2010-11-14

Multiplexing

• Multiplexing is the name given to techniques, which allow more

than one message to be transferred via the same
communication channel. The channel in this context could be a
transmission line, e.g. a twisted pair or co-axial cable, a radio
system or a fibre optic system etc.

• A channel will offer a specified bandwidth, which is available for

a time t, where t may  . Thus, with reference to the channel
there are 2 ‘degrees of freedom’, i.e. bandwidth or frequency
and time.

Multiplexing

CHANNEL
BL BH freq

BH

Multiplexing is a technique which allows k

users to occupy the
channel for the duration in time that the
channel is available.

BL
Frequency

Time t

Now consider a signal v s (t )  Amp cos(t   )

The signal is characterised by amplitude, frequency, phase and time. 2

1
 2010-11-14

Multiplexing

• Various multiplexing methods are possible in terms of the channel bandwidth and time,
and the signal, in particular the frequency, phase or time. The two basic methods are:

1) Frequency Division Multiplexing FDM

FDM is derived from AM techniques in which the signals occupy the same physical
‘line’ but in different frequency bands. Each signal occupies its own specific band of
frequencies all the time, i.e. the messages share the channel bandwidth.

2) Time Division Multiplexing TDM

TDM is derived from sampling techniques in which messages occupy all the channel
bandwidth but for short time intervals of time, i.e. the messages share the channel time.

• FDM – messages occupy narrow bandwidth – all the time.

• TDM – messages occupy wide bandwidth – for short intervals of time.
3

Multiplexing

These two basic methods are illustrated below.

time

time
M1
M2 BL
BL M3 M5
M4 M3 M4
B M5 M1 M2

BH t
BH
freq freq

t
BH
BL
M1

M2

B M3 M1 M2 M3 M4 M5

M4

M5
BL
BH
t t
FDM TDM
4

2
 2010-11-14

Frequency Division Multiplexing FDM

• FDM is widely used in radio and television systems (e.g.

broadcast radio and TV) and was widely used in
multichannel telephony (now being superseded by digital
techniques and TDM).

• The multichannel telephone system illustrates some

important aspects and is considered below. For speech,
a bandwidth of  3kHz is satisfactory.

• The physical line, e.g. a co-axial cable will have a

bandwidth compared to speech as shown next

Frequency Division Multiplexing FDM

3kHz

freq

GHz

From AM we have noted:

m(t)

freq
m(t) DSBSC
B

carrier DSBSC
cos( c t ) freq

fc
6

3
 2010-11-14

Frequency Division Multiplexing FDM

In order to use bandwidth more effectively, SSB is used i.e.

SSB
m(t) Filter
SSBSC

carrier
cos( c t ) freq

fc
We have also noted that the message signal m(t) is usually band limited, i.e.

Speech Band m(t)

Limiting SSB SSBSC
Filter
Filter
300Hz – 3400Hz
cos( c t ) 7

Frequency Division Multiplexing FDM

The Band Limiting Filter (BLF) is usually a band pass filter with a pass band 300Hz to
3400Hz for speech. This is to allow guard bands between adjacent channels.

f f f
300Hz 3400Hz 300Hz 3400Hz
10kHz

Speech m(t) Convention

4
 2010-11-14

Frequency Division Multiplexing FDM

For telephony, the physical line is divided (notionally) into 4kHz bands or channels, i.e.
the channel spacing is 4kHz. Thus we now have:

Guard Bands

Bandlimited
Speech
f

4kHz

Note, the BLF does not have an ideal cut-off – the guard bands allow for filter ‘roll off’
in order to reduce adjacent channel crosstalk.
9

Frequency Division Multiplexing FDM

Consider now a single channel SSB system. m(t) DSBSC SSBSC

SSB
BLF Filter
The spectra will be
fc

m(t)
300Hz 3400Hz
freq

DSBSC

freq
fc

freq
fc 10

5
 2010-11-14

Frequency Division Multiplexing FDM

Consider now a system with 3 channels

m1(t)

f
SSB
BLF
Filter
fc1 f1

FDM
m2(t) Signal
SSB 
BLF
Filter M(t)
f
fc2 f2

SSB
BLF
m3(t) Filter
fc3 f3
f
FDM Transmitter
Bandlimited 11
or Encoder

Frequency Division Multiplexing FDM

Each carrier frequency, fc1, fc2 and fc3 are separated by the channel spacing
frequency, in this case 4 kHz, i.e. fc2 = fc1 + 4kHz, fc3 = fc2 + 4kHz.
The spectrum of the FDM signal, M(t) will be:

4kHz 4kHz 4kHz

M(t)

Shaded areas are to

show guard bands.

f1 f2 f3
freq
fc1 fc2 fc3
12

6
 2010-11-14

Frequency Division Multiplexing FDM

Note that the baseband signals m1(t), m2(t), m3(t) have been multiplexed into adjacent
channels, the channel spacing is 4kHz. Note also that the SSB filters are set to select
the USB, tuned to f1, f2 and f3 respectively. A receiver FDM decoder is illustrated below:

SSB
LPF m1(t)
Filter

f1
fc1 Band
Limited
SSB
M(t) Filter
LPF m2(t)
FDM Back to
Signal f2 baseband
fc2

SSB
LPF m3(t)
Filter

f3
fc3 13

Frequency Division Multiplexing FDM

• The SSB filters are the same as in the encoder, i.e. each one
centred on f1, f2 and f3 to select the appropriate sideband and reject
the others. These are then followed by a synchronous demodulator,
each fed with a synchronous LO, fc1, fc2 and fc3 respectively.

• For the 3 channel system shown there is 1 design for the BLF (used
3 times), 3 designs for the SSB filters (each used twice) and 1
design for the LPF (used 3 times).

• A co-axial cable could accommodate several thousand 4 kHz

channels, for example 3600 channels is typical. The bandwidth used
is thus 3600 x 4kHz = 14.4Mhz. Potentially therefore there are 3600
different SSB filter designs. Not only this, but the designs must
range from kHz to MHz.

14

7
 2010-11-14

Frequency Division Multiplexing FDM

60kHz
For ‘designs’ around say 60kHz, Q  = 15 which is reasonable.
4 kHz
However, for designs to have a centre frequency at around say 10Mhz,
10,000kHz
Q gives a Q = 2500 which is difficult to achieve.
4 kHz

To overcome these problems, a hierarchical system for telephony used the FDM
principle to form groups, supergroups, master groups and supermaster groups.

15

Basic 12 Channel Group

The diagram below illustrates the FDM principle for 12 channels (similar to 3 channels)
to a form a basic group.

m1(t)

m2(t)

m3(t) Multiplexer
freq

12kHz 60kHz
m12(t)

i.e. 12 telephone channels are multiplexed in the frequency band 12kHz  60 kHz in
4kHz channels  basic group.

16

8
 2010-11-14

Basic 12 Channel Group

A design for a basic 12 channel group is shown below:

Band Limiting Filters
SSB Filter
DSBSC
4kHz
CH1 8.6  15.4kHz 12.3  15.4kHz
m1(t)

300Hz 3400kHz
f1 = 12kHz

4kHz

12.6  19.4kHz 16.3  19.4kHz

CH2
m2(t)
300Hz 3400kHz

f1 = 16kHz


Increase in 4kHz steps

FDM OUT
12 – 60kHz
4kHz

CH12 52.6  59.4kHz

m12(t)
56.3  59.4kHz
300Hz 3400kHz

f12 = 56kHz
17

Super Group

These basic groups may now be multiplexed to form a super group.

BASIC
12 SSB
GROUP
Inputs FILTER
12 – 60kHz

420kHz

BASIC
12 SSB
GROUP
Inputs FILTER
12 – 60kHz

468kHz

BASIC
12 SSB
Inputs
GROUP
FILTER 
12 – 60kHz

516kHz

BASIC
SSB
12 GROUP
FILTER
Inputs 12 – 60kHz

564kHz

BASIC
SSB
12 GROUP
FILTER
Inputs 12 – 60kHz

612kHz
18

9
 2010-11-14

Super Group

5 basic groups multiplexed to form a super group, i.e. 60 channels in one super group.
Note – the channel spacing in the super group in the above is 48kHz, i.e. each carrier
frequency is separated by 48kHz. There are 12 designs (low frequency) for one basic
group and 5 designs for the super group.

612 kHz
The Q for the super group SSB filters is Q   12 - which is reasonable
48kHz

Hence, a total of 17 designs are required for 60 channels. In a similar way, super groups
may be multiplexed to form a master group, and master groups to form super master
groups…

19

Time Division Multiplexing TDM

TDM is widely used in digital communications, for example in the form of pulse code
modulation in digital telephony (TDM/PCM). In TDM, each message signal occupies
the channel (e.g. a transmission line) for a short period of time. The principle is
illustrated below:
1
m1(t) 1
m1(t)
2 2
m2(t)
m2(t)
3
m3(t) 3
Tx Rx m3(t)
4 SW1 SW2
m4(t) 4
m4(t)
5 Transmission
m5(t) 5
Line m5(t)

Switches SW1 and SW2 rotate in synchronism, and in effect sample each message
input in a sequence m1(t), m2(t), m3(t), m4(t), m5(t), m1(t), m2(t),…
The sampled value (usually in digital form) is transmitted and recovered at the ‘far end’
to produce output m1(t)…m5(t). 20

10
 2010-11-14

Time Division Multiplexing TDM

For ease of illustration consider such a system with 3 messages, m1(t), m2(t) and m3(t),
each a different DC level as shown below.

m 1(t) V1

0 t

m 2 (t) V2

0 t
m 3(t) V3

0 t

SW1
‘Sample’

t
Position 1 2 3 1 2 3 21

Time Division Multiplexing TDM

V3
V2
V1
t
m1(t) m2(t) m3(t) m1(t) m2(t) m3(t) m1(t)

Channel
Time
Slots

1 2 3 1 2 3 1

Time slot
22

11
 2010-11-14

Time Division Multiplexing TDM

• In this illustration the samples are shown as levels, i.e. V1, V2 or V3.
Normally, these voltages would be converted to a binary code before
transmission as discussed below.

• Note that the channel is divided into time slots and in this example, 3
messages are time-division multiplexed on to the channel. The sampling
process requires that the message signals are a sampled at a rate fs  2B,
where fs is the sample rate, samples per second, and B is the maximum
frequency in the message signal, m(t) (i.e. Sampling Theorem applies). This
sampling process effectively produces a pulse train, which requires a
bandwidth much greater than B.

• Thus in TDM, the message signals occupy a wide bandwidth for short
intervals of time. In the illustration above, the signals are shown as PAM
(Pulse Amplitude Modulation) signals. In practice these are normally
converted to digital signals before time division multiplexing.

23

Time Division Multiplexing TDM

A schematic diagram to illustrate the principle for 3 message signals is shown below.
‘PAM’
m1(t) BLF S/H
1

fs1 Multiplexing
Analogue
‘PAM’ Serial output
m2(t) S/H
To
BLF 2 Digital Binary digital
Convertor data d(t)
fs2

‘PAM’
m3(t) BLF S/H
3

fs3

Sample and Hold Multiplexing ADC

Band limiting
Sample rate fs Converts each input
Filter 0  B Hz
fs  2B Hz in turn to an n bit code.

Again for simplicity, each message input is assumed to be a DC level.

24

12
 2010-11-14

Time Division Multiplexing TDM

25

Time Division Multiplexing TDM

26

13
 2010-11-14

Time Division Multiplexing TDM

• Each sample value is converted to an n bit code by the ADC. Each n bit code ‘fits into’
the time slot for that particular message. In practice, the sample pulses for each
message input could be the same. The multiplexing ADC could pick each input
(i.e. a S/H signal) in turn for conversion.

• For an N channel system, i.e. N message signals, sampled at a rate fs samples per
second, with each sample converted to an n bit binary code, and assuming no
additional bits for synchronisation are required (in practice further bits are required) it is
easy to see that the output bit rate for the digital data sequence d(t) is

Output bit rate = Nnfs bits/second.

27

School of Electrical, Electronics and

Computer Engineering
University of Newcastle-upon-Tyne

Baseband digital Modulation

Prof. Rolando Carrasco

Lecture Notes
University of Newcastle-upon-Tyne
2005

14
 2010-11-14

Baseband digital information

Bit-rate, Baud-rate and

Bandwidth

B denotes the duration of the 1 bit 1

bits per second
Hence Bit rate =  B

All the forms of the base band signalling shown transfer data at the same bit rate.
E denotes the duration of the shortest signalling element.
Baud rate is defined as the reciprocal of the duration of the shortest signalling element.
1
Baud Rate = baud
E
In general Baud Rate ≠ Bit Rate

For NRZ : Baud Rate = Bit Rate

RZ : Baud Rate = 2 x Bit Rate
Bi-Phase: Baud Rate = 2 x Bit Rate

AMI: Baud Rate = Bit Rate

15
 2010-11-14

Non Return to Zero (NRZ)

The highest frequency occurs when the data is 1010101010…….

i.e.

This sequence produces a square wave with periodic time   2 E

Fourier series for a square wave,

If we pass this signal through a LPF then the maximum bandwidth would be 1/T
Hz, i.e. to just allow the fundamental (1st harmonic) to pass.

Non Return to Zero (NRZ)

(Cont’d)

The data sequence 1010……

could then be completely
recovered

Hence the minimum channel bandwidth

1 1 Baud Rate 1
Bmin    Since  Baud Rate
T 2 E 2 E

16
 2010-11-14

Return to Zero (RZ)

Considering RZ signals, the max frequency occurs when continuous 1’s are transmitted.

  2 E
This produces a square wave with periodic time
Baud Rate
Bmin  f U 
2
If the sequence was continuous 0’s, the signal would be –V continuously, hence

f L  ' DC '

Bi-Phase

Maximum frequency occurs when continuous

1’s or 0’s transmitted.

This is similar to RZ with

1
Baud Rate = = 2 x Bit rate
 E

Baud Rate
Bmin  f U 
2
The minimum frequency occurs when the sequence is 10101010…….
e.g.
In this case B = E
Baud Rate = Bit rate

Baud Rate
Bmin  f L 
2

17
 2010-11-14

Digital Modulation and

Noise
The performance of Digital Data Systems is dependent on the bit error rate, BER, i.e.
probability of a bit being in error.
Prob. of Error or BER,
No of Errors E
P as N  
Total bits N
Digital Modulation

There are four basic ways of sending

digital data

The BER (P) depends on several factors

• the modulation type, ASK FSK or
PSK
• the demodulation method
• the noise in the system
• the signal to noise ratio

Digital Modulation and

Noise

Amplitude Shift Keying ASK

18
 2010-11-14

Digital Modulation and

Noise

Frequency Shift Keying FSK

Digital Modulation and

Noise

Phase Shift Keying PSK

19
 2010-11-14

System Block diagram for

Analysis

DEMODULATOR – DETECTOR – DECISION

For ASK and PSK

Demodulator-Detector-Decision

FOR FSK

20
 2010-11-14

Demodulator

Demodulator Cont’d)

1
RC 
Vout   VIN dt

Hence design RCT

21
 2010-11-14

Detector-Decision

V1 - V0 is the voltage difference

between a ‘1’ and ‘0’.

 V1  V2
(VREF   )
2 2

Detector-Decision (Cont’d)

ND is the noise at the Detector input.

Probability of Error,

1  
  1  erf 
2  2 2 N D 

Hence

22
 2010-11-14

Probability density of binary signal

v0 0 v1 v

P(v0)

- 0  vn

Probability density function of noise

 2  ND
( v0  v1 ) 2
1 1 
P0 (vn )  e 2 2
P1 (vn )
2  2

v0 v1 vn

 ( v n  v0 ) 2
1 
Pe1  
v0  v1  2
e 2 2
dv n (*)

2
vn  v0
Using the change of variable x
2

23
 2010-11-14


1
e
 x 2 dx
This becomes Pe1 
 v1  v0
(**)
2 2

The incomplete integral cannot be evaluated analytically but can be recast as a

complimentary error function, erfc(x), defined by

2
e
x2
erfc ( z )  dx
 z erfc( z )  1  erf ( z )
Equations (*) and (**) become 1  v1  v0 
Pe1  1  erf  
2  2 2 
1  v v 
Pe1  erfc 1 0  v0  v1
( vn  v1 ) 2
 2 2 
2
2 1 
Pe 0  
  2
e 2 2
dvn

It is clear from the symmetry of this problem that Pe0 is identical to Pe1 and the
probability of error Pe, irrespective of whether a ‘one’ or ‘zero’ was transmitted, can
be rewritten in terms of v = v1 – v0

1  v 
Pe  1  erf  
2  2 2 

for unipolar signalling (0 and v)

v
for polar signalling (symbol represented by voltage 
2

24
 2010-11-14

Detector-Decision (Cont’d)

1 S IN  
ASK e  1  erf  

2 4 N IN  
OOK 



1 S IN  
FSK  e  1  erf  
 For Optimum ASK , FSK , PSK
2 2 N IN  


1 S IN   SNR in watt
PSK e  1  erf 
  ASK FSK PSK
2 N IN  SNR in dB 10 SNR in dB/10 
 Pe Pe Pe
 Linear gain
PRK 
0 1.00 0.2398 0.1587 0.0786
2 1.5849 0.1867 0.104 0.0375
4 2.5119 0.1312 0.0565 0.0125
6 3.9811 0.0791 0.023 0.0024
8 6.3096 0.0379 0.006 0.0002
10 10.00 0.0127 0.0008 0
12 15.8489 0.0024 0 0

Detector-Decision (Cont’d)

Probability of Symbol Error

1.00E+00

1.00E-01
Probability of Symbol Error

ASK
1.00E-02 FSK
PSK

1.00E-03

1.00E-04
0 2 4 6 8 10 12 14
SNR in dB

25
 2010-11-14

FM/ FSK Demodulation

One form of FM/FSK demodulator is shown below

In general VIN (t) will be

V IN ( t )  V c Cos  IN t
Where  IN is the input frequency (rad/sec)  IN  2  f IN 
V x  V IN t V IN t   
V x  V c Cos  IN t . V c Cos  IN ( t   )
1
Since CosA CosB  Cos  A  B   Cos  A  B 
2
V c2
Vx  Cos  IN t      IN t   Cos  IN t      IN t 
2

FM/ FSK Demodulation (Cont’d)

V c2
Vx  Cos  IN t   IN    IN t   Cos  IN t   IN    IN t 
i.e 2
2
V x  c Cos 2  IN    IN t   Cos  IN  
V
2
Thus there are two components
V c2
2
  
Cos 2  IN t  
2
           (1)
2
Vc
and Cos  IN t            (2)
2

Component (1) is at frequency 2 fIN Hz and component (2) is effectively a ‘DC’ voltage if
 IN is constant.
The cut-off frequency for the LPF is designed so that component (1) is removed and
component (2) is passed to the output.

Vc2
VOUT  Cos IN t
2

26
 2010-11-14

FM/ FSK Demodulation (Cont’d)

The V/F characteristics and inputs are shown below

Analogue FM

 f c   Vm
ym xc
f out   V IN  f 0
V IN V DC  m (t )
V IN V DC  V m Cos  m t
i.e. f out   V DC   V m Cos  m t  f 0
1
f c   V DC , Tc 
fc

f c Vm
Modulation Index   
fm fm

FM/ FSK Demodulation (Cont’d)

The spectrum of the analogue FM signal depends on  and is given by


FM Vs (t ) Vc  J n ( ) Cosc  nm t
n1

27
 2010-11-14

Digital FSK

ym xc
f out   V IN  f 0
V IN V DC  m ( t )
V IN V DC  V 1 for 1 ' s
V IN V DC  V 0 for 0 ' s
f 1   V DC   V 1  f 0 for 1 ' s
f 0   V DC   V 0  f 0 for 0 ' s
1
f c   V DC , Tc 
fc
Normalized frequency Deviation ratio

f1  f 0
h i.e. Modulus f1  f 0
Rb
The spectrum of FSK depends on h

Digital FSK (Cont’d)

28
 2010-11-14

FM/ FSK Demodulation (Cont’d)

V c2
Consider again the output from the demodulator V OUT  Cos  IN 
2
1
The delay  is set to Tc
4
where Tc 
fc
and fc is the nominal carrier frequency
Hence  2  f IN    f IN 
Vc2 Vc2
VOUT  Cos   VOUT  Cos  
2  4 fc  2  2 fc 

FM/ FSK Demodulation (Cont’d)

The curve shows the demodulator F/V characteristics which in this case is non linear.

29
 2010-11-14

Practical realization of F/V process

The comparator is LIMITER – which is a zero crossing detector to give a ‘digital’ input to
the first gate.

This is form of ‘delay and multiply’ circuit where the delay  is set by C and R with

 = CR

Practical realization of F/V process (Cont’d)

30
 2010-11-14

Practical realization of F/V process (Cont’d)

Consider now
f IN ≠ f c

Practical realization of F/V process (Cont’d)

AE f IN
VOUT  Plotting Vout versus f IN (Assuming A=1)
4 fc

31