DIGITAL COMMUNICATIONS
Zafi Shah Pulse Amplitude Modulation and basics of detection – part 1
Fair Use Notice
The material used in this presentation i.e., pictures/graphs/text, etc. is
solely intended for educational/teaching purpose, offered free of cost to
the students for use under special circumstances of Online Education
due to COVID-19 Lockdown situation and may include copyrighted
material - the use of which may not have been specifically authorised
by Copyright Owners. It’s application constitutes Fair Use of any such
copyrighted material as provided in globally accepted law of many
countries. The contents of presentations are intended only for the
attendees of the class being conducted by the presenter.
This lecture is based on the Chapters 5 and 6 from the book
Digital Communications by Glover and Grant. This book should be
available in the library.
I was fortunate enough to be a student of Prof. Ian Glover at
Strathclyde University
Natural and Flat-top sampling
There are obvious similarities between sampling and PAM
In fact, in many cases, the two processes are indistinguishable
if the pulse duration of the PAM signal is very short
So, two processes, both referred to as “sampling”
Natural sampling
Flat-top sampling
Noise over the transmission channel
As the signal propagates over the channel, it will incur noise (also lose
strength)
In most cases, one can model the noise as Gaussian or approximately
Gaussian
True if thermal noise dominates
Even in presence of multiple noise source due to operation of the central
limit theorem
Noise over the transmission channel
It is common to model noise in communication systems as Gaussian, esp.
when the objective is to undertake bit error rate analyses
𝑟 𝑡 = 𝑠 𝑡 + 𝑛(𝑡)
Baseband Center Point Detection
The detection of digital signals involves two processes
Reduction of each received voltage pulse (i.e. symbol) to a single numerical
value
Sample?
Comparison of this value with a reference voltage to determine which
symbol was transmitted
Comparator?
Probability of error in Gaussian noise
Probability of error in Gaussian noise
If symbols are equi-probable, the optimum decision level can be set
as: 𝑉0 + 𝑉1
2
Suppose bit 0 is transmitted with voltage v0 at the transmitter
Given that we have Gaussian noise, we can derive an equation for the
probability of error
Probability of error in Gaussian noise
The probability of error, that is the received signal voltage will exceed
the detection threshold is the area under the shaded part of the curve
∞
1 𝑣𝑛 −𝑉0 2
−
𝑃𝑒1 = න 𝑒 2𝜎2 𝑑𝑣𝑛
𝜎 2𝜋
𝑉0 +𝑉1
2
Probability of error in Gaussian noise
Now, using change of variables
𝑣𝑛 −𝑉0
𝑥=
𝜎 2
We get
∞
1 −𝑥 2
𝑃𝑒1 = න 𝑒 𝑑𝑥
𝜋
𝑉1 −𝑉0
2 2𝜎
We can now recast this equation in terms of the complimentary error
function
Probability of error in Gaussian noise
Complimentary error function
∞
2 2
erfc(𝑧) ≜ න 𝑒 −𝑥 𝑑𝑥
𝜋
𝑧
Hence
1 𝑉1 − 𝑉0
𝑃𝑒1 = 𝑒𝑟𝑓𝑐
2 2 2𝜎
Probability of error in Gaussian noise
As obvious, the error function (erf) and the complimentary error
function are related as
erfc(𝑧) = 1 − erf(𝑧)
Therefore,
1 𝑉1 − 𝑉0
𝑃𝑒1 = 1 − 𝑒𝑟𝑓
2 2 2𝜎
Probability of error in Gaussian noise
Why define in terms of error function or the comp. error function?
This function has been extensively tabulated
Probability of error in Gaussian noise
Similarly, for the bit 1
𝑉0 +𝑉1
2
1 𝑣𝑛 −𝑉1 2
−
𝑃𝑒0 = න 𝑒 2𝜎2 𝑑𝑣𝑛
𝜎 2𝜋
−∞
Due to symmetry, the probability of error will remain same, regardless
of bit 1 or bit 0 being transmitted
Probability of error in Gaussian noise
In fact, note that the probability or error depends on the difference
between the symbol voltage levels, therefore, we have
1 𝑉1 − 𝑉0
𝑃𝑒1 = 1 − 𝑒𝑟𝑓
2 2 2𝜎
1 ∆𝑉
𝑃𝑒 = 1 − 𝑒𝑟𝑓
2 2 2𝜎
Probability of error in Gaussian noise
The equation for prob. of error is valid for both unipolar and polar
signaling methods as long as ΔV represents the voltage difference at
sampling instant
