Mod 3

MOD 3
Source coding theorems I and II (Statements only). Waveform coding.

Sampling and Quantization.
Pulse code modulation, Transmitter and receiver. Companding. Practical
15 level A and mu-law companders.
DPCM transmitter and receiver. Design of linear predictor. Wiener-Hopf
equation.
Delta modulation. Slope overload.
2/9/2022 Nithya M,Asst.Prof, Dept of ECE 1

Coding
• Purpose of coding is to improve the efficiency of the communication

system.
• Procedure to map a given set of msgs [m1…..mN] into a new set of
encoded message [C1….CN] in such a way that the transformation is
one to one.
• ie; for each msg , there is only one encoded message.
• Source Coding
• Conversion of the output of a DMS (Discrete Memoryless Source )into a
sequence of binary symbols is called as source coding.
• Obj: minimise the average bit rate reqd for representation of source by
reducing the redundancy of the information source.
• By using source coding,
• Improve the transmission efficiency.
• Reduce the probability of error by detecting & correcting the errors.
• Communication system with an encoder & decoder
Source Encoder Channel Decoder Receiver
• At the txr,the msgs are encoded by using encoder ,txd via channel .
• At the receiving end, the msgs are first decoded in the decoder & the
original msgs are recovered.

Terminologies used with coding
• Letter ,symbol or character – any individual member of the alphabet.
• Message or word- a finite sequence of letters in an alphabet.
• Length – no of letters in a message.
• Encoding or enciphering-conversion of symbols to binary 1’s and 0’s , & the
group of bits is called code word.For each symbol, there will be a unique code
word
• Decoding & deciphering – reverse process of encoding
• Irreducibility or prefix property-when no encoded words can be obtained
from each other by the addition of more letters,the code is said to be
irreducible or of a prefix property.
• Uniquely decipherable/seperable encoding & decoding- The correspondence
of all possible sequence of words is one to one & no space b/w words .
• When a code is irreducible,it is decipherable , but vice versa is not true.
• eg, C1=0,C2=10 & C3= 110.
• if we receive 01101010001011011010, it is decoded as
• C1C3C2C2C1C1C2C3C3C2
• We are getting the same msg by addition of 0.
• So its irreducible & decipherable .
• Let C1= 0,C2= 01,C3= 011
• Eg.00011011010010101011 can be decoded as
C1C1C3C3C2C1C2C2C2C3 .The same code is not irreducible as codes
C2 and C3 are formed by adding 1 to codes C1 & C2 respectively.

Blk dgm of source coding
xi Binary seq
DMS Source Encoder
X={x1…..xm}
Let X be a DMS( discrete memoryless source) with finite entropy H(x) & an
alphabet {x1….xm} with probabilities P(xi).
Let a binary code assigned to symbol xi by the encoder having length ni is
measured in bits.

• Code Length & Code Efficiency
• The length of a codeword is the number of binary digits in the
codeword.
• The avg code word length L = σ𝑚 𝑖=1 𝑃 𝑥𝑖 𝑛𝑖
• L- avg no of bits/symbol ,ni –number of symbols
Lmin
• Code efficiency 𝜂 = .
L
• For efficient transmission, L should be minimum.
• Let H(x) be the entropy of the source in bits/msg.
• Also let M be the max average information associated with each
letter .
• Lmin= H(x)/log M
𝐻 𝑥 𝐻 𝑥
• The coding efficiency 𝜂 =
𝐿 𝐿ത log 𝑀
• Redundancy =1- 𝜂

Source coding Theorem 1
• Given a DMS of entropy H(x) ,the average code word length for any
distortionless source encoding scheme is bounded as L ≥ H(x).
• According to the theorem, the source coding H(x) represents a
fundamental limit on the average number of bits / source symbol
necessary to represent a DMS source , not smaller than entropy.
• When Lmin= H(x) , ƞ= H(x)/L.

• Eg let M= [m1,m2,m3,m4] ,P(M)= [1/2,1/4,1/8,1/8] assuming a noise free
channel , Without coding ,η= H(x)/log2M=
• = - [1/2log2+1/4log4+1/8log8+1/8log8]/log4 = 87.5%
• Let us use binary code for coding.the code letters r 0 & 1.M=2
message Code Length No of

of code zeroes
m1 C1= 00 n1=2 2
m2 C1=01 n2=2 1
m3 C2=10 n3 =3 1
m4 C3=11 n4=4 0

• L= σ𝑚
𝑖=1 𝑃 𝑥𝑖 𝑛𝑖
• σ4𝑖=1 𝑃 𝑥𝑖 𝑛𝑖 =
𝐻 𝑥
• 𝜂= =
𝐿ത log 𝑀
• The coding is not improving efficiency.

Message Code Length of
code
m1 C1=0 n1= 1
m2 C2=10 n2=2
m3 C3=110 n3=3
m4 C4=111 n4=3
L = σ𝑚
𝑖=1 𝑃 𝑥𝑖 𝑛𝑖 =
𝐻 𝑥 𝐻 𝑥
𝜂= =
𝐿 𝐿ത log 𝑀

• The second coding method is better compared to first .
• In first method,
• P(0)= σ𝑚 𝑖=1 𝑃 𝑘 𝑐𝑘0 /σ 𝑃 𝑥𝑖 𝑛𝑖
• Where Cko is the no of zeros in the kth coded message
• P(0)= 1/2*2+1/4*1+1/8*1+1/8*0/2= 11/16
• P(1)= σ𝑚 𝑖=1 𝑃 𝑘 𝑐𝑘1 /σ 𝑃 𝑥𝑖 𝑛𝑖
• Where Ck1 is the no of ones in the kth coded message
• P(1)= 5/16

• For the second method, P(0)= σ𝑚 𝑖=1 𝑃 𝑘 𝑐𝑘0 /σ 𝑃 𝑥𝑖 𝑛𝑖 =
• P(0) = [1/2*1+1/4*1+1/8*1+1/8*0]/ 7/4 =1/2
• P(1)=1-P(0)= ½
• The coding efficiency is more when P(0)=P(1) & less when the prob are
different .
• When c1= 111,c2=110,c3=10& c4= 0 gives 66.7% efficiency
• While coding ,the rule followed is that
• Encode a message with a high prob in a short code word,then only avg
length decrease resulting in increased efficiency.
• Decrease of L can be considered as data compression

Source Coding Theorem II
• Aka channel coding theorem.
• Assume the DMS source has the source alphabet x & entropy H(x)
bits /source symbol .
• The source emits symbols once every Ts seconds.
• Hence the average information rate, R = r H
ie; R= H(x)/Ts
• The discrete memoryless channel has a channel capacity equal to C
bits /use of the channel.
• Assume that the channel is capable of transmitting once every Tc
seconds
• Hence the channel capacity /unit time is C/Tc bits/second.

• The channel coding theorem can be stated as
• Let there be a discrete memoryless source with an alphabet x have
entropy H(x) & produce symbol every Ts seconds.Let a discrete
memoryless source have capacity C & can be used once every Tc
seconds.then ,
if H(x)/Ts ≤ C/Tc,--------------(1)
there exists a coding scheme for which the source output can be
transmitted over the channel with a small probability of error

• Conversely, if
H(x) /Ts ≥ C/ Tc ,
it is not possible to transmit information over the channel & reconstruct
it with an arbitrarily small probability of error.
The theorem specifies channel capacity C as a fundamental limit on the

rate of transmission of reliable error-free messages can take place over a
discrete memoryless channel.
Note :The channel coding theorem does not show how to construct a good
code.
• If eqn A is satisfied, then good codes do exist.

Waveform Coding Techniques
• Analog to Digital Conversion
• A digital signal is superior to an analog signal because it is more robust to
noise and can easily be recovered, corrected and amplified. For this
reason, the tendency today is to change an analog signal to digital data.
• In this section we describe two techniques, Pulse Code Modulation and
Delta Modulation.
▪ Pulse Code Modulation (PCM)
▪ Delta Modulation (DM)

PCM
• PCM consists of three steps to digitize an analog signal:
1. Sampling
2. Quantization
3. Binary encoding
▪ Before we sample, we have to filter the signal to limit the maximum
frequency of the signal as it affects the sampling rate.
▪ Filtering should ensure that we do not distort the signal, ie remove
high frequency components that affect the signal shape.

Figure 4.21 Components of PCM encoder

Sampling theorem
• A bandlimited signal of finite energy with a maximum frequency component of fm Hz
is described by its sample values at uniform intervals less than or equal to 1/2fm
second apart.
• A bandlimited signal of finite energy with a maximum frequency component of fm Hz
is completely recovered from the knowledge of its samples taken at rate of 2fm
samples per second.
• Combining the two parts,
• A CT signal may be completely represented in its sample and recovered
successfully if fs ≥ 2fm,where
fs-sampling frequency, fm-max frequency in the signal

• Analog signal is sampled every TS secs.
• Ts is referred to as the sampling interval.
• fs = 1/Ts is called the sampling rate or sampling frequency.
• There are 3 sampling methods:
• Ideal - an impulse at each sampling instant
• Natural - a pulse of short width with varying amplitude
• Flattop - sample and hold, like natural but with single amplitude
value
• The process is referred to as pulse amplitude modulation PAM and
the outcome is a signal with analog (non integer) values

a ) input
b) impulse
c) spectrum of input
d) Sampled s/g
e) e) spectrum
a0
b) d)
c) e)
Nyquist rate and Nyquist Interval
• When the sampling rate becomes exactly equal to 2fm samples per second, it
is called Nyquist rate.
• It is also called minimum sampling rate fs=2fm
• Maximum sampling interval is called Nyquist interval Ts=1/2fm
• When the CT band limited signal is sampled at Nyquist rate, fs=2fm, the
sampled spectrum Y(w) contains non overlapping Y(w) periodically.
• But the successive cycles of Y(w) touches each other .
• So original spectrum X(w) can be recovered from the sampled spectrum by
using a LPF with cut off frequency wm.

3 types of Sampling Techniques
• Impulse sampling
• Natural Sampling
• Flat top Sampling
• 1. Impulse sampling/ Ideal sampling method

• Impulse sampling can be performed by multiplying input signal x(t) with
impulse train
• δTs(t) = Σ∞n=−∞ δ(t−nTs) period 'Ts'.
• the amplitude of impulse changes with respect to amplitude of input signal x(t).
• The output of sampler is given by y(t)=x(t) . Impulse train.
• ckt- switching sampler used
• generation of such impulse train is not possible practically.
• cannot use this practically because pulse width cannot be zero
2. Natural Sampling
• Natural sampling is similar to impulse sampling, except the
impulse train is replaced by pulse train of period T.
• i.e. multiply input signal x(t) to pulse train Σ∞n=−∞ P(t−nT)
• Uses natural sampler
• Output y(t)=x(t) . Pulse train
2/9/2022 Nithya M,Asst.Prof, Dept of ECE 4.26

3. Flat Top Sampling
• Aka rectangular pulse sampling .

• During transmission, noise is introduced at top of the
transmission pulse which can be easily removed if the pulse is in
the form of flat top.
• In flat top sampling, top of the samples are flat i.e. they have
constant amplitude.
• Flat top sampling makes use of sample and hold circuit.
• complex

• the sampled signal can be obtained by convolution of rectangular
pulse p(t) with ideally sampled signal say yδ(t)
2/9/2022 Nithya M,Asst.Prof, Dept of ECE

Figure 4.22 Three different sampling methods for PCM

Example 4.6
For an intuitive example of the Nyquist theorem, let us

sample a simple sine wave at three sampling rates: fs = 4f
(2 times the Nyquist rate), fs = 2f (Nyquist rate), and
fs = f (one-half the Nyquist rate). Figure 4.24 shows the
sampling and the subsequent recovery of the signal.
It can be seen that sampling at the Nyquist rate can create

a good approximation of the original sine wave (part a).
Oversampling in part b can also create the same
approximation, but it is redundant and unnecessary.
Sampling below the Nyquist rate (part c) does not produce
a signal that looks like the original sine wave.
Figure 4.24 Recovery of a sampled sine wave for different sampling rates

Performance Comparison of various sampling techniques

Aliasing
• fs=2fm is Nyquist rate.

• Upon sampling at this rate, sampled spectrum has non-overlapping repeating
periodics.
• But successive cycles touch each other
• Original spectrum is then recovered by using a LPF of cut off fm.
• When a CT signal is sampled at a rate less than Nyquist rate, ie fs<2fm, then
successive cycle of spectrum of sampled signal overlaps with each other.
• This overlapping is referred to as aliasing.
• Aliasing means a higher frequency component in the spectrum of the signal taking an
identity of a lower frequency component in the spectrum of sampled signal

• To combat the effects of aliasing in practice
an anti aliasing filter is used prior to sampling to attenuate the high frequency
component of the signal that does not convey any information.
Filtered signal is sampled at a rate slightly higher that the Nyquist rate
Reconstruction Filter (LPF )
• Use of sampling rate higher than Nyquist rate helps in the design of
reconstruction filter
• Reconstruction filter is used to recover the original signal from its sampled
version.
• reconstruction filter is a low pass with a passband extending from –W to W
which is itself determined by anti aliasing filter
• Filter has a transition band from W to fs –W;
• The reconstruction filter has a well defined transition band which makes it
physically realisable.
Reconstruction filter :LPF
• LPF is used to recover original signal from its samples

• Process of reconstructing CT signal from its samples is called interpolation
• Also known as interpolation filter
• In contrast to ideal LPF, a practical LPF has an amplitude response which
decreases gradually to zero .
• This transition band in practical LPF helps in reconstruction of original signal.

• Sampling freq is --------
1. fs=fm
2. fs≥2fm
3. fs≤fm
Nyquist rate is
1. fs=2fm
2. fs≥fm
3. fs≤fm

Quantization
• Original CT signal may be approximated by a signal constructed of discrete

components selected on a minimum error basis.
• Sampling results in a series of pulses of varying amplitude values ranging
between two limits: a min and a max.
• The amplitude values are infinite between the two limits.
• We need to map the infinite amplitude values onto a finite set of known
values.
• This is achieved by dividing the distance between min and max into L zones,
each of height 
= (max - min)/L
Quantisation is defined as the process of transforming the sample amplitude
x(nTs) of a msg signal x(t) at time t=nTs into discrete amplitude y(nTs) taken from
a finite set of possible amplitude.
qk : { xk <x≤ xk+1} k=1,2,3,…L
• quantisation is memoryless and instantaneous.
• Discrete amplitude xk is called decision levels or threshold.
• y(k) k=1,2,3,… L are representation or reconstruction levels
• Spacing between 2 adjacent representation levels are called step size or quantum (∆)
• The midpoint of each zone is assigned a value from 0 to L-1 (resulting in L values)
• Each sample falling in a zone is then approximated to the value of the midpoint.
• Both sampling and quantization result in
the loss of information
• The discrete amplitudes of the quantized
output are called as representation levels
or reconstruction levels.

Types of Quantization
• Uniform Quantization and Non-uniform Quantization.
• The type of quantization in which the quantization levels are uniformly
spaced is termed as a Uniform Quantization. (step size remains same )
• The type of quantization in which the quantization levels are unequal and
mostly the relation between them is logarithmic, is termed as a Non-
uniform Quantization. (step size remains varies according to the input s/g
values )
• Types of Uniform Quantization
• Mid-tread
• Midrise
• The Mid-Rise type is so called because the origin lies in the middle of a
raising part of the stair-case like graph. The quantization levels in this type
are even in number.
• The Mid-tread type is so called because the origin lies in the middle of a
tread of the stair-case like graph. The quantization levels in this type are odd
in number.
• Both the mid-rise and mid-tread type of uniform quantizers are symmetric
about the origin.

Quantization Noise
• It is a type of quantization error,
which usually occurs in analog audio signal,
while quantizing it to digital.
• For example, in music,
• the signals keep changing continuously, where
• a regularity is not found in errors.
• Such errors create a wideband noise called as
Quantization Noise
∈= xq(nTs-x(nTs)
Quantisation noise

Quantization Zones
• Assume we have a voltage signal with amplitutes Vmin=-20V
and Vmax=+20V.
• We want to use L=8 quantization levels.
• Zone width  = (20 - -20)/8 = 5
• The 8 zones are: -20 to -15, -15 to -10, -10 to -5, -5 to 0, 0 to
+5, +5 to +10, +10 to +15, +15 to +20
• The midpoints are: -17.5, -12.5, -7.5, -2.5, 2.5, 7.5, 12.5, 17.5

Assigning Codes to Zones
• Each zone is then assigned a binary code.
• The number of bits required to encode the zones, or the
number of bits per sample as it is commonly referred to, is
obtained as follows:
nb = log2 L
• Given our example, nb = 3
• The 8 zone (or level) codes are therefore: 000, 001, 010, 011,
100, 101, 110, and 111
• Assigning codes to zones:
• 000 will refer to zone -20 to -15
• 001 to zone -15 to -10, etc.
Figure 4.26 Quantization and encoding of a sampled signal

Pulse Modulation Techniques
• The different analog pulse modulation techniques are
• PAM( Pulse Amplitude Modulation)

• PWM(Pulse Width Modulation)- PDM,PLM
• PPM(Pulse Position Modulation)- PTM

Pulse Amplitude Modulation (PAM)
• Pulse Amplitude Modulation (PAM) is an analog

modulating scheme in which the amplitude
of the pulse carrier varies proportional
to the instantaneous amplitude of the
message signal.

Pulse Width Modulation
• In PWM,the width of the pulse proportional
to the amplitude of the modulating signal.
• Pulse width is maximum during when signal

reaches maximum positive.
• Pulse width is minimum during when signal
reaches minimum negative.

• Pulse Position Modulation (PPM) is an analog
modulating scheme in which the amplitude and
width of the pulses are kept constant,
while the position of each pulse, with reference to
the position of a reference pulse varies according to the
instantaneous sampled value of the message signal.
• Starts from negative trailing edge of a PWM signal.
Nithya M,Asst.Prof, Dept of ECE

2/9/2022 53
Pulse Code Modulation
• Digital pulse modulation technique
• Complex compared to PAM,PWM &PPM (more no of operations)
• The essential operations are sampling , quantising & encoding.
• Analog s/g is sampled according to sampling theorem resulting
• in a discrete –time signal.
• quantising & encoding is performed in a single ckt known as ADC(Analog
to Digital Converter).
• At receivers,DAC are used.
• To minimise the accumulated effects of distortion & noise, regenerative
repeaters are used.
• PCM output is digital pulse modulation,it is a form of digital pulses of
constant amplitude ,width & position.
PCM Transmitter

PCM Transmitter
• In PCM generator, the signal is first passed through LPF of cutoff
frequency fm
• So bandlimited to fm
• The sample & hold ckt samples the signal at a rate of fs(fs ≥ 2fm).
• Sampling freq is properly chosen to avoid aliasing
• The output of sample & hold ckt samples the signal at a rate of fs
• The output is x(nTs) which is discrete in time & continuous in
amplitude.
• A q level quantizer compared the input with its fixed digital levels.The
output of quantiser is xq(nTs).
• xq(nTs) is given to the binary encoder .This encoder is known as
digitizer .
PCM Transmission Path

• Path b/w PCM txr & rcvr.
• Aim is its ability to control the effects of distortion & noise when PCM
travels in the channel .
• Accomplished by using a chain of regenerative repeaters
• These are placed close to each other in the transmission path.
• These regenerative performs equalization, timing & decision making.
• So the repeater will reproduce a clean noise free PCM signal from the
PCM distorted signal by the channel noise.
• This improves the performance in presence of noise .

Blk dgm of Regenerative Repeater

• The amplitude equalizer shapes the distorted PCM so as to compensate for
the effects of amplitude & phase distortion .
• The timing ckt produces a periodic pulse train which is derived from input
PCM pulses . The pulse train is applied to the decision ckt .
• This decision device uses the pulse train for sampling the equalized PCM
pulses.
• Sampling is carried out at instants where SNR is maximum.
• The decision device makes a decision whether the equalized PCM wave at its
input has a 0 or 1 corresponding to the instant of sampling .
• The decision is made by comparing equalized PCM with a reference level
called decision threshold.
• At the output of the decision device, we get a clean PCM without any trace of
noise .

PCM Receiver

• Regenerator at the start of PCM receiver reshapes the pulse & removes
the noise.This signal is converted to parallel digital words for each
sample.
• The digital word is converted to its analog value denoted as xq(t) with
the help of a sample & hold circuit.
• The output of sample & hold ckt is allowed to pass through a LPF to get
the appropriate message signal x(t).
• It is impossible to extract original signal b’coz of permanent quantisation
introduced during quantization.
• The quantization error can be reduced by increasing the binary levels,
increasing ‘n’ will increase the signaling rate(r) & txn b.w also increases.
• Choice should be in such a way so as to reduce the quantization noise .
(tolerable limits )
Proof

COMPANDING
• Companding is nonuniform quantisation
• It is derived from two words compression & expansion.
• It is required to improve the SNR of weak signals.
• ( can affect the quality of signals ) I/P Compressor
Uniform
Expander
Quantiser O/P
• Np= ∆ /12
2
• It is difficult to implement non uniform quantisation b’coz it is not known how

the s/g is going to advance.
• Weak signals are amplified & strong signals r attenuated before applying
them to a uniform quantiser. This process is called compression .
• A non uniform quantizer is equivalent to passing the base band signal through
a compressor and then applying the compressed signal to uniform quantizer.
• At Receiver, the opposite operation followed is called expansion. The ckt
used is called as expander .
• The compression of s/g at txr & expansion at the receiver is called
companding.

Types of Compressor Characteristics
• We need a linear Compressor characteristics for small amplitudes &
logarithmic characteristic elsewhere.
• This can be achieved by two methods .
❖1. µ-law companding
❖2.A-law companding
• 1. µ-law companding
In µ-law companding,it is approximately linear for small values of input
levels & logarithmic for high input levels.
The µ-law compressor chara can be expressed as

• µ=0 uniform quantisation
• µ law used for speech &
• music signals
• used for PCM telephone

systems in US, Canada &
Japan.

2. A- law Companding
• In A- law Companding , the compressor characteristic is piece wise ,
made up of a linear segment of low level inputs & logarithmic segment
for high level inputs .
• Linear for low level inputs & logarithmic for high level inputs.
• Use in PCM telephones system in Europe.
• A= 87.56
• A=1 uniform quantization

NOISE IN PCM
• 2 main sources of noise
• Channel noise: anywhere between the txtd output and receiver input.
• Quantisation noise: introduced in transmitter and carried all the way along to
the receiver output.
• Channel noise
• - introduce bit errors
• 1 as 0 and 0 as 1
• More bit errors means more dissimilarity between output received signal and
input msg signal
• Avg probablility of symbol error(BER) is the prob that the reconstructed symbol
at receiver differs from the transmitted binary symbol on avg
• Different symbol errors needs to be weighted differently
• To reduce BER model the channel as AWGN
• Quantisation noise
• Designers control
• Quantisation noise can be reduced by increasing the number of
levels.
• No of levels increases the bit rate (r) & txn bandwidth.

Differential Pulse Code Modulation(DPCM)
• When a voice or a video signal is sampled at a rate higher than the Nyquist rate as
in DPCM, the resulting sample has high degree of correlation with its adjacent
samples.
• High correlation means on an average, the signal does not change rapidly from on
sample to the other.
• The difference between the near by sample variance is much lesser than the signal
variance itself
• This samples on encoding results in redundant information where necessary
symbols are not generated.
• Removing the redundant bits to attain efficient coded signal is the basic idea of
DPCM
• DPCM is a procedure of converting an analog into a digital signal in which an
analog signal is sampled and then the difference b/w the actual sample value
& its predicted value (predicted value is based on previous sample or samples) is
quantized and then encoded forming a digital value.
• The samples at 4Ts, 5Ts & 6Ts are
encoded to the same value of (110).
• The information can be carried
only by one sample value.
• But three samples are carrying the
same information means redundant.
• At 9Ts & 10Ts, the difference b/w the
samples r the last bit & first two bits
are redundant since they do not change.
So by reducing the redundant information , no of bits can be reduced.
• It is an intelligent decision to take a predicted sampled value, assumed from its
previous output and summarise them with the quantized values. Such a process is
called a Differential PCM (DPCM) technique.
DPCM TRANSMITTER
• By encoding the quantizer output ,a variant of PCM is obtained as DPCM.
• e(n) = m(n) - m^(n)
• Quantizer output is eq[n] = e[n] + q[n] where q[n] is the quantisation error
• The prediction filter i/p is obtained by adding the quantised output to the predicted value
m^[n] , ie mq[n] = m^[n] + eq[n]
• On substituting eq[n] in above equation we get
mq[n] = m^[n] + e[n] + q[n]
• Also m^[n] + e[n]= m[n] therefore
mq[n] = m[n]+ q[n]
• That means irrespective of the properties of prediction filter, the quantised sample mq[n] at
the prediction filter differs from the original sample m[n] by quantisation error q[n].
• With a good prediction filter, the variance of the error is smaller than variance of the
original signal so that the quantisation noise over a number of levels can be considerably
reduced
2/9/2022
in DPCM than compared to PCM.
Nithya M,Asst.Prof, Dept of ECE 4.79
DPCM RECEIVER
• Decoder reconstructs the quantised error signal.

• At the receiver, the quantised version of the original signal is developed from the
decoder using the same prediction filter used in transmitter.
• In the absence of channel noise, encoded signal at receiver input is same as that
at the transmitter output.
• Accordingly the receiver output is equal to the mq[n], which differs from the
original input m[n] only by quantisation error q[n] due to quantising the
prediction error e[n].
Applications
• DPCM used in compression of images and video signals.
• DPCM conducted on signals with correlation between successive samples
leads to good compression ratios.
Design of linear predictor. Wiener-Hopf equation.

Prediction method used to estimate the behavior of a signal at an instant of
time.

LINEAR PREDICTION METHOD

Design

• Toeplitz Matrix
• All elements of autocorrelation matrix Rx are Rx(0)=𝜎𝑥2
• Elements on any other diagonal parallel to main diagonal are equal
• The matrix having Toeplitz property can be defined by a set of
autocorrelation values Rx(0) ,Rx(1)… Rx(p).
• The p filter coefficients of the optimum filter are uniquely defined by
variance Rx(0)=𝜎𝑥2 & p values of the autocorrelation function for
timelag of Ts,2Ts…..pTs

Delta Modulation
• PCM requires larger signalling rate and transmission channel bandwidth.
• To over come this, DELTA modulation is employed.
• The incoming messages are oversampled to increase the correlation b/w adjacent
samples .
• This is done to permit the use of quantizing strategy for constructing the encoded signal.
• The present sample value is compared with the previous samples & quantized to 2 levels
±Δ corresponding to positive and negative differences.
• If the diff is positive,the approximated s/g is increased by one step. ie; +Δ Transmit 1
• If the diff is negative,the approximated s/g is reduced by one step. ie; - Δ transmit 0
• Input is approximated to step s/g by DM. This step size is fixed .
• Provided that the s/g does not change too rapidly from sample to sample
• The type of modulation, where the sampling rate is much higher and in which the step
size after quantization is of smaller value Δ, is termed as Delta Modulation.

Features of DM
• An over-sampled input is taken to make full use of a signal correlation.

• The quantization design is simple.
• It transmits one bit /sample.
• The design of the modulator and the demodulator is simple.
• The stair-case approximation of output waveform.
• The step-size is very small, i.e., Δ (delta).
• DM is 1 bit version of DPCM
• The bit rate can be decided by the user.
• It requires simpler implementation.
• The quality is moderate.

• m(n)=m(nTs) where Ts is the sampling period n= 0, ±1, ±2….
• The principle of delta modulation can be explained by the following
eqns.
• e(n)= m(n) - mq(n-1)
where e(n) is an error signal representing the difference b/w the
present sample m(n) & the last approximation mq(n-1).
• eq(n) = Δ sgn(e(n))
eq(n) is the quantized version of e(n)
• mq(n)= mq(n-1)+eq(n)
The quantizer o/p is coded to produce the delta modulation signal.

• It can be generated by applying the sampled version of signal to a
modulator that involves a comparator, quantizer & an accumulator.
• The block labelled z-1 inside accumulator represents a unit delay
• (1 delay = 1 sampling period)
• the comparator computes the diff b/w the 2 inputs
• The quantizer consists of a hardlimiter with an input output relation
which is a scaled version of signum function.
The quantizer o/p is applied to the accumulator producing the result
𝑛
𝑛 𝑒𝑞(𝑖)
mq(n)= Δ෌𝑖=1 sgn 𝑒 𝑖 = ෍
𝑖=1
If m(n) > mq(n), a positive increment +Δ is applied to the approximation.
If m(n) < mq(n), a negative increment -Δ is applied to the approximation.
DM Receiver
• At the receiver ,decoder reconstructs the quantized error signal
• Uses the same prediction filter used in the txr.
• The delay circuit provides the previous samples
• High frequency out of band quantisation noise is removed with the help
of LPF.
• Two types of Quantization Error
• 1 slope overload distortion (startup error)
• 2. Granular noise (Hunting)

1.Slope overload distortion
• Distortion arises due to large dynamic range of input signal
• The rate of rise of input signal is so high that the staircase signal cannot
approximate it, ie; the step size Δ becomes too small for staircase signal
to follow the steep segment of the input signal.
• Thus there is a large error b/w staircase approximated signal & original
input signal
• This error is called as Slope overload distortion
• To reduce this error, the step size is increased ,when the slope of input
signal is high.

2.Granular Noise
• Occurs when the step size is too large to small variations in the input signal
• For small variations in the input signal, the staircase signal is changed by a
large amount because of large step size.
• The error b/w the input signal & the approximated signal is called granular
noise.
• The solution to this problem is to make the step size small .
• These errors can be overcome by adaptive delta modulation
• ie; The step size is made adaptive to variations in the input signal , known
as Adaptive Delta Modulation

• For steep rise in input signal the step size must be large – else
it results in slope overload distortion
• If there is no much difference in input the step size must be

made small.- else it results in granular noise.
• APPLICATION
• voice telephony applications

Mod 3

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Mod 3

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mod 3

Uploaded by

Copyright:

Available Formats

MOD 3

Source coding theorems I and II (Statements only). Waveform coding.

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 1

• Purpose of coding is to improve the efficiency of the communication

Source Encoder Channel Decoder Receiver

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 3

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 5

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 6

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 8

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 9

message Code Length No of

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 10

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 11

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 12

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 13

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 14

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 15

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 16

The theorem specifies channel capacity C as a fundamental limit on the

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 17

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 18

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 19

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 20

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 21

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 22

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 24

• 1. Impulse sampling/ Ideal sampling method

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 4.26

• Aka rectangular pulse sampling .

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 27

2/9/2022 Nithya M,Asst.Prof, Dept of ECE

2/9/2022 Nithya M,Asst.Prof, Dept of ECE

For an intuitive example of the Nyquist theorem, let us

It can be seen that sampling at the Nyquist rate can create

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 4.31

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 4.32

• fs=2fm is Nyquist rate.

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 4.34

• LPF is used to recover original signal from its samples

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 38

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 39

• Original CT signal may be approximated by a signal constructed of discrete

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 41

2/9/2022 Nithya M,Asst.Prof, Dept of ECE

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 4.45

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 4.47

2/9/2022 Nithya M,Asst.Prof, Dept of ECE

• PAM( Pulse Amplitude Modulation)

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 50

• Pulse Amplitude Modulation (PAM) is an analog

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 51

• Pulse width is maximum during when signal

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 52

Nithya M,Asst.Prof, Dept of ECE

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 57

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 59

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 60

2/9/2022 Nithya M,Asst.Prof, Dept of ECE 61