Communication Channels

I wires (PCB trace or conductor on IC)

I optical fiber (attenuation 4dB/km)
I broadcast TV (50 kW transmit)
I voice telephone line (under -9 dbm or 110 µW)
I walkie-talkie: 500 mW, 467 MHz
I Bluetooth: 20 dBm, 4 dBm, 0 dBm
I Voyager: X band transmitter, 160 bit/s, 23 W, 34m dish antenna

http://science.time.com/2013/03/20/humanity-leaves-the-solar-system
-35-years-later-voyager-offically-exits-the-heliosphere/
Communication Channel Distortion
The linear description of a channel is its impulse response h(t) or
equivalently its transfer function H(f ).
y(t) = h(t) ∗ x(t) ⇐⇒ Y (f ) = H(f )X(f )
Note that H(f ) both attenuates (|H(f )|) and phase shifts (∠H(f )).
Channels are subject to impairments:
I Nonlinear distortion (e.g., clipping)
I Random noise (independent or signal dependent)
I Interference from other transmitters
I Self interference (reflections or multipath)
Channel Equalization
Linear distortion can be compensated for by equalization.
1
Heq (f ) = ⇒ X̂(f ) = Heq (f )Y (f ) = X(f )
H(f )
The equalization filter accentuates frequencies attenuated by channel.
However, if y(t) includes noise or interference,
y(t) = x(t) + z(t)
then
Z(f )
Heq (f )Y (f ) = X(f ) +
H(f )
Equalization may accentuate noise!
Channel Equalization Example
h(t) = u(t)e−t , x(t) is square wave, y(t) = h(t) ∗ x(t).
1 2

0.8
1

0.6
0
0.4

−1
0.2

0 −2
0 2 4 6 −50 0 50

0.8
1 0.6

0.5 0.4

0.2
0
0
−0.5 −0.2

−1 −0.4

−0.6
0 1 2 3 4 5 0 2 4 6
Channel Equalization Example (cont.)
Equalizing filter has transfer function 1 + j2πf , approximates differentiator.
350 2

300
1
250

200
0
150

100
−1
50

0 −2
−50 0 50 0 2 4 6

120 1.5

100 1

80 0.5

60 0

40 −0.5

20 −1

0 −1.5
0 2 4 6 0 2 4 6
Signal Energy and Energy Spectral Density
Parseval’s theorem for an energy signal g(t) is
Z ∞ Z ∞
2
Eg = |g(t)| dt = |G(f )|2 df
−∞ −∞

Signal has same total energy Eg in the time domain frequency domain.
The essential bandwidth is the range of frequencies with “most” of the
signal’s energy of the signal. The definition of ”most” depends on the
application. One choice might be 90%.
If G(f ) is a lowpass signal, and EB is the energy from −B to B, then
Z B
EB = |G(f )|2 df
−B

Then the essential bandwidth is the B such that

EB /Eg = 0.9
This is illustrated below:

90% of the energy

G(f )

−B B f
10% of the energy

Other definitions of width

I 95% or 99% energy
I Half amplitude width
I Half power width
I 50% energy
Autocorrelation and Energy Spectral Density
The autocorrelation of a signal g(t) is
Z ∞
ψg (t) = g(τ )g ∗ (t + τ ) dτ
−∞

You’ll show in your homework that

F {ψg (t)} = |G(f )|2 = Ψ(f )
This is the energy spectral density or ESD. It reflects where the energy of
the signal is located.

|G(f )|2 df |G(f )|2

Note that Z ∞ Z ∞
Eg = |G(f )|2 df = Ψ(f )df
−∞ −∞
Energy Spectral Density Example
Let g(t) = Π(2t)
g(t) = Π(2t)

−2 −1 0 1 2 t

The autocorrelation ψ(t) is

1
ψg (t) = ∆(t)
2

−2 −1 0 1 2 t

The energy spectral density is then

1 !π "
Ψ(f ) = sinc2 f
4 2

−4 −2 0 2 4 f
Autocorrelation and Power Spectral Density
For power signals, we normalize the ESD by the duration, to produce the
power spectral density or PSD.
The autocorrelation for a power signal g(t) is defined as
Z
1 T /2
Rg (t) = lim g(τ )g ∗ (t + τ )dτ
T →∞ T T /2

This has the Fourier transform

1
F {Rg (t)} = lim Ψg,T (f ) = Sg (f )
T →∞ T
Sg (f ) is the power spectral density, PSD.
Again, this shows the frequency distribution of the power of the signal.
Power Spectral Density Example
Let g(t) be a random binary sequence of rectangle pulses g1 (t) = Π(2t)
! g1 (t) = Π(2t)
g(t) = an g1 (t − n)
n

−2 −1 0 1 2 t

For small displacements, the autocorrelation looks like

−2 −1 0 1 2 t

After normalizing by the interval T , this is the same a for a single pulse.
For large displacements, overlaps are just as likely to be ±1, and will cancel.

−2 −1 0 1 2 t

This will go to zero as T gets large. The autocorrelation is then

1
Rg (t) = ∆(t)
2
−2 −1 0 1 2 t

The power spectral density is then

1 !π "
Sg (t) = sinc2 f
4 2

−4 −2 0 2 4 f
Baseband Communication
The baseband is the frequency band of the original signal.
I Telephones: 300–3700 Hz
I High-fidelity audio: 0–20 KHz
I Television (NTSC) video: 0–4.3 MHz
I Ethernet (10 Mbs): 0–20 MHz

Baseband communication usually requires wire (single, twisted pair, coax).

Multiple baseband signals cannot share a channel without time division
multiplexing (TDM).
Carrier Communication
Carrier communication uses modulation to shift spectrum of signal.
I Wireless communication requires frequencies higher than baseband
I Multiple signals can be sent at same time using different frequencies:
frequency division multiplexing (FDM)

In carrier communication, the signal modulates a sinusoidal carrier.

The signal modifies the amplitude, frequency, or phase of carrier.


s(t) = A(t) cos 2πfc (t)t + φ(t)

I amplitude modulation: A(t) is proportional to m(t)

I frequency modulation: fc (t) is proportional to m(t)
I phase modulation: φ(t) is proportional to m(t)
Frequency and phase modulation are called angle modulation.
Double-Sideband Amplitude Modulation
The simplest modulation method is multiplication by sinusoid:
x(t) = m(t) cos(2πfc t + φ)
We usually set phase φ to 0 to simplify mathematical discussion.
The Fourier transform of the modulated signal is
X(f ) = 12 (M (f + fc ) + M (f − fc ))
Double-Sideband Amplitude Modulation (cont.)
This scheme is called double-sideband, suppressed-carrier (DSB-SC).
Signal Bandwidth vs. Carrier Frequency
Transmitters can radiate only a narrow band without distortion. Thus we
choose the carrier frequency such that
fc B
 1 ⇐⇒ 1
B fc
Examples:
I AM radio: B = 5 KHz, 550 ≤ fc ≤ 1600 KHz
⇒ 100 < fc /B < 320
I FM: B = 200 KHz, 87.7 ≤ fc ≤ 108.0 MHz
⇒ 43 < fc /B < 54
I US television: B = 6 MHz, 54 ≤ fc ≤ 862 MHz
⇒ 9 ≤ fc /B ≤ 142
Digital TV uses the same frequency bands as analog TV.
Demodulation of DSB-SC Signals
Demodulation uses a multiplier and a low-pass filter.
e(t) = x(t) cos(2πfc t) = m(t) cos2 (2πfc t) = 12 m(t) + 21 cos(4πfc t)

The low pass filter does not have to be very sharp. But it should be flat
over the signal baseband.
DSB-SC Example
Modulating a sinusoid is an important way to test the system. Let
m(t) = cos(2πfm t)
Then
M (f ) = 12 δ(f + fm ) + 12 δ(f − fm )
and
ϕDSB-SC (t) = m(t) cos(2πfc t) = cos(2πfm t) cos(2πfc t)


= 12 cos((fc + fm )t) + cos((fc − fm )t)
The transform of the modulated signal contains two impulse pairs separated
by 2fc .
DSB-SC Example: Frequency Domain
Modulation and demodulation of cosine.
DSB-SC Example: Time Domain
x(t) = m(t) * cos(2*pi*fc*t)
1

0.5

−0.5

−1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

e(t) = x(t) * cos(2*pi*fc*t)

1

0.5

−0.5

−1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

e(t) low−pass filtered

0.5

−0.5

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Types of Modulators
I Multiplier modulators using variable gain amplifiers.
I Nonlinear modulator. Suppose the input-output characteristic is
y(t) = ax(t) + bx2 (t)
Let
x1 (t) = cos(2πfc t) + m(t)
x2 (t) = cos(2πfc t) − m(t)
It we apply x1 (t) and x2 (t) to the nonlinear modulator and look at the
difference
y1 (t) − y2 (t) = a(cos(2πfc t) + m(t)) + b(cos(2πfc t) + m(t))2
− a(cos(2πfc t) − m(t)) − b(cos(2πfc t) − m(t))2
= 2a m(t) + 4b m(t) cos(2πfc t)
Convince yourself this is true!
Types of Modulators (cont.)
From the previous page
y1 (t) − y2 (t) = 2a m(t) + 4b m(t) cos(2πfc t)
This has the term we want at ωc = 2πfc , plus another copy of the message
at baseband.
The unwanted baseband component is blocked by bandpass filter. This
could be the antenna or the amplifier.

Or we can just forget about the baseband signal, it won’t propagate!

Types of Modulators (cont.)
Switching modulator: multiply message by a simple periodic function.
Suppose w(t) is periodic with a fundament frequency fc :
∞
X
w(t) = Dn ej2πfc nt
n=−∞

This weighted sum of complex exponentials that are impulses at all

multiples of fc . Then
∞
X
m(t)w(t) = Dn m(t)ej2πfc nt
n=−∞

By the convolution theorem, the spectrum of m(t)w(t) consists of M (f )

shifted to ±fc , ±2fc , ±3fc , . . . Suppose w(t) is a square wave centered at
t = 0. Then from Lecture 3,
∞
1 1 X 1 j2πfc nt
w(t) = + e , n odd
2 π n=−∞ n
Switching Modulator
Ring Modulator
Frequency Converter
Multiplying a modulated signal by a sinusoidal moves the frequency band to
sum and difference frequencies.

Super-heterodyning: ωmix = ωc + ωI .
Sub-heterodyning: ωmix = ωc − ωI .
Demodulation of DSB-SC Signals
Both modulator and demodulator use a multiplier by carrier signal.

I Modulator uses bandpass filter

I Demodulator uses lowpass filter

The carrier used by the demodulator must be in phase with the transmitter
carrier (taking into account transmission delay).
Such a receiver is called synchronous, coherent, homodyne.
The receiver has a local oscillator that must be adjusted to stay in phase
with the received signal.
A voltage-controlled oscillator (VCO) that is controlled by a phase-locked
loop (PLL) is commonly used.
The phase of the carrier in the received signal must be extracted.
Demodulation of DSB-SC Signals (cont.)
Suppose that the signal is not ideal,


r(t) = Ac m(t − t0 ) cos 2πfc (t − t0 )


= Ac m(t − t0 ) cos 2πfc t − θd
where θd = 2πfc t0 .
e2(t) = x(t) * sin(2*pi*fc*t)
0.5

−0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

e2(t) low−pass filtered

0.1

0.05

−0.05

−0.1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

We can end up transmitting with a cosine, and receiving with a sine. These
are orthogonal, and we get nothing!
Commercial AM
I If the goal is cheap receivers, then we can eliminate the PLL by
transmitting the carrier signal along with the modulated message.
ϕAM (t) = A cos(2πfc t) + m(t) cos(2πfc t)
= (A + m(t)) cos(2πfc t)
I The tone A cos(2πfc t) contains the desired carrier in correct phase.
I As long as A is larger than |m(t)|, then we can recover m(t) from
ϕAM (t), as we will show next time.
Next time
I Commercial AM, and power
I Single Sideband AM (SSB)
I Vestigial Sideband AM (VSB)
I Quadrature Amplitude Modulation (QAM)