Tele 2060

Noise

• Definition: Noise is unwanted signal energy in the passband of a

communications device
• Types
 White Noise (Thermal)

 Shot Noise

‰ Due to Active Devices

‰ Electrons Moving Across Boundaries

Impluse Noise


 Crosstalk

• Effect is Signal Corruption

Martin B.H. Weiss

Noise and Systems - 1 University of Pittsburgh

Tele 2060

Theoretical Basis

• Random Signal is Added to Desired Signal

• The Characteristics of this Random Signal Depends on the Noise
Type
 Thermal Noise has a Gaussian Amplitude Distribution

 Shot and Impulse Noise has a Poisson Arrival Distribution

• The Noise Signal has an Average Frequency Characteristic

 Power Spectral Density, P(f), which is the Fourier Transform

of the Autocorrelation function

 Noise can be filtered just like any other signal

Martin B.H. Weiss

Noise and Systems - 2 University of Pittsburgh
Tele 2060

Thermal Noise

• Due to Random Transitions of Electrons in Materials

• Proportional to Temperature
• White Noise
 Noise that is Not Dependent on Frequency

 Assume Power Spectral Density is Constant

Martin B.H. Weiss

Noise and Systems - 3 University of Pittsburgh

Tele 2060

Characterization of Thermal Noise

• Let n(t) be the Noise Signal in the Time Domain
 The Peak Voltage at Any Given Time is Random

‰ Assume that the Distribution of Peak Voltage is Gaussian

‰ Assume the Mean Voltage is Zero

The Noise Voltage at Time t1 is Independent of the Noise



Voltage at Time t2 2
−x2
• Gaussian Distribution: P[v(t ) < x ) = σ 2 π e
1 2σ

• Graphical Representation:
v(t)

t
t1 t2

Martin B.H. Weiss

Noise and Systems - 4 University of Pittsburgh
Tele 2060

Model of Noise in the Communication Channel

• At Any Time ti, the Noise Voltage is Drawn From a Random

Variable
• Thus, We Need a Different Random Variable for Each Point in
Time for Which n(t) has a Value
 Imagine n(t) has a Value for m Points in Time

 We Could Imagine m Identical Noise Generators Operating

Simultaneously, Continuously, and Independently

 Each is Sampled at the Appropriate Time

‰ At Time t1, Noise Generator 1 is Sampled

‰ At Time ti, Noise Generator i is Sampled
‰ At Time tm, Noise Generator m is Sampled

• Such a Collection of Random Variables is Called a Random

Process

Martin B.H. Weiss

Noise and Systems - 5 University of Pittsburgh

Tele 2060

Ergodic Random Process

• The Mean Value of the Random Variables are Identical

• The Variance of the Random Variables are Identical
• The Average Across Random Variables in the Ensemble (the
Ensemble Average of the Random Process) is Identical to the
Time Average of Any Random Process in the Ensemble

Martin B.H. Weiss

Noise and Systems - 6 University of Pittsburgh
Tele 2060

Description of an Ergodic Random Process

• Mean m = <n(t)> = E[n(t)]

• Variance σ2 = <[n(t) - <n(t)>]2> = <n2(t)> - <n(t)>2
• Covariance
 m τ) = <[n(t) - <n(t)>][n(t+ττ) - <n(t+ττ)>]>
XY(t,t+τ
 or, mXY(t,t+ττ) = <n(t)n(t+ττ)> - <n(t)><n(t+ττ)> = <n(t)n(t+ττ)>
 Where

‰ X Refers to the Value n(t) may Take on at Time t

‰ Y Refers to the Value n(t) may Take on at Time t+τ

Martin B.H. Weiss

Noise and Systems - 7 University of Pittsburgh

Tele 2060

Description of an Ergodic Random Process

• <n(t)> is the DC Component

• <n(t)>2 is the DC Power
• σ2 is the AC Power
• <n2(t)> is the Total Power
• <n(t)n(t+ττ)> is the Autocorrelation Function, R(ττ)
 This is Only a Function of τ

 For Ergodic Processes, this is independent of t

∞
 In General, R n ( τ ) =< n ( t ) n ( t + τ ) >= ∫ n ( t ) n ( t + τ ) dt
−∞
2
 Note That R(0) = <n (t)> = Total Power

Martin B.H. Weiss

Noise and Systems - 8 University of Pittsburgh
Tele 2060

Power Spectral Density of Noise

• The Power Spectral Density of Noise, Pn(f) = F{R(ττ)} (Watts/Hz)

• If the Noise is White
 It is Uniform Across all Frequencies
N0
 That is, 3 ( f ) =
2
 Thus, R(τ τ) = F-1{Pn(f)}
N0
‰ Or, R(τ) = 2 δ ( τ )
1 τ = 0
‰ Where the Function δ ( τ ) = 
0 τ ≠ 0
 White Noise is Completely Uncorrelated with Itself for Any
Time Delay

Martin B.H. Weiss

Noise and Systems - 9 University of Pittsburgh

Tele 2060

Filtering Noise

• Noise Behaves Like a Signal

ω)=|H(ω
• Therefore, Nout(ω ω)|Nin(ω
ω)
• Assume a Low Pass Filter
1
 H (ω ) =
1 + j ω RC
 ω) = N/2
Nin(ω
N 1
 N out ( ω ) =
2 1 + ( ω RC ) 2
• In the Time Domain,
 R(ττ) = F-1{Pn(f)}
τ
N −
 R out ( τ ) = e RC
4 RC N
 Recall That R(0) = Total Power = 4 RC
 Note that Output Noise is No Longer Uncorrelated

Martin B.H. Weiss

Noise and Systems - 10 University of Pittsburgh
Tele 2060

Characterization of Thermal Noise

• P(f) = kT
 P(f) = Power Spectral Density (Watts/Hz)

o o
 T = Temperature of the Conductor (in K = 273 + C)

-23
 k = Boltzman's constant = 1.38*10 joule/oK
 Example

‰ Noise Power at Room Temperature (290oK)

‰ P(f) = 1.38*10-23*290 = 4*10-21W/Hz

• In a Particular Bandwidth B
 The Noise Power N = P(f) B = kTB

 N is in Watts

Martin B.H. Weiss

Noise and Systems - 11 University of Pittsburgh

Tele 2060

Characterization of Thermal Noise

• The Existence of Power Presupposes

 A Generator

 A Circuit

 A Load

• Circuit Diagram

Vn RL

Martin B.H. Weiss

Noise and Systems - 12 University of Pittsburgh
Tele 2060

Characterization of Thermal Noise

• If the “Source” Impedance is Matched to the “Load” Impedance

for Maximum Power Transfer
 In General, P = V /4R
2
s s
 Let R = R (i.e., the Resistor Under Investigation)
s
 Let P = kTB, and V
2 2
s = Vn
‰ Vn2/R=kTB
‰ So Vn2 = 4RkTB, by Substitution

Also, Vn2 /B = 4kTR is the Voltage Spectral Density



• Generalization
 This Applies to Imperfect Conductors as Well

 Conductors Have Resistance

Martin B.H. Weiss

Noise and Systems - 13 University of Pittsburgh

Tele 2060

Thermal Noise Example

• Given
 R = 10KΩ Ω
o
 T = 290 K

 B = 1MHz

• Vn2 = 4(4*10-21)(104)(106) = 16*10-11

Martin B.H. Weiss

Noise and Systems - 14 University of Pittsburgh
Tele 2060

Series Resistance

• Procedure
 Sum the Resistances

 Sum the Noise Powers

• Thus,
 For Each Resistor, V
2
n = 4RkTB
 The Total Noise is V
2 2 2
n = Vn1 + Vn2 + ...
 Or, V
2
n = 4(R1+R2+ ...) kTB
• A Similar Argument Holds for Parallel Circuits
 Rely on Noise Current Instead of Noise Voltage

 Use Conductances Instead of Resistances

Martin B.H. Weiss

Noise and Systems - 15 University of Pittsburgh

Tele 2060

Reactance

• Ideally, Reactance Does not Dissipate Heat

• Hence, No Noise Contribution
• Noise Bandwidth is Affected

Martin B.H. Weiss

Noise and Systems - 16 University of Pittsburgh
Tele 2060

Noise Bandwidth

• In General,
 S
n0 = |H(ωω)|2kT
 B π/2)(B3dB)
eff = (π
• This is Why Bandpass Filters are Often Included in the Front
End of Communications Systems

Martin B.H. Weiss

Noise and Systems - 17 University of Pittsburgh

Tele 2060

Correlated Noise

• Result of System Non-Linearities

• Harmonic Distortion
• Intermodulation Distortion

Martin B.H. Weiss

Noise and Systems - 18 University of Pittsburgh
Tele 2060

Harmonic Distortion

• Result of Non-Linearities
• Occurs When the Result of a A Single Input Frequency Passed
Through a System Contains the Fundamental and Its Harmonics
• %Total Harmonic Distorion = (Vhigher)/(Vfundamental) * 100

v(f) Input v(f) Output

fA f fA 2fA 3fA 4fA f

Martin B.H. Weiss

Noise and Systems - 19 University of Pittsburgh

Tele 2060

Intermodulation Noise
• Product of Two or More Signals are Passed Through a Non-
Linear System
• Common Test is Second Order IM Distortion
 % IMD=(RMS of Second-Order Cross Products)/(Total RMS

Amplitude of Input Frequencies) * 100

• Test
 Use Four Frequecies

‰ “A” Band: fa1 and fa2

‰ “B” Band: fb1 and fb1

 Second Order Cross Products (2A-B)

‰ (2fa1 - fb1)
‰ (2fa1 - fb2)
‰ (2fa2 - fb1)
‰ (2fa2 - fb2)
‰ (fa1 + fa2) - fb1
‰ (fa1 + fa2) - fb2 Martin B.H. Weiss
Noise and Systems - 20 University of Pittsburgh
Tele 2060

Noise and Amplifiers

• General
 Amplifiers Amplify Noise as well as Signal

 Amplifiers Add Noise to the Signal

• Signal to Noise Ratio

 Definitions

‰ Voltage Ratio: S/N = VS2/VN2

‰ Power Ratio: S/N = PS/PN

Common Form: SNR = 10 log (S/N)



• Common Mode Rejection

 Applies to Differential Amplifiers

 Differential Amplifiers Amplify the Difference Between the

Inputs
 The Elements of the Signal that the Inputs Have in Common

Should Not Be Amplified

Martin B.H. Weiss
Noise and Systems - 21 University of Pittsburgh

Tele 2060

Noise Figure

• Qualitative
 Measure of the Amount of Noise Added by a System

 May be Frequency Dependent

• Analytical Definition
 Ratio of S/N at Input to S/N at Output


(
Noise Factor: F = S
S )
N i
( N)


 o

 Noise Figure (NF)

‰ Noise Ratio with S/N in Decibels (SNRdB = 10 log10 (S/N))
‰ Thus, SNRdB(i) = SNRdB(o) - NF

Martin B.H. Weiss

Noise and Systems - 22 University of Pittsburgh
Tele 2060

Noise Figure

• NF ≥ 0
 Perfect Systems Add No Noise

 Imperfect Systems Add Noise (i.e., the SNR Decreases)

 Analog Systems Cannot Remove Noise Due to its Random

Nature
• More Specifically,
 Gain, G = S /S
o i
 NR = F = (N /N )(S /S ) = N /GN
o i i o o i
 Or N = FGN
o i
 But N = kTB, so
i
 N = FGkTB Si GSi Si
o
=
Ni G GNi + Nd Ni + Nd G
Nd

Martin B.H. Weiss

Noise and Systems - 23 University of Pittsburgh

Tele 2060

Amplifiers in Cascade

• Given a Series of Amplifiers, a1, a2, a3, . . .

• With Gains G1, G2, G3, . . .
• And With Noise Ratios F1, F2, F3, . . .
• Then the Noise Ratio for the Series is Given by Friiss's Formula
F2 − 1 F3 − 1
F = F1 + + +
G1 G1G2

Martin B.H. Weiss

Noise and Systems - 24 University of Pittsburgh
Tele 2060

Noise Temperature

• General
 Another Way of Measuring Noise

 Since Noise is Associated with Temperature,

‰ Its Level Can be Represented as an Equivalent Temperature

‰ i.e., the System Seems to be Operating Hotter Than It Is

Used Frequently in Microwave Systems with Receiver Input



Noise
• Analytical Representation
 Let kT B = (F - 1)kT B
e 0
 Or, T = (F - 1)T
e 0

Martin B.H. Weiss

Noise and Systems - 25 University of Pittsburgh

Tele 2060

Implications for Communications Systems

• In Analog Systems,
 The Repeaters Have a Noise Ratio > 1

 A Repeatered Network Consists of Amplifiers in Cascade

‰ This Implies that Noise Multiplies

‰ This Limits the Length of Transmission Systems

Thus, in Addition to Amplifying the Signal and the Noise,



they add Noise

• In Digital Systems, the Regenerative Repeaters Do Not Add
Noise
 That is, F = 1

 Thus, Noise Does not Accumulate

 In Addition, The Repeaters Can Repeat Only Signal

Martin B.H. Weiss

Noise and Systems - 26 University of Pittsburgh
Tele 2060

Sample Noise Calculation

• Problem
 Three Telephone Circuits in Series Have SNR = 44dB

 A Fourth Circuit is Added with SNR = 34dB

 What is the Overall SNR?

• Solution
 Recall that SNR
dB = 10 log10[Ps/Pn]
 The Signal Remains Constant Across All Circuits
 Ps 
 Thus, S N R d B = 1 0 lo g 1 0  
 Pn + Pn + Pn + Pn 
1 2 3 4

 We Can Compute the Noise Power for Each Circuit:

Ps SNR dB
‰ = 10 10

Pn P 44

‰ Thus, For the First Three Circuits, s = 10 10

≅ 25 × 10 3
Pn
P
‰ For the Last Circuit, s = 10 ≅ 2 . 5 × 10 3
34
10

Pn
 Combining,  Ps   Ps 
S N R d B = 1 0 lo g 1 0   = 1 0 lo g 1 0   = 3 3d B
 3 Pn a + Pn b   P
 s [ 3
25×10 3 +
1
2 .5 × 1 0 3 ]

 Martin B.H. Weiss
Noise and Systems - 27 University of Pittsburgh