Вы находитесь на странице: 1из 10





Submitted By
Submitted to


In communications channel, refers either to a physical transmission medium
such as a wire, or to a logical connection over a multiplexed medium such as a
radio channel, optical fibres and magnetic tapes, flash drives. A channel is used
to convey an information signal, senders (or transmitters) to one or multiple
As channel has multiple environment factors involved with it so it always has
some randomness ad effects caused by this randomness which most of the time
leads to degradation/corruption of signal which is known as noise and every
channel has some randomness or noise which is additive in nature.

ed signal


Gain Function



Figure 1 : A block diagram for channel

Channel Study Mechanism

Channel study refers to understand the parameter of channel which is causing
variation in the signal output at the receiving end and how to minimise that
variation so that quality of transmission increases.
We always try to realise a simple channel at first as it has lesser no of complexity
levels and a solution can be derived out of that, After archiving the effective
solution for a simpler channel we do add complexity as per our daily life
scenarios and parameters of the channel to the simpler one and try to resolve it
via similar techniques used in previous complexities. Direct approach to a
complex or real life channel may lead us to absurd results because our
assumptions and techniques may be wrong but applying the assumption
archived through solving the simpler channels gives the direct approach.

Aman Mehta(ASU2011010100024)

Page 1

Types of Noise affecting the Signal

Additive Noise
Which gets added to signals and the noise is intrinsic to the system, being
intrinsic means similar i.e having the same natural background as to the original
White Noise
Noise having uniform power spectral density i.e. the signal is uniformly
distributed over the signal.
Gaussian Noise
This is a type of statistical noise having a probability density function equal to
the normal distribution. Probability density function is given by

Here z is random Gaussian variable.

Figure 2 : Random White Gaussian noise

Pink Noise
This is the noise which is inversely proportional to the frequency of the signal
also known as flicker noise. Pink noise is also known as 1/f noise.
Black Noise
This noise is dependent on black body radiation also known as silent noise.
Brown Noise
Aman Mehta(ASU2011010100024)

Page 2

This is also known as 1/f2 noise because this is indirectly proportional.

Contaminated Gaussian Noise

PDF is linier mixture of Gaussian PDF
Power law noise
Noise having components form all frequencies is known as power law noise.PSD
per unit bandwidth to 1/fb.
Cauchy noise/distribution
This is the distribution of a random variable that is the ratio of two independent
standard normal variables and has the probability density function

f ( x : 0,1 )=

(1+ x 2 )

Multiplicative Noise
This type of noise implies itself over signal frequencies and modulates it.
Quantisation error
The conversion of analogous signal to discreet type signal there are losses in
signal considered as quantisation error.
Shot Noise
Caused by static electricity discharge occurs in bust mode/type.
Transient Noise
This occurs because of short pulse due to discharging in oscillation.
Bust noise
The effect capacity of this noise is very high but it occurs for very small type.
Poisson Noise
This is the noise which just occurs at every fixed interval of time to the signal.

Considering that the all sort of noise add some error in the signal and that error
is collectively represented as e and this error is being studied in the channels.
The error might be specific to bit or specific to frequencies so all these studied
Aman Mehta(ASU2011010100024)

Page 3

with different aspects, here are the different type of channel models where
different types of errors are being considered.

Channel Models
Binary symmetric channel
In this type of channel lets the probability of error in reception is p and this error
is independent of the signal which is transmitted the error will always be p.

Figure 3: Binary Symmetric Channel

Binary Channel
In this type of channel the probability of reception of a bit is different for different
bit but each time same for the same bit i.e. 1 will be received with an error
probablity of q each time and 0 will of p each time.

Figure 4: Binary Channel

Aman Mehta(ASU2011010100024)

Page 4

Binary Erasure Channel

Considering that the binary channel had been added one more complexity/noise
type in which the error caused by channel is such that the bit is either 1 not 0
it is received as e or erased bit hence its binary channel the probability of
occurrence of such error is different for different bit but if the erasure is derived
from BSC then the probability of a bit getting erased in a transmission will be
same. This type of channel is used to save the bandwidth, where we intentionally
puncture/erase the bits of signal and at the receiver end the error correction
algorithms detects that the bit is being punctured and add the corrected bit to
the signal.

Figure 5: Binary Erasure Channel (derived from BSC)

Binary error and erasure channel

Considering the channel type where noise of BC and BEC both occurs to the
transmitted signal then our channel will be of type binary error and erasure
channel. Here, the error of receiving a bit as to other bit is q and the error that
the bit will be erased out of the transmission is p.

Figure 6: Binary Error and Erasure Channel

Aman Mehta(ASU2011010100024)

Page 5

Channels discussed up to now are of Discreet Memory less channels by

saying discreet it defines that this some specific set of values and there will be
discreet signals which will be travelling from these channels and being memory
less means it doesnt remember what is happing to signal when it is being
transmitted through it.
Discreet Memory less channels
Considering any above of channels i.e. having common factor in them is that all
are binary type hence the signal which will be traveling from them will be either
0 or 1. For a binary type channel 0, 1 are the symbols in which the signal is
travelling, and for different type of channel there could be different numbers of
symbols. Now consider a channel having n no of symbols represented by x1, x2
. xn and after passing through the channel gain and noise it can receive n
outputs represented by y1, y2. yn.

Input to the channel:

x i={x 1 , x 2 , x n }
y i={ y1 , y 2 , y n }

Received (Output of channel):


x i , y i are the sets of input and output of channel a general equation of

channel to can be written as

x n+ [ channel gain+ errors ]= y n

If the probability of reception of

y i if

x i is input to the channel is given by:

p( y i / x i)

( )

Then, p x = pr {Y = y i X=x i }
Considering the equation our transition matrix can be given as:

p( y 1x 1 ) p ( y 2|x 2 ) ..
p( y 1x 2 ) p ( y 1| x2 )
p( y 1x n ) p ( y 1| xn )

p( y nx 1 )
p( y nx 2 )
p( y nx n )

Aman Mehta(ASU2011010100024)

Page 6

Now these are channels for discreet values similarly channels could be for
continuous values which are known as continuous valued channel.

Additive white Gaussian noise channel for discreet values (AWGN)

This is most basic type of noise model in communication as the name suggest
the noise in this channel is assumed to be of intrinsic and additive in nature, also
being the white type of nature the noise has uniform power over the whole
frequency band i.e. the effect of this noise will be same for all types of signals
passing through this channel, whereas the Gaussian refers to the normal
distribution of the time domain with average value of zero.
The noises occurring in white band may be of thermal or short or black body
radiation coming from sun and various other sources
If St is input and Rt is output and the Nt is the additive white noise then we can

R(t)=S (t)+ N ( t )
Also noise is of white type i.e. this autocorrelation of this signal will not exist.

Figure 7: AWGN channel

Capacity of AWGN is given by

Here P is maximum channel power and N is noise variance.

AWGN channel with Memory
This channel is AWGN channel but the noise in this channel is not completely
white type this leads to have an autocorrelation of a signal which makes the
signal to remind its past values hence this is AWGN channel with memory. As
autocorrelation is function of same signal with two different time intervals
Linear Filter Channel
Aman Mehta(ASU2011010100024)

Page 7

The filter channels are designed to keep signals in the specified bandwidth
assigned for that communication and this is also necessary so that signals do not
interfere with the other bands, such channels are designed with additive noise
and known as linear filter channel.Considering the r(t) as the output and h(t) as a
linear filter gain/response having additive noise of n(t), then

r ( t ) =s ( t )h ( t ) +n ( t )

h ( ) s ( t ) d +n ( t )

Figure 8: LFC with additive noise

As we can see auto correlation of signal will exist hence this is also memory

Aman Mehta(ASU2011010100024)

Page 8


NPTEL lectures and website notes

C. E. Shannon, A mathematical theory of communication, Bell System
Technical Journal, vol. 27, pp. 379423 and 623656
Book Signals and Systems By A.V.Oppenheim

Aman Mehta(ASU2011010100024)

Page 9