You are on page 1of 52

ELE 324

Telecommunication Theory I

Noise,
Noise in AM Receivers, Noise in FM Receivers

Cenk Toker
Hacettepe University
Department of Electrical and Electronics Engineering

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 1


Noise
 Electrons in a (semi-)conductor can move freely in the medium.
 Motion of each electron is independent from others and in a random direction/velocity.
◦ Velocity depends on the ambient temperature. Higher temperature means faster mobility.

 Motion of each electron results in a minute change in the voltage across the terminals of the component
of concern.
 If the medium is assumed to be linear, voltage across the terminals due to the motion of each electron is
added together
◦ Central Limit Theorem.

 Noise can be considered as a random electrical signal added on top of the desired signal,
◦ out of the control of the system designer.

 There are many potential sources of noise


◦ Thermal noise (internal to the receiver)
◦ Atmospheric noise, cosmic noise, man-made noise (external to the receiver).

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 2


Thermal Noise
 Arises from the random motion of electrons in a (semi-) conductor.
 The mean-square value of the thermal noise voltage, VTN, across a resistor in a bandwith of Δf is

where k = 1.38 x 10-23 Joules/oK is the Boltzman constant. T is the absolute temperature in oK, and R is
resistance in ohms.
Eventually, a resistor can be modeled by

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 3


Thermal Noise
 Central Limit Theorem states that Thermal Noise is Gaussian with zero-mean.
 Under matched conditions, i.e. Rload = R, a noisy resistor produces «available noise power» equal to

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 4


White Noise
 When observed in a period of time, noise can be considered to be a random process.
 Motion of electrons are assumed to be physically independent. Eventually time samples taken from the
noise process are uncorrelated. Since the process is Gaussian, they are also statistically independent.

 No is the noise power spectral density, referred to the input of a receiver, and

Te: equivalent noise temperature of the receiver.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 5


White Noise
 Example
 Suppose a white Gaussian noise, w(t), of zero mean and PSD, Sw(f)=No/2, is applied to an ideal LPF of
bandwidth B and passband amplitude response of unity.

 Since

Then

 Since w(t) is Gaussian, then the bandlimited noise, n(t), at the output is also Gaussian.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 6


White Noise
 Example
 Consider white noise, w(t), of zero mean, and PSD, Sw(f)=No/2, is applied to the following LPF

with transfer function

and PSD

with

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 7


Noise Equivalent Bandwidth
 Suppose there is a source of white noise of zero mean and PSD, Sw(f)=No/2, connected to a LPF of transfer
function H(f). The resulting average output noise power is therefore,

(two sided PSD: No/2, one sided PSD: No)


 Consider the same noise is applied to the input of an ideal LPF of dc-response H(0) and bandwidth B.
 The average output noise power is

 Therefore, we may define the noise equivalent bandwith as

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 8


Narrow-band Noise
 We have assumed that the communication signal of concern is narrowband (i.e. B << fc).
 Therefore, receiver has narrowband filters whose bandwidth is just large enough to pass the desired signal
undistorted, but not so large to pass excessive noise.
 The noise appearing at the output of the receiver is called narrow-band noise.
 With the filters centered at fc, this noise, n(t), appears as a sinusoidal wave of frequency fc with slowly
changing (random) amplitude and phase angle, amplitude noise and phase noise.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 9


Narrow-band Noise
 Consider n(t) at the output of a narrow-band filter with input w(t), a white Gaussian noise with zero mean
and PSD SW(f) = 1.
 If the transfer function of the filter is H(f), then w(t) n(t)
SW(f) SN(f)

 We would like to represent n(t) in terms of its in-phase and quadrature components.
 Let 𝑛+ 𝑡 and 𝑛෤ 𝑡 denote the pre-envelope and complex envelope of n(t).
 Assuming the spectrum SN(f) is centered around ±𝑓𝑐 ,

and
We also know that,
hence,

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 10


Properties of Narrow-band Noise
 Property 1:
 nI(t) and nQ(t) have zero mean.
 Proof: Since n(t) and 𝑛ො 𝑡 have zero mean, then their weighted sums, nI(t) and nQ(t), also have zero mean.

 Property 2:
 If n(t) is Gaussian, then nI(t) and nQ(t) are jointly Gaussian.
 Proof: Since 𝑛ො 𝑡 is obtained from n(t) by linear filtering, then n(t) and 𝑛ො 𝑡 are jointly Gaussian.
 Eventually, nI(t) and nQ(t) are jointly Gaussian since they are weighted sums of jointly Gaussian processes.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 11


Properties of Narrow-band Noise
 Property 3:
 If n(t) is w.s.s., then nI(t) and nQ(t) are also jointly w.s.s.
 Proof: Since n(t) is stationary, so is 𝑛ො 𝑡 . Then

using the properties

we can write

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 12


Properties of Narrow-band Noise

 A similar derivation applies to RNQ(τ) also, which yields RNQ(τ) = RNI(τ).


 The cross-correlation functions of nI(t) and nQ(t) are

 Since nI(t) and nQ(t) have zero mean and RNI(τ), RNQ(τ), and RNINQ(τ) depend only on τ, nI(t) and nQ(t) are
w.s.s.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 13


Properties of Narrow-band Noise
 Property 4:
 nI(t) and nQ(t) have the same PSD which is related to the PSD, SN(f), of n(t) as:

assuming SN(f) = 0 outside the interval fc – B ≤ f ≤ fc + B.


Proof: Using the property

and
we have

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 14


Properties of Narrow-band Noise
 Now, if SN(f) sgn(f)
1
f
f
- fc fc -1
2W 2W
SN(f - fc)
sgn(f - fc)
1
f
f fc
-B 0 B 2fc -1
SN(f + fc)
sgn(f + fc)
1
f
f - fc
-2fc -B 0 B -1

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 15


Properties of Narrow-band Noise
 Observations:
1. For frequencies -B ≤ f ≤ B

then,
2. For 2fc - B ≤ f ≤ 2fc + B

then,
3. For -2fc - B ≤ f ≤ -2fc + B

then,
4. Outside these intervals,

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 16


Properties of Narrow-band Noise
 We may extract nI(t) and nQ(t) from narrowband noise n(t) using

and, we may generate n(t) using the following synthesizer

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 17


Properties of Narrow-band Noise
 Property 5:
 nI(t) and nQ(t) have the same variance as n(t).
 Proof: Total area under SNI(f) or SNQ(f) is the same as SN(f). Since nI(t) and nQ(t) have zero mean, their
variances are the same.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 18


Properties of Narrow-band Noise
 Property 6:
 Cross-spectral densities of the quadrature components of a narrowband noise are purely imaginary.

 Proof:

 Using the previous procedure, we complete the proof.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 19


Narrow-band Noise
 Property 7:
 If n(t) is Gaussian with zero mean and SN(f) is locally symmetric, then nI(t) and nQ(t) are statistically
independent.

 Proof: If SN(f) is locally symmetric around ±𝑓𝑐 , then

and, from Property 6, cross-spectral densities are zero which means

that is, they are orthogonal.

 nI(t) and nQ(t) also have zero mean, then they are independent since they are Gaussian.
 Their joint pdf is σ2 is the variance of n(t).

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 20


Narrow-band Noise
 Example: Ideal band-pass filtered white noise. w(t) n(t)

 Consider a white Gaussian noise of zero mean and PDF SW(f) = No/2, passing through an ideal BPF with
response equal to one, midband frequency fc, and bandwidth 2B.

Thus,

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 21


Narrow-band Noise
 Example: White Noise through a high-Q, tuned filter.
 Consider the RLC bandpass filter L

w(t) n(t)

The resonant frequency is and Q-factor is


Then, the transfer function is

If Q>>1, then
No/2

with PSD No/4 fc


Q

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 22


Narrow-band Noise
 Example: cont.d
 Therefore
SNI( f ) = SNQ( f )

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 23


Noise in CW Modulation Systems
 Receiver Model:

 w(t) is the front-end receiver noise and the received signal is s(t) + w(t).
 The bandwidth of the BPF is BT, just enough to pass s(t) through without distortion.
 It is assumed that w(t) is white Gaussian and SW(f) = No/2.
 The filtered signal is x(t) = s(t) + n(t), where

 The average noise power is NoBT (after the BPF) (two sided noise spectrum: No/2).

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 24


Noise in CW Modulation Systems

 Definitions:
◦ Input Signal-to-Noise Ratio (SNR)I : Ratio of the power of the modulated signal, s(t), to the power of n(t).
◦ Channel Signal-to-Noise Ratio (SNR)C : Ratio of the power of s(t) to the power of noise in the message bandwidth at
the receiver input.
◦ Output Signal-to-Noise Ratio (SNR)O : Ratio of the power of the demodulated message signal to the noise power at
the receiver output.

 Figure-of-Merit is defined as

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 25


Noise in DSB-SC Receivers

oscillator

 Assume that the local oscillator is synchronized both in phase and in frequency with the carrier of s(t).
 The filtered signal is
 Its DSB-SC component is
where m(t) is the message signal, and c is a scaling factor.
 We assume that, m(t) is the sample function of a stationary zero mean process, and has PSD, SM(f), with
bandwidth W. The average power of the message signal is

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 26


Noise in DSB-SC Receivers
 The carrier wave is statistically independent of m(t). For the sake of simplicity, we omit the uniformly
distributed phase, θ, of the carrier.
 The average power of s(t) is

 Since SW(f) = No/2, the average noise power in message bandwidth is

 Then, the channel SNR is Channel Signal-to-Noise Ratio (SNR)C :


Ratio of the power of s(t) to the power
of noise in the message bandwidth at
the receiver input.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 27


Noise in DSB-SC Receivers

The filtered signal is oscillator

 Then,

LPF

 Therefore, the receiver output is

 Note that,
◦ 1. m(t) and nI(t) appear additively,
◦ 2. nQ(t) is completely rejected.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 28


Noise in DSB-SC Receivers
 The message signal component power is
Output Signal-to-Noise Ratio (SNR)O :
Ratio of the power of the demodulated
where P is the average power of m(t).
message signal to the noise power at
the receiver output.

 Now, the average noise power of n(t) is

 The average power of the (lowpass) component nI(t) is the same as that of n(t), therefore

 Hence, the output SNR is

 And, the Figure-of-Merit is


Input Signal-to-Noise Ratio (SNR)I :
Ratio of the power of the modulated
 Also, note that, signal, s(t), to the power of n(t).

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 29


Noise in SSB Receivers
 Assume that we use a coherent detector

oscillator

 Consider a Lower-SideBand (LSB) SSB signal (similar discussion applies to USB also)

 Note that
◦ i. 𝑚 𝑡 and 𝑚 ෝ 𝑡 are orthogonal to each other. m(t) is assumed to have zero-mean, hence 𝑚 𝑡 and 𝑚 ෝ 𝑡 are
uncorrelated. Hence their PSDs are additive (cross spectral density is zero).
◦ ii. Since 𝑚
ෝ 𝑡 is obtained by passing 𝑚 𝑡 through a linear filter with H( f ) = -j sgn( f ), 𝑚 𝑡 and 𝑚
ෝ 𝑡 have the same
2
PSD, 𝑆𝑀෡ 𝑓 = 𝐻 𝑓 𝑆𝑀 𝑓 = 𝑆𝑀 𝑓 .

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 30


Noise in SSB Receivers
𝑐 2 𝐴2𝑐 𝑃
 In-phase and quadrature components of s(t) contribute an average power of each. Therefore, the
8
average power of s(t) is

which is half of that in DSB-SC. (since SSB occupies half the bandwidth of DSB-SC.)
The average noise power in the message bandwidth W is WNo as in a DSB-SC receiver.
Therefore, the channel SNR is (half of that in DSB-SC)

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 31


Noise in SSB Receivers
 The PSD of n(t) in a LSB-SSB receiver is

oscillator

-fc -fc + W fc - W fc
-fc + W/2 fc - W/2

 We may express n(t) as

 The output of the coherent detector is

hence

 Unlike the case of DSB-SC modulation, the quadrature component of n(t) appears at the output now.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 32


Noise in SSB Receivers
 The message component is (1/4)cAcm(t), and its average power is

 The noise component is .


 Note that
SNI ( f ) = SNQ ( f )
◦ i.
No

f
-W/2 W/2

◦ ii. The PSD of 𝑛𝐼′ 𝑡 = 𝑛𝐼 𝑡 cos 𝜋𝑊𝑡 , and the PSD of 𝑛𝑄′ 𝑡 = 𝑛𝑄 𝑡 sin 𝜋𝑊𝑡 are
SNI ( f ) = SNQ ( f )

𝑊
2𝜋 𝑡
2 No/4

f
-W W

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 33


Noise in SSB Receivers
𝑊𝑁𝑜
 Then, the average power of 𝑛𝐼′ 𝑡 or 𝑛𝑄′ 𝑡 is and the average noise power at the output is
2

 Thus, the output SNR is

and

 We observe that, an SSB receiver has the same FOM as a DSB-SC receiver.
 Furthermore, in both cases, the noise performance is the same as that obtained by transmitting the
message signal itself.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 34


Noise in AM Receivers
 Consider an envelope detector receiver

 The AM signal is

𝐴2𝑐
 The average power of the carrier in AM signal is and the average power of the information component
2
𝐴𝑐 𝑘𝑎 𝑚 𝑡 cos 2𝜋𝑓𝑐 𝑡 is

 Then, the average power of s(t) is

 The average power of noise in message bandwidth is WNo. Then, the channel SNR is

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 35


Noise in AM Receivers
 The filtered signal is

 Then, the output signal is

 If 𝐴𝑐 1 + 𝑘𝑎 𝑚 𝑡 is large enough compared to nI(t) and nQ(t), we may approximate y(t) as

 Ignoring the d.c. term Ac, the output has a form similar to the output of a DSB-SC receiver, and the output
SNR is
 Thus, Figure-of-Merit is

 The FOM of an AM receiver is always less than unity!, meaning that noise performance of an AM receiver
is always inferior to those of DSB-SC and SSB.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 36


Noise in AM Receivers
 Example (Single tone modulation)
 Consider
then,

where 𝜇 = 𝑘𝑎 𝐴𝑚 .
 The average power of m(t) is

 Then

 When 𝜇 = 1, 𝐹𝑂𝑀 = 1Τ3.


 This means, for the same noise performance an AM system must transmit three times as much power as a
DSB-SC system.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 37


Treshold Effect in AM Receivers
 When the carrier-to-noise ratio is small compared to unity, the noise term dominates. Then, we may write

 Recall, 𝑥 𝑡 = 𝑠 𝑡 + 𝑛 𝑡 , and consider the phasor diagram


n(t)

Too much noise,


less signal.

The output signal can be written as

 y(t) contains m(t) multiplied by cos(ψ(t)), the phase of n(t), which is uniformly distributed over 2π rad.s.
Hence, message is distorted and lost in noise. This is called the Threshold Effect.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 38


Threshold Effect in AM Receivers
 Threshold is a function of carrier-to-noise ratio below which performance deteriorates rapidly.

DSB-SC and SSB


AM (envelope detector)

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 39


Noise in FM Receivers
 Consider the FM receiver model

 Here, w(t) is additive white Gaussian noise with zero-mean and SW(f) = No/2.
 The received signal s(t) has a carrier frequency of fc and transmission bandwidth of BT.
 Assume that the BPF is ideal. Then, its output is

where n(t) is narrowband noise.


 The limiter consists of an amplitude clipper and a BPF to round off the resulting rectangular wave.
Eventually, the output is a sinusoidal with constant amplitude.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 40


Noise in FM Receivers
 The discriminator consists of two stages:
◦ 1. A slope circuit (differentiator) that produces a wave in which both amplitude and phase vary according to the
message signal.
◦ 2. An envelope detector.

 The LPF removes the out-of-band noise at the discriminator output.


 Therefore, we may write

and

𝑡
where 𝜙 𝑡 = 2𝜋𝑘𝑓 ‫׬‬0 𝑚 𝑥 𝑑𝑥

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 41


Noise in FM Receivers
 Then,

or, in the polar coordinates

where .
 The envelope is of no interest since it is made constant by the limiter.
1 𝑑𝜃 𝑡
 The discriminator output is proportional to .
2𝜋 𝑑𝑡

 Under the assumption that the carrier-to-noise ratio is high enough (Ac>>|r(t)|), so that

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 42


Noise in FM Receivers
 Then, the discriminator output is

where .
 Since noise phase ψ(t) is uniformly distributed over 2π rad.s, we may assume that ψ(t) - φ(t) is also
uniformly distributed over 2π rad.s. Then, we may simplify the equation as

 We also know that the quadrature component of n(t) is

eventually,

 The message component of the output is kf m(t). Hence, the output signal power is kf2P.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 43


Noise in FM Receivers
 The output noise component is proportional to the derivative of nQ(t). Since the transfer function of a
differentiator is H(f) = j2πf, we can write

Hence, the PSD of nd(t) is

 Since 𝑆𝑁𝑄 𝑓 is lowpass,

 After LPFing

 and, the average noise power is

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 44


Noise in FM Receivers
𝐴2𝑐
 Note that, the noise power is inversely proportional to the carrier power, . Hence, increasing the carrier
2
power has a «noise-quieting» effect.
 The output SNR is

𝐴2𝑐
 The power of s(t) is and the noise power in the message bandwidth is WNo. Hence, the channel SNR is
2

and,

𝑘𝑓 𝑃1/2
 The deviation ratio, D, is proportional to . The FOM is a quadratic function of D. Also remember that,
𝑊
in wideband FM, BT is proportional to D.
 Hence, an increase in BT provides a corresponding quadratic increase in the (SNR)o,FM or FOM.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 45


Noise in FM Receivers
 Example (Single-Tone Modulation)
 Consider
then,
where Δf = kmAm.
 The power of m(t) is

 The output SNR is

 Then,
 Note that, W is the bandwidth of the post-detection LPF.
 For an AM system:
3 1 2
 2 𝛽2 ≈ 3 → 𝛽 ≈ = 0.471, narrowband FM ≈ AM. The noise performance of ~WBFM is better than AM.
3

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 46


Capture Effect in FM
When there is a weak FM signal (interference) near the carrier frequency of the desired signal, it is
suppressed.
However, if the interference signal is stronger, then the receiver locks onto the stronger interference signal
and suppresses the desired signal.
If they are of equal strength, the receiver bounces back and forth between the signals.
This phenomenon is known as the capture effect.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 47


FM Threshold Effect
 The expression

is valid only if the carrier-to-noise ratio is high enough.


 When the carrier-to-noise ratio decreases, after a point (a threshold), the above equation begins to fail.
𝐴2
 The loss of message is negligible if the carrier-to-noise ratio is 𝜌 = 2𝐵 𝑐𝑁 > 20 (13 𝑑𝐵).
𝑇 𝑜

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 48


Pre-Emphasis and De-Emphasis
 Recall that the noise at the output of an FM receiver is

 The following is a typical message PSD

 We observe that, at high frequencies, signal level is low, and noise level is high.
 In order to improve the noise performance, we may use pre-emphasis and de-emphasis filters.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 49


Pre-Emphasis and De-Emphasis
 The high-frequency components of m(t) are «emphasized» prior to modulation and an inverse operation is
performed after demodulation:

We know that,

 The modified PSD at the de-emphasis filter output is |Hde(f)|2SNd(f).


 The noise power is then

 The message power is not affected.


 Then, the improvement in (SNR)o is

2𝑁𝑜 𝑊 3
 Since the output noise power without pre,de-emphasis is 3𝐴2𝑐

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 50


Pre-Emphasis and De-Emphasis
 Example:
 A simple pre, de-emphasis filter has
|Hpe( f )|

which can be realized with the following circuits.

1
f
fo

1
𝑅 ≪ 𝑟, 2𝜋𝑓𝐶𝑟 ≪ 1, 𝑓𝑜 = 2𝜋𝐶𝑟 |Hde( f )| fo f

 In commercial FM, fo = 2.1 kHz, W = 15 kHz. Then 𝐼 = 22 ≈ 13 dB.

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 51


AM, DSB, FM Comparison

FM, β = 5
with pre-,de-emphasis
FM, β = 2

DSB, SSB
AM, μ = 1

C. Toker ELE 324 - TELECOMMUNICATION THEORY I 52

You might also like