Noise

Noise is every undesirable signal superimposed to the signal of interest which leads fluctuations on
amplitude and time measurement

Noise is invariably present in any electronic circuit and it is unavoidable. Noise is experienced in
radio as a “hiss sound” and in television as “snow”.
Noise signals are random and must be treated by statistical means. Even though we cannot predict
the actual noise waveform, we can predict the statistics such as the mean (average) and variance.
The average value of the noise waveform is zero 〈 ()〉 =  ()  = 0

Mean square noise voltage 〈 ()〉 =  ()  ≠ 0, which signifies that noise carriers power. So,
the variance, is non-zero.
In analog signal, noise is amplified along with the original signal in subsequent amplification stages
of the amplifier. Digital signals have two levels, so noise will hardly affect the digital output.

Thermal noise (Johnson noise)

This type of noise is typical of resistors. It is caused by the random thermal motion of charge
carriers (electrons) within resistor. It is independent of a DC current.
If a sensitive electronic ac voltmeter is connected across a resistance at finite temperature in
absence of any voltage source connected to it, fluctuating voltage will be recorded as a function of
time. The fluctuating voltage can be monitored on CRO screen also.

A noisy resistor is equivalent to a noiseless resistor connected with a noise voltage source in series.
Let a noiseless load  is connected with the voltage source.
〈 〉
Noise power across the load:  = ( Maximum noise power delivered to the load when
  )


 =  ; and the Maximum noise power:  = 〈 〉/4 (i)

A resistor kept at T oK creates e. m. waves with wide frequency spectrum randomly travelling along
the transmission line. This is equivalent to heated object emitting photons of wide range of
frequency spectrum following Boson distribution.

Noise power is written as  =  /!" # ∆%. For low frequency, ℎ% ≪ ();  = ()∆%…..(ii)

∆% represents the noise band width. Comparing (i) & (ii) 〈 〉 = 4 ()∆%
r.m.s noise voltage: *〈 〉 = *4 ()∆% We assumed that the resistor is dimensionless.
,〈 〉
Power spectral density is defined as noise power per unit of bandwidth: + (%) = = 4 ()
,
〈 〉 〉
〈.
Eq. (i) can be rewritten as:  = =  ; implies that a noise current source shunted with a
- -
〈 〉 〉
〈.
resistance  . =  *〈 〉 = *4 ()∆%
- -
- 0 ∆ 〉
〈. -0 ∆
 = =  〈1 〉 = = 4()2 ∆%. Power spectral
- - 
〉
,〈.
density of noise current source: +. (%) = = 4()2
,

In both cases power spectral density is independent of frequency. So,

thermal noise is a white noise.
Significance of white noise: In such cases the noise waveform is totally unpredictable as a
function of time. In other words, there is absolutely no correlation between the noise waveform at
time t and some later time t+δ, no matter how small we make δ.
Say,  = 10(Ω, ) = 300 8, ∆% = 19:;.
4() = 4 ∗ 0.026?@ = 4 ∗ 0.026 ∗ 1.6 × 10# B
= 1.66 × 10#

C〈 〉 = ,DEF = *4 ()∆% = *1.66 × 10# ∗ 10 × 10G ∗ 10H = 1.3 ∗ 10#I @ = 13J@

So, a voltage greater than 13J@ in the specified band width can be resolved from the noise
background. If the resistors are placed in series, mean square noise voltage is given by
〈 〉 = 4()( +  + ⋯ )∆% = 〈 〉 + 〈
 〉
+⋯
Uncorrelated noise sources must be added in quadrature 〈MNM
 〉
= 〈  〉 + 〈 〉 + 〈G 〉 + ⋯. i.e. The
noise powers add, not the noise voltages.
Likewise, for resistors in parallel, we can add the mean square currents.
〈1 〉 = 4()(2 + 2 + ⋯ )∆% = 〈1 〉 + 〈1
 〉
+⋯
This holds for any pair of independent noise sources (zero correlation).
In case of ideal capacitors and inductors, there is no dissipation (current leads or lags by 90o), and
as a result of dissipation-fluctuation theorem there will be no Noise.
Power spectral density is independent of frequency and also linear in thermal energy for ℎ% ≪ ().

If ℎ% ≥ (); then → ? #/0 ; spectral density decreases beyond some frequency. The roll off
  /!" #
0 . H∗ .H× QRS
starts roughly at %E = 
= H.HT× QUV
= 6.3 × 10 
= 6.3):;. This frequency is well beyond
the capacity of measuring apparatus. So, in the detectable range, thermal noise is white.

Noise for Passive Circuits: Thermal noise is sometimes called Johnson noise after its discoverer.
The physical and mathematical descriptions were provided by Nyquist, who also proved that any
electrical system in thermal equilibrium has an open circuit thermal noise voltage corresponding to
the real part of the impedance, equivalently a short-circuit current corresponding to the real part of
the admittance.
For a general linear circuit, the mean square noise voltage (current) at any port is given by the
equivalent input resistance (conductance) 〈W
 〉
= 4()?{Y(%)}[%
This is the “spot” noise. Unlike resistors, L’s and C’s do not generate noise. They do shape the noise
due to their frequency dependence.

Consider a network with set of resistive and reactive elements. The mean square noise voltage is
expressed over the frequency range where the impedance variation is appreciable:
\
〈 〉 = 4() ?{Y(%)} %
 
The impedance looking into the output is Y = ]^_
; ?`Y(%)a = (^_)
\  -0 \  -0 \ _ -0 b 0
〈 〉 = 4() %= c = b_ c = b_  =
(^_) b (^_) (^_) _
0 b b
〈 〉 = 4()∆%; 4()∆% = So, noise bandwidth: ∆% = -_ = b_  =  ∗ %G,d = 1.56%G,d
_
Noise band width in RC circuit is 1.56 times than 3dB bandwidth.
Since mean square noise voltage is proportional to R and the bandwidth is inversely proportional
to R, the final result for noise voltage is independent of R.
The dielectric medium of an ideal capacitor is insulator. So, in the equivalent circuit, the effect of its
resistance parallel to the capacitance itself is insignificant and the impedance is purely imaginary.
By Nyquist theorem, the noise voltage due to capacitor is thus zero. Similar will appear for ideal
inductor also.
0
r.m.s noise charge f = g*〈 〉; 〈f 〉 = g  〈 〉 = g  _
= ()g: called ()g noise. Although it is
independent of R, KTC noise arises in R only as capacitance does not contribute to noise.

Signal-to-noise (S/N) ratio: is defined as the signal power F to the noise power  at the same
h j   /  
point. = jk = k / = l k m . F ,  are the rms. value of signal voltage and noise voltage
i   
respectively. S/N ratio should be kept as high as possible to obtain a good signal.
In amplifier circuit, both signal power S and noise power N are amplified by the power gain factor G
of the amplifier. Noise from the detector (source) n. is amplified to 2n. ; and that from the amplifier
is amplified to 2no .
Total noise output, nN = 2n. + 2no = 2(n. + no ). Then, (n. + no ): equivalent noise input.
NMoq NrMsrM N.F sNtD w(ix iy ) i
Noise factor: p = urMsrM N.F sNtD ,r MN FNrDv (,MvMND) = = 1 + iy > 1
wix x

The maximum available noise power per bandwidth occurs when noise resistance is same as the
source resistance:  = F . Noise power from the detector due to the source resistance,
  -0 k d iy | d
n. = = = (){, B: bandwidth. So, p = 1 + =1+ =1+ 
; As the amplifier adds
-k -k 0 d 0 d
more noise, we can think of an equivalent noise temperature of amplifier ) > ).
Equivalent noise temperature: ) = (p − 1)), say p = 2.5, ) = 4508, larger than ambient
temperature.
Consider output signal is +N due to amplification of input signal +. . So, +N = 2+. .
 
w(ix iy ) i~ whx i~ hx /ix jkx /jx l m .srM l m Do.N
p= = = = = = =
€ x €
wix wix h~ wix h~ /i~ jk~ /j~  
l m NrMsrM l m Do.N
€ ~ €

Significance of noise factor: The S/N ratio available at the output is always less than that at the
input, since any amplifier or network adds noise. So, the noise factor is a measure of the amount of
noise added. F is always greater than unity.

Noise in cascade amplifier:

i
p = 1 + iy no = (p − 1)n.
x

1st stage input noise, n. + no ; output: 2 (n. + no )

2nd stage input: 2 (n. + no ) + no ;
i
2nd stage output: 2 2 (n. + no ) + 2 no = 2 2 (n. + no ) + l wy m‚ = 2 2 `(n. + no ) + no
ƒ a
R
iy
When 2nd stage is connected to the 1st stage, input noise for the 2nd stage is no
ƒ
=
wR

Input of the 3rd stage: 2 2 (n. + no ) + 2 no + noG

3rd stage output: 2G 2 2 (n. + no ) + 2G 2 no + 2G noG
iy i i
= 2G 2 2 `(n. + no ) + + w yU a = 2G 2 2 `n. + no + no
ƒ
+ noG
ƒ
a noG
ƒ
= w yU
wR w R w  R

Similarly the equivalent input noise due to noG is noG

ƒ
as if it were an equivalent source at the input.
Therefore the collective equivalent input noise due to the three stages would be;
iy i („ # )ix („U # )ix
no = no + no
ƒ
+ noG
ƒ
= no + wR
+ w yU
w
= (p − 1)n. + wR
+ w wR
 R
(„ # ) („ # ) („ # ) („U # )
no = (p − 1)n. ; (p − 1) = (p − 1) + wR
+ wU w p=p +
wR
+
w wR
 R
(„ # ) („U # ) („ # )
In general for n-stage cascading; p = p + + + ⋯…+ w
wR w wR QR….. w wR

This is known as Friss’ formula. So, the 1st stage plays an important role in contributing to the
noise factor of the amplifier system. So, that should be minimized.
In terms of noise temperature; ) = (p − 1))
(„ # ) („U # )
(p − 1) = (p − 1) + + ⇒ ) = ) + 
+ w U
wR w wR wR w
 R

Shot Noise: Current is due to the transport of discrete electrons. As electrons suffer collisions with
the impurity, phonons, their individual kinetic energy fluctuates. This causes fluctuation in the
arrival rate of electrons. So, for a given interval of time number of electrons reaching at a node
where current is measured, fluctuate. This causes a randomly varying noise current superimposed
on the direct current (†,v ) of the output electrode.
Shot noise results from the flow of current across a potential barrier. This is a statistical effect of
random emission of electrons (and holes) or the production of photoelectrons. It is found in
vacuum tubes, transistors, and diodes where activated processes are involved. †,v can be due to
signal, bias currents, background radiation (photodetectors), and leakage.
Suppose, ‡ˆ (∆) number of electrons are crossing a given node in a very short time interval ∆ for
‰-th event. If N events are followed, the numbers ‡ˆ ’s will fluctuate around an average value 〈‡〉.
Average current due to average 〈‡〉 is †,v = 〈1〉 = 〈‡〉Š/, and 1ˆ (∆) = ‡ˆ Š/∆. The RMS
fluctuations in n, ∆‡DEF , would give rise to fluctuations in the current, or “current noise”. For N
number of large events,
/
∆‡DEF = i ∑i
ˆŒ (‡ˆ − 〈‡〉) ‚

. As current is proportional to number of charges concerned, rms
/
fluctuation also has the same form. 1FNM = i ∑i
ˆŒ (1ˆ − 〈1〉) ‚


Poisson statistics is to be taken in account for random fluctuation. So, we have, ∆‡DEF = *〈‡〉.
Electrons have spin and the arrivals are correlated in pairs, so the square variance is doubled. Thus,
/ ∆Žk W *〈〉W
∆‡DEF = *2〈‡〉 ˆŒ (1ˆ − 〈1〉) ‚
1FNM = i ∑i 
= = ;
M M

*〈.〉MW 〈.〉W
〈1〉 = 〈‡〉Š/ 1FNM = =C = *2〈1〉Š{ = *2†,v Š{ B: bandwidth
√WM M
〉
,〈.
Power spectral density: +. (%) = ,d
= 2Š†,v , which does not depend on frequency. So, shot noise
is spectrally white, like Johnson noise, up to frequency %N of the order of the electron arrival rate,
‘ QS
when the averaging argument breaks down. The frequency %N ≈ W = QRS ~10B :;~2:;, for a
.H×
current of 1nA. Spectral density depends on current, but independent of temperature. Hence shot
noise is non-thermal white noise. Consider a reversed-biased diode, with leakage I = 1 nA. Power
spectral density: +. (%) = 2Š† = 2 ∗ 1.6 × 10# B ∗ 10#B = 3.2 × 10#“ ” /:;
Prob: Calculate the shot noise current present on a direct current of 13mA for a noise bandwidth of
7 MHz. [0.171 µA]

Flicker noise (1/f noise): It is associated to random trapping and recombination of charge
carriers in the semiconductors, typically caused by imperfections at the interface regions.
,〈 〉 ‘y
Power spectral density: + (%) = ,
= 8 •. I is dc current; 8 is a constant (varies from device
to device); a ~ 0.5 – 2; b ~ 1. It depends on frequency and dominates at low frequencies. Due to
low frequency weighting, it is often referred to as ‘pink noise’. If, c > c the flicker noise is
negligible compared to white noise. For frequency below c flicker noise contributes much.
===============================================