2.

Representation of Signals
Signals may be classified as predictable or as unpredictable, as analog or digital and of
finite or infinite duration. Deterministic signals are defined exactly as a function of time.
They can be periodic (such as a sinewave or squarewave) or they can be an aperiodic “one
shot” signal. Deterministic signals contain no information because their future is completely
predictable by the receiver. These signals are easy to model and are useful since they can
provide a reasonably accurate evaluation of communication system performance. Stochastic
signals, on the other hand, are unpredictable and can communicate information. Although the
time waveform of a stochastic signal is random, the signal power may be predictable.
Examples of stochastic signals are thermal noise in electronic circuits, (i.e. “background
hiss”) and information signals such as voice or music.
An analog signal, denoted x(t), is a continuous function of time and is uniquely
determined for all t. When a physical signal such as speech is converted to an electrical signal
by a microphone, we have an electrical analog of the physical waveform. A discrete-time
signal, denoted as x(kT), exists only at discrete instants. It is characterized by a sequence of
values that exist at specific times, kT, where k is an integer and T is normally a fixed time
interval. On the other hand, a discrete amplitude signal is restricted to a discrete set of
amplitudes but is continuous in time. A signal that is discrete in both time and amplitude is
referred to as a digital signal. Furthermore, these discrete digital signal amplitudes can be
represented by a set of numbers (codes) and, as such, can be stored in a computer memory.
Pulse code modulation (PCM) is an example of a digital signal. These categorizations are
illustrated in Figure 2-1.
Continuous Time Discrete Time

Height of Patrons
Entering a Theatre

Temperature vs time
Analog Signal Discrete Time Signal

Quantized Height
200
190
180
170
160
cm
Number of Telephone Calls
170 160 180 190

Discrete Amplitude Signal Digital Signal - discrete amplitude

- discrete time

Figure 2-1 Concept of Analog and Digital Signals

Chapter 2: Representation of Signals 4

An energy waveform has finite amplitude and it either exists for a finite duration or it
decays to zero over time. An energy signal has finite energy but zero average power. A power
waveform has finite amplitude and semi-infinite duration thus it has finite average power but
semi-infinite energy. The signal is considered to be continuous over the observation period.
Energy and power classifications are mutually exclusive; a signal must be one or the other.
Periodic signals are classified as power signals.

2.1 SIGNAL POWER

Instantaneous power is the voltage-current product at a specific time, p(t) = v(t) i(t),
while average power is P = 〈p(t)〉 = 〈v(t) i(t)〉 where 〈•〉 indicates time average. With
normalized power, the load resistance is assumed to be 1 Ω, current is numerically equal to
voltage and average normalized power is simply the time average of voltage squared, PN =
〈v(t)2〉. Root-mean-square (rms) voltage is the square root of normalized power, Vrms = PN .
For example, the 12 volt supply in an automobile has normalized power of 144 ‘watts’.

Example 2.1 – Integrate to find the time average normalized power of v(t) = Acosωot (period To = 2π/ωo).
Solution: The instantaneous power is pN(t) = A2cos2ωot = A2(1+cos2ωot)/2
T 2
A 2 2 T0 2 A 2 T0 2 A 2 ⎡ sin 2ω0 t ⎤ 0
Average Power PN = p N (t) = ∫ (1 + cos 2ω0 t )dt = [t]−T0 2 +
T 0 −T0 2 2T 0 2T 0 ⎢⎣ 2ω0 ⎥⎦−T 2
0

Since ωo = 2π / T0 , then sin (2ωoTo/2) and sin (-2ωoTo/2) are both equal to zero and
A2 ⎡T0 −T0 ⎤ A
2
PN = ⎢⎣ 2 − =
2T0 2 ⎥⎦ 2

Drill Problem 2.1 (Normalized Power) - Enter normalized power for each signal in the column. The column
total is provided as a checksum. A calculator should not be required for this problem.

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 5

2.2 SIGNAL GAIN MEASUREMENTS

Signal gain is a measurement of an amplifying device or a transmission medium over
which the signal propagates. The ratio of input and output power (or voltage) is usually
expressed in logarithmic units and this is convenient for analyzing a series cascade of
elements.

2.2.1 Power Gain in decibels

The ear responds to sound levels in an approximately
logarithmic manner and thus acoustic power gain or loss ratios are “Rules of thumb”
expressed in a base 10 logarithmic units. A power ratio of 10 is -3 dB 1/2 power
defined as one Bel (to honor Alexander Graham Bell) but since human -10 dB 1/10 power
ears can distinguish sound power differences that are less than one Bel, the adopted unit is the
decibel (dB). To determine the amplification of a system we apply an input signal, measure
the input and output signal power and then express the dimensionless ratio in dB.
⎛P ⎞
Power Gain (dB) = 10 log 10 ⎜ o ⎟ .
⎝ Pin ⎠

Virtual Laboratory 2.1 - estimate the minimum difference in dB that your

ears can detect. Consider an absolute measurement (for example, one day to
the next) and a relative measurement (one second to the next).

Since power measurement with a traditional wattmeter is not practical for low power
communication signals, measurements are made with a rms responding voltmeter. Power ratio
is then calculated using circuit impedance values.

Figure 2-2 Amplifier input and load impedances

Example 2.3 - an amplifier has an input of 3 watts and output of 600 watts, what is the power gain?
⎛P ⎞
Answer: Power gain (dB) = 10 log10 ⎜ o ⎟ = 10 log10 (200) = 23 dB
⎝ Pin ⎠

2.2.2 Voltage Gain

The decibel unit of power gain has been generalized to “Rules of Thumb”
describe the voltage ratio in systems where the input and output Gain Factor Voltage ratio
impedances are not defined. For example, an ideal transformer -0.5 dB 0.95 (-5%)
with turns ratio 1:2 will have no power gain (0 dB) but will -1 dB 0.9 (-10%)
have a voltage gain of 2 (6 dB). Voltage gain is useful to -3 dB 0.707 (-30%)
describe electronic amplifiers which have high input impedance -6 dB 0.5
-10 dB 0.31
and low output impedance. In this case, voltage gain is constant
-20 dB 0.1
but the power gain varies with the load resistance. As -40 dB 0.01
illustrated in the "rules of thumb" table, a 1 dB change

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 6

represents approximately 10% change in voltage. When the gain is less than unity, we refer to
the device as having loss or attenuation and thus the dB values are not expressed as negative.
⎛ Vo2 ⎞ ⎛ Vo ⎞
Voltage gain (dB) = 10 log10 ⎜ 2 ⎟ = 20 log10 ⎜ ⎟ .
⎝ Vin ⎠ ⎝ Vin ⎠

Voltage gain vs frequency - in audio systems, gain

is usually specified at 1000 Hz and its variation 0

with frequency is known as the frequency 2

response or sometimes the amplitude response or
4
magnitude response. In telephone systems, for
example, gain is reduced at low frequencies by 6
transformer coupling and reduced at high 8
frequencies by transmission line loss as illustrated 0 1000 2000 3000 4000
Figure 2-3. A more complete specification is Frequency (Hz)
given by the Transfer function H(f) which includes Figure 2-3 Gain vs frequency
both amplitude (i.e gain) and phase angle.
Phase vs frequency response is a companion to Figure 2-3. An ideal system will have
constant delay between input and output and is known as a linear phase system since the
output phase increases in proportion to frequency.
Example 2.4 - an amplifier has input of 2 mV rms and output of 600 mV rms. Determine the voltage gain and
also the voltage gain expressed in dB.
Answer: Voltage Gain = 600mV / 2 mV = 300 Gain (dB) = 20log10300 = 49.5 dB

Drill Problem 2.2 (dB) – The gain properties of four amplifiers are listed in the table below. Complete the
missing table entries and verify using the checksum.

2.2.3 Signal Distortion (Linear and non-linear)

A transmission system or amplifier with inductive, capacitive and resistive elements
will have variable gain and phase shift over the frequency range of interest. Mathematically,
this is expressed as a transfer function, G(f). A signal, such as a pulse waveform, is distorted
in this type of system and since the distortion is caused by linear circuit elements, we refer to
this as “linear” distortion. The signal waveform can be restored by compensating this gain and
delay (i.e. phase) distortion with a device known as an equalizer. In this type of signal
distortion, no new frequencies are generated.

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 7

Non-linear distortion occurs in amplifiers that have reduced gain at higher amplitudes.
This effect is known as saturation and a graph of output vs input voltage is not a straight line.
This input-output amplitude non-linearity introduces “harmonic distortion” in single
frequency signals and introduces “inter-modulation distortion” in signals with several
frequency components (i.e. new frequencies are generated). Non-linear distortion is often
correctable, but with somewhat more difficulty than for linear distortion.

2.3 SIGNAL LEVEL MEASUREMENTS

2.3.1 dBm
The absolute power of a signal is commonly expressed in dB relative to one-milliwatt.
Since wattmeters are not able to measure small powers, voltage-measuring instruments have
been calibrated to read the power that would be delivered to a 50 Ω, 75 Ω or 600 Ω load. For
example, a voltmeter reading of 0.775 Vrms may be calibrated to read 0 dBm which
corresponds to 1 mW dissipation in a 600 Ω load. Function generators and meters calibrated
for 50 Ω systems read 0 dBm for 0.225 Vrms signals. For larger signals, the calibration may
be in dBW (i.e. the reference power is one watt). To properly measure complex waveforms, a
dBm meter should respond in proportion to the “true rms” voltage.
⎛ P ⎞
Level (dBm) = 10 log10 ⎜⎜ o ⎟⎟ .
⎝ 1 mW ⎠

Example 2.5 - The decibel unit can be used to express signal amplitude as compared to a reference power or
voltage. The dBm unit has a one milliwatt reference and is normally used to quantify the level (or amplitude) of
communication signals. Calculate the signal level (in dBm) of a 2 Vp-p sinusoid which is present at a 600 Ω
load.

( )
2
Vrms 2 1V / 2 ⎛ 0.8333 mW ⎟⎞
Pin = = = 0.8333 mW P(dBm) = 10 log10 ⎜ = − 0.79 dBm
R 600Ω ⎝ 1 mW ⎠

Example 2.6 - The rms voltage across the 300 Ω antenna input terminals of an FM receiver is 7.0 μV. Find the
input power (watts), evaluate the input power as measured in decibels below 1 mW (dBm).
(7.0 μV)2
watts (163.3 fW)
−15
Pin = = 163.3 x 10
300Ω
⎛163.3x10−15 W ⎞
P(dBm) = 10 log10 ⎜ ⎟ = − 97.9 dBm
⎝ 1mW ⎠

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 8

2.3.2 dBV
This unit expresses voltage level as a ratio to one volt rms. No impedance level is
specified (although 1000Ω impedance could be assumed to make dBm = dBV). Since this
unit is clearly a voltage measurement, it is a preferred unit for systems with electronic
amplifiers. While most radio frequency (RF) test equipment if for 50-ohm impedance, cable
television (CATV) systems are based on 75-ohm interfaces, so to avoid confusion, the cable
television industry uses the dBmV unit (1 mV reference). The dBV meter has a single
calibration scale and there is no confusion about readings taken on 50Ω, 75Ω or 600Ω
systems.
⎛ Vo ⎞
Level (dBV) = 20log10 .
⎝ 1 Vrms ⎠

Drill Problem 2.3 (dBW) - Observe the sample calculation and complete the tables to one decimal place
accuracy. A calculator should not be required.
P norm Level P norm Level
Signal (watts) (dBW 1Ω) Signal (watts) (dBW 1 Ω)
. . . .
2 volts rms 4 1 volt rms 1 0
20 volts rms 400 10 volts rms
2 V p-p sinusoid 0.5 0.9 Vrms
4V cos 2 π 1000t 8 +9 2 Vrms .
Checksum 412.5 38 Checksum 105.8 +25

Example 2.7 - dBm, dBW and dBV values depend on the measured voltage and the load impedance of the
system being measured. Note the 30 dB difference between dBm and dBW

2.3.3 Volume Unit (Vu) *

The volume unit was developed empirically to measure compound signals in a way
that corresponds to the human ear. The purpose is to present consistent loudness to the listener
and to avoid transmission system overload. It is used to continuously monitor transmission
levels of voice and music. To model the response of the human ear, the meter has a 30-15,000
Hz frequency response and it responds with approximately 170 ms averaging time constant.
Vu meters are incorporated in high-quality audio recorders.
The volume unit meter responds to rms voltage and is calibrated to read 0 Vu when a 0
dBm (0.775 Vrms at 600 Ω) sine wave is measured. The meter is calibrated in logarithmic
units and the peak observed needle deflection is taken as the reading. The observation period
is 60 seconds for music and 10 seconds for telephone conversations.

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 9

2.3.4 Cascaded Gains

In systems with cascaded elements, the dBm
(or dBV) level of the output is easily
calculated by adding all dB gains to the dBm
(or dBV) level of the input. It is standard
practice to express all levels, gains, signal to
noise ratios etc. in logarithmic dB units.
Figure 2-4 Cascaded Gain Elements
2.3.5 Summation of Non-Coherent Tones
While Coherent tones with the same frequency and phase “add on a voltage basis”,
sinusoids of the same frequency with phase difference of 90 degrees “add on a power basis”.
In power addition, the resulting voltage amplitude is calculated as the square root of the sum
of the component amplitudes squared. Same frequency signals with an arbitrary phase
difference can be resolved into in-phase components and quadrature components. These are
added separately on a voltage basis then the two resulting sums are combined using the square
root of the sum of the squares (power addition).
Non-Coherent tones of different frequency may be considered as two signals with a
steadily increasing phase difference. Since, over time, the phase difference is uniformly
distributed in the range 0 ≤ θ < 2π, the signals will add on a power basis. The following
relations summarize the linear addition of sinusoidal signals.
coherent Pt = (Vp1 + Vp2)2 / 2 when f1 = f2 and θ1 = θ2
non-coherent Pt = P1 + P2 when f1 ≠ f2 or when f1 = f2 and θ1 = θ2 ±π/2

Virtual Laboratory 2.2 - For two equal amplitude sinusoids of the same
frequency, each with power P = A2/2, in-phase addition results in P = 2A2
(quadruple power), while in-phase subtraction results in P = 0. Addition
at an arbitrary phase angle results in intermediate values of power. For
uniform distribution of phase shift in the range 0 ≤ θ < 2π, the average
power is P = A2.

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 10

2.3.6 Summation of Signals that are Quantified in dB

When two signals of different frequency are summed, the resulting level is not simply
the sum of the constituent dBm levels. Instead, the sum of the input signal powers must be
used. The level LT of two added signals with dBm levels L1 and L2 is calculated by summing
signal powers as follows:
⎛ ⎛ L1 ⎞ ⎛ L2 ⎞ ⎞
LT = 10 log10 ⎜ log -1 + log -1
⎝ ⎝ 10 ⎠ ⎝ 10 ⎠ ⎠
Example 2.8 - Two tones at 900 and 1100 Hz are added to form a signal. The tones have levels, -6 dBm and -9
dBm respectively. Calculate the dBm level of the combined signal.
⎛ P ⎞
10log10 ⎜ 900 ⎟ = − 6 dBm ⇒ P900 = log-1 -0.6 = 0.251 mW
⎝1 mW ⎠
P1100 = log-1 -0.9 = 0.126 mW
⎛P ⎞
Ptotal = 0.377 mW ⇒ 10log10 ⎜ total ⎟ = − 4.24 dBm
⎝1mW ⎠

Example 2.9 - Two "in phase" tones at 1000 Hz are added and both tones have level -2 dBm. Calculate the
dBm level at the combined signal.
⎛ −2 ⎞
P1 = P2 = log− 1⎜ ⎟ = 0.631 mW and A1 = A2 = 2 P1 = 3.55x10-2 V
⎝ 10 ⎠
PT = (A1 + A2)2/ 2 = 2.52x10-3 W = +4.02 dBm (i.e. –2 dBm + “6 dB” = 4 dBm)

Drill Problem 2.4 (dBm/dBV) - complete the following tables using one decimal place accuracy.
Level Level P norm Level
Signal (dBm 600Ω) (dBV) Signal (Watts) (dBV)
. . . .
0.775 volts rms 4 volt rms
10 volts rms 4V cos2π100t + 3V cos2π110t
5 mV rms 4V cos2π100t + 3V cos2π100t
10 V p-p sinusoid 4V cos2π100t + 3V cos2π200t
Checksum -8.4 -17.2 Checksum 65.5 +47.9

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 11

2.4 MATHEMATICAL REPRESENTATION OF A SINUSOID

Sinusoids are the most fundamental of all signals and a thorough understanding of
three forms of representation is essential for the study of communication systems.
1. Trigonometric Representation - sinusoidal signals are most simply represented in
cosine form. A phase shift can be included as part of the argument.

The polar form v(t) = A cos(ωot + θ ) can also be represented in Cartesian form (with in-
phase and quadrature components) as v(t) = x1cosωot - y1sinωot where x1 = A cosθ and y1 =
A sinθ . Conversion from Cartesian to polar form is accomplished using θ = tan-1(y1/x1) and
A = x 12 + y 12 .
2. Sum of complex exponentials - Sinusoidal signals may also be represented by sum
of two complex exponentials. Similarly, a phase shifted waveform v(t) = Acos(ωot + θ ) is
broken into Cartesian cosine and sine terms, v(t) = x1cosωot - y1sinωot, then expressed as the
jθ − jθ
sum of complex exponentials using the Euler relations 2 cosθ = e + e and 2j sinθ =
jθ − jθ
e −e .

When the sinusoid is expressed in trigonometric form, the normalized power has been
shown to be A2/2. When represented as the sum of two complex exponentials, each
exponential has constant amplitude and power equal to (A/2)2. The normalized power sum in
the two exponentials is thus (A/2)2 + (A/2)2 = A2/2.

3. Real part of a complex exponential - Sinusoidal signals may also be represented by

the real part of the complex exponential (i.e. a rotating vector). The multiplying term Ae j θ is
known as a phasor since it defines the phase and amplitude of v(t) with respect to the
reference vector e j ωt .

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 12

2.4.1 Complex Envelope Representation of a Modulated Sinusoid

A sinusoid that has time variation in both amplitude and phase can be represented as

{
v(t) = Re A(t)e jθ (t)e jω ct } = A(t)cos[ω ct + θ (t)] .
The complex exponential is multiplied by a time varying phasor or “complex envelope”
where A(t) represents amplitude (or “envelope”) variation and θ(t) represents phase variation.
We define the complex envelope in polar form as
g(t) = A(t)e jθ (t ) .

{ }
The time varying sinusoid v(t) = Re g(t)e jω ct can also be represented in Cartesian form by
expanding g(t) = x(t) + jy(t) and this results in
v(t) = Re{x(t)e jω ct + jy(t)e jω ct} = x(t)cos ω c t − y(t )sin ω c t .

Virtual Laboratory 2.3 - Complex Exponential

a) When viewed from the end of the positive time axis, does the
vector ejwt rotate clockwise or counter-clockwise?
b) The real (horizontal) portion of ejwt represents sine or cosine?
c) Complex envelope – is it amplitude or phase modulation?

Example 2.10 – Complex Envelope - Representation of amplitude modulation using complex envelope
(phasor) notation.

Example 2.11 – Complex Envelope – Representation of phase modulation using complex envelope notation.

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 13

2.5 FREQUENCY SPECTRUM REPRESENTATION OF PERIODIC WAVEFORMS

A periodic waveform is composed of repeated copies of a waveform segment where
the segment length is equal to the repetition period. These waveforms are classed as power
signals since the duration is assumed to be infinite and there is no decay. Examples of
periodic waveforms are a sinusoid, a square wave and a triangular wave. A lengthy example
might be a continuously repeating musical recording. We are particularly interested in pulse
waveforms since they are used in sampling, an essential foundation of digital communication.
2.5.1 Spectral Representation of a Sinusoid
A sinusoid of the form v(t) = 100 mV cos 2π2000t can be observed with an
oscilloscope, an instrument that displays voltage versus time. The sinusoid can also be
observed using a spectrum analyzer where the horizontal scale calibrated in hertz and vertical
scale calibrated in Vrms, dBV or in dBm (50Ω). Figure 2-5 shows oscilloscope and spectrum
analyzer displays of the sinusoid described above.

Figure 2-5 Corresponding oscilloscope and spectrum analyzer displays

This spectrum analyzer employs a narrow band-pass filter that is slowly swept over the
frequency range of interest. The filter output voltage drives the vertical deflection of the
screen while the horizontal deflection is proportional to filter frequency. The bandwidth of
the filter is normally adjustable, for example 10 Hz, 30 Hz, 100 Hz and so on. Narrow
bandwidth gives good spectral resolution but the sweep speed must be reduced in proportion
to the square of the resolution bandwidth. Swept-frequency spectrum analyzers are
commonplace at high frequencies, however, modern analyzers now use FFT processing for
low frequency signals.
2.5.2 Which Symbol for Frequency - ω or f ?
Radio frequency allocations, equipment specifications, laboratory instruments and
non-sinusoidal signals are described cycles per second or hertz. While we sometimes use
radian frequency ω (radians per second) to aid compactness, the variable f (hertz) will be
used throughout our discussion of communication systems. The two frequency variables are
simply related by ω = 2πf.

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 14

2.5.3 Repetitive Waveforms - Fourier Series Representation

A real periodic signal w(t) with period T0 and frequency f0 = 1/ T0 can be represented
as a series of sinusoids having frequencies at integer multiples (or harmonics) of the
fundamental frequency f0 . This is known as a Fourier series and the sinusoids, when added
together, produce a signal equivalent to the original signal w(t). The zeroth harmonic
represents a dc component, the average value of the waveform. In Cartesian trigonometric
form, the Fourier series consists of cosine and sine terms which are orthogonal functions.
This quadrature trigonometric form has cosine and sine terms (separated by 90°). In polar
trigonometric form we have a series of cosine terms, each with magnitude and phase shift.
Thus we have
∞
w(t ) = a0 + ∑ an cos nω 0t + bn sinnω 0 t (Cartesian trigonometric form)
n =1
∞
and w(t ) = D0 + ∑ Dn cos(nω 0t + θn ) (polar trigonometric form)
n=1
where the coefficients are related as D0 = a0, Dn2 = an2 + bn2 and θn = tan−1 (−bn an ). Plots of
these coefficients versus frequency are the magnitude spectrum and phase spectrum. Using
Euler’s identity, each cosine term can be expressed as 2cos(ω t + θn ) = e j (ωt +θ n ) + e− j(ωt+θ n ) ,
and this leads to the Fourier series in complex exponential form and a two-sided
representation in the frequency domain. We then have polar exponential form and an
equivalent Cartesian exponential form (or quadrature form) where coefficients are complex.
These are expressed as
1 ∞ 1 ∞
w(t ) = D0 + ∑
2 n =1
Dn e jθ n + jnω 0 t
e + ∑
2 n =1
Dn e− jθn e − jnω 0 t (polar exponential form)

∞ −∞
and w(t ) = c0 + ∑ cn e jnω 0 t
+ ∑ cn e jnω t 0
(Cartesian exponential form) .
n =1 n =−1

1 1
Using Eulers identity c 0 = a0 , c n = (an − jbn ) and c− n = (an + jbn ) with c− n = c *n and
2 2
∞
w(t ) = ∑ c ne jnω t 0
(compact Cartesian exponential form)
n =−∞
By substituting this compact expression into the calculation of the cn coefficients,
a+T0 a+T0
1 − jnω 0 t 1
cn = ∫a w(t)e dt simply results in cn = ∫a c ne jnω 0 t e− jnω 0 t dt = cn and we verify the
T0 T0
jnω 0 t jmω 0 t
correctness of the coefficient calculation since ∫e
T
e dt = 0 when n ≠ m.
o

If w(t) is real and even (symmetric about t = 0) then cn is only real, but if w(t) is real
and odd (inverse symmetric about t = 0), then cn is only imaginary. To clarify, we number
quadrants in a counter-clockwise direction about the origin starting with the top right as
quadrant #1. In even symmetry, the waveform is flipped about the vertical axis, quadrants 1

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 15

and 2 are mirror images across the t = 0 axis and so are quadrants 3 and 4. In odd symmetry,
the waveform is flipped around the vertical axis then flipped around the horizontal axis.
Example 2.12: Symmetry - illustration of even and odd symmetric functions. Even functions are symmetric
across the boundary between quadrants 1 and 2. Odd functions are inverse symmetric across the boundary.

t t Quadrants
cosωot sinωot
2 1
t=0 t=0
3 4
t t

Even Fuctions Odd Functions

2.5.4 Repetitive Pulse Waveforms - Time Series Representation

A repetitive waveform can be expressed as the sum of delayed versions of a waveform
segment. As a simple example, we consider a single rectangular pulse waveform p(t). The
single pulse is mathematically described by rect(t/τ) or by Π(t/τ) (this character has been
chosen to resemble the time waveform). The expression for a periodic rectangular pulse
sequence, w(t), is given below.

Figure 2-6 Expression for a periodic pulse waveform

Sampled signals form the basis for modern communication systems thus we are
interested in their analysis. A convenient model is multiplication of the input signal by a
periodic pulse signal with magnitudes 0 and 1. To aid the analysis of sampled signals, we
study time and frequency representations of periodic (repetitive) pulse waveforms.

2.5.5 Repetitive Pulse Waveform - Fourier Series Representation

Consider a repetitive pulse waveform of amplitude A with period T0 = 5 ms and pulse
width τ = 1 ms (duty cycle = 0.2). The waveform, w(t), can be expressed as the sum of
∞
Fourier series components as w(t) = ∑c e n
jnω ot
where ωo = 2πfo = 2π/To . This waveform is
n=−∞
real and thus cn = c-n . Also we select the t = 0 axis to give an even function (symmetric about
t = 0) and thus all cn are real (no imaginary portion). Since the waveform is repetitive and of
infinite duration, we have a “line spectrum” composed of the repetition frequency and integer
multiples of that frequency. For the rectangular pulse waveform, we calculate cn as

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 16

T 2 τ 2
1 o 1
∫ w(t)e − jnω ot dt = Ae− jnω o t dt
To −τ∫ 2
cn =
To −T 2
o

− jnω o τ 2 jnω τ 2
A ⎡e −e o ⎤
cn =
A
− jn ω o To
[ e ]
− jnω o t τ 2
−τ 2
=
n2π fo To ⎢⎣ −j ⎥
⎦

A Aτ ⎡ sin π nfoτ ⎤ ⎛ τ ⎞ ⎡ sin nπ (τ To )⎤

cn =
nπfo To
[sin nω o τ 2] =
⎢
To ⎣ πnfo τ ⎦ ⎥
= A⎜ ⎟
⎝ To ⎠ ⎢⎣ nπ (τ To ) ⎥⎦
.

For the specific rectangular waveform with A = 1 V, T = 5 ms and τ = 1 ms, the Fourier
coefficients are illustrated as spectral lines versus frequency. Vertical units are in volts (peak
not rms).
Two-sided Amplitude Spectrum
exponential components 1 ⎛ sin 2π / 5 ⎞
∞ c2 = ⎜ ⎟ = 0.151
∑c n e 0.187

0.187
jnω 0 t 5 ⎝ 2π / 5 ⎠

0.2
0.151
0.101

0.101
n=−∞ 1 ⎛ sin 4π / 5 ⎞
c4 = ⎜ ⎟ = 0.0468
0.047

5 ⎝ 4π / 5 ⎠
0.025
0.017

-0.021
-0.038
-0.043
-0.031

6f0 8f0 f
12f0 -10f0 -4f0 -2f0 0 2f0 4f0 10f0 12f0

D 2 = c2 + c − 2 = 0.303

D3 = 0.202
One-sided Amplitude Spectrum
0.2
sinusoidal components D 4 = c4 + c− 4 = 0.0935
∞
D0 + ∑ Dn cos(nω 0 t + θ n ) 0.050
n=1 6f0 8f0 f

0 2f0 4f0 12f0

400 Hz 800 Hz 2400 Hz

Figure 2-7 Fourier Coefficients on one-sided and two-sided frequency scales.

Example 2.13 – Pulse Spectrum: Calculate the first eight Fourier series coefficients for unit amplitude pulse
sequences with duty cycles 3/4, 1/2, 1/3, 1/4, 1/5, 1/6, 1/7 and 1/8.
Answer:

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 17

Virtual Laboratory 2.4 - Observe the square wave approximation that has
been generated using a multi-sinusoid waveform synthesizer.
a) The desired square waveform has amplitude two-volt peak-to-peak
and zero average voltage. What is the normalized power?
b) What are the theoretical peak-to-peak voltages of the first three
components? (checksum = 3.905)
c) What is the normalized power sum of the first three synthesizer
components? (checksum = 0.933)

Drill Problem 2.5 (Fourier Series) - Determine the peak, peak-to-peak, average and rms voltages of the
following two signals x(t) and y(t). Also determine the normalized power, fundamental frequency, period and the
first six coefficients of the complex "exponential" Fourier series. Complete the table to two decimal places.
1V 40 ms y(t) 10 ms

x(t) = 7.07V cos 2¹ 50t 0

-1V 0

Vp Vp-p Vave Vrms PN f0 T0 C ±5 C ±4 C ±3 C ±2 C ±1 C0 Check-

volts volts volts volts watts Hz ms volts volts volts volts volts volts
sum

x(t) 7.07 0.00 25.00 124.75

y(t) 2.00 0.15 69.83

2.5.6 Sinc Function and Sampling Function

The term sin(πft)/πft appears frequently in communication
theory and, to simplify notation, many textbooks have
defined sinc(x) = sin(πx)/πx while others have defined the
sampling function Sa(πx) = sin(πx)/ πx. Not all books are
consistent and, to avoid possible confusion, these
notations have not been applied in this workbook. The
expression sin(πfτ)/πfτ is written in full. The sinc Figure 2-8 Sinc wavefrom
function diminishes with time and represents an energy
waveform.

2.5.7 Periodic Signals - Spectral Power and Parseval’s Theorem

A periodic signal, w(t), is a power waveform and may be represented as a series of
component signals at frequencies harmonically related to the fundamental frequency. For
these signal components, the power in the total signal is the sum of the powers associated with
2 2 2 2 2
each component as given by Pf = •• • + c −2 + c−1 + c 0 + c1 + c 2 + • •• (a sum to
infinity). Parseval' s theorem states that this calculation of signal power in the frequency
domain should equal Pt , the time average of instantaneous power. Thus for periodic signals
a+T o +∞
1
P=
To
∫ 2
w(t) dt = ∑ cn 2 (i.e. P = Pt = Pf . )
a n=−∞

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 18

Example 2.14 – Parseval’s Theorem Illustrate the time waveform of a ±2 V squarewave with frequency 2
kHz. Calculate the normalized power in the waveform. Also illustrate Fourier series coefficients on a 2 sided
frequency scale for frequencies in the range ±15 kHz. Calculate the power sum of the frequency components up
to 15 kHz. Do your results show Pf ≈ Pt as consistent with Parseval's theorem?
volts Pt = (2V)2 = 4 Watts
2V
Example spectral component
⎛ 1 ⎛ sin 5π ( 500 ⎞
)
250 t (us)
c 5 = 4V ⎝ ⎞⎠ ⎜
-2V
⎟ = 0.254 V 0 125 375 625
( 500 )
2 ⎝ 5π 250 ⎠
volts
1.273 1.273
0.254 0.254
Sum of power spectral components
-14 -6 6 14
f (kHz)
Pf ≈ (0.0)2 + 2[(1.273)2 + (-0.424)2 -10 -2 0 2 10
-0.182 -0.182
+ (0.254)2 + (-0.182)2] = 3.796 Watts -0.424 -0.424

2.5.8 Repetitive Pulse Waveform – Spectral Effects of τ and T

Variation in pulse width and pulse period change the pulse waveform’s spectral
components as illustrated in Figure 2-9. In the initial spectrum, components occur at 200,
400, 600 Hz …. , (multiples of 1/To.). Spectral nulls occur at 1 kHz, 2 kHz, 3 kHz …. ,
(multiples of 1/τ).

Figure 2-9 Fourier coefficients of a rectangular pulse rain when T and τ are varied.

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 19

2.6 FREQUENCY SPECTRUM CHARACTERIZATION OF NON-REPETITIVE SIGNALS

2.6.1 Energy Signals – Fourier Transform and Rayleigh’s Energy Theorem*
We now consider “one-shot” energy signals that have insignificant amplitude after
some finite duration. First, we assume the waveform repeats at interval To and thus we
initially obtain a Fourier series representation
∞ ∞ ⎡ 1 T0 /2 ⎤
w(t) = ∑ [c n ]e j2 πnf 0t = ∑ ⎢T ∫ w( τ )e
− j 2 πnf 0τ
d τ ⎥ e
j2 πnf 0 t
.
n=−∞ n=−∞ ⎣ −T 0 / 2
0 ⎦
To analyze the single waveform, we let To approach infinity so fo (i.e.1/To) becomes
the differential frequency df, and nfo becomes the continuous variable f. We replace
summation with integration and the calculation of w(t) becomes the inverse Fourier
transform integral
∞ ∞
⎡ +∞ ⎤
∫ [W ( f )]e
j2π ft
w(t ) = df = ∫ ⎢ ∫ w(τ )e − j2π f τ dτ ⎥ e j 2π ft df .
f =−∞ f = −∞ ⎢
⎣ τ =−∞ ⎥⎦

The portion in square brackets is defined as amplitude spectral density W(f). While Fourier
series coefficients cn have units of volts, W(f) has units of volt-seconds or volts/hertz (since
the calculation does not include the factor 1/To ). W(f) is two-sided with frequency ranging
over ± infinity. Calculation of amplitude spectral density W(f) from the time waveform is
known as the Fourier transform while the reverse process is the inverse Fourier transform.
∞
W ( f ) = ℑ [w(t)] = ∫ w(t)e− j 2πft dt (Fourier Transform)
t=−∞
∞
j2 πft
w(t) = ℑ−1[W( f )] = ∫ W( f )e df . (Inverse Fourier Transform)
f = −∞

Table 2-1 Fourier Transform Pairs.

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 20

Also of interest is the (normalized) energy spectral density calculated as

E ( f ) = W ( f ) with units of joules/hertz . Total energy in an energy waveform may be
2

calculated as the integral of energy spectral density over the entire frequency range.
Rayleigh's energy theorem (similar to Parseval's theorem ) equates the energy calculated in
the time domain with that calculated from the signal's Fourier transform.
∞ ∞
E= ∫ 2
w(t) dt = ∫ W ( f ) 2 df .
−∞ −∞

Example 2.15 – Fourier Transform of a Rectangular Pulse - Calculate the amplitude spectral density W(f)
and the total spectral energy of a single 4 volt, 5 s pulse. Verify Rayleigh’s (Parseval’s) theorem by calculating
time domain energy.

t +∞ 2
w(t) = A Π ⎛ ⎞ where τ = 5 s and A = 4 volts Spectral = ∫ A 2τ 2⎡ sin πτf ⎤ df
⎝τ⎠ ⎢ ⎥
Energy −∞ ⎣ πτ f ⎦
∫ −ττ/2/2 Ae− j 2πft dt = − j2Aπ f [e− j 2π ft ]−τ /2
τ /2
W(f) = +∞ 2
A 2τ 2 ⎡ sin πτf ⎤
τ /2
= 2
πτ
∫ ⎢ πτ f ⎥ dπτf
⎡e − j 2 πft
−e + j 2π ft
⎤ ⎡sin πτf ⎤ 0 ⎣ ⎦
W ( f ) = Aτ ⎢ ⎥⎦ = Aτ⎢
⎣ − j2πft ⎣ π τf ⎥⎦ = 2
80 J ⎛ π ⎞
= 80 J
−τ / 2
π ⎝2⎠
= 20[sin(5π f )/5π f ] volts/Hz

Time Domain Energy = (4 V)2 5 s = 80 joules (demonstrates RayleighÕs theorem)

Example 2.16 – Fourier Transform of a Sinc Pulse - A sinc waveform in the time domain and a rectangular
spectrum form a Fourier transform pair. Calculate the Fourier transform of w(t) = 12Vsin(3πt)/3πt and show
that the normalized energy in the waveform is 48 Joules.

W(f) is a complex function of frequency which can be written in Cartesian form as

W(f) = X(f) + jY(f) or in polar form as W(f) = |W(f)| ejθ(f). The symmetry properties of W(f) in
the Fourier Transform are similar to those of cn in the Fourier series.
If w(t) is real, W(f) = W*(-f) i.e. θ(f) = -θ(-f)
If w(t) is real and even, then W(f) has only a real part.
If w(t) is real and odd, then W(f) has only an imaginary part.

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 21

In summary, we consider w(t) to be a real voltage waveform that might be displayed

on an oscilloscope. The voltage fluctuates with time and thus there are associated frequency
(spectral) components. The Fourier transform gives an amplitude spectrum showing the
strength of one frequency component relative to others.

2.6.2 Noise – a continuous non-repetitive power signal

Communication systems are impaired by distortion, interference and noise. While
distortion can often be reversed and interference somewhat mitigated by filters, noise remains
as a system limitation. To succeed in battle against noise, we must understand the enemy.
Noise, in many cases, is an added random voltage with Gaussian amplitude distribution and
uniform (“white”) power spectral density (PSD) over the frequency range of interest. For
example, Johnson noise results from thermal agitation and has white spectral density.
Virtual Laboratory 2.5 – Observe the waveform, spectrum, and sound of
the sum of 16 sinusoids in the frequency range 2500 – 3000 Hz. Compare
with white noise bandlimited to 2500 – 3000 Hz.
Pop up a trace of the two waveforms. Can you see any significant
difference in the signal waveforms?

White noise can be viewed as an infinitely large number of equal-amplitude sinusoids

separated by an infinitely small frequency increment. Fourier transform yields a continuous
frequency distribution. To allow application of the Fourier transform, the aperiodic semi-
infinite random noise waveform, w(t), is truncated to form the energy waveform, wE (t ) .
Application of the Fourier transform yields the amplitude spectrum WE ( f ) in volts/hertz and
2
the energy spectral density WE ( f ) in joules/hertz. Dividing by T, yields the power spectral
density, SW ( f ) , in units of watts/hertz.
1
SW ( f ) = lim WE ( f )
2
(power spectral density)
T →∞ T

For the range of frequencies used in communication systems, thermal noise has
uniform spectral density, SN ( f ) = N 0 , so noise power is proportional to transmission
bandwidth. In double-sided spectra, power spectral density is denoted N0 / 2 .
Thermal noise is present in all electronic components including resistors. Thermally
excited electrons in the conducting medium undergo collisions resulting in an abrupt change
in current flow. If these collisions are modeled as impulses, the energy spectral density from
each impulse is “flat” (uniform spectral density vs frequency) and, since the ensemble of
impulses occur randomly, the power spectrum of the total noise is also flat.
Amplitude spectral density of noise is computed as the square root of SW ( f ) and has
units of volts hertz . As an example, the TL081 operational amplifier is specified to have
17nV hertz equivalent input noise at frequencies near 1 kHz.
In a swept-frequency spectrum analyzer, a narrow band-pass filter is slowly swept
over the frequency range of interest and the filter output voltage drives the vertical deflection
of the screen. If the filter bandwidth is increased, more frequency components are included
and the output noise voltage increases in proportion to the square root of bandwidth. When
measuring audio signals, the filter bandwidth (a.k.a. resolution bandwidth) might be 30 Hz.

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 22

Modern spectrum analyzers use the discrete Fourier transform (DFT) or the fast Fouier
transform (FFT) to compute the spectral display of a sampled segment of the signal. This
segment is repeated to form a periodic signal so the resulting transform is discrete as in the
Fourier series.

Virtual Laboratory 2.6 - measurement of lowpass filtered noise. Filters do not

have sharp cutoff so measurements deviate may slightly from theory.
a) Determine the meter reading in volts when using 10 kHz bandwidth. _____
b) Estimate the meter reading in volts when using 25 kHz bandwidth. _____
c) The display amplitude would reduce by what factor if resolution _____
bandwidth is reduced from 100 Hz to 1 Hz ? CHECKSUM 11.18

Virtual Laboratory 2.7 - measurement of bandpass filtered noise.

a) Find the normalized noise power spectral density (in mW/Hz) _____
b) Estimate the meter reading in volts for 5 - 15 kHz bandwidth. _____
c) What is the noise level in dBm (600 ohms) when the meter _____
reads 2.2 Vrms for 6 – 14 kHz bandwidth? CHECKSUM 13.7

Drill Problem 2.6 (Spectral Density) - Voltage and normalized power of three signals are measured directly
and then measured after passing through 50 Hz and 80 Hz ideal lowpass filters. Complete the table.

T = 50 ms, τ = 25 ms 10 cos 2π60 t No/2 = 20 mW/Hz

+10 V +10 V
Check-
-10 V
-10 V sum
Pn (W) ° °
VLPF 50 (Vrms) 1.41 10.41

PLPF 50 (W) 81.06 83.06

VLPF 80 (Vrms) 9.49 18.35

PLPF 80 (W) 50 143.26

Checksum 289.61 107.07 °

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 23

2.7 SIGNAL AMPLITUDE CHARACTERIZATION*

To overcome the effects of noise, it is desirable to transmit the largest possible signal
power. However, to avoid signal clipping, peak voltages must be constrained within the
system voltage limits. This tradeoff is especially problematic in OFDM multicarrier systems
such digital subscriber line (DSL) systems and wireless computer access to the Internet.

2.7.1 Peak to rms ratio (Crest Factor)

When compared to a sinusoid,
information signals such as voice, music
and DSL voltages have much larger peak
values. Thus, when the peak voltage is
rm s v o l t a g e

Signal V olt age

r m s v o lt a g e

constrained (by amplifier supply voltage

for example), information signals will t

have lower rms voltage and lower signal

power. Peak to rms ratio (or crest factor)
gives a rough characterization of
amplitude variation. A more complete Figure 2-10 Peak and rms voltages for different waveforms
characterization is given by the amplitude
probability density function (pdf).

2.7.2 Histogram and probability

density function (PDF) of a sinusoid
Amplitude probability density function (PDF) is introduced by considering the
amplitude histogram of the signal. Assume that the voltage range of a signal is divided into a
number of voltage “bins” and a histogram is calculated to show the probability of the signal
voltage being within the range of each voltage bin. As the number of bins is increased and the
range of each bin approaches zero, the probability of bin occupancy approaches zero.
However, by dividing the probability of bin occupancy by the bin range, we obtain the
amplitude probability density function expressed in units of probability/volt.

Figure 2-11 Sinusoid with amplitude bins, occupancy histogram and probability density

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 24

In the illustration Figure 2-11, we consider a 20-bin histogram of a unit amplitude

sinusoid. The sinusoid “takes a long time to turn around” at the peaks and therefore spends a
large fraction of each cycle in the uppermost and lowest voltage bins. If the bin width is
reduced and the number increased, the shape of the histogram will approach that of the
probability density. Probability density is proportional to the reciprocal of the waveform slope
and approaches infinity at the waveform peaks.

2.7.3 Sum of independent sinusoids, convolution and Gaussian distribution

Numerical convolution is now illustrated with an example. Using a computer, two
unit amplitude sinusoids were added and the amplitude histogram was calculated using 0.1
volt bins over the voltage range ±2 V. The sinusoid frequencies were 1000 Hz and 1220 Hz.
The empirical histogram below has a maximum around zero volts with the probability of bin
occupancy diminishing at higher voltages.

Figure 2-12 Sum of two sinusoids showing empirical histogram of bin occupancy.

As an alternative to empirical measurement, we use numerical convolution to

calculate the combined signal histogram from the component histograms of x(t) and y(t). In
the following demonstration, we consider the specific cases that result in an output of 1.8
volts. The three required voltage sums are 0.8+1.0, 0.9+0.9, and 1.0+0.8. As illustrated in
Figure 2-13, we “flip” one of the component histograms so as to align pairs of bins that sum
to 1.8 volts. If sinuslids x(t) and y(t) are independent, the joint probabilities for these three
events are 0.048(0.141), 0.062(0.062) and 0.141(0.048) respectively and the output 1.8 volts
occurs with probability 0.0174. Probability for an output of 1.7 volts is found by “sliding”
the “flipped” histogram so that 4 pairs of bins align with each pair summing to 1.7 volts.
Calculation of the complete output histogram by this method is known as discrete
convolution of the component histograms which is expressed as
n=0.95
P(z=m ) = ∑ P( x=n) P(y=m−n ) . As bin size is reduced, we approach continuous convolution
n=−0.95
v =+1
Pz (v = V ) = ∫ Px (v )Py (V − v ) dv = Px (v )* Py (v ) .
v=−1

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 25

Figure 2-13 Probability histogram computed by discrete convolution of component histograms.

2.7.4 Noise – Amplitude Characterization

When a large number of independent signals are summed, the resulting PDF
approaches the “bell shaped” curve of a Gaussian density function. This property is formally
discussed in the central limit theorem.
This Gaussian amplitude density function is important in the study of
communication systems since many types of noise (particularly thermal noise) have
approximately Gaussian amplitude densities. From the above study of two added sinusoids,
it can be appreciated that the sum of many equal amplitude sinusoids (as in “white” noise),
would approximate a Gaussian amplitude density.

Virtual Laboratory 2.8 - see how amplitude probability density (PDF)

can be indirectly observed on a spectrum analyzer through the use of a
voltage to frequency converter (note that the spectrum analyzer
normally shows amplitude spectral density).
Observe the probability density when 2, 4 and 8 sinusoids are added and
investigate the bell-shaped amplitude spectral density for bandlimited
noise

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 26

2.8 MEAN, STANDARD DEVIATION, CORRELATION AND CONVOLUTION

2.8.1 Signal mean

The mean of a signal (its dc value) is given by the time average of the signal or by the
Fourier component W(0). The time average is computed over a finite observation interval as
t2
1
mw = W (0) = w (t ) =
t2 − t1
∫ w( t) dt .
t1
For periodic waveforms, t2 -t1 is set equal to the period and for non-periodic waveforms, the
average is obtained as t2 -t1 approaches infinity.

2.8.2 Variance and Standard Deviation

The variance σ 2 of a signal is a measure of the variation about its mean. The
standard deviation σ is simply the square root of the variance. For voltage signals with zero
mean, the variance equals the signal power and the standard deviation is equal to the root-
mean-square (rms) value.
t
1 2
σ w = (w (t ) − mw ) =
2 2
t2 − t1 t
∫ (w (t) − mw )2 dt
1

Drill Problem 2.7 (Amplitude Characterization) – Consider the sum of four equal amplitude “raised cosine”
voltages of the form x(t) = 10 V + 10 V cos2πfnt where the four frequencies are unrelated. For the total waveform,
compute the following:
Mean voltage m = _______ V, Variance σ2 = _________ V2 , standard Deviation σ = _________ V
Normalized Power = ___1800_W, rms voltage = ___________V Checksum = 2096.57

2.8.3 Correlation coefficient

The correlation coefficient indicates a linear relationship between signals x(t) and y(t).
When x(t) changes, is there a corresponding change in y(t)? It tells us, on average, how likely
each sample of x(t) is proportional to y(t) and the square of the correlation coefficient
indicates how well a straight line can be fit to the sample values of x(t) and y(t). For
continuous signals the correlation coefficient is given by
μxy (x ( t) − mx )(y (t) − my )
ρ= =
σ xσ y
(y (t) − my )
2
(x (t ) − mx )2
where m x and m x are the mean values of x(t) and y(t) . Including σ xσ y in the denominator
normalizes the expression so ρ falls in the range −1≤ ρ ≤ 1 . A similar expression can be
written for sampled data.
∑i (xi − m x )(yi − my )
ρ=
∑i (xi − mx )2 ∑i (y i − my )
2

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 27

If an x-y display of x(t) and y(t) is a line with positive slope extending from the first to
the third quadrant, the correlation coefficient is ρ = 1.0 . If the x-y display is a line with
negative slope then the correlation coefficient is ρ = −1.0 . A circular x-y display of the two
variables indicates correlation coefficient ρ = 0.0 .

Example 2.17 – Calculate the correlation coefficient of the first four digits in the numbers π and e.
π
1,4 me = 4.5
2,3 m¹ = 2.25

7,1 8,1
e
(3 − 2.25)(2 − 4.5) + (1− 2.25)(7 − 4.5) + (4 − 2.25)(1− 4.5) + (1− 2.25)(8 − 4.5)
ρ=
(3− 2.25)2 + (1− 2.25)2 + (4 − 2.25) 2 + (1− 2.25)2 • (2 − 4.5)2 + (7 − 4.5)2 + (1− 4.5) 2 + (8 − 4.5) 2

− 1.875 − 3.125 − 6.125 − 4.375

ρ= = −0.98
6.75 • 37

Virtual Laboratory 2.9 - Correlation of audio signals can be seen on

an oscilloscope and can also be observed using stereo headphones.
a) Observe and listen to X-Y noise. Does the “fuzzy ball”
correspond to two correlated or uncorrelated noise sources?
b) Observe and listen to X-Y stereo music. In which case does
the sound seem to be in the middle of your head?

2.8.4 Cross-correlation
Radar, sonar, and medical ultrasound use correlation to determine the magnitude and
delay of a returned echo signal and thus estimate the size and distance of a reflecting object.
Since the echo signal y(t) is displaced in time from the transmitted signal x(t) , the simple
correlation coefficient will not suffice. To address this problem, we introduce the variable
time-shift τ to advance the echo signal and then compute correlation as a function of this
time-shift. Cross-correlation is thus defined for power signals as
t1 +T
1
∗
Rxy (τ ) = x (t)y(t + τ ) = lim
T →∞ T
∫ x ∗ (t)y(t + τ )dt .
t1
While the correlation coefficient is a single value, the cross-correlation function is a function
of relative delay. Also note that mean values are not eliminated from the calculation of
Rxy (τ ) and there is no normalization.

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 28

2.8.5 Autocorrelation
The autocorrelation function Rxx (τ ) measures the similarity between a signal x(t) and
a time-advanced copy of the same signal x(t + τ ) . The two signals match perfectly when
τ = 0 and, at that point, autocorrelation is equal to signal power. If the signal contains echos,
some degree of matching will occur at other vaues of τ . Time variation of a physical
waveform is thus evaluated by comparing the signal with various delayed versions of itself.
For a periodic signal, autocorrelation will reach a maximum for time differences that are an
integer multiple of the period. For example, autocorrelation of outside temperature will show
maximums at 24-hour intervals. Since a signal and it’s time-shifted version are identical, we
have symmetry about τ = 0 and thus Rxx (− τ ) = R xx (τ ) . If the signal has consistent
(stationary) mean and variance, Rxx (τ ) will be independent of starting time t1. For a power
signal, the autocorrelation function is defined as
t1 +T
1
Rxx (τ ) = lim
T →∞ T
∫ x(t)x(t + τ )dt (power signal - watts) .
t1
And for the energy signal xT (t) (subscript T indicates truncation to the interval T), the
autocorrelation function is
∞
RExx (τ ) = ∫ xT (t)xT (t + τ )dt (energy signal - joules) .
−∞

Example 2.18 A ship-to-ship communication system transmits repeated radio frequency (RF) bursts. For a
single burst, with rectangular amplitude envelope, x(t), the distant receiver receives two bursts, one from the
direct point-to-point path and the second from a reflection off the water. The received envelope is described
by y(t). Observe the cross-correlation Rxy (τ ) and the auto-correlation Ryy (τ ) .

50 ns 100 ns -20 ns 0 ns 20 ns 40 ns 60 ns
10 V
0.1 W
Rxy(τ )
x(t)

0V 0W
t τ
3 mW Ryy (τ )
200 mV y(t)
0V 0W
t τ

2.8.6 Convolution
In the discussion of probability density of summed signals in Section 2.7.3 we
introduced the concept of convolution. In subsequent chapters, convolution will be used to
calculate the output spectrum when two or more signals are multiplied together. Similarly, for
cascaded elements, where the system frequency response is the product of the individual
element frequency responses, the time response is calculated through convolution of
individual impulse reponses. Convolution g(τ ) of two energy signals is calculated as a
function of displacement τ between signal xT (t) and a time-reversed version of the other
signal, hT (t) .
∞
g(τ ) = xT (t) ⊗ hT (t) = ∫ xT (t)hT (τ − t)dt
−∞

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 29

Taking the Fourier transform of both sides shows that multiplication in the frequency domain
is equivalent to convolution in the time domain. It can be similarly shown that multiplication
in the time domain is equivalent to convolution in the frequency domain.
∞ ∞
⎡ ⎤
ℑ{x T (t) ⊗ hT (t)} = ∫ ∫ x T (t)hT (τ − t) dt e− j ωτ dτ
−∞ ⎢
⎣−∞ ⎥⎦
∞ ∞
⎡ ⎤
ℑ{x T (t) ⊗ hT (t)} = ∫ xT (t) ∫ hT (τ − t)e − jω (τ − t )dτ e− jω t dt
−∞ ⎢⎣−∞ ⎥⎦
∞
ℑ{x T (t) ⊗ hT (t)} = ∫ xT (t)H ( f )e− jωt dt
−∞

ℑ{x T (t) ⊗ hT (t)} = X( f )H ( f ) .

2.8.7 Autocorrelation and power spectral density*

To obtain an estimate of the power spectral density SN ( f ) of a random noise signal,
we select a finite-length sample nT (t ) of the waveform. While we could calculate the
Fourier transform of this finite duration “energy” signal, a mathematical expression of the
waveform segment is too complex for analytical purposes. However, the autocorrelation of
the waveform segment is essentially quite short and very suitable for Fourier transform which
yields the energy spectrum of the wavefrom segment. Dividing by the segment length T,
provides an estimate of the noise power spectral density. We begin with the previous
expression for convolution
⎧∞ ⎫
ℑ⎨ ∫ xT (t)hT (τ − t) dt⎬ = X( f )H ( f )
⎩−∞ ⎭
and replace xT (t ) with nT (t ) and also replace hT (τ − t ) with a delayed version of the noise
sample waveform nT ( t − τ ) . Because of the time reversal in this latter replacement, we then
replace H( f ) with N ∗( f ) . (Note that magnitude spectrum is independent of time reversal).
∞
⎧ ⎫
ℑ⎨ ∫ nT (t)nT (t − τ )dt ⎬ = N( f )N ∗ ( f )
⎩−∞ ⎭
We recognize the right side as the energy spectral density E n ( f ) and, by recalling
Rxx (− τ ) = R xx (τ ) , we recognize the left side as the Fourier transform of signal
autocorrelation.
∞
⎧ ⎫
ℑ⎨ ∫ nT (t)nT (t − τ )dt ⎬ = E n ( f ) .
⎩−∞ ⎭
This is a simplified case of the Wiener-Khintchine theorem. To estimate power spectral
density Sn ( f ) we calculate time average (expected value) by dividing both sides by T and let
T → ∞.
ℑ{Rnn (τ )} = Sn ( f ) .

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 30

2.9 CHAPTER SUMMARY

Practical signals are real-valued functions that can be displayed on an oscilloscope.
We define signals to be single valued functions of time. For mathematical convenience, we
sometimes consider signals to be complex valued. Signals may be classified into basic
categories: energy or power, periodic or non-periodic, deterministic or random.
System outputs occur in response to applied inputs and, if the system is linear,
superposition applies and no new frequencies are generated. We say that the system is time-
invariant, for example, if the input-output response is consistent at different times of day.
System gains and signal amplitudes are expressed in dB for convenience. A series connection
of several elements is analyzed by simply adding the dB values (instead of a lengthy
multiplication). On the other hand, for the summation of signals with quadrature phase or
different frequencies, the resulting power and rms voltage must be computed by adding the
powers of the constituent signals.
An arbitrary waveform can be represented as a summation of sinusoids (our most basic
waveform) and these sinusoids can be represented in trigonometric form or in complex
exponential form. Periodic waveforms are represented by a Fourier series of sinusoids with
frequencies at integer multiples of the fundamental (periodic) frequency. Plots of the sinusoid
amplitude (in volts) and phase coefficients (in radians) versus frequency are called the
amplitude spectrum and the phase spectrum.
To analyze the spectral content of non-repetitive signals, we adapt the Fourier series
calculation and this results in the Fourier Transform. For energy signals (one-shot signals),
we do not divide by the period T so the amplitude spectrum has units of volt-seconds or
volts/hertz; the energy spectrum has units of joules/hertz. Since the waveform is not
repetitive, the spectrum is continuous (i.e. no discrete lines).
For continuous non-repetitive (random) signals, we truncate the waveform to the
interval T, calculate the energy spectral density then divide by T to get the power spectral
density in units of watts/hertz. Amplitude spectral density is calculated by taking the square
root of the power spectral density and has units of volts per root hertz.
In amplifiers and A/D converters, to improve the ratio of signal to noise, we try to
maximize signal power and yet we need to avoid peak clipping. We therefore characterize
signals in terms of peak to rms ratio to quantify the signal power for a given peak voltage.
Full amplitude characterization is given by the probability density function (PDF) which takes
on a Gaussian distribution when a large number of independent signals are summed together.
We also characterize signals by similarity (correlation) to another signal or to delayed
versions of itself.

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 31

Problems
p2.1 Can an energy signal be periodic? Yes No
The following expressions describe energy or power waveforms. Circle the power waveforms.
-10t -j20t
5V-10Vu(t), 10Vδ (t) , 5Vcos1000t, 5Vu(t)e , 5Ve , 5Vsin(10t)/10t

p2.2 A test signal is specified as 10V cos 377 t.

a) What is the normalized power of this signal?
b) If the signal exists for exactly 30 sinusoidal cycles, determine the normalized energy.

p2.3 Two sinusoids are added. Calculate the average normalized total power and rms voltage when the
first sinusoid is 8V cos2π1000t and a) the second sinusoid is 6V cos 2π1000t , b) the second
sinusoid is 6V cos (2π1000t + π/2) and c) the second sinusoid is 6V cos (2π1200t).

p2.4 A signal is measured at 4.0 Vrms in a 50 ohm radio frequency cable. What is the signal level in
dBm and in dBW?

p2.5 An amateur radio signal traveling on a 50 ohm co-axial cable has sinusoidal waveform with peak to
peak voltage 7.07 volts. What is the signal level in dBm and dBV?

p2.6 An average reading power meter reads 3.5 W when connected at the output of a CATV amplifier.
The amplifier is driving a 75 ohm resistive load.
a) What is the power in dBW?
b) What is the power in dBm?
c) What is the level in dBV?

p2.7 Several things are wrong with the following statement. “A system with gain 20 dBm has input level
+5 dBm and output level -15 dB”. Make the corresponding correct statement.

p2.8 Calculate the following to one decimal place.

a) an audio amplifier has voltage gain of 400. What is the voltage gain in dB?
b) a microwave amplifier has power gain of +13 dB. What is the ratio of output power to input
power?

p2.9 A signal is composed of three linearly added tones at 200 Hz, 500 Hz and 1000 Hz. All three
components have level of -50 dBm. What is the average power of the combined signal?

p2.10 Three signals of different frequencies have levels -3, -5, and -10 dBm. What is the resulting level if
the three signals are combined? Note that signals of different frequencies add on a power basis.

p2.11 A speech signal with average power 0 dBm is added to a 60 Hz tone with level -6 dBm. What is the
total resultant power expressed in mW?

p2.12 A +20 dBm signal is added to a +30 dBm signal of different frequency. What is the sum in dBm?

p2.13 A video disk player outputs a -17 dBm, channel 3 signal on a 75 ohm cable while a video cassette
player outputs a -12 dBm, channel 4 signal. Both signals are linearly added without loss to form
signal X on a 75 ohm cable. Determine the rms voltage levels of signals 3, 4 and X.

p2.14 The decibel unit can be used to express signal amplitude as compared to a reference power or
voltage. The dBm unit has a one milliwatt reference and is normally used to quantify the level (or
amplitude) of communication signals.
a) A 2v p-p sinusoid is present at a 600 ohm load. Calculate the signal level in dBm.

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 32

b) Two tones at 900 and 1100 Hz are added to form a signal. The tones have levels, -6 dBm and -9
dBm respectively. Calculate the dBm level of the combined two tone signal.
c) Two "in phase" tones at 1000 Hz are added. Both tones have level -2 dBm. Calculate the dBm
level of the two-tone signal.

p2.15 A telephone station set generates two tones at 941 Hz and 1204 Hz when the “*” button is pressed.
The expression v(t ) = 0.8 cos(2π * 941t ) + 1.0 cos(2π *1204t) describes the generated voltage (in
volts) and the waveform is illustrated below. Calculate the following:
a) The normalized average power (in watts) of: i) the 941 Hz signal component, ii) the 1204 Hz
signal component and iii) the total combined signal.
b) The actual power (in dBm) of: i) the average combined signal and ii) the instantaneous peak
power of the combined signal. Assume that the telephone line impedance is 600 ohms.
c) The “beat” frequency (i.e. the envelope variation frequency).
2.0 V

1.5 V

1.0 V

0.5 V

0.0

-0.5 V

-1.0 V

-1.5 V

-2.0 V
0 2 4 6 8 10 12
Time (ms)

p2.16 When viewed from the end of the positive time axis, which way does the vector e j t rotate? Is e j t
analogous to a “right hand thread” or a “left hand thread.” Is this the same thread as the screw cap
on a pop bottle?

p2.17 A modulated signal can be expressed in complex envelope form as v(t) = Re{g(t) e jωt} where the
complex envelope g(t) = 4 + 3 cos 377t and the signal frequency ω = 2000π. Sketch the first 20 ms
of this waveform.

p2.18 Four rectangular waveforms sw, sx, sy and sz are shown; a) indicate the frequency on each
waveform, b) what is the Fourier coefficient c3 for the waveform sy and c) which waveform has
coefficients: c0 = 0.250, c1 = c-1 = 0.225, c2 = c-2 = 0.159.
1.0 1.0
sw sy
0
1μs 9μs 11μs 1μs 5μs 7μs

1.0 1.0
sx sz
0
1μs 7μs 9μs 1μs 3μs 5μs 7μs 9μs
t=0 t=0
p2.19 Review Example 2.13 and compute one more decimal place for all values of c5.

p2.20 Determine the “exponential” Fourier series frequencies and amplitudes (up to 40 kHz) for an 8 kHz
pulse signal y(t) with 1/10 duty cycle and levels 0 and +50 volts. Assume the waveform has even
symmetry. Verify your calculation using Parseval's theorem.

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 33

p2.21 Determine the Fourier components of a pulsed voltage waveform with repetition rate of 1 kHz and
duty cycle 1/3. The waveform has base voltage 0 volts with pulses to +6 volts. Do not compute
components for frequencies higher than 9 kHz. Verify your result using Parseval's theorem.

p2.22 In the illustration below, the signal x(t) is a 60 Hz, 12 volt (rms) cosine wave. This signal is passed
through an ideal full wave rectifier and then a highpass filter that removes the dc component and
frequencies up to 2 Hz. Sketch the waveforms at points A, B, and C. Determine the normalized
power of signals x(t), |x(t)| and y(t). Hint: signals of different frequency add (subtract) on a power
basis; this includes signals at zero frequency (dc).

x(t) = 17Vcos(2π60t) |x(t)| 2 Hz y(t)

Full Wave
Rectifier HPF
A B
p2.23 Using calibrated axes, sketch the time waveform of w(t) = sinc(2t) = sin(2πt)/2πt.

p2.24 A single rectangular pulse in the time domain and a sinc function in the frequency domain are a
Fourier transform pair as illustrated in Example 2.15. Similarly, a rectangular spectrum and a sinc
function in the time domain form a Fourier transform pair as illustrated in Example 2.16. Calculate
the Fourier transform of the waveform w(t) = 6V sin(3πt)/3πt. Show that the normalized energy in
the waveform is 12 joules.

p2.25a The signal m(t) with Fourier transform M(f) enters a signal processing device which has the output
M 2 ( f ) = M( f )e j5 f . What specific function is performed by the signal processor? ( 1 pt)

p2.26 A noise-like signal, v(t), is composed of sinusoidal components each with amplitude 1 Vrms and
spaced at one-tenth hertz intervals. Each component has normalized power equal to one watt. The
one-sided spectrum is illustrated below.
1 Hz Filter
n = +∞
v (t) = ∑ 2 cos(0.2 π nt)
n=0
0 0.5 1.0 1.5 2.0 f (Hz)
a) What is the power spectral density (watts/Hz) and the amplitude spectral density ( volts hertz )
of the noise-like signal?
b) The noise-like signal passes through a filter with bandwidth two hertz. What is the normalized
power and the rms voltage of the filter output signal?
c) Sketch the amplitude density versus voltage (i.e. PDF) for the filter output and indicate the
standard deviation (in volts).

p2.27 Draw lines between each time waveform and the appropriate frequency domain representation(s).
i) Fourier series (V) a) v(t) = 10 sgn (sin 377 t)
ii) Energy spectral density (J/Hz) b) v(t) = 5 sin(30πt)/30πt
iii) Amplitude spectral density (V/Hz) c) v(t) = 5 u(t)exp(-6t)
iv) Power spectral density (W/Hz) d) v(t) = 20 sin(2π1000t)
v) Amplitude spectral density V/ Hz( ) e) v(t) = white noise (PSD=N0)

p2.28 A rms responding meter reads 1.5 volts when measuring a white noise source that is bandlimited by
a 20 kHz lowpass filter. Determine the meter reading when the filter cutoff is reduced to 5 kHz.

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 34

p2.29 The (normalized) single sided power spectral density, S(f), of a telephone (μW/Hz) S(f)
30
modem signal is illustrated. Calculate normalized power and the level in dBV
for the signal. 300 3300 f (Hz)

p2.30 The noise in an audio system has amplitude spectral density 0.045 V / Hz .
a) calculate the rms noise voltage is the noise is bandlimited to 10 kHz.
b) determine the two-sided power spectral density (PSD), N0/2 .
p2.30b A FM receiver is tuned to a distant station and the audio output has excessive white noise. The
noise is measured at 100 mVrms (Pn = 10 mW). Using the tone control, the output bandwidth is
reduced from 10 kHz to 2 kHz. What are the new values of noise voltage and noise power?

p2.31 The input to the system shown below is zero mean, Gaussian white noise n(t) with normalized two-
sided power spectral density (N0/2) equal to 5 mW/Hz. The system is a 10 Hz ideal lowpass filter.
Determine the normalized power at the output y(t), sketch the amplitude probability density (i.e.
probability density vs voltage) and determine the mean and standard deviation. Hint: the standard
deviation is equal to the rms voltage.

n(t) 10 Hz y(t)
LPF

p2.33 The differentiator and 10 Hz ideal lowpass filter illustrated below have equivalence to noise
processing in a FM receiver. The input is a zero mean, Gaussian white noise n(t) with normalized
two-sided power spectral density (N0/2) equal to 5 mW/Hz. Determine the normalized power at the
output y(t). At each frequency, the differentiator has voltage gain 2πf and power gain (2πf ) 2.

n(t) d x(t) 10 Hz y(t)

dt LPF

p2.34 Peak to RMS ratio is a convenient way to determine maximum signal power when the peak-to-peak
voltage is limited by amplifier or A/D voltage limits. Determine, from first principles, the peak to
rms ratio of a triangular waveform with zero mean. (Integration is required)

p2.35 The following 3 waveforms have zero average value and peak to peak voltage of 20 volts. Given the
peak-to-rms ratio (crest factor), calculate the normalized power in each case .
Waveform Peak/rms Norm. Power
Square wave 1
Sine wave 2
Triangular wave 3

p2.36 A periodic waveform w(t) consists of four 1 ms segments each with one cycle of sinewave. The
sinewave amplitudes in the 4 successive segments are 2v, 4v, 2v, and 1v. The periodic waveform
has a repetition rate of 250 Hz and has no discontinuities. Determine the average normalized power
< x 2 (t) > and peak to rms ratio of the waveform.
4v

2v 2v 2v

1v

4 ms

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 35

p2.37 Determine the normalized power and peak to rms ratio of the waveform below.
4v

2v 2v 2v

t
4 ms

p2.38 Consider a triangular waveform with zero average and peak voltages of ± 0.5 volts.
a) Illustrate the probability histogram assuming 0.1 volt bins.
b) Assume the addition of two such waveforms with identical amplitude but unrelated frequency.
Illustrate the histogram of the combined signal.
c) Roughly sketch the amplitude PDF of four added waveforms (recall the central limit theorem).

p2.39 Consider a speaker with the specifications approximately reproduced below. i) Assuming 80 watts
peak instantaneous power, what average power might be expected for a music signal with peak to
rms ratio of 4.0. ii) The specifications state the frequency range as 45 Hz to 22,000 Hz. What is
your estimation of the ±3 dB frequency range of the speaker?

p2.40 A wavelength division multiplexed optical transmission system has 64 unmodulated optical carriers
with 100 GHz precise spacing starting at 193 THz. Each carrier has rms amplitude 0.707Ep and
peak amplitude Ep. Assuming that, at some point in time, all carriers simultaneously reach their
peak value, what is the peak/rms ratio of the combined optical signal.

p2.41 Modern communication systems such as DSL Internet access and IEEE 802.11b wireless computer
modems use multicarrier OFDM (orthogonal frequency division multiplexing). In these systems it
is advantageous to phase the precisely spaced carriers to minimize the crest factor (peak/rms ratio).
To demonstrate the effect of phasing, consider the transmitted signal
s(t ) = cos(2π 300t + θ 3 ) + cos(2π 400t + θ 4 ) + cos(2π 500t + θ5 )
and use a computer to evaluate the crest factor when a) θ 3 = 0, θ 4 = 0, and θ5 = 0 and b)
θ 3 = π / 2, θ 4 = 0, and θ5 = π / 2

p2.42 Calculate the correlation coefficient of the digits in the numbers 365 and 352.

p2.43 For continuous signals, an expression for the

μxy (x(t) − mx )(y(t) − my )
correlation coefficient is shown. Given the signals ρ= =
σ xσy
(y(t) − my )
2
illustrated below, sketch x(t) − m x and y(t) − m y and (x(t) − mx )2
their product and then calculate the correlation
coefficient.

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11

Chapter 2: Representation of Signals 36

+1
x (t ) +1
y (t )

1 2 2
0 t 0 0.5 t

-1 -1

p2.43b A service technician is investigating the source of excessive noise in a FM stereo receiver. The right
and left channel outputs are connected to an X-Y oscilloscope to display correlation of the two noisy
outputs. There is no received signal. Select the oscilloscope display for each case below.
a) noise originating only in the left channel amplifier. Circle one: d), e), f)
b) noise in the receiver power supply, affecting both channels equally. Circle one: d), e), f)
c) noise originating in the L-R amplifier (see FM Stereo, Chapter 3). Circle one: d), e), f)
Y (left) Y (left) Y (left)

X (right) X (right) X (right)

d) e) f)

p2.44a Random binary information is transmitted by driving a laser with a series of pulses at the rate of 500
MHz. Logic 1 is represented by a 4-volt pulse and logic 0 by a 2-volt pulse. Between the pulses,
the voltage is 1 volt. Sketch the autocorrelation of this information waveform over the range
−7 ns < τ < 7 ns . What are the units for the vertical axis.
4V Example random
data waveform
Ņ1Ó Ņ1Ó Ņ1Ó
2V
1V Ņ0Ó Ņ0Ó

0 2 ns 4 ns 6 ns 8 ns

p2.44b A half wave cosine wave with amplitude 4 volts and frequency 60 Hz is centered about the t = 0
axis. Calculate the first c0-c4 Fourier series coefficients of the half wave signal by using
convolution of a cosine wave spectrum with that of a unit amplitude square gating waveform.
1.0 1.0
p2.45 Two rectangular pulse signals with levels 0 and +1 as
illustrated are convolved to produce a pulse waveform *
with peak value +2.0 •10-6. Illustrate the resulting -10-6 +10-6 t -10-6 +10-6 t
t=0 t=0
waveform.

p2.47 In the display of frequency spectra, several different units are used for the vertical axis depending on
the characteristics of the time waveform. Possible units are volts, volts per hertz, volts per root
hertz, watts, joules per hertz and watts per hertz.
a) What are the vertical axis units of the amplitude spectrum of a time function consisting of a
single pulse, not repeated?
b) What are units for the amplitude spectrum if the time function is a periodically repeating pulse?
c) What are the units for the amplitude spectrum if the time function is a repeating pulse with
random amplitude.

Ch2-Sig-11g
Introduction to Communication Systems – A Multimedia Workbook 37

p2.48b Virtual Laboratory 2.4 - Observe the square wave

approximation that has been generated using a multi-
sinusoid waveform synthesizer. a) The desired square
waveform has amplitude two-volt peak-to-peak and zero
average voltage. What is the normalized power? b) What
are the theoretical peak-to-peak voltages of the first three
components? (checksum = 3.905) c) What is the
normalized power of the first three components summed and
what rms voltage is expected? (0.933 W and 0.966 V).

p2.49a Virtual Laboratory 2.6 - Access the virtual experiment on

lowpass filtered noise. Observe the images of noise
bandwidth and measured noise voltage. a) determine the
normalized power spectral density, N0, of the noise source
(in mW/Hz) b) estimate the meter reading in volts and in
dBm when the bandwith is increased to 25 kHz.

p2.50 Design Problem – (the first 7 pages of Chapter 3 will aid the solution of this problem) A traveling
rock music group wants a portable instrument to analyze the echo acoustics of each venue prior to
their concert. Using a random noise source x(t) on stage and a microphone y(t) at various places
in the audience, results are to be recorded on a laptop computer with a 4-channel data acquisition
capability. a) Prepare a block diagram of your design that includes an electronically controlled
delay element, a 4-quadrant multiplier, and a sliding window averaging filter. b) Sketch the
predicted test results if the microphone is located 100 m from the stage and there is a 20 ms delay in
the first (and only) echo. Design criteria are as follows:
i) simultaneous measurement of cross-correlation and autocorrelation for delays up to 400 ms in
steps of 2 ms. Bandwidth of signal processing 20 Hz to 10 kHz.
ii) Averaging filter (LPF) to minimize variation in the output due to double frequency products and
60 Hz interference.
iii) Measurement time less than 60 seconds. You will need to consider consider that the response
time of the LPF is approximately the reciprocal of its bandwidth.

Extra Problems

p2.0c Find the mean value (time average) and the power for the following signals:
(a) v(t) = 3V + 3V cos(4000πt),
(b) v(t) = 3V sin(40t) + 4V cos(4000πt)

p2.0d Sketch the waveform and find the energy of the following signals:
⎧sin(t ) for 0 < t ≤ 2π
(a) g(t ) = ⎨
⎩0 otherwise
⎧3Vsin 4π t + 4Vcos 8π t for 0 < t ≤ 0.5
(b) v(t ) = ⎨
⎩0 otherwise

Ch2-Sig-11g © D.E. Dodds, Saskatoon, Canada 1/11/11