UCCN2043 Lecture Notes (Prepared by Lee HengYew)

7.0

Multiplexing

In a communication system, the costliest element is the transmission medium. To make the
best use of the medium, we have to ensure that the bandwidth of the channel is utilized to its
fullest capacity.
Multiplexing is the technique used to combine a number of channels and send them over the
medium to make the best use of the transmission medium. We will discuss the various
multiplexing techniques in this lesson.

7.1

Multiplexing and De-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 fiber 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 two degrees of freedom, i.e. bandwidth
(frequency) and time.

Page 1 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

Use of multiplexing technique is possible if the capacity of the channel is higher than the data
rates of the individual data sources. Consider the example of a communication system in
which there are three data sources.
As shown in Figure 7.1, the signals from these three sources can be combined together
(multiplexed) and sent through a single transmission channel. At the receiving end, the
signals are separated (de-multiplexed).

Figure 7.1: Multiplexing and de-multiplexing.

At the transmitting end, equipment known as a multiplexer (abbreviated to MUX) is

required.
At the receiving end, equipment known as a de-multiplexer (abbreviated to DEMUX) is
required.
Conceptually, multiplexing is a very simple operation that facilitates good utilization of the
channel bandwidth.
The group of multiplexing technologies may be divided into several types, all of which have
significant variations:
Frequency Division Multiplexing (FDM)
Time Division Multiplexing (TDM)
Wavelength Division Multiplexing (WDM)
Code Division Multiplexing (CDM)
Discrete Multitone (DMT)
(combination of FDM and QAM modulation)
Space Division Multiplexing (SDM)

Page 2 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

7.2

Frequency Division Multiplexing (FDM)

Frequency-division multiplexing (FDM) is derived from amplitude modulation (AM)

techniques and is inherently an analog technology. The transmission medium is divided into
several communication channels where each communication channel is assigned with a
carrier frequency fc.
FDM achieves the combining of several digital signals into one medium by sending signals in
several distinct frequency ranges (bands) centered around fc over that medium.
Each signal occupies its own specific band of frequencies all the time, i.e. the messages share
the channel bandwidth.

One of FDM's most common applications is cable television. Only one cable reaches a
customer's home but the service provider can send multiple television channels or signals
simultaneously over that cable to all subscribers. Receivers must tune to the appropriate
frequency (channel) to access the desired signal.
FDM is widely used in radio and television systems (e.g. broadcast radio and TV) and was
widely used in multichannel telephony. However, it is prone to noise problems, and has been
overtaken by Time Division Multiplexing which is better suited for digital data.

The multichannel telephone system illustrates some important aspects and is considered
below. For speech, a bandwidth of 3 kHz is satisfactory. The physical line, e.g. a co-axial
cable will have a bandwidth compared to speech as shown below.

Page 3 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

From AM, we have noted:

where
Carrier frequency = fc
DSB-SC Double Sideband Suppressed Carrier
In order to use bandwidth more effectively, Single Sideband (SSB) is used i.e.

SSB
Filter

m(t)

SSBSC

carrier
cos( c t )

freq

fc

Note the Upper Sideband (USB) has been selected.

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

Speech

Band
Limiting
Filter

m(t)

SSB
Filter

SSBSC

300Hz 3400Hz

cos( c t )

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.

Page 4 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

f
300Hz

3400Hz

f
300Hz

3400Hz

10kHz

Convention

m(t)

Speech

For telephony, the physical line is divided (notionally) into 4 kHz bands or channels, i.e. the
channel spacing is 4 kHz. 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 inorder to reduce adjacent channel crosstalk.

Consider now a single channel SSB system.

DSBSC

m(t)
BLF
fc

300Hz

3400Hz

Page 5 of 21

SSB
Filter

SSBSC

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

The spectra will be

m(t)
freq

DSBSC
freq
fc

freq
fc

Consider now a system with 3 channels

m1(t)
f

SSB
Filter

BLF
fc1

m2(t)

f1

SSB
Filter

BLF

FDM
Signal

fc2

M(t)

f2

SSB
Filter

BLF
m3(t)
fc3

f3

FDM Transmitter
or Encoder

Bandlimited

Each carrier frequency, fc1, fc2 and fc3 are separated by the channel spacing frequency, in this
case 4 kHz, i.e
fc2 = fc1 + 4 kHz,
fc3 = fc2 + 4 kHz.

Page 6 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

The spectrum of the FDM signal, m(t) will be:

4kHz

4kHz

M(t)

4kHz
Shaded areas are to
show guard bands.

f1
fc1

f3

f2
fc2

freq

fc3

Note that:

The baseband signals m1(t), m2(t), m3(t) have been multiplexed into adjacent channels,
the channel spacing is 4 kHz.

The SSB filters are set to select the USB, tuned to f1, f2 and f3 respectively.

A receiver FDM decoder is illustrated below:

SSB
Filter
f1

M(t)
FDM
Signal

LPF
fc1

SSB
Filter
f2

Band
Limited
LPF

m2(t)
Back to
baseband

fc2

SSB
Filter
f3

m1(t)

LPF

m3(t)

fc3

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

Page 7 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

For the 3 channel system (Transmitter and Receiver) shown, there is:
1 design for the BLF (used 3 times)
3 designs for the SSB filters (each used twice)
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 4 kHz = 14.4 MHz. Therefore there
are 3600 different SSB filter designs. Not only this, but the designs must range from kHz to
MHz.
Consider also the Q of the filter, where Q is defined as Q =

centre frequency
.
bandwidth

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

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

Channel Grouping
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

m12(t)

12kHz

60kHz

i.e. 12 telephone channels are multiplexed in the frequency band 12kHz 60 kHz in 4 kHz
channels basic group. A design for a basic 12 channel group is shown below:

Page 8 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

Band Limiting Filters

SSB Filter
DSBSC

4kHz

CH1
m1(t)

8.6 15.4kHz

300Hz

12.3 15.4kHz

3400kHz

f1 = 12kHz

4kHz

16.3 19.4kHz

12.6 19.4kHz

CH2
m2(t)
300Hz

3400kHz

f1 = 16kHz

Increase in 4kHz steps

FDM OUT
12 60kHz

4kHz

52.6 59.4kHz

CH12
m12(t)
300Hz

56.3 59.4kHz

3400kHz

f12 = 56kHz

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

12
Inputs

BASIC
GROUP
12 60kHz

SSB
FILTER
420kHz

12
Inputs

BASIC
GROUP
12 60kHz

SSB
FILTER
468kHz

12
Inputs

BASIC
GROUP
12 60kHz

SSB
FILTER

516kHz

12
Inputs

BASIC
GROUP
12 60kHz

SSB
FILTER
564kHz

12
Inputs

BASIC
GROUP
12 60kHz

SSB
FILTER
612kHz

5 basic groups multiplexed to form a super group, i.e. 60 channels in one super group.
Page 9 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

Note the channel spacing in the super group in the above is 48 kHz, i.e. each carrier
frequency is separated by 48 kHz. There are 12 designs (low frequency) for one basic group
and 5 designs for the super group.
612 kHz
12 - which is reasonable. Hence, a
48kHz
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

The Q for the super group SSB filters is Q =

7.3

Time Division Multiplexing (TDM)

TDM is a digital technology derived from sampling techniques. It is widely used in digital
communications. In TDM, messages occupy all the channel bandwidth but for short time
intervals of time, i.e. the messages share the channel time.
In comparison with FDM,
FDM messages occupy narrow bandwidth all the time.
TDM messages occupy wide bandwidth for short intervals of time.

TDM involves sequencing groups of a few bits or bytes from each individual input stream,
one after the other, and in such a way that they can be associated with the appropriate
receiver. If done sufficiently and quickly, the receiving devices will not detect that some of
the circuit time was used to serve another logical communication path.
In TDM, each message signal occupies the channel (e.g. a transmission line) for a short
period of time. The principle is illustrated as follow:

1
1

m1(t)
2

m2(t)
m3(t)
m4(t)
m5(t)

m1(t)
2

Tx
4

Rx
SW2

SW1
5

Transmission
Line

Page 10 of 21

m2(t)
3

4
5

m3(t)
m4(t)
m5(t)

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

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

m1(t)

V1
t

0
m2(t)

V2
0

m3(t)

V3
0

SW1
Sample
t
Position

V3
V2
V1
t
m1(t)

m2(t)

m3(t)

m1(t)

m2(t)

m3(t)

m1(t)

Channel
Time
Slots

1
t

Time slot

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 timedivision multiplexed on to the channel.

Page 11 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

The sampling process requires that the message signals are 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.
This process is illustrated below.
A schematic diagram to illustrate the principle for 3 message signals is shown below.

m1(t)

S/H

BLF

PAM
1

Multiplexing
Analogue
To
Digital
Convertor

fs1
m2(t)

S/H

BLF

PAM
2

Serial output
Binary digital
data d(t)

fs2
m3(t)

S/H

BLF

PAM
3

fs3
Band limiting
Filter 0 B Hz

Sample and Hold

Sample rate fs
fs 2B Hz

Multiplexing ADC
Converts each input
in turn to an n bit code.

Again for simplicity, each message input is assumed to be a DC level. (see next page)
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
synchronization 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.

Page 12 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

m1(t) V1
t

m2(t) V2
t

m3(t) V3
t

fs1
Sample
pulses
fs2

fs3

PAM 1
S/H

PAM 2
S/H
PAM 3
S/H
001

011

110

001

011

110

d(t)
e.g. n = 3 bits
m1(t) m2(t) m3(t) m1(t) m2(t) m3(t)

There are two types of Time-division multiplexing:

Synchronous Time-division Multiplexing (Sync TDM)
Statistical Time-division Multiplexing (Stat TDM)
Page 13 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

Synchronous Time Division Multiplexing

Transmitters take turns to transmit in round-robin order.
An extra bit is inserted at the beginning of each frame. The extra bit alternated between
zero and one.

Used by the de-multiplexor to detect a synchronization error.
The following figure illustrates the synchronous TDM system used by the telephone
system in which a framing bit precedes each round of slots.

Copyright 2009 Pearson Prentice Hall, Inc.

In synchronous TDM, every possible sender has a reserved time slot, whether it needs it or
not. This may lead to underutilization of the transmission channel.
The following figure Illustrates a synchronous TDM system leaving slots unfilled when a
source does not have a data item ready in time.

Copyright 2009 Pearson Prentice Hall, Inc.

Page 14 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

Statistical Time Division Multiplexing

Select items for transmission in round-robin order.
But if a senders data is not ready, skip that sender and move to the next one.
All slots will be filled as long as some sender has some data ready to send.
But now each slot must also contain an identifier to indicate who the receiver is.
The following figure Illustrates how statistical TDM avoids unfilled slots and takes less
time to send data.

Copyright 2009 Pearson Prentice Hall, Inc.

7.4

Wave Division Multiplexing (WDM) (http://www.rp-photonics.com/wavelength_division_multiplexing.html)

Wavelength division multiplexing is a technique where optical signals with different

wavelengths are combined, transmitted together, and separated again. It is mostly used for
optical fiber communications to transmit data in several (or even many) channels with
slightly different wavelengths.
Through WDM, the transmission capacities of fiber-optic links can be increased strongly, so
that most efficient use is made not only of the fibers themselves but also of the active
components such as fiber amplifiers.
The following figure illustrates the use of prism to combine and separate wavelengths of light
in WDM technologies

Page 15 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

Copyright 2009 Pearson Prentice Hall, Inc.

Theoretically, the full data transmission capacity of a fiber could be exploited with a single
data channel of very high data rate, corresponding to a very large channel bandwidth.
However, given the enormous available bandwidth (tens of terahertz) of the low-loss
transmission window of silica single-mode fibers, this would lead to a data rate which is far
higher than what can be handled by optoelectronic senders and receivers.
Also, various types of dispersion in the transmission fiber would have very detrimental
effects on such wide-bandwidth channels, so that the transmission distance would be strongly
restricted.
WDM solves these problems by keeping the transmission rates of each channel at reasonably
low levels (e.g. 10 Gbit/s) and achieving a high total data rate by combining several or many
channels.
Two different versions of WDM, defined by standards of the International
Telecommunication Union (ITU), are distinguished:
Coarse wavelength division multiplexing (CWDM, ITU standard G.694.2)
Dense wavelength division multiplexing (DWDM, ITU standard G.694.1)

Course Wave Division Multiplexing (CWDM)

Uses a relatively small number of channels, e.g. four or eight, and a large channel spacing
of 20 nm.
The nominal wavelengths range from 1310 nm to 1610 nm.
The wavelength tolerance for the transmitters is fairly large, e.g. 3 nm, so that unstabilized DFB lasers can be used.
The single-channel bit rate is usually between 1 and 3.125 Gbit/s.

Page 16 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

Dense Wave Division Multiplexing (DWDM)

Extended method for very large data capacities, as required e.g. in the Internet backbone.
Uses a large number of channels (e.g. 40, 80, or 160), and a correspondingly small channel
spacing of 12.5, 25, 50 or 100 GHz.
All optical channel frequencies refer to a reference frequency which has been fixed at
193.10 THz (1552.5 nm).
The transmitters have to meet tight wavelength tolerances. Typically, they are
temperature-stabilized DFB lasers.
The single-channel bit rate can be between 1 and 10 Gbit/s, and in the future also 40 Gbit/s.

7.5

Code Division Multiplexing (CDM)

CDM allows signals from a series of independent sources to be transmitted at the same time
over the same frequency band.
This is accomplished by using orthogonal Code division multiplexing codes to spread each
signal over a large, common frequency band. At the receiver, the appropriate orthogonal code
is then used again to recover the particular signal intended for a particular user.
The key principle of CDM is spread spectrum. Spread spectrum is a means of communication
with the following features:
1. Each information-bearing signal is transmitted with a bandwidth in excess of the
minimum bandwidth necessary to send the information.
2. The bandwidth is increased by using a spreading code that is independent of the
information.
3. The receiver has advance knowledge of the spreading code and uses this knowledge to
recover the information from the received, spread-out signal.
Spread spectrum seems incredibly counterintuitive. Weve spent quite some lecturing hours
studying ways to transmit information using a minimum of bandwidth. Why should we now
study ways to intentionally increase the amount of bandwidth required to transmit a signal?
If you understand CDM, you will see that spread spectrum is a good technique for providing
secure, reliable, private communication in an environment with multiple transmitters and
receivers. In fact, spread spectrum and CDM are currently being used in a number of
commercial cellular telephone systems and satellite communications.
Application of CDM (in this case CDMA) in cellular telephony

Each mobile device is assigned unique 64-bit code (called Chip spreading code)
To send a binary 1, mobile device transmits the unique code
To send a binary 0, mobile device transmits the inverse of code
Page 17 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

At the receiver
Gets summed signal
Multiplies it by receiver code
Adds up resulting values
Interprets as a binary 1 if sum is near +64
Interprets as a binary 0 if sum is near 64

CDM / CDMA Application Example

For simplicity, assume 8-chip spreading codes

3 different mobiles use the following codes:

- Mobile A: 10111001
- Mobile B: 01101110
- Mobile C: 11001101
Assume
- Mobile A sends a 1
- Mobile B sends a 0
- Mobile C sends a 1

Signal code: 1-chip = +N volt; 0-chip = -N volt

Three signals transmitted:

Mobile A sends a 1, or 10111001, or +-+++--+
Mobile B sends a 0, or 10010001, or +--+---+
Mobile C sends a 1, or 11001101, or ++--++-+

Summed signal received by base station: +3, -1, -1, +1, +1,

-1, -3, +3

Base station decode for Mobile A:

Signal received:
+3, -1, -1, +1, +1, -1, -3, +3
Mobile As code:
+1, -1, +1, +1, +1, -1, -1, +1 (10111001)
Product result:
+3, +1, -1, +1, +1, +1, +3, +3
Sum of Product results: +12
Decode rule:
For result near +8, data is binary 1
Base station decode for Mobile B:
Signal received:
+3, -1, -1, +1, +1, -1, -3, +3
Mobile Bs code:
-1, +1, +1, -1, +1, +1, +1, -1 (01101110)
Product result:
-3, -1, -1, -1, +1, -1, -3, -3
Sum of Product results: -12
Decode rule: For result near -8, data is binary 0
Base station decode for Mobile C:
Signal received:
+3, -1, -1, +1, +1, -1, -3, +3
Mobile Cs code:
+1, +1, -1, -1, +1, +1, -1, +1 (11001101)
Product result:
+3, -1, +1, -1, +1, -1, +3, +3
Sum of Product results: +8
Decode rule:
For result near +8, data is binary 1
Page 18 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

7.6

Discrete Multitone (DMT)

DMT is a form of multicarrier modulation that encodes bits in the frequency domain. It is
now used in certain wireless communication systems (802.11) and Digital Subscriber Line
(DSL) technologies.
Through the use of DMT, DSL technology enables very high speed connections from
individual computers to switching stations over a standard copper telephone line.
DMT places the data onto 247 separate sub-channels, each 4 KHz wide. This is like
having 247 different dial-up lines connected to a computer all at the same time!
On top of that DSL allows a subscriber to be able to receive phone calls over the same line at
the same time, without risk of disconnection or data loss.
The existing local loop*1 can handle up to 1.1 MHz bandwidth
The first 4 KHz bandwidth is used for regular telephone voice service
Rest of the bandwidth is divided into 256 channels each occupying a bandwidth of
4.312 KHz
Each sub-channel can carry up to 60 Kbps data rate. (4 KHz 15 bits/Signal Change) = 60
Kbps
Allocation of the 1.1 MHz bandwidth (refer to Figure 7.6 below)
Channel 0 is used for voice
Channel 1 to 5 are not used to allow a gap between voice and Data
Channels 6 to 30 (25 channels) are used for up stream transmission and control. One
for Control and 24 for data. Thus upstream data rate is: 24 4 KHz 15 bits/ Signal
change = 1.44 Mbps
Channel 31 to 255 (225 channels) are used for downstream transmission, one for
control and 224 for data. Thus the downstream data rate is : 224 4 KHz 15 bits/
Signal change = 13.4 Mbps

Voice
Ch 0
0

Upstream
Ch 6 to 30
4

26

Downstream
Ch 31 to 255
108 138

1104

KHz

Figure 7.6
Note that these are theoretical maximum bandwidth. The actual practical data rate is usually
lower and depends on:
The S/N Ratio of the link
The distance between the customer location and the Central Office*2 (CO)
Note: 1.Central Office (CO) is the physical building where the local telephone switching
equipment is located. All telephone lines in a town lead to the CO.
Note: 2. the pair of wires (twisted pair) connecting a telephone subscriber to a CO.

Page 19 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

The actual data rates are as follows:

Upstream:
64 Kbps to 1.5 Mbps
Downstream: 500 Kbps to 9 Mbps

Being an adaptive technology, the ADSL modems, when turned ON, check the quality of line
and automatically adjust the data rate.
The noise level of each sub-channel is monitored; if a sub-channel with a particular
frequency becomes too noisy, data will be reallocated from the noisy sub-channel to
others with less noise.
As the frequency response of the channel changes with time, the ADSL system
constantly shifts data from one sub-channel to another, searching for the frequency
distribution that allows for an optimal data rate.

7.7

Space Division Multiplexing

In wired communication, space-division multiplexing simply implies different point-to-point

wires for different channels.
One example is an analogue stereo audio cable, with one pair of wires for the left
channel and another for the right channel.
However, wired space-division multiplexing is typically not considered as multiplexing.
In wireless communication, space-division multiplexing is achieved by multiple antenna
elements forming a phased array antenna. Examples are:
Multiple-input and multiple-output (MIMO) multiplexing
Single-input and multiple-output (SIMO) multiplexing
Multiple-input and single-output (MISO) multiplexing
For example, a IEEE 802.11n wireless router with N antennas makes it possible to
communicate with N multiplexed channels, each with a peak bit rate of 54 Mbit/s, thus
increasing the total peak bit rate with a factor N.
Different antennas would give different multi-path propagation (echo) signatures, making it
possible for digital signal processing techniques to separate different signals from each other.
These techniques may also be utilized for space diversity (improved robustness to fading) or
beam-forming (improved selectivity) rather than multiplexing.

Page 20 of 21

UCCN2043 Lecture Notes (Prepared by Lee HengYew)

7.8

Comparison of Different Multiplexing Techniques

Page 21 of 21