0% found this document useful (0 votes)
18 views43 pages

Pulse Modulation Lecture Notes 02

The document outlines the Pulse Modulation module (EENG524) at Fourah Bay College, detailing the course objectives, assessment criteria, and key topics such as Pulse Analogue Modulation, Time Division Multiplexing, and Information Theory. It emphasizes the importance of understanding modulation theory, noise effects, channel capacity, and coding schemes in communication systems. The module includes practical applications and theoretical concepts essential for engineering students in the Electrical and Electronic Engineering Department.

Uploaded by

Umarr A Sesay
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
18 views43 pages

Pulse Modulation Lecture Notes 02

The document outlines the Pulse Modulation module (EENG524) at Fourah Bay College, detailing the course objectives, assessment criteria, and key topics such as Pulse Analogue Modulation, Time Division Multiplexing, and Information Theory. It emphasizes the importance of understanding modulation theory, noise effects, channel capacity, and coding schemes in communication systems. The module includes practical applications and theoretical concepts essential for engineering students in the Electrical and Electronic Engineering Department.

Uploaded by

Umarr A Sesay
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 43

FOURAH BAY COLLEGE

INSTITUTION

ELECTRICAL AND ELECTRONIC ENGINEERING


DEPARTMENT

PULSE MODULATION
MODULE TITLE

EENG524
MODULE CODE

PROF JONAS A S REDWOOD-SAWYERR


MODULE LECTURER
Mobile : +232 76 670904, +232 88 001019, +232 25 273401
Email : [email protected],
CONTACTS
[email protected], [email protected]
LECTURE HOURS/LAB.
3-0-3
PRACTICALS/CREDIT HOURS
Examination will account for 70% of the overall grade whilst
continuous assessment will account for 30%. Question sets at the
ASSESSMENT end of each lecture must be returned for grading one week after
receipt of lectures and will account for 10% of the continuous
assessment grade.
OBJECTIVES AND
GENERAL OBJECTIVES:
OUTCOMES
1. To understand the theory of Pulse Modulation (i.e. Pulse
Analogue modulation and Pulse Code Modulation) in contrast to
other forms of modulation schemes discussed in earlier years.
2. To understand the theory of Time Division Multiplexing (TDM)
in Pulse modulation systems while introducing the concepts of
sampling and quantisation.
3. To understand the theory of modulation in the presence of
noise and the computation of signal-to-quantisation noise ratio
in communications systems.
4. To understand the concepts defining Information Theory, while
discussing the properties of sources and channels in the
context of the Shannon and Hartely laws of information.
5. To understand the theory of Information content, and Average
Information or entropies of sources and channels.
6. To understand how channels are modelled for analysis and how
to calculate the average information transmission rate over
discrete channels in terms of source and destination entropy
rates.
7. To define channel capacity in terms of mutual information and
Information transmission rates.
8. To analyse the fundamental theorems of Information theory in
terms of information Rate and Channel capacity.
9. To understand aspects of coding to match sources to channels
by making reference to different coding schemes.
10. To derive the Syndrome vector [S] in error detection and
correction and to understand the function of the parity check
matrix [H] in the discussion of transmission efficiency and
redundancy.
Pulse Modulation: Pulse Analogue Modulation, Pulse code
Modulation.
Signal to Quantisation Noise: Linear codes.

Information Theory
Information Content, Entropy or Average Information Content,
Entropy Rate, Entropy of Joint Event, Redundancy & Information
Efficiency or Relative Entropy, Channel Representation, Transition
Probability Matrix, Noise-free Channel, Channel Probability or Joint
Probability Matrix, Information Transmission Rate, Channel Capacity
MODULE DESCRIPTION C = Max { I (X : Y)} bits/symbol, where I (X : Y) = Trans-information
or Mutual Information = H (X) – H (X/Y) = H (Y) – H (Y/X), Channel
Capacity and Information Transmission Rate, i.e. C = Max {H’(X) –
H’(X/Y)} bits/sec., Hartley – Shannon Law of Information, Shannon’s
Law for Channel Capacity C = 2W log 2 {(S + N)/N}½ = B log2(1+S/N)
bits/sec., Fundamental Theorems of Information Theory:
R ≤ C … (1);
R > C … (2).

Coding:
Basic Principles, Binary Coding, Fano Coding Scheme, Huffman’s
Coding Scheme, Coding Efficiency, Algebraic Codes/Block or (n, k)
Codes, Convolution Code, Error Detection and Correction, the
Syndrome (S) = [H][R], Parity Check matrix, Hamming Codes.
References
SUGGESTED 1. Ziemer and Tranter. Principles of Communications. Systems,
Modulation and Noise. Houghton Miffin
READINGREFERENCE
2. A Bruce Carlson. Communication Systems. An introduction to
TEXTS/MANUALS/WEBSITES Signals and Noise in Electrical Communications. McGraw Hill.
02 – Pulse Analogue Modulation
LECTURE NUMBER
LECTURE DURATION 01
(HOURS)
At the end of this lecture the students should be able to :
SPECIFIC INSTRUCTIONAL 1. Differentiate between the various forms of pulse modulation
OBJECTIVES AND LEARNING 2. Apply the basic theoretical expressions to solve related
OUTCOMES engineering problems
3. Analyse the concepts of bandwidth and signal recovery in pulse
analogue modulation schemes

The Sampling Principle – Ideal Sampling

For this analysis our sampling signal is the unit impulse ∂(t) which is represented in Figure 15.
v(t)

(t+To) (t-To) (t-2To)


(t+2To)
(t )

t
2To To 0 To 2To

Fig. 16. Impulse train

Let us first of all recall the properties of the delta function (t)
(i) (t) = 0 except at t = 0

(ii)   t dt 1
 i.e. area under the delta function is equal to its weight or strength

The periodic impulse train can be written as:



v t    t  kTo 
k  
1.0
The exponential form of its Fourier series is given by :


v t )   Vn e jnws t
n 
where
Ts
2
1 1
V n= ∫
T s −T
∂(t ) e
− jn w t
dt= ; ∀ n
s

Ts
s
2

...2
2π 1
Also w s= ∧fs=
Ts Ts

The ideal sampled signal with baseband signal v m(t) is given by


v s (t )=v m (t) ∑ δ (t−k T s)
k=−∞

∞ ∞
1
¿ v m (t) ∑ V n e jn w t =¿❑ v m (t)
s

T
∑ e jn w t ¿
s

k=−∞ k=−∞ ❑

..3
Taking a general term say
1 jnw t
v (t) e s

Ts m
and invoking the Frequency shift theorem we can write its spectrum as

1 jnw t 1
v (t) e s
↔ V m (w−n ws )
Ts m Ts
...4
Giving

1
V s ( w )= ∑ V (w−w s )
T s n=−∞ m
...5

Spectrum of Vm(w)
|V m (f )|

f
-fs -B B fs
Fig. 17. a. Baseband signal spectrum
|V s (f )|

2fs -fs -B B fs 2fs f


b Output spectrum

Hence provided fs ≥ 2B, we will obtain the spectrum as above. We note that a true replica of the
baseband spectrum centred at ±nfs is obtained unlike the case of using a pulse train as the switching
or gating signal discussed earlier. It can be seen that filters can be designed centred at any ±nfs to
recover the original signal. This is therefore the ideal situation.

TRUE PAM SIGNALS

S(t)
Balanced Modulator
1.0 τ
v(t) xs(t) = v(t)s(t)
t
T
s(t) v(t)

Figure 18. a. Balanced modulator xs(t)


b Switching waveform τ
c Baseband signal
d vPAM(t) t
T
Analysis
From the Fourier analysis we can write the switching function in its exponential form as


S ( t )= ∑ f s τSinc (n ¿ f s τ )e j 2 πn f τ ¿
s

n=−∞

...6
Alternatively we can write


S ( t )=C 0+ ∑ 2C n cosn w s t ..
n=1

7
where
C n=f s τsinc( n f s τ ); ωs =2 π f s
...8
Hence noting that the output

xs(t) = v(t). S(t) ..9

we obtain

x s ( t ) =C o v ( t ) +2 C1 v ( t ) cos ωs t+2 C 2 v ( t ) cos 2 ω s t +…


..10
Giving its spectrum as
X s ( f )=C 0 V ( f ) +C 1 {( V ( f −f s ) +V ( f + f s ) }+C 2 {V ( f −2 f s ) +V ( f +f s ) }+…
..11
where
v(t) V(f)
..12
Once again we note that the spectrum is centred on the dc term as well as on ±nf s. Hence filtering can
be used to recover the original spectrum of the baseband signal provided the Nyquist condition is met.
Bipolar Choppers

The circuit in Fig. 19a shows a schematic representation of a Bipolar Chopper, where the baseband
signal and its inverted form are alternatively transmitted using a switching function sampling at a rate
of fs. This effect can be realised by applying a bipolar switching waveform alternating between ±1.
Fig 19b shows the output of the chopper in the time domain.

X1 xs(t) m(t)
m(t)
fs
SPDT_OPEN

t
a. Bipolar Chopper xs(t)
x(t)
S(t)

1.0
t t

-1.0

Ts Ts/2
c Bipolar switching signal d. Bipolar modulation
PAM
Ts/2
Figure 19. Timing waveforms
Let us assume the spectrum of the baseband signal as shown in Fig. 20a and the bipolar switching
waveform as shown in Fig. 20b. We can therefore write in Fourier series expansion:

[ 4 1
3
1
5 ]
1
S ( t )= cos ω s t− cos 3 w s t+ cos 5 ws t− cos 7 w s t+ …
π 7
..13
From which we obtain the output xs(t) as

[ 4 1
3
1
5
1
]
x s ( t ) = x (t)cos ω s t− x (t )cos 3 ws t+ x (t)cos 5 w s t− x (t)cos 7 ws t+ …
π 7
..14
We note the DSBSC effect of the multiplication operation with the spectra centred on ±nfs shown in
Fig. 20c.
NB.
1. The Fourier series of the switching waveform has no DC term, hence the PAM waveform has its
first term centred on ± fs.
2. Only odd harmonics are present in the analysis
3. Bandpass filtering is required to extract the baseband signal from the modulation with the filters
centred on nfs where n is an odd integer.
4. This form of PAM forms the basis for the production of the baseband multiplexed signal for FM
stereo.

| X (f )| | X s (f )|

f f
fs 3fs 5fs
a. Baseband signal spectrum c. vPAM(t)
S(t)
1.0
t

-1.0 b. Bipolar switching signal Figure 20

Quasi PAM or ‘Sampled and Held’ signal


Chopper modulation techniques provide a convenient method of generating PAM as discussed.
However in preparation for encoding the samples in readiness for Pulse Code Modulation, Flat top
Sampling or alternatively, Quasi PAM techniques offer a more convenient method. We shall now
discuss the process of Quasi PAM. Fig 21 shows the timing waveform of the Sampled and Held’ signal
The following steps summarize the process:

Samples
v(t)

Hold
Hold period
Period Ts t

Figure 21. S & H signal Ts

1. The signal is sampled in accordance with the Sampling Principle at intervals of Ts.
2. The samples are ‘held’ at their values for a period of time up to Ts seconds long, when the next
sample is taken.
3. The new value of the sample will either rise or fall to the instantaneous value of the baseband
signal at the sampling time.

The S & H circuit principle

A rudimentary circuit is shown in Fig 22 demonstrating the Sample and Hold principle.

In this circuit the switch S is momentarily closed, i.e. it samples v(t), and causes the capacitor C to
charge to the sample value at the sampling instant. S is kept open for Ts seconds before closing again
for an instant to take another sample of v(t) and the cycle repeats itself.
Ideally the source resistance R s → 0, e.g. a voltage follower circuit could be used for this purpose, and
R → ∞, i.e. a tantalum capacitor followed by another voltage follower circuit. S is an electronic switch,
i.e. a diode bridge or a FET as discussed earlier.

The S & H sample is then ready for conversion to binary form during the time interval available
between samples, i.e. Ts using analogue to digital techniques.

S
R1
V1
C1
R2

Figure 22. Simple S & H circuit


Flat top sampling

Sampling switch Discharge switch


Q1

Q2 vPAM(t) = xp(t)
C1 x(t) x(kTs)
x(t) G1(p(t-kTs)) v0(t) =
G2 xp(t)
τ
kTs t

Figure 23. a. Rudimentary Circuit


b. Flat top sampling waveform

Noting that an n-channel FET is triggered by a positive signal, the circuit is activated by the closing of
the sampling switch by a gate pulse on G1, applied briefly thereby charging capacitor C to the sampled
value. This value is held until discharged by a pulse applied at gate G2. In both instances when a
positive gating signal is applied to the FETs the Drain –Source presents a near short circuit. In the case
of G1, the sampled value is transmitted to the capacitor C1. In the case of G2 the short circuit causes
the capacitor to discharge quickly to zero, awaiting the next sample value.

Periodic gating of the S & H circuit generates the sampled wave given by


x p (t )=∑ x(k T s¿ ) p(t−k T s )¿
k

..15
Each output pulse of duration τ, represents the instantaneous value of the sampled signal.
Let us now analyze the waveforms using Signal theory.
Now
p ( t−k T s )= p ( t )∗∂ (t−k T s)
...16

i.e. the convolution of a function with the delta function produces the original function at the location of
the delta function
The sampled output becomes

x p (t )=∑ x ( k T s ) [ p ( t )∗∂ ( t−k T s ) ]
k

...17

¿ p ( t )∗∑ x ( k T s ) ∂ ( t−k T s )
k

¿ p ( t )∗x ∂ (t)
...18
Taking the Fourier transforms we obtain the output transform as


X p ( f )=P ( f ) {f s ∑ X ( f −n f s ) }=P ( f ) X ∂ (f )
n

...19
Note: The Fourier transform of the convolution of two functions is the product of their individual
transform.

Practical Sampling

The following differences are observed between ideal and practical sampling:

1. The sampled wave consists of pulses having finite amplitude and duration rather than impulses.
τ
t

2. Practical reconstruction filters are not ideal filters


3. The messages to be sampled are time-limited signals whose spectra are not and cannot be
strictly bandlimited.

For example let us consider the gate pulse as our data


G(ꞷ)
gT(t)

A


 4  2 2 4
   
- /2  /2 t

Fig.a Time limited Gate function Fig. b The spectrum of g(t)

4. Also it can be noted that the shape of the sampling pulses used is not very critical as the basic
process reduces to a multiplication operation.

Figure 24 provides a description of the limitations of practical filters in contrast to ideal filters and ideal
sampling. Let us assume a baseband spectrum as shown in Fig. 24a. The output PAM waveform is
shown in Fig 24 b where a practical reconstruction filter response is shown. It can be seen that in this
case the filter response does not drop off sharply as in the case of an ideal filter. In filtering out the
baseband spectrum at the dc level it is seen that additional spectral components will appear at the
output of the filter at frequencies greater than f s – W. This occurrence manifests itself as a hissing high
frequency noise in audio systems and is outside the message band. Although an undesirable
phenomenon, it is not critical since careful design of the roll off of the filter can reduce this situation,
which in any case occurs outside the message band of the baseband signal and is heavily attenuated
compared to the message signal. Furthermore these components are a function of the baseband
signal x(t). If x(t) = 0 they will disappear. For x(t) ≠ 0, the spectral amplitudes within W are large
enough to reduce the effect of the components at the tail end of the filter response and so making
their effect insignificant.

Effect of Aliasing

Figure 24c shows the spectrum of a message signal which has components well beyond the nominal
bandwidth, W, of the message. These are referred to as pre-cursor and post-cursor components.
Here we note that spectral amplitudes above W are insignificant hence the stated bandwidth of W for
the signal. The sampled output is shown in Fig. 24d.
Xs(f)
Spectrum of baseband
Filter response

w
-W W
a. Baseband spectrum f
-fs -W 0 W fs-W fs
b. Practical reconstruction filter
X(f)
Xs(f)
Bandlimiting the baseband
Prior to sampling fs

f f1
-f1 -W 0 W f1
c. Spectrum of time limited message -fs -WW fs f
f s - f1< W
d. Sampled message spectrum showing overlaps

Figure 24. Practical reconstruction filter characteristics

It can be seen that a sidelobe which originally had a peak at ±f 1, now appears at fs – f1, well within the
bandwidth of the first sample centred at the zeroth frequency. This will lead to distortion of the
message signal, making it very difficult to extract the original spectrum using a BPF, without any
distortion. This phenomenon occurs whenever a frequency component is undersampled and when fs <
twice the largest significant frequency of the baseband signal, in this case f s < 2f1. Furthermore
frequencies such as f1 > W will now become fs – f1 < W as shown in Figures 24 c and d. The above
phenomenon of downward frequency translation, where frequencies which were originally outside
the nominal message band appear at the filter output in the form of much lower frequencies is referred
to as aliasing. Aliasing is more critical than the earlier situation where the unwanted frequency
components occurred outside the message band.

Remedy

1. The combination of careful filter design, as well as creating adequate guard bands between
samples by ensuring that fs > 2W makes practical filters approach the ideal case;
2. Introducing pre-filtering before sampling of the message or baseband signal and also if necessary
sampling at a rate much higher than the Nyquist rate combats aliasing and
3. Also although the amplitudes at f1 in the spectrum of the baseband signal may appear
insignificant it makes practical sense to apply a sampling rate fs >2f1 to ensure a reduction in
the effect of aliasing.

Xs(f)

f
-fs f1 fs

A practical example

The average voice signal normally extends beyond 10 kHz. However its energy is concentrated
between 100 and 600 Hz. In practice a bandwidth of 3 kHz is provided to transmit intelligible speech.
If a voice signal is pre-filtered by a 3.3 kHz LPF and then sampled at f s = 8 kHz, the aliased
components are typically less than 30 dB (i.e. less than1000 units) below the desired signal and will go
un-noticed by the listener. These are the standards in voice telephone system. Discuss.

PULSE CODE MODULATION


BASIC PRINCIPLES
Although the three pulse analogue modulation systems (i.e. PAM, PDM and PPM) possess the potential
for multiplex operation, they suffer (like all analogue modulation schemes) from the degrading effect of
noise and distortion which can be irreversible. This can be demonstrated through the following
diagrams in Figure 25. The waveforms consider a single channel PAM system. The degradation or
corruption in the detected baseband signal, whilst due to noise and distortion, is a consequence of
transmitting analogue sampled values of the baseband signal.

In PCM systems, the effects of noise and distortion are combated by encoding the sampled values that
constitute the PAM signal.
In principle, each sampled value is converted into a corresponding binary code of ‘on-off’ pulses (i.e.
binary 1’s and zero’s), which must be transmitted during the interval between successive sampling
instants.

At the receiver, these On-Off pulses will also be corrupted by noise. However, the receiver has only to
decide whether a pulse is present or not and hence there is greater immunity to noise.

Finally a means must be found for reconstructing the sampled values of the original baseband signal
from the binary coded Signal. Let us now discuss the following points.

m(t) vPAM(t) v’PAM(t) m’(t) (close estimate of m(t))

Sampling signal corrupted PAM signal


m(t) vPAM(t) v’PAM(t) m’(t)

t t t t
Figure 25. Effect of noise and distortion on a single channel PAM

It must be stated that it is impracticable to represent every possible sampled value of the baseband
signal in binary form as this would require an infinite-length binary code. However if we divide the
voltage range which encompasses all possible sample magnitudes of the baseband signal into a
suitably large number of levels, we can then ‘round off’ every sampled value to a nearby level. This
process of ‘rounding off’ the sampled values to a nearby pre-defined level is referred to as
quantising and the subsequent operation of generating the binary codeword appropriate to the
particular quantised sample is known as encoding. The inverse operation of recovering the quantised
sampled value from the coded binary message is called decoding. Due to the binary nature of the
codeword, there must be a definite relationship between the length of the codeword and the number of
quantising levels.

Thus for a 5-digit code the number of quantisation levels will be 2 5 = 32. Thus level 1 will be
represented by say, 00000, level 2 will be represented by 00001, and so on until level 32 which is
11111.

A block schematic of a single channel PCM system incorporating the above features is shown in Fig 26
The quantising and encoding operations are usually performed by a single circuit referred to as an
Analogue-to-Digital Converter (ADC). The decoder is also referred to as a DAC. The waveforms at
various points in the system of Figure 26 are demonstrated in the illustrative example. Here a 3 digit
code is used which means that the number of quantisation levels is 2 3 = 8.
We note that although at first sight we appear to overcome the problem of noise by encoding the
quantised PAM signal {v2(t)} into a digital form, the quantisation process performed on the original
analogue PAM signal [ v1(t)], results in some loss of signal definition. This is analogous to introducing
some ‘noise’ into the system. The induced noise is referred as quantisation noise. Also by increasing
the number of quantisation levels (and therefore the length of the binary codeword), the quantisation
noise will be reduced, and the decoded and filtered signal will approach a faithful replica of the original
baseband signal.

2022-07-21
Discuss Simulation Assignment (https://www.youtube.com/watch?v=rg6enk_PcOk;
https://www.youtube.com/watch?v=ibnz5UjQ4u0.)

Illustrative example
A single channel PCM system samples the input baseband waveform resulting in the following results:

No. Sampled
values
1 2.8 V
2 7.0 V
3 7.7 V
4 4.4 V
5 2.3 V

a. Discuss the PCM system using a suitable block schematic


b. Illustrate the waveforms v1, v2, and v3 using a 3-bit coding scheme

Solution

a. Block schematic
LPF

A/D Converter m’1(t)


m1(t) D/A Converter
v1 v2 v3 v2

Rotating switch/Decommutator
Rotating Switch/Sampler/ vPAM(t) vPCM(t) vPAM(t)
Commutator
Figure 26. A single channel PCM system

b. Waveforms

Volts (v1)
Code Level No. Quantisation
Levels (Volts)
111 8 7.5 8
7.7 V
110 7 6.5 7 7.0 V
6
101 6 5.5
5
100 5 4.5
4 4.4 V
011 4 3.5
3 2.8 V
010 3 2.5
2 2.3V
001 2 1.5
1
000 1 0.5
0
T

Fig. Samples and quantisation evels


Volts (v2)
Code Level No. Quantisation
Levels (Volts)
111 8 7.5 8

110 7 6.5 7

101 6 5.5 6

100 5 4.5 5

011 4 3.5 4
3
010 3 2.5
2
001 2 1.5
1
000 1 0.5
0
t
T
Fig. Flat top PAM (Sampled values are rounded off to nearest quantisation level.)

Sampled Quantised Quantised Binary code word


values values levels
2.8 V 2.5V 3 010
7.0 V 6.5 V 7 110
7.7 V 7.5 V 8 111
4.4 V 4.5 V 5 100
2.3 V 2.5 V 3 010

v3(t)

0 1 0 1 1 0 1 1 110 0 0 1 0
t
Fig. Binary coded sampled values

QUANTISATION

Practically, the number of pulses used in coding any level of baseband signal v i(t) is finite and an
integer value.
Let the number of quantisation levels chosen = n
Let the number of pulses in a group or the length of the codeword = m
Let the number of possible amplitudes of pulses = μ.

We therefore obtain
m
μ ≥n
..20
Or
m ≥ log μ n
..21
Example

Given a binary system with the number of quantisation levels = 128, find the length of the code word
used.

Solution: n = 128 and μ=2, i.e. a binary system.


This implies that

m = log2128 = 7, i.e. 7 pulses per codeword or a 7-bit binary code.


Quantisation error
In practice the next lowest quantisation level is chosen at the transmitter. At the Receiver, half a
quantisation step or quantum is added after the signal has been reconstructed as a quantised signal.
In effect this is equivalent to choosing the nearest level at a + or – level at the Transmitter. Hence the
greatest quantising error is half a quantum step.

Time Division Multiplexing PCM system


The PCM system previously described relates to a single channel operation, i.e. only one baseband
signal is sampled, quantised, encoded , etc. This is not typical of practical systems as in most cases
more than one baseband signal is processed and transmitted employing Time Division Multiplexing
techniques. Figure 27 demonstrates such a system. Multiplex operation in PCM is achieved via a
similar process as we discussed in PAM multiplex systems, i.e. by sequentially sampling the various
baseband signals. The multiplex PAM signal is then quantised and encoded. At the receiving end the
inverse operation of decoding is effected to give the quantised PAM multiplex signal. A demultiplexer,
shown here as a rotating switch on the receiving end, then separates the individual PAM pulse trains
corresponding to the various baseband signals. Finally, each baseband signal is recovered from the
corresponding PAM waveform by low-pass filtering.

Bank of LPFs

A/D Converter m’1(t)


m1(t) D/A Converter
v1 v2 v3 v2

m’2(t)
m2(t) Demultiplexer/Decommutator

Multiplexer/Sampler/Commutator vPAM(t) vPCM(t) vPAM(t)


Figure 27. A two channel TDM/PCM system

NB
The final output m1’(t) is slightly different from m1(t) due to the quantisation process. The same
applies to m’2(t).

Illustrative Example

A PCM multiplexing system permits the combining of two baseband signals, m 1(t) and m2(t). The
signals are band-limited to 4 kHz and 6 kHz respectively and have voltage amplitudes in the range of 0
– 8 Volts. Each signal is sampled, quantised into 8 levels and finally represented by a 3-digit code.
The binary pulses occupy 60% of their assigned time slots.
i. Draw a block schematic of the system and illustrate with waveforms.
ii. Determine the minimum sampling rate of the system
iii. What is the maximum bit duration allowable?
iv. Sketch the frame of the PCM output when the sampled values are 3.2 V and 6.8 V respectively.
v. Find the Gross bit rate of the system if one bit or marker pulse is assigned to each channel for
synchronisation.

Solution

i. Same diagram as in Fig. 27. The waveforms will be illustrated later and are as shown below.

ii. The Nyquist rate for m1(t) = 2 x 4 kHz = 8 kHz

The Nyquist rate for m2(t) = 2 x 6 kHz = 12 kHz

Therefore the minimum sampling rate required for the multiplexed system = f smin = 12 kHz.
Refer to the general rule for determining fsmin for multiplexed signals having a common sampling rate.

iii. The figure below shows the multiplexed PAM signal

vPAM(t)

T t
T/2 T/2 NB. T = common sampling interval

Equal assigned
Time slots for each binary symbol

Now we note that the assigned time slot for each binary digit can be written as

T
No. of channels T /2 T T
Assigned time slot = = = =
No of digits∈code No . of digits∈code 2 x 3 6

Noting that we are using a 3-digit code.


Now the maximum allowable value for the sampling interval T = 1/f smin seconds
i.e.
1
T max= 3
=83.3 μs
12 x 10
Now each binary digit occupies 60% of its assigned time slot. Hence we obtain the maximum
bit duration given by

83.3
τ max=0.6 x =8.33 μs .
6

iv.The frame of the PCM system is the collection/arrangement of digits representing the samples of the
signals within the sampling interval, T, of the PCM system.

Let us now consider the waveforms and the coding process

Waveforms

Volts
Code Level No. Quantisation 6.8 V: Rounded off as 6.5 V
Levels (Volts) encoded as 110
111 8 7.5 8
110 7 6.5 7 m2(t) m1(t)
6
101 6 5.5
5
100 5 4.5
4 m1(t) m2(t)
011 4 3.5
3
010 3 2.5
2
001 2 1.5
1
000 1 0.5
0 t
Figure P. Waveforms and quantisation process T
3.2 V: This is rounded off to 3.5 V and
Hence encoded as binary code 011

m 1(t) sample m2(t) sample

0 1 1 1 1 0

Start of next code for m 1(t)


t
60% of assigned
Time slot
T
Assigned Time slot for binary digit

v. Find the Gross bit rate of the system if one bit or marker pulse is assigned to each channel
for synchronisation.

0 1 1 1 1 0

Marker

t
60% of assigned
Time slot
T
Assigned Time slot for binary digits

Gross bit rate.

This is defined as
Total number of bits∈a frame (T ) 2 ( 3+1 )
GBR= = =96 kbits/ sec
Time intervalT 83.3 x 10
−6

NB.
1. When a marker pulse is inserted at the end of each rotation of the commutator for
synchronisation, the frame is given by the interval between the markers. This is referred to as
frame synchronization.
2. When markers are inserted after each channel this is referred to as channel synchronization.
3. In some cases, both frame and channel synchronisation are used to facilitate the separation or
decoding of the channels.

Channel synchronisation pulses

01 1 1 1 0

Frame Marker
(synchronization)

t
60% of assigned
Time slot
T
Assigned Time slot for binary digit

TUTORIAL EXAMPLES (See Tutorial 2021)


Modulation in the presence of noise (PCM)

Although PCM is attributed with an impressive noise immunity due to the nature of the binary coded
data that is transmitted and the process of decision making over only two levels at the receiver, the
process of quantising introduces errors during the rounding off (quantisation noise), as well as the
additive noise due to the inherent imperfection of the transmission channels. Hence two categories of
noise are inherent in PCM systems.
For this analysis however we shall concentrate on the impact of the quantisation noise on the
transmission efficiency of the signal.

Quantisation Noise

Consider an analogue signal having a peak-to-peak voltage 2V quantised into q equal levels. Let the
voltage spacing be ∆ v volts which for linear quantisation is given by
2V
∆ v=
q−1
.22
Refer to Waveforms in Fig. 28

We note that the error ε between an analogue voltage and its quantised value is always less than ½∆V.

Now let us assume that all quantisation levels have an equal probability of occurrence over a long
period of observation. The mean square error voltage ϵ 2 is then given by
1
∆V
2
1
ϵ 2=
∆V
∫ ϵ 2 dϵ
−1
∆V
2

.23
[ ]
1
1 ε3 2
∆V
1 2 ∆V3 ∆V2
¿ = . . =
∆V 3 −1
2
∆V ∆V 8 3 12
..24
Waveforms

Volts
Code Quantisation. Quantisation
Level No. Levels (Volts) 4.0
111 8 3.5
3.0
110 7 2.5
2.0 ∆v
101 6 1.5 1.0
100 5 0.5
0 t
011 4 -0.5 m1(t)
- 1.0
010 3 -1.5
-2.0
001 2 -2.5
-3.0
000 1 -3.5
-4.0

Figure 28. Waveforms for bipolar sampling process


Thus the noise power dissipated in a load of RΩ due to the error voltage ε will be
2
∆V 1
Nq= . watts
12 R
..25

Signal Power dissipated in R (assuming no noise present)

Let us assume bipolar levels are used, i.e. the quantisation levels are [Refer to Fig. 28]

1 3 5 1
± ∆ V , ± ∆V , ± ∆ V … , ± ∆ V ( q−1 ) ,
2 2 2 2

Also we assume that in a long message, each level is equally likely to occur with a probability of 1/q.
We will now add the powers associated with each quantisation level to obtain

[ ]
2 2 2 2
2 1 3 5 1
S= ( ∆ V ) +( ∆ V ) +( ∆ V ) +… ( (q−1) ∆ V )
qR 2 2 2 2
..26
1
Where the factor of 2 occurs because of the bipolar nature of the levels (e.g. ± ∆ V ), and the load
2
resistance is R

NB
Vpp =( q-1)∆ V ..27

= q∆ V , for q large ..28

Simplifying Eqn. 26 further we can write


2
∆V 2 2 2
S= [ 1 + 3 + 5 + …(q−1)2 ]
2 qR
..29
It can be shown that the series

[ ]
2
2 2 2 q −1 2
1 +3 +5 + …(q−1) =q
6
Hence we obtain
2 2
∆V (q −1)
S=
12 R
.30
The Signal power to Quantisation noise ratio is therefore given by

2 2
S ∆ V (q −1)/12 R 2
= 2
=( q −1)
N ∆V /12 R
..31
which for a large number of quantisation levels we obtain

S 2
=q
N
..32
Exercise
Table.
No. of S/N(db) No. of bits
levels 10log[(S/N)] (Codeword)
16 24 4
32 30 5
64 36 6
128 42 7
In practice non-linear quantisation is often used, i.e. the quantising intervals are not equally spaced
because the amplitude levels in speech and noise are not equally distributed. Logarithmic laws are
used that dictate the quantisation levels and step or quantum size, such as the A-law and the μ-law.

S/N versus Bandwidth

Consider an n-digit codeword. Also let the Bandwidth or the highest frequency of the signal be B.

Therefore the number of pulses transmitted per second = n/T


Where T = Sampling interval
But T = 1/fs = 1/2B
Hence the number of pulses transmitted per second = n/T = 2nB, where the sampling rate is 2B.
The width of each pulse τ is given by

Sampling interval ( T ) T
No of channels 1 1 1
τ= = = =
No . of bits∈T n bits f s n 2 nB
…33
v PAM

Fig. 29. Illustrative t


waveforms
T
n equally-spaced
binary pulses (τ)

The bandwidth W required to transmit a pulse of length τ is given by [See diagram below for
explanation]

1
W= Hz

..34
G(f)
gT(t)(i.e. binary pulse) Here trans. bandwidth
Aτ reqd. W = 1/τ. In practice W =
1/2τ is used.
A

2 1 1 2
    f
- /2  /2 t

Fig. 30 a. Gate or pulse function b. The spectrum of g(t)

Giving
W = nB ….35
1
Noting that τ = from Eqn 33
2nB
It has been shown that

S 2
=q
N

This becomes
S n❑ 2
=( 2 ) =2
2n
N
..36
Resulting in
2W
s B
=2
N
..37
i.e. an exponential increases in S/N with required transmission bandwidth W. This implies that
increasing the transmission bandwidth W increases the S/N, exponentially. This is better than the case
for analogue modulation.

Implications of increasing W and S/N (Students participation)

1. Considering the expression for the required Transmission bandwidth we obtain

1
W =nB=

…38
2. Increasing W implies a reduction in τ, i.e. allowing more bits to be transmitted within the
assigned time slot and hence a larger codeword can be used.
3. A larger code word implies an increased number of quantisation levels which in turn reduces the
quantisation error in encoding the signal [Refer to Tutorial]. This implies an increase in the S/N
and hence greater transmission efficiency. However this is at the expense of equipment
complexity and cost, a trade-off that will have to be made by the engineers.

Modulation of PCM on to a carrier

Three techniques are available for modulating PCM, now a sequential series of binary 1’s and 0’s on to
a carrier, i.e. PSK, ASK and FSK as shown in the schematic in Fig. 31

m(t) A B

Demodulator
Modulator

Figure 31. Digital Baseband system


m’(t)

Error rates

Error Rate (P)


1.0
10-1

10-3
x PSK FSK ASK

10-5
S/N
-2 0 4 6 10 12 14

NB
1. The probability of error of transmission is a function of the SNR and acceptable error limits
depends on the application.
2. An error rate of 10-6 implies 1bit error in 106 or 1,000,000 bits transmitted.
3. The average number of errors in a message of m bits length is given by:

Average No. of errors = m x Error probability ..38

Refer to Tutorial
ASSIGNMENT

1. 24 voice channels are to be transmitted via a multiplexed PAM system with a marker pulse for
frame synchronisation. The sampling frequency is 8 kHz and the TDM signal has a 50% duty
cycle. Calculate the Gross Signalling Rate, pulse duration and minimum transmission bandwidth.

2. The signal v(t) = 4cosw0t + 2cos2w0t + 2cos3w0t is to be converted into a PAM train by
multiplying v(t) by a unit amplitude impulse train given by:
2 1
s(t )= [ +coswt+ cos 2 wt + coswt+… .]
T 2


Where T = is the period of theimpulse train .
w
The signal is to be reconstituted from the PAM train by passing through a Low Pass Filter.

a. If v(t) is to be recovered without distortion, determine the minimum value of w and the minimum
bandwidth of the LPF.
b. If v(t) is sampled at 5w0, and passed through an ideal LPF with cut-off frequency of 3w 0,
determine the expression for the output signal.
MODE AND DEADLINE FOR SUBMISSION OF Paper form submitted to the Secretary,
ASSIGNMENT One week from date of posting

You might also like