NOISE AND INTERFERENCE IN

COMMUNICATION SYSTEMS
EEE 3307 (Communications Theory)
Professor Raqibul Mostafa, PhD (EEE Dept, UIU)
OVERVIEW OF NOISE
 Noise: It refers to undesirable electrical signal that
falls within the passband of a desired signal. Noise is
usually additive and corrupts the desired signal by
getting added to it
 Categories of noise: In broad sense it can be
categorized as
 Correlated noise
 Uncorrelated noise
 Correlated noise: It implies a noise that is related to
the desired signal. Thus noise is present when signal
is present and vice versa
 Uncorrelated noise: Here noise has no relationship
with signal and is always present. This is random in
nature.
 Our main focus will be on uncorrelated noise
ADDITIVE NOISE
 Channel imperfections or
distortions:
 Additive noise in the
channel
 Nonlinear distortion

 Additive noise:
 𝑦 𝑡 =𝑥 𝑡 +𝑛 𝑡
 𝑛 𝑡 is a random noise
process that is described by
 Distribution of the random
process: probability density
function (pdf)
 Power spectral density
function (psd)
UNCORRELATED NOISE
 It has the following subcategories
 External noise: Noise originating from outside the device and
originates from either atmosphere(e.g. lightning), solar (sun-spot
activity) or galactic radiation or it can be man made (through sparks
in motors, switches etc.)
 Internal noise: It is internal to the device due to the random motion of
electrons in conductors. This is often classified as thermal noise and is
present in all devices.
 Thermal noise:
 Thermal noise power density is expressed as 𝑁𝑜 = 𝑘𝑇; watts/Hz.
 Total noise power within a given bandwidth B is 𝑃𝑛 = 𝑘𝑇𝐵
 Here k= Boltzmann’s constant= 1.3810−23 Joules/0Kelvin
 T= Kelvin
 B= Bandwidth in Hz
 For example thermal noise power at room temperature within a
bandwidth of 100 kHz system is given by 𝑃𝑛 = 𝑘 × 300 × 100 × 103 =
4.14× 10−16 watts.
 In dBm it’s 10*log10(Pn)=-153 dBW = -123 dBm
THERMAL NOISE
 Additive thermal noise are described as AWGN or
Additive White Gaussian Noise
 White: Noise spectral components are present in all
frequencies
 Gaussian: The probability density function (pdf) is
described by Gaussian distribution as
𝑛2
1 −
 𝑝 𝑛 = 2
𝑒 2𝜎2 . It’s a zero mean signal with variance 𝜎 2
2𝜋𝜎
 It’s a symmetric distribution with the following properties
∞

−∞
𝑝 𝑛 𝑑𝑛 = 1
0 ∞

−∞
𝑝 𝑛 𝑑𝑛 = 0
𝑝 𝑛 𝑑𝑛 = 0.5
 Probability that noise falls within the values between a and b
𝑏
is given by 𝑎 𝑝 𝑛 𝑑𝑛
 PDF OF NOISE RANDOM VARIABLE
 The noise process is described as
AWGN
 pdf is assumed to be Gaussian
 The psd is white in nature
 Probability that noise is below x
𝑥
is given by Φ 𝑥 = −∞ 𝑝 𝑛 𝑑𝑛
 To compute this tabulated values
of Q-function can be used
∞
 𝑄 𝑥 =1−Φ 𝑥 = 𝑥
𝑝 𝑛 𝑑𝑛
 Here p(n) is the normal
distribution with zero mean and
unit variance
 RV needs to be normalized as
𝑛−𝜇
𝑥 = 𝜎 before using the table
Q-FUNCTION TABLE
 Q(0)= 0.5
 Probability
noise is greater
than 1 is
𝑄 1 =
1.58 × 10−1
 Find the
probability
that noise falls
between 0 and
1
SPECTRAL CHARACTERISTICS: WHITE NOISE
 White process: All frequencies with equal
power
SPECTRAL CHARACTERISTICS: PRACTICAL
CASE

 Quantum mechanical analysis

 Though precisely not flat but can be considered so for
practical ranges up to 21012 Hz.
 At room temperature 𝑆𝑛 𝑓 drops to 90% of the
maximum at 21012 Hz
𝑁0
 𝑆𝑛 𝑓 = where 𝑁0 = 𝑘𝑇 watts/Hz
2
NOISE EQUIVALENT
BANDWIDTH
NOISE
EQUIVALENT
BANDWIDTH
NOISE PSD AND PDF
PSD(f)

No/2

f
White Noise PSD

Gaussian pdf
SIGNAL TO NOISE RATIO (SNR)
 This is defined as a ratio between signal power
𝑃
and noise power as 𝑆𝑁𝑅 = 𝑠 (unit-less number)
𝑃𝑛
 In dB: 10*log10(SNR)
 Let 𝑥 𝑡 = 𝐴𝑐𝑜𝑠(𝜔𝑡) be corrupted with noise. This
signal is filtered with a LPF with a BW of B Hz
𝐴2
 Here 𝑃𝑠 = watts and 𝑃𝑛 = 𝑘𝑇𝐵 𝑤𝑎𝑡𝑡𝑠
2
𝐴2
2
 𝑆𝑁𝑅 =
𝑘𝑇𝐵
 Filter BW needs to be as small as possible to pass the
signal
 SNR CALCULATION FOR TONE
DEMODULATION
 Let the input to the demodulator:
x(t)= cos(10t)cos(500t)
 𝑥 𝑡 = 2𝑚 𝑡 𝑐 𝑡 =
LPF
2cos(10 t)cos(500 t) mV
H()
 Let the filter BW be 10 kHz
 After LPF we have 𝑥 𝑡 = 𝑚 𝑡 + 𝑛(𝑡) cos(500t)
(10−3 )2
 Message power, 𝑃𝑠 = 2
watts Y()

 Noise power, 𝑃𝑛 = 𝑘𝑇𝐵 𝑤𝑎𝑡𝑡𝑠 = 1.38 ×

10−23 × 300 × 103 𝑤𝑎𝑡𝑡𝑠

𝑃 5.0000𝑒−05
𝑆𝑁𝑅 = 𝑃𝑠 = 4.1400𝑒−17 = 1.2077 × 1012 1
𝑛
𝑁
 In dB= 10log10(SNR)= 120 dB Noise level: 20

 Note: For very high values of carrier 

frequency, filter BW needs to be high -50  50 
as well to design a practical filter
 This will increase noise power and reduce
SNR
NOISE VOLTAGE
 Noise source can be represented as a volatge
source in an equivalent circuit as Vn with series
resistance Rn connected to an external load of R.
 Assuming worst case scenario of maximum power
transfer, the voltage across R is Vn/2
𝑉𝑛 2
 The noise power at load is /𝑅 = 𝑘𝑇𝐵
2
 Thus 𝑉𝑛 = 4𝑅𝑘𝑇𝐵 volt
𝑉𝑠
 Here SNR=20𝑙𝑜𝑔10 dB
𝑉𝑛
CHANNEL DISTORTION
 Linear distortion: When a signal x(t) passes
through a linear time invariant channel with
impulse response h(t), the output is y(t)= x(t)*h(t)
 In frequency domain Y(f)= X(f)H(f)
NON-LINEAR DISTORTION
 Here the output of the channel is given as a
power series of input
 𝑦 𝑡 = 𝑎𝑜 + 𝑎1 𝑥 𝑡 + 𝑎2 𝑥 2 𝑡 + ⋯
 Let 𝑦 𝑡 = 𝑥 𝑡 + 0.1𝑥 2 𝑡 and 𝑥 𝑡 = cos 𝜔𝑜 𝑡
 Thus 𝑦 𝑡 = 0.05 + cos 𝜔𝑜 𝑡 + 0.05 cos 2𝜔𝑜 𝑡

 At the output there are two additional terms of a

dc value and a second harmonic term
 Example 3.15 from text book
CHANNEL DISTORTION
CHANNEL DISTORTION
MULTIPATH CHANNEL
SIMULATION STUDY OF NOISE IN
MATLAB
NOISE SIMULATION IN MATLAB
 Noise
 Random number
generated from
randn()
 Mean= 0, Standard
Deviation= 0.1
Sine wave Noisy Signal
 Signal
 Sine wave of
amplitude 1 Noise
 Frequency 100 Hz

 Sound playback
 Fs= 8192
samples/sec. (default
in Matlab)
 NOISE AND SIGNAL + NOISE PLOTS
0.4

0.3

0.2

0.1
Noise amplitude

-0.1

-0.2

-0.3

-0.4

-0.5
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Time, sec.

1.5

1
Signal+Noise amplitude

0.5

-0.5

-1

-1.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Time, sec.
SPECTRAL CHARACTERISTICS OF SIGNAL +
NOISE

5000
Signal Spectrum amplitude (linear)

4000

3000

2000

1000

0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Frequency, Hz

4
10
Signal Spectrum amplitude (dB)

3
10

2
10

1
10

0
10

-1
10
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Frequency, Hz
 PROBABILITY DENSITY FUNCTION (PDF) OF
NOISE

400

350

300

250
Noise pdf

200

150

100

50

0
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
Noise amplitude
MATLAB CODE
 figure
 clear all
 subplot(2,1,1)
 close all
 plot(t,noise)
 clc
 xlabel('Time, sec.')
 Fs= 8192;
 ylabel('Noise amplitude')
 Ts= 1/Fs;
 subplot(2,1,2)
 F= 100;
 plot(t,y)
 A=1;
 xlabel('Time, sec.')
 N= 1e4;
 ylabel('Signal+Noise amplitude')
 t= (0:N-1)*Ts;
 sound(y)
 noise_std= 0.1;
 pause(5)
 noise_mean=0;
 figure

 histfit(noise)
 noise= normrnd(noise_mean,noise_std,1,N);%randn(1,N);
 xlabel('Noise amplitude')
 x= A*sin(2*pi*F*t);
 ylabel('Noise pdf')
 y= x+noise;
 figure

 subplot(2,1,1)
 Pn= var(noise);
 plot(F_scale,(abs(Y_f)))
 Ps= A^2/2;
 xlabel('Frequency, Hz')
 SNR= 10*log10(Ps/Pn)
 ylabel('Signal Spectrum amplitude (linear)')
 Y_f= fft(y);
 subplot(2,1,2)
 Noise_f= fft(noise);
 semilogy(F_scale,(abs(Y_f)))
 F_scale= (0:N-1)*Fs/N;
 xlabel('Frequency, Hz')
 ylabel('Signal Spectrum amplitude (dB)')