterion for Distortionless Baseband
Transmission
• Transfer function of channel and transmitted pulse shape are
specified, and we need to determine the transfer functions of
transmit and receive filters to reconstruct original data {𝑏𝑘 }
• Control of the overall pulse 𝑝(𝑡)
1 ,𝑖 = 𝑘
𝑝 𝑖𝑇𝑏 − 𝑘𝑇𝑏 =ቊ (1)
0 ,𝑖 ≠ 𝑘
• If 𝑝(𝑡) satisfies the above condition, then receiver output
𝑦 𝑖𝑇𝑏 = 𝑎𝑖 for all 𝑖
1

→Zero ISI -2 -1
0
0 1 2 3 4

-1

8 Modern Digital Communication Systems: Dr. Muqaibel

 Derivation of Nyquist’s Criterion
• Consider the sequence of samples 𝑝 𝑛𝑇𝑏 , 𝑛 = 0, ±1, ±2, ±3, … and recall the sampling in time results in periodicity in
frequency domain.
• Fourier transform of infinite periodic sequence of pulses
∞
1
𝑃𝛿 𝑓 = 𝑅𝑏 ෍ 𝑃(𝑓 − 𝑛𝑅𝑏 ) 𝑅𝑏 : bit rate, 𝑅𝑏 = 𝑇
𝑛=−∞ 𝑏

∞ ∞

𝑃𝛿 𝑓 = ℱ[𝑝𝛿 𝑡 ] = න ෍ [𝑝 𝑚𝑇𝑏 𝛿(𝑡 − 𝑚𝑇𝑏 )] exp −𝑗2𝜋𝑓𝑡 𝑑𝑡

−∞ 𝑚=−∞

1 ,𝑖 = 𝑘
• For zero ISI and based on (1), 𝑝 𝑖𝑇𝑏 − 𝑘𝑇𝑏 = ቊ (1)
0 ,𝑖 ≠ 𝑘
▪ Remember that for 𝑚 = 𝑖 − 𝑘, if 𝑚 = 0 ⇒ 𝑖 = 𝑘 and likewise 𝑚 ≠ 0 ⇒ 𝑖 ≠ 𝑘
∞
𝑃𝛿 𝑓 = න 𝑝 0 𝛿(𝑡) exp −𝑗2𝜋𝑓𝑡 𝑑𝑡 = 𝑝 0 = 1
−∞
▪ Here, 𝑝(𝑡) is normalized, i.e. 𝑝 0 = 1.

9 Modern Digital Communication Systems: Dr. Muqaibel

 ∞
𝑛
Nyquist’s Criterion ෍ 𝑃 𝑓−
𝑛=−∞
𝑇𝑏
= 𝑇𝑏

The condition for zero ISI is

∞

𝑅𝑏 ෍ 𝑃 𝑓 − 𝑛𝑅𝑏 = 1
𝑛=−∞
∞
1 𝑛
෍ 𝑃 𝑓− =1
𝑇𝑏 𝑇𝑏
𝑛=−∞
𝑛
σ∞
𝑛=−∞ 𝑃 𝑓 − 𝑇 = 𝑇𝑏 (2)
𝑏

• Nyquist Criterion
The necessary and sufficient condition for 𝑝(𝑡) to satisfy (1) is that its Fourier transform 𝑃(𝑓) satisfies
∞
𝑛
෍ 𝑃 𝑓− = 𝑇𝑏
𝑇𝑏
𝑛=−∞

10 Modern Digital Communication Systems: Dr. Muqaibel

Nyquist’s Criterion
• Recall that
𝑃 𝑓 =𝐺 𝑓 𝐻 𝑓 𝐶 𝑓

• and let 𝑊 be the bandwidth of 𝑃(𝑓). Then,

𝑃 𝑓 = 0 for all 𝑓 > 𝑊
• There are two cases:
1. Case 1
1 𝑅𝑏
𝑊= = or 𝑅𝑏 = 2𝑊, the simplest way to satisfy (2) is to specify
2𝑇𝑏 2
the frequency 𝑃(𝑓) to be in the form of a rectangular function:

1
= 𝑇𝑏 , −𝑊 < 𝑓 < 𝑊 1 𝑓
𝑃 𝑓 = ቐ2𝑊 = rect (3)
2𝑊 2𝑊
0, 𝑓 >𝑊

𝑝 𝑡 = sinc(2𝑊𝑡) (4)

11 Modern Digital Communication Systems: Dr. Muqaibel

 Nyquist’s Criterion
• The special case of bit rate 𝑅𝑏 = 2𝑊 is called Nyquist rate, and
𝑊 is called Nyquist bandwidth Example:
For binary sequence 1011010
• Ideal baseband pulse transmission in (3) [frequency domain] or
in (4) [time-domain] is called ideal Nyquist channel

• Ideal Nyquist achieve economy in bandwidth (it solves the

problem of zero ISI with the minimum possible bandwidth)

• There are two practical difficulties associated with sinc 2𝑊𝑡 :

1. 𝑃(𝑓) is physically unrealizable (abrupt transition at ±𝑊)    

2. sinc(2𝑊𝑡) decreases as 1/|𝑡| for large |𝑡|, resulting in a
slow rate of decay. No margin of error in sampling times.

12 Modern Digital Communication Systems: Dr. Muqaibel

 Raised Cosine
2. Case 2
• For 𝑅𝑏 < 2𝐵𝑇 , we may overcome the practical
difficulties encountered with ideal Nyquist
channel by extending the bandwidth from 𝑊 to
an adjustable value 𝑊 < 𝐵𝑇 ≤ 2𝑊
• Transmission bandwidth 𝐵𝑇 = 𝑊 1 + 𝛼 ,
• 𝛼 is called the rolloff factor defined as
𝑓1
𝛼 =1−
𝑊
• Particular form of 𝑃(𝑓) with many desirable
features provided by a raised cosine spectrum.

13 Modern Digital Communication Systems: Dr. Muqaibel

 𝑅𝑏
0 ≤ 𝑓𝑥 ≤
Raised Cosine 2

P(f) fx fx
f1
1
= 𝑇𝑏 , 0 ≤ 𝑓 < 𝑓1
2𝑊
𝑃 𝑓 = 1 𝜋( 𝑓 − 𝑊)
1 − 𝑠𝑖𝑛 𝑓1 ≤ 𝑓 < 2𝑊 − 𝑓1
4W 2𝑊 − 2𝑓1 f
0 𝑓 ≥ 2𝑊 − 𝑓1
Rb/2
• Remember that 𝐵𝑇 = 𝑊 1 + 𝛼 = 2𝑊 − 𝑓1 and
Rb/2 + fx
1
𝑃 𝑓 + 𝑃 𝑓 − 2𝑊 + 𝑃 𝑓 + 2𝑊 = , |𝑓| ≤ 𝑊 where 𝑓𝑥 is the excess bandwidth. It defines
2𝑊
how much bandwidth required above the
minimum bandwidth of a sinc pulse, where

14 Modern Digital Communication Systems: Dr. Muqaibel

 Raised Cosine : full roll-off factor
The special case with 𝛼 = 1 (𝑓1 = 0) is known as the full-cosine rolloff characteristic, and 𝑃(𝑓) becomes
1 𝜋𝑓 P(f) fx fx
1 + cos( ) , 0 < 𝑓 < 2𝑊
𝑃 𝑓 = ൞4𝑊 2𝑊
0, |𝑓| ≥ 2𝑊
f
And Rb/2

Rb/2 – fx
sinc(4𝑊𝑡) Rb/2 + fx
𝑝 𝑡 =
1−16𝑊 2 𝑡 2
In addition to faster decay, this time response exhibits two interesting properties:
𝑇𝑏 1
1. At 𝑡 = ± =± , we have 𝑝 𝑡 = 0.5
2 4𝑊
3𝑇𝑏 5𝑇
2. There are zero crossings at 𝑡 = ± , ± 𝑏,… in addition to the usual zero crossings at the sampling
2 2
times 𝑡 = ±𝑇𝑏 , ±2𝑇𝑏 , …

These two properties are extremely useful in extracting a timing signal from the received signal for synchronization.

→The price is the use of a channel bandwidth double of ideal Nyquist channel (𝛼 = 0)

15 Modern Digital Communication Systems: Dr. Muqaibel

 Extremes of Raised-Cosine Spectra
P(f) fx = Rb/2
fx = 0
“Sinc”
𝛼 is the roll-off factor
Rb/2 Rb/2

Designing for a specific pulse shape

might have practical limitations,
especially that the designed pulse
f should be at the receiver side.
Rb/2 Rb
=1/2Tb =1/Tb

𝐸𝑥𝑐𝑒𝑠𝑠 𝐵𝑎𝑛𝑑𝑤𝑖𝑑𝑡ℎ 𝑓𝑥 2𝑓𝑥 For baseband transmission

𝛼= = =
𝑀𝑖𝑛𝑖𝑚𝑢𝑚 𝐵𝑎𝑛𝑑𝑤𝑖𝑑𝑡ℎ 𝑅𝑏 𝑅𝑏 𝐵𝑊 =
𝑅𝑏
+𝛼
𝑅𝑏
2 2 2
16 Modern Digital Communication Systems: Dr. Muqaibel
 Raised-Cosine Pulses
1.5

0.5

Amplitude
0

-0.5

𝛽 = 𝑟 𝑖𝑛 𝑡ℎ𝑖𝑠 𝑔𝑟𝑎𝑝ℎ
-1

-1.5
0 5 10 15 20 25 30
Time

𝛼 = 0.2 & 0.5 which is which?

0.5 (blue curve) to 0.2 (red curve)

17 Modern Digital Communication Systems: Dr. Muqaibel

 Exercise: chose the correct answer
• Compared to a sinc pulse, the raised-cosine pulse with full roll-off factor (𝛼 = 1) has
the advantage of:
a) Less transmitted power.
b) Less bandwidth.
c) faster rate of decay.
d) less DC value.
e) All of the above.
• Data at a rate of 7 𝑘𝑏𝑖𝑡/𝑠 is to be transmitted over a baseband channel of bandwidth 4 𝑘𝐻𝑧
using binary Nyquist criterion pulses. Determine the maximum value of the roll-off factor 𝛼 that
can be used.
a) 4/7
b) 7/4
c) 1/4
d) 1/7
e) 3/7
• Is it possible to design a system with zero ISI and symbol rate greater than twice the bandwidth?
Justify your answer in view of Nyquist zero ISI criterion.

18 Modern Digital Communication Systems: Dr. Muqaibel

 Correlative Decoding, Optimum Receiver &
Eye Pattern

If you have any comments or corrections on

the slides , you may send to Dr. Ali Muqaibel,
[email protected]
1 Modern Digital Communication Systems: Dr. Muqaibel
 Outline

• Correlative-level coding
– Duobinary
– Modified Duobinary

• M-ary PAM
• Optimum linear receiver
• Eye pattern
2 Modern Digital Communication Systems: Dr. Muqaibel
 Correlative-Level Coding
• Correlative-level coding (or partial-response signaling) is the addition of ISI to the transmitted
signal in a controlled manner (why?)
• To reduce the bandwidth (contrary to raised cosine scheme), the pulse is widened in time
resulting in ISI but in a controlled manner.
• Introduced ISI at transmitter is known; hence, it can be removed at receiver side in
deterministic way.
• Duobinary Signaling
▪ Binary input sequence {𝑏𝑘 } is applied to pulse amplitude modulator resulting in short pulses of
amplitude
+1 , 𝑏𝑘 = 1
𝑎𝑘 = ቊ
−1 , 𝑏𝑘 = 0
▪ The sequence {𝑎𝑘 } is applied to duobinary encoder 𝑐𝑘 = 𝑎𝑘 + 𝑎𝑘−1

3 Modern Digital Communication Systems: Dr. Muqaibel

 Correlative-Level Coding
• Block Diagram Input Ideal Output
two-level Channel
sequence sequence
𝐻𝑁𝑦𝑞𝑢𝑖𝑠𝑡 (𝑓) 𝑐𝑘
𝑎𝑘

• Duobinary coder output 𝑐𝑘 = 𝑎𝑘 + 𝑎𝑘−1

• The overall frequency response of the filter and ideal channel is
𝐻𝑂 𝑓 = 𝐻𝑁 𝑓 1 + e−𝑗2𝜋𝑓𝑇𝑏
= 𝐻𝑁 𝑓 e+𝑗𝜋𝑓𝑇𝑏 + e−𝑗𝜋𝑓𝑇𝑏 e−𝑗𝜋𝑓𝑇𝑏
= 2𝐻𝑁 𝑓 cos(𝜋𝑓𝑇𝑏 )e−𝑗𝜋𝑓𝑇𝑏

4 Modern Digital Communication Systems: Dr. Muqaibel

 Duobinary Frequency Response
𝐻𝑂 𝑓 = 2𝐻𝑁 𝑓 cos(𝜋𝑓𝑇𝑏 )e−𝑗𝜋𝑓𝑇𝑏
• Magnitude frequency response and phase response 𝐻𝐼 𝑓 = 2cos(𝜋𝑓𝑇𝑏 )e−𝑗𝜋𝑓𝑇𝑏

• Note the continuity in band edges for 𝐻𝑂 𝑓 which results in easier approximation of the frequency response in
practice.
• The corresponding impulse response,
𝑇𝑏2 sin(𝜋𝑡/𝑇𝑏 )
ℎ𝑂 𝑡 =
𝜋𝑡(𝑇𝑏 − 𝑡)

5 Modern Digital Communication Systems: Dr. Muqaibel

 Duobinary Time Response
𝑇𝑏2 sin(𝜋𝑡/𝑇𝑏 )
• Impulse response ℎ𝑂 𝑡 = 𝜋𝑡(𝑇𝑏 −𝑡)

• The response of input pulse is spread over two signaling intervals

• It decays as 1/ 𝑡 2
• Estimate of transmitted 𝑎𝑘 at receiver
𝑎ො𝑘 = 𝑐𝑘 − 𝑎ො𝑘−1
• Problem in decoding duobinary signals?
• Error Propagation

6 Modern Digital Communication Systems: Dr. Muqaibel

 Correlative-Level Coding
• To avoid error propagation in duobinary scheme, we use precoding before encoding the binary sequence
Input Binary
Sequence Output
{𝑏𝑘 } {𝑑𝑘 } {𝑎𝑘 }
three-level
Sequence
{𝑏𝑘 }

𝑑𝑘 = 𝑏𝑘 ⨁𝑑𝑘−1
• Decision rule for detecting original binary sequence 𝑏𝑘
• If |𝑐𝑘 | > 1, then 𝑏𝑘 = 0
• If 𝑐𝑘 < 1, then 𝑏𝑘 = 1

7 Modern Digital Communication Systems: Dr. Muqaibel

Example Correlative-Level Coding with Precoding
• Encode and decode the binary sequence 0010110 using precoded duobinary scheme

Original {𝒃𝒌 } 0 0 1 0 1 1 0

Precoded 1 1 1 0 0 1 0 0
𝑑𝑘 = 𝑏𝑘 ⨁𝑑𝑘−1
sequence {𝑑𝑘 }
Two-level +1 +1 +1 -1 -1 +1 -1 -1
sequence {𝑎𝑘 }
Duobinary +2 +2 0 -2 0 0 -2
coder {𝑐𝑘 } 𝑐𝑘 = 𝑎𝑘 + 𝑎𝑘−1

Decoded 0 0 1 0 1 1 0 If |𝑐𝑘 | > 1, then 𝑏𝑘 = 0

If 𝑐𝑘 < 1, then 𝑏𝑘 = 1

9 Modern Digital Communication Systems: Dr. Muqaibel

 Modified Duobinary and Generalized Form of
Correlative-Level Coding
Modified Duobinary to remove DC Generalized Form of Correlative-Level Coding

Modified duobinary signaling scheme.

Frequency response of the modified duobinary conversion filter. (a) Magnitude

response. (b) Phase response.

10 Modern Digital Communication Systems: Dr. Muqaibel

 Baseband M-ary PAM Transmission
• In baseband 𝑀-ary PAM system, the pulse amplitude modulator can produce 𝑀
different amplitude levels (𝑀 > 2)
• For symbol duration, 𝑇, the signaling rate for the system is 1/𝑇 [symbols/second]
• Relationship between the bit and symbol durations
𝑇 = 𝑇𝑏 log 2 𝑀 or 𝑅𝑏 = 𝑅 log 2 𝑀
→ In M-ary PAM, the information is transmitted at a rate log 2 𝑀 faster than binary PAM
• In general, what will you sacrifice in M-ary PAM to have higher transmission
rate?
• Transmission power, e.g. for the same average symbol error probability M-ary PAM
requires more transmitted power (must be increased by a factor 𝑀2 / log 2 𝑀).

11 Modern Digital Communication Systems: Dr. Muqaibel

 Optimum Linear Receiver
• Receiver design for baseband transmission system,
1. AWGN (alone) → Matched filter receiver
2. ISI (alone) → pulse-shaping, raised cosine,…etc
3. AWGN + ISI → ?
• For combined AWGN and ISI effects, there are many receiver designs
- Zero-forcing (ZF) equalizer
• (easy to design, forcing ISI to zero, neglect additive noise)
• → leads to performance degradation (noise enhancement)
- Minimum mean-square error (MMSE) equalizer
• (additional complexity in design, reduce both effects of noise and ISI)
• → outperforms ZF equalizer

12 Modern Digital Communication Systems: Dr. Muqaibel

 Optimum Linear Receiver
• Output of the receive filter
∞
𝑦 𝑡 = න 𝑐 𝜏 𝑥 𝑡 − 𝜏 𝑑𝜏
−∞

𝑐 𝑡 : impulse response of receive filter

𝑥 𝑡 : output of the channel
• For perfect operation of receiver, the sampled received signal at 𝑖 after filtering is 𝑦 𝑖𝑇𝑏 = 𝑎𝑖
• Here, the error signal is
𝑒𝑖 = 𝑦 𝑖𝑇𝑏 − 𝑎𝑖
• We need to find 𝑐(𝑡) such that the mean-square error, 𝐸[𝑒𝑖2 ], is minimized

13 Modern Digital Communication Systems: Dr. Muqaibel

 Optimum Linear Receiver
• Optimum linear receiver 𝐶 𝑓 (in minimizing MSE sense) has two cascaded
components:
1. Matched filter
2. Tapped-delay-line equalizer (Transversal)
• If the channel characteristics are unknown or time-varying, how
would you eliminate/reduce the noise and ISI effects?
• Adaptive Equalizers

14 Modern Digital Communication Systems: Dr. Muqaibel

 Eye Pattern
• Eye pattern (or eye diagram) is synchronized superposition of all possible realizations of
the signal of interest (received signal, receiver output) viewed within a particular signaling
interval

15 Modern Digital Communication Systems: Dr. Muqaibel

 Eye Pattern
• The preferred time for sampling is the instant of time at which the eye is open the widest
• Height of the eye opening define the noise margin of the system

16 Modern Digital Communication Systems: Dr. Muqaibel

 Eye Pattern Effect of Noise

(a) Eye diagram for noiseless quaternary system

(b) Eye diagram for quaternary system with SNR = 20 dB. (c) Eye diagram for quaternary system with SNR = 10 dB.
17 Modern Digital Communication Systems: Dr. Muqaibel
 Eye Pattern Effect of Bandwidth (noiseless)
• Example
• A quaternary PAM system (𝑀 = 4), channel follows low-pass Butterworth filter
2 1
𝐻 𝑓 = 𝑓 2𝑁
, 𝑁 is the order of filter, 𝑓0 is 3 dB cut-off frequency.
1+
𝑓0

SNR = 20 dB (𝒇𝟎 = 𝟎. 𝟗𝟕𝟓 Hz) For band-limited system with 𝒇𝟎 = 𝟎. 𝟓 Hz

18 Modern Digital Communication Systems: Dr. Muqaibel