Dm1 e

DM1
DM1 DIGITAL MOBILE TELEPHONY

Radio School
DM1 Digital Mobile Telephony
Mobile Telephone Generation 2 (G2)

Modulator Detector

Channel coder Channel decoder

Speech coder Speech decoder

RCU
Core Unit Radio Systems and Technology

1


Ericsson Radio Systems 2000

2


Digital Mobile Telephone DM1
Mobile Telephone Generation 2 (G2)

Contents Page Contents Page
1 Overview 4 3 D-AMPS, original system 79
3.1 Overview 79
2 GSM, original system 8 3.2 System background 81
2.1 Background to GSM 8 3.3 Radio specification 83
2.1.1 System specification, introduction 8 3.4 Speech coding 91
2.1.2 System technology development 11 3.5 Comparison of the GSM and
D-AMPS systems 101
2.1.3 System options for GSM 18
2.2 Overview of the radio system 27
4 PDC, generation 1 106
2.2.1 Introduction 27
2.2.2 TDMA structure of traffic channels 29
5 Cordless Telephone 108
2.2.3 Structure of data bursts in a TDMA
time slot 30 5.1 Overview 108
2.2.4 Multiframe with SACCH 33 5.2 DECT 112
2.2.5 Duplex arrangement 34 5.3 PHS 117
2.2.6 Diversity against fast fading 35
2.2.7 Background to choice of radio 6 Further development of NMT 119
system parameters 38
6.1 Shut-down of NMT 900 119
2.3 Detailed systems description 39
6.2 Modernization of NMT 119
2.3.1 Introduction 39
2.3.2 Signalling, TDMA structure 41
7 Further development of GSM
2.3.3 Channel coding and interleaving 52 and D-AMPS 121
2.3.4 Radio modem 59 7.1 Improved speech coding 121
2.3.5 Channel equalization 66 7.2 Digital signal channel for D-AMPS 124
2.4 Radio performance 70 7.3 Adaptation to data transmission 129
2.5 The fixed network 71
2.5.1 The speech path 71 8 Cell structures 148
2.5.2 Switching and control 74 8.1 Additional frequency bands 148
8.2 Need for hierarchical cell
structures 148
8.3 Land mobile satellite
communication 150
Appendix. Follow-up questions 156

3


1. Overview
The first generation of mobile telephone (“G1”) comprised of systems based
on analog speech transmission. One example is NMT, which is overviewed in
module G1. The second generation of mobile telephone (“G2”) or first gene-
ration of digital mobile telephone, is based on digital transmission of speech.
One of the main reasons for the introduction of digital speech was improved
frequency economy through reduced radio bandwidth per speech channel
and/or reduced cluster size. Examples of G2 systems are GSM (Europe), D-
AMPS (USA) and PDC (Japan). GSM, D-AMPS and PDC are discussed in
sections 2 to 4. The dominating G2 system is GSM with 60% of the market
and 230 milj. users at the end of 1999.
The cellular systems above have wide-area coverage, with nearly full
coverage of large regions. Another type of G2 system, with local service
areas, is cordless telephone. See section 5. The main cordless telephone sys-
tem in Europe is DECT (Digital European Cordless Telephone - new name
Digital Enhanced Cordless Telecommunications). In Japan, PHS (Personal
Handy Phone System), is used extensively in areas with high traffic density.
GSM and to some extent D-AMPS have been exported to different parts of the
world. The G1 and G2 generations were originally optimized for speech trans-
mission - data transmission was a secondary service, used to a very small
extent.
The third generation of mobile telephone (“G3”) or second generation of digi-
tal mobile telephone is optimized for a mixture of different services such as
speech, data and video, incl. multi media. This is covered in module DM2.
User data rates up to 2 Mb/s shall be accomodated. The dominating
transmisssion mode will be based on Internet (packet oriented, IP protocol).
Due to the large penetration of G2 systems there has been a strong motivation
to modernize them to partially accomodate the new services which will be
handled by the G3 systems. See section 7. The major development in this
respect is the introduction of EGPRS (EDGE-based GPRS) in networks based
on GSM and D-AMPS. For good radio connections (high C/N and C/I values)
user rates up to 384 kb/s are possible. (384 kb/s is the lower data rate
specified for the European G3 alternative UMTS).
Even before GPRS and EDGE there has been a gradual evoluation of the G2
systems, i.e. introduction of improved speech codecs and low-bandwidth data
services. One example is the SMS (Short Message Service). D-AMPS was
originally an add-on to Analog AMPS, sharing the same radio channels and
relying on A-AMPS for the signalling and control functions to set up a traffic
channel.
Later D-AMPS became self sufficient by adding a digital control channel.
See section 7.2.
A major difference between the G1 and G2 generations is that FDMA is used
for G1 and TDMA (combined with FDMA) for G2. The G3 systems are
mainly based on CDMA for multiple access to the radio medium, even if
CDMA is often complemented by TDMA and FDMA. One G2 system (IS-95)
is based on CDMA. For that reason it is included in module DM2, which
covers the G3 systems.
A summary of the main characteristics of the different generations of cellular
systems is given in figure 1.1.

4


Generation of cellular systems
G1 G2 G3- G3 G3 + (G4)
Generation of 1 2
digital cellular

System examples NMT GSM, PDC GPRS+ WCDMA HIPERLAN/2
AMPS D-AMPS, cdma One EDGE TDCDMA WLAN
TACS DECT, PHS MCCDMA WATM

Maximum user 9.6 kb/s 384 kb/s 2 Mb/s 20 Mb/s
data rate

Analog Digital Internet Multimedia High speed
Dominating Service
speech speech Data Internet data
Speech IP-telephony using IP

Multiple Access FDMA TDMA TDMA CDMA OFDM
(CDMA) (TDMA) (CDMA)

Duplex arrangement FDD FDD, TDD FDD, TDD FDD, TDD FDD,TDD

Introduced 1982 1992 2000 2002 2002

In module G1 DM1 DM1 DM2 DM2

Figure 1.1

The systems, mentioned above, are terrestrial cellular networks, which do not
cover areas with very low user density due to economic considerations. Also,
there are limited roaming possibilities between cellular systems in different
regions. Therefore mobile satellite networks with word wide or regional
coverage are established. See section 8.3. The satellite services are considera-
bly more expensive than terrestrial cellular systems due to the difficult link
budget, but anyway a considerable market is foreseen, mainly combined with
and as gap-fillers to the terrestrial systems (using dual-mode or triple mode
terminals). The first worldwide land-mobile satellite network IRIDIUM started
commercial service late 1998, and before year 2001 additional systems are
established. However, there are indications that the market will be less than
foreseen.
Due to the rapid expansion of the mobile telephone market, large
improvements are necessary with respect to frequency economy and
geographic availability. One important system feature is hierarchical cell
structures, incl. good hand-over capabilities between different hierarchical
layers. The highest layer is satellite cells, the lowest layer is indoor pico cells.
Cell structures are discussed in section 8.
The G3 systerns have maximum user data rates up to 2 Mb/s in pico cells
(indoors) and possibly also in micro cells (hot spot outdoors). However there
will be a need for even higher data rates - up to 10 times higher. Systems with
this capability are generally wireless extensions of high speed LANs (Local

5


Area Networks), see column G3+ in figure 1.1. They could also form part of
the next generation (G4) of mobile telephone, which is not yet clearly defined.
However, the general concept of G4 is a closely integrated cluster of several
systems in a hierarchical structure. The highest layer (satellite networks)
would. have world-wide coverage but small bandwidth capabilities, perhaps
50 kb/s. The lowest layer contain G3+ systerns with around 10 Mb/s maxi-
mum user data rate but these systerns could only provide coverage indoors
and of very small outdoor areas. This layer might cover around I % of a re-
gion. The middle layers contains the G3 and the G3- systerns, G3- providing
substantial coverage of a region, and G3 metropolitan areas with high traffic
density, which motivates the use of small cells. The hierarchical structure is
sketched in figure 1..

G4. Hierarchical concept

Maximum
user data rate

20 Mb/s
G3+

Broadcast
LMDS
satellites
(downlink)
2 Mb/s

384 kb/s G3
(BRAN)

G3-
150 kb/s (UMTS)
(GSM+)
(DAMPS+)
(DECT+)
50-100
kb/s Mobile
Bluetooth satellite
systems
Coverage
Mobility

Figure 1.2

Other key concepts of G4 are the use of packet/IP based transmission for all
types of services and the interaction between intelligent networks and termin-
als (software-controlled). The terminals are continuously connected to the op-
timum system within the hierarchical structure, considering the coverage situa-
tion and its current need for bandwidth. Store-and-forward capabilities make it
possible to transfer a large amount of data, when the terminal moves through
the coverage area of a G3+ system. Outside of metropolitan areas, the maxi-
mum data rate would be quite limited, which still could be useful, by
compressing the source data rates and keeping only the essential information.

6


Additional systems could be added to this structure. The combination of large
coverage and broadband could be provided in the outward direction by exten-
sion of satellite-based digital TV and radio. In the other direction G3 or G3-
systerns could be used (asymmetric service). Other related systerns could
point-to-multipoint fixed networks (MDS) and moderate-rate, short-range
networks (i.e. Bluetooth).

7


2 GSM, original system
2.1 Background to GSM
2.1.1 System specification, introduction
The initiative for a digital mobile telephone system came from the Scandina-
vian Telecommunication Administrations, which submitted in1981 a joint
proposal to CEPT for the specification of a pan-European mobile telephone
system, conceivably to be based on digital transmission. The reason for
proposing serious consideration of a digital transmission system was based on
the findings of studies conducted by a Scandinavian working group. In 1982,
CEPT appointed the GSM group (Groupe Special Mobile), whose members
consisted of representatives from a number of countries in Western Europe, to
investigate the idea.
Following system studies coordinated by the GSM group, a decision was ta-
ken in 1985 to draw up a goal specification for a digital system. The general
criteria stipulated were that the new system should provide at least the same
speech quality and spectrum efficiency as the existing analog mobile
telephone systems. Another requirement was that the estimated cost of the
fully developed system, when in mass production, should be lower than that
of the existing analog ones.
In addition, the system must be able to interface with the ISDN on the fixed
side, even if some services requiring wide bandwidth might not be available
due to the frequency shortage. Also, a number of GSM specific services were
required, see figure 2.1a.

GSM Services
GSM Services = GSM Specific Services + ISDN Services

GSM Specific services:
PAN-European roaming

Authentication (fraud control)

Ciphering (speech, data, signalling information)

User confidentiality
(Ciphered subscriber number on radio path)

Figure 2.1a

At that time, simulations and experiments of digital-speech transmission
systems based on FDMA had progressed far enough to predict with
considerable certainty that a new system based on digital transmission would
be able to offer higher performance than existing analog systems. However, it
seemed likely that further development work could result in alternative forms
of multiple access to FDMA with improved system performance. Due to
technical uncertainties it was not yet possible to recommend any other
multiple-access arrangement. The main unknown factor was if it would be

8


possible to suppress strong intersymbol interference caused by time dispersion
in wideband radio transmission. The GSM group therefore decided that an
evaluation should be made of systems based on other types of multiple access
than FDMA.
Nine R & D groups in Western Europe designed test systems, which were
evaluated in Paris in autumn 1986 by means of laboratory evaluations,
employing fading simulators and field tests. It was very much on the basis of
these comparative tests that the GSM group recommended in spring 1987 that
a joint pan-European mobile telephone system should be developed, based on
digital speech transmission and Narrowband TDMA (NTDMA). The system
would be called GSM. This was followed by a Memorandum of Understan-
ding signed by 13 countries, under which they agreed to introduce GSM by
July 1991.
Key features of the outlines specification of 1987 were TDMA with 8 time
slots in a time frame of 4.6 ms, an advanced version of a RELP speech coder
with a data rate of 13 kb/s, convolution coding for error correction, and
GMSK modulation with 200 kHz channel spacing.
A comprehensive specification, drawn up by a consolidated GSM group (the
permanent nucleus), was ready by the end of 1988. The extensive
documentation covered not only the different radio subsystems but also the
network services to be offered and interfaces to the fixed network. However, a
great deal of work still remained on the fine details of the design, and this was
made the responsibility of the European Telecommunication Standards Insti-
tute (ETSI). The work on developing the GSM as a commercial product pro-
ved to require considerably more resources than had been foreseen. In
consequence, the project overran the original time plan by about a year. The
first large scale introduction of GSM was in Germany in 1992-93, where the
capacity of the existing analog mobile telephone system had inadequate traffic
capacity and high costs for subscribers.
The growth in the number of GSM subscribers during the same period was
slower in the Nordic countries. The main explanation for this was that the
NMT network had still not reached its capacity limit, and many mobile
telephone subscribers held the opinion that the service offered by NMT, with
wide coverage in Scandinavia, was adequate and relatively low priced.
However, a sharp upturn in the number of GSM subscribers came in the
beginning of 1994. In middle 95 the Swedish frequency administration
authority decided that part of the frequency band used by NMT 900 should be
given over GSM. As more and more users prefer GSM, the NMT 900 service
will be shut down in a few years time. See section 6. GSM 900 will then have
access to a 2x25 MHz wide spectrum.

9


Frequency allocation in western Europe for the 900 MHz cellular systems

Mobile to base
890 899 915
MHz
Analog systems
GSM

Base to mobile
935 944 960
MHz
Analog systems
GSM

Figure 2.1b

Systems of the GSM type are also used at 1800 MHz in Europe and at 1900
MHz in the US (the PCS band).
When the system was introduced, its full name was changed to Global System
for Mobile Communications, which meant that the abbreviation GSM could
still be used.

10


2.1.2 System technology development
General

Development of digital mobile system
1. Military systems
Motivated by requirements on secure enciphering
Otherwise marginal transmission performance
2. FDMA
GMSK gives reasonable compromize between
implementation complexity and frequency economy
Channel coding necessary to counteract fast fading
Low-rate speech codecs
3. TDMA
Implementation advantages
MAHO
Certain flexibility in demand assignment of bandwidth
4. DS-CDMA (FH-CDMA)
(interference-limited systems)
Interference averaging of cochannel interference
Bandwidth on demand (fast dynamic allocation)
Improved frequency diversity (bandwidth expansion)
5. OFDM
for systems with very high user data rates
(simplified channel equalization)
low bandwidth expansion
(2 and 3 channel-limited systems, i.e. system capacity
limited by number of radio channels per cell)

1. Circuit switching
2. DSI, Statistical Multiplex, Asynchronous TDMA
3. Packet transmission

Figure 2.2a

The origin of digital speech transmission in mobile systems was military and
police applications, where digital speech was motivated by the need for very
secure enciphering. One example from Sweden is given the block diagram in
figure 2.2b. In other respects, these early systems had very marginal
performance in comparation with the corresponding analog systems.
Available technology made it extremely difficult to transmit digital speech
within the channel width specified for the corresponding analog systems. An
extensive development of digital radio transmission technology was therefore
necessary before digital G2 systems could compete with the analog systems
with respect to system cost, speech quality and spectrum.

11


The introduction of digital speach was originally motivated by
requirements for very secure encryption

SEMIDIGITAL MOBILE RADIO SYSTEM (also including normal analog speech)
From the 1980-1985 period

Traditional analog
Speech FM-radio

(TX) (TX) (TX) Transmitter
10 10 0-6 kHz
kb/s kb/s
Speech Enciphering Baseband
codec unit Modem

(Rx) (Rx) (Rx) Receiver
Speech

25 kHz
db
0

Adjacent
channel
Problem:
Marginal speech quality at 10 kb/s (adaptive delta modulation)
Marginal receiver sensitivity (data modem for 10 kb/s
that complies with 70 dB adjacent selectivity)
15 kHz
High complexity, using VLSI technologies of the 1980s
-70 f
Interference on
Transmitter spectrum adjacent channel

Figure 2.2b

FDMA
Development of the fundamental technology for digital cellular systems
started in Sweden at the beginning of around 1980, and some years later in
other western European countries. Initially, the Swedish studies focused on the
simplest system configuration based on FDMA (see Figure 2.3). The principal
subsystems were speech coding, channel coding and radio modem.

12


Digital mobile telephone system (FDMA).

Data
Radio transmitter
Radio
Speech Encryp- Channel modem Power
Speech encoder tion unit encoding 16 kb/s (T) Amplifier
10 kb/s

(Speech Key (Channel
codec) Key codec)
Radio receiver
Speech Speech Radio Front-
Decryp- Channel
decoder tion unit decoding modem end
(R)

Data

Figure 2.3

The main finding of the work was that digital radio transmission eventually
could provide better speech quality and higher spectrum efficiency. The
conclusion was based on the results of a combination of computer
simulations, laboratory tests using Rayleigh fading simulators and field tests.
The improvement in spectrum efficiency compared with analog mobile
telephone systems was largely due to major advances in speech coding and
channel coding. The Swedish FDMA system incorporated an early RELP-type
16-kb/s speech coder with a permissible bit error rate of 1%, and channel
coding optimized to suppress the effect of fading dips caused by the fast fa-
ding due to multi path propagation. Additional facilities to deal with fading
were soft channel decoding and interleaving (see section 2.3.3). These
measures to counter fading yielded a significant reduction of the required
protection ratio. The GMSK modem also contributed to the good spectrum
efficiency through its combination of moderate protection ratio and fairly
narrow modulation spectrum.
The difference in spectrum efficiency between analog and digital transmission
is shown in Figure 2.4, which compares three cellular system alternatives:
a) Analog speech transmission and FM modulation
b) Digital speech transmission without channel coding
c) Digital speech transmission with FEC channel coding
The alternatives b and c use 16 kb/s speech coders.
In all three cases it is assumed that a duplex band of 2 x 10 MHz is
available. In alternative b), the input data rate to the modulator is 16 kb/s,
which corresponds to a necessary channel separation of 15 kHz. In
alternative c), the data rate, going through the channel coder, increases from
16 to 27 kb/s, which requires a channel spacing of 25 kHz – in other words,
the network in this case has available 400 two-way speech channels. The
overall spectrum efficiency is also determined by the required reuse distance
between co-channel cells (cluster size), which, in turn, depends on the local
mean of the protection ratio, KI over the fast fading.

13


A typical requirement for analog cellular systems (without diversity) is
KI = 18 dB. Results from lab tests indicate that KI = 20 dB is required in
option b) and KI = 13 dB in option c). The improvement going from b) to
c) can be explained by considerable diversity gain from channel coding.
The protection ratio is the required power ratio between the wanted signal C
and the co-channel interference I needed for adequate transmission quality,
e.g.
KI = (C/I)min

Comparison of spectrum efficiency

Analog system Digital system
(companded FM) with without
channel coding

Data rate, speech encoder - 16 kb/s 16 kb/s

System data rate - 16kb/s 27kb/s

Channel spacing 25 kHz 15 kH/z 25 kH/z

Protection ratio 18 dB 20 dB 13 dB
(local mean)

Cluster size 3x7 3x9 3x3

Spectrum efficiency:

Channels per MHz per cell 1.9 2.4 4.4

Traffic per cell for 12.4 e 17.2 e 35.1 e
10MHz system

C and I subject to Rayleigh fading
120° sector antennas. Each base station serves 3 cells

Figure 2.4

Besides by KI, the required normalized reuse distance, D/d (D: reuse
distance, d: cell radius), is also determined by the distance-dependence of
the global propagation attenuation (propagation exponent), the structure of
the shadow fading (variance of the log-normal distributions and the
correlation between the fading in C and I), and the required area availability
(the proportion of the area of a cell in which the local mean of C/I exceeds
the protection ratio). The required cluster size is determined by D/d. The
cluster size is the number of cells with different channel allocations that is
required to enable co-channel cells to be adequately separated. The cluster
size of 3 x 3, shown in figure 2.5, is often used for the GSM. (The local
mean of the protection ratio for GSM is 9-10 dB). Each base station site
serves three cells.

14


Cell structure with clusters
Cluster size 3 x 3

4 7 4 7

5 8 1 5 8 1 5 8

2 6 9 2 6 9 2

4 7 3 4 7 3 4 7

D 5
1 8 1 5 8 1

9
6 2 6 9 2 6 9

3 4 7 3 4 7 3

d
1 1

D: reuse distance d: cell radius

Figure 2.5

Simulations based on typical propagation characteristics gave the relationship
shown in Figure 2.6. The same characteristics between the protection ratio
(local mean) and the geographic availability applies also to other types of
multiple access. The figure gives a rough indication of the needed protection
ratio and thus the cluster size for several MA alternatives. Previously 90 %
geographic availabily was considered marginally acceptable. In the future,
higher availability would be required, which tends to increase the necessary
cluster size.

15


Probability distribution for (C/I < KI) for different cluster sizes

Availability Probability
C/I < KI
1.0

20% 0.8
Cluster 1 3 9
size 12
0.6

50% 0.5

0.4
21 27

80% 0.2
90%
90% 0.1 availability
KI=(C/I)min
100% 0.0 local average
DS-CDMA FH-CDMA 20 30 dB
-10 10
GSM NMT
KI : protection ratio

Figure 2.6

The figure shows, for instance, that for 90% availability (C/I > KI over 90%
of the cell) and KI = 18 dB (local mean), a cluster size of 21 is required.
(Each site serves three ells.) Accordingly, the total number of radio channels
available to the system must be distributed over 21 cells. (See also module
S4).
The number of traffic channels available per cell is thus derived from the clus-
ter size, the channel spacing (per traffic channel) and the total frequency band
available. The average number of speech channels per cell that can be
serviced during busy hour is less than the number of radio channels, otherwise
too much traffic is lost due to traffic overload (blocking) during traffic peaks.
If a loss system (Erlang B) with 2% permissible blocking is assumed, the
carried traffic per cell will be that shown in Figure 2.4. (See also module S3).
As is evident from Figure 2.4, systems using digital speech transmission with
FEC channel coding might achieve a spectrum efficiency three times higher
than the cellular systems of the first generation (G1). Although channel
coding implies an increased input data rate to the modulator – in other words,
wider channel spacing than in a corresponding system without channel coding
– this is more than compensated for by the considerable reduction in the
required protection ratio. A significant improvement in the overall spectrum
efficiency is thus obtained.

16


TDMA

The use of digital transmission means that other forms of multiple access can
be used besides FDMA. The most readily available option is TDMA, possibly
combined with time duplex (TDD: time division duplex). This offers further
advantages in terms of system performance and cost savings. A summary of
the advantages of TDMA is presented infigure 2.7. At TDMA, Time Division
Multiple Access, each radio channel is time-shared between several transmis-
sion channels. See figure 2.22 (basic frame) and 2.26.

Advantages of TDMA (+ TDD)

Fewer radio units and simpler antenna filters at the base
No duplex filter needed at terminals
Mobile Assisted Handover (MAHO)
Wide radio channels reduce requirements on frequency stability and selectivity
TDM instead of FDM replaces analog high-Q filters with digital VLSI

Figure 2.7

The possibility of listening or transmitting in other frequency or time slots in
idle periods during each frame affords important system benefits. Such
periods can be used for system signalling, preparing for handover and if
antenna diversity is used at the terminals, to select and connect a suitable
antenna to the receiver input before the reception time slot occurs.
An important facility is Mobile Assisted Hand Over (MAHO), i.e. the informa-
tion needed by the system control to determine when hand-over shall take
place comes both from the terminal and the base. The mobile must measure
C/I and C/N for signals from adjoining cells and transmit this information to
the base.
TDMA also allows a terminal to transmit and receive in different time slots
(time duplex). This eliminates relatively expensive and bulky duplex filters at
the terminals.
These advantages often outweigh the disadvantages of TDMA. The
drawbacks are listed in figure 2.8.

17


Drawbacks of TDMA

Higher transmitter peak power level for a given mean power level
(determines range)

Wide modulation bandwidth can result in intersymbol interference
due to multipath propagation (need for adaptive channel
equalization)

Greater equipment complexity (requires advanced VLSI with low
power consumption)

Increased channel spacing which reduces flexibility of frequency
planning

Figure 2.8

2.1.3 System options for GSM
The choice of multiple-access arrangement for GSM was largely based on the
results of the evaluations made in Paris in late 1986 and early 1987.
The majority of the test systems were based on TDMA. The main contenders
for GSM were Narrowband TDMA (NTDMA) and Wideband TDMA
(WTDMA). Several versions of Narrowband TDMA were evaluated by the
GSM group. One French proposal (SHF-900) combined TDMA with low-rate
channel coding supported by frequency hopping. The proposal (MAX 2) from
the Swedish Telecom Administration was for 8-PSK and as little as four time
slots in the TDMA frame. This gave such a narrow modulation bandwidth that
channel equalization would only have been necessary for very difficult
propagation conditions.
However, the TDMA option specified by the GSM group corresponded most
closely to the experimental system based on NTDMA developed by Ericsson
(DMS 900). The main competitor was WTDMA, (CD900) which, in terms of
performance, for the most part was on a par with the best NTDMA
alternatives.
An interesting finding of the Paris tests was that several systems achieved
roughly the same spectrum efficiency (see figure 2.9).
The differences among the systems with regard to the radio bandwidth per
speech channel were offset by different protection-ratio requirements. As
described earlier, different protection ratios result in different cluster sizes.
Roughly speaking, if a system can cope with half the channel width per
speech channel, an increase in the cluster size by a factor of two can be
allowed without any impact on the total spectrum efficiency.
The CD 900 system (SEL, Germany), based on wideband TDMA,
incorporated very powerful, low-rate channel coding, which increased the
required bandwidth but resulted in a much lower protection ratio than
narrowband TDMA. At the other end, MAX 2, was designed for the narrowest
possible channel width per speech channel, which resulted in a fairly high
protectionratio.

18


Spectrum efficiency for different mobile telephone systems
Equivalent bandwidth
per speech channel Curves corresponding to
constant spectrum ef ficiency

CD 900 100 a) Existing analog systems

b) 2 x a
SHF 900 50

US-(A-AMPS)
DMS 90 25 x UK-TACS
GSM

MAX 2 12,5

Protection
5 10 15 20 25 ratio

Cluster size N=3 N=9 N=27 (for 90% area availability)

Figure 2.9

Several of the experimental systems achieved better spectrum efficiency than
the analog mobile telephone systems. This was one of the criteria stipulated at
the outset that the digital systems would have to meet. The evaluation also
seemed to show that digital systems can provide better speech quality also
during fast fading.

Main features of DMS-90 (NTDMA, ERA proposal)

A block diagram of the ERA’s test system is shown in Figure 2.10, and the
system’s multiple-access structure in figure 2.11. The channel coding was
supported by interleaving and frequency hopping. This gave a considerable
diversity gain – i.e. low protection ratio – even for portable terminals
(quasistationary propagation channel). The interleaver splitted up a 384-bit
block from the channel coder into four sub blocks of 96 bits, which were
distributed among four time slots. (Each time slot could accommodate 2 x 96
user bits, i.e. contained blocks from two of the 384 bit blocks). The time dis-
persion of the propagation channel was handled by the adaptive equalizer,
which also could give multi-path diversity if the propagation channel had
fairly large time dispersion. The impulse response of the radio channel was
determined with the help of a training sequence at each time slot of the TDMA
frame. The training sequence is used also for frame synchronization, see
figure 2.22.

19


Narrowband TDMA DMS-90

Speech Channel coder Radio Pulsed trans-
encoder Interleaving modulator mitter stage

Frequency-
hopping
synthesizer

Quality

Speech Deinterleaving Adaptive A/D Demo- HF
decoder Channel decoder equalizer dulator IF
Binary
signal

Correlator T raining
sequence

Figure 2.10

To improve the performance of the channel decoder, the quality (estimated
ber) of each bit to the decoder is an additional input to the decoder. This
procedure is called soft decoding.

DMS -90. Channel coding, interleaving and frequency hopping

32 ms

From
16 kbps 256 256 speech coder

24 kbps 384 384 From (12.8)
RS channel coder

From TDMA
192

192

192

192

192

192

340 kbps interleaver
(interleaving depth 4)
From
192

192

f1 f1 frequency hopper
192

192

f2 f2
192

f3
192

f4

Figure 2.11

20


Since the initial development phase and the Paris evaluations, further
development of the GSM radio transmission system has resulted in
substantially better performance than DMS 90. The final GSM specification
and implementation gave a protection ratio of KI < 10 dB even for portable
terminals, when frequency hopping is used. The improvement is mainly
achieved through further refinement of the channel coding. Further rapid
advances in speech coding have also made it possible to reduce the data rate
from the speech coder with more or less the same speech quality. In 1994 a
half-rate speech coder was standardized, further improving the spectrum
efficiency by a factor of two. (Some people have expressed the opinion that,
instead of reducing the data rate, the advances made in speech coding could
better be used to improve the speech quality.)
The relationship between C/I or C/N on the one hand and subjective speech
quality on the other differs between analog and digital transmission (see figure
2.12). (TACS is the British analog mobile telephone system.)

Speech quality in GSM and TACS

Speech quality

TACS

Acceptable
speech GSM
quality

(K I )GSM (K I )TACS
C
10 20 30 40 dB I

KI: protection ratio

Figure 2.12

With digital transmission using FEC channel coding, the speech quality is al-
most constant down to a threshold that corresponds to the error-correction li-
mit of the channel decoder. If the input signal to the receiver falls below this
level, the error-correction fails and speech quality rapidly degrades. If the
quality of the input signal to the receiver is high, analog mobile telephone
systems are superior, since the speech coder causes some quality degradation
even if the C/I and C/N are high enough for no transmission errors to occur.

Main features of CD-900 (WTDMA, German-French proposal)

One of the demonstration systems evaluated in Paris was the CD-900
WTDMA system, which was developed by a consortium led by the German
organization SEL. The system concept was based on an earlier military

21


project – AUTOTEL. The technical performance and spectrum efficiency of
the CD-900 system were on a par with the best NTDMA systems. Because the
published information is limited, to gain a general idea of the system we need
to examine the combined available information on Autotel and CD-900.
The main system characteristic of wideband TDMA was a very wide modula-
tion bandwidth through a combination of many time slots per TDMA frame
and a substantial bandwidth expansion through low-rate channel coding. The
channel coding is based on near-orthogonal codes, i.e. optimum soft decoding
can be based on matched filters implemented by a correlation procedure. The
principle is shown in figure 2.13, which applies to AUTOTEL. A group of
four information bits is coded into 16 chips.
In the CD-900 system, five information bits are coded into 32 chips, and an
additional sixth bit is transmitted via the polarity of the chip sequence. A
group of 32 chips can be considered to form a symbol in an alphabet of size
26 = 64. If the QAM arrangement is included, the size of the symbol alphabet
is 128.

Correlator-based matched receiver (CD-900)

4 bits 16 chips Correlator

(Block code: 16, 4)

∑

Filter matched to binary sequence

Figure 2.13

The powerful channel coding produces a high coding gain, i.e. a substantial
reduction in the required C/I with respect to co-channel interference. A further
reduction in the required protection ratio in a rapid fading situation is obtained
from a considerable gain from frequency diversity. The reason is the wide mo-
dulation bandwidth which, for most propagation conditions, is much greater
than the correlation bandwidth of the propagation channel. The combined
coding and diversity gains enable the local mean of the protection ratio to be
brought down as low as (C/I)min ≈ 4dB. This means that a cluster size of three
is adequate.

22


Wideband TDMA (CD-900). Modem diagram.

I channel
6-bit
D/A
12-bit
block ≈900 MHz
Baseband Transmitter
+ amplifier
modulator π/2
D/A Multipath
6-bit propagation
Q channel
Linear
amplifier
I channel N
6-bit C
A/D
12-bit
block Baseband ≈900 MHz Icoch
demodulator HF & IF
π/2
A/D
6-bit
Q channel

Figure 2.14

To achieve reasonable spectrum efficiency despite the large bandwidth expan-
sion due to channel coding, the WTDMA system uses linear modulation
(QAM), i.e. a linear transmitter amplifier has to be used (see figure 2.14). Each
coded radio symbol carries 12 information bits, 6 bits on each of the I and Q
channels. To obtain a sufficient modulation bandwidth both for accurate
measurement of the impulse response (for setting of the channel equalizer)
and for a high frequency diversity gain, a TDMA arrangement with as many
as 63 time slots was used (60 traffic channels and three channels for signal-
ling). The time slots were used jointly by the three sector cells belonging to
each base-station site. Thus, each cell was allocated an average of 20 traffic
channels. (Cluster size of 3 means that all base-station sites can use the same
radio channels. This simplifies the cell-frequency planning).
Advanced digital signal processing is used for the channel equalization (see
figures 2.15 and 2.16). Each data burst, which comprises a number of blocks
(radio symbols) each carrying 12 information bits, starts with a synchroniza-
tion/training sequence. The wide modulation bandwidth allows accurate
measurement of the impulse response of the propagation channel. This infor-
mation is used by the advanced channel equalizer, which in an optimum way
adds together the signal power from the different propagation paths. The
channel equalization is placed before the symbol detector. It is a filter matched
to the impulse function h(t) of the radio channel. The filter convolves the
received burst (excluding the training sequence) with h(T-t). This gives opti-
mum, coherent addition of the signals from propagation paths with different
delays, thereby eliminating intersymbol interference and at the same time
achieving frequency (or multi-path) diversity.
After this initial signal processing, detection takes place by determining from
which of the 32 matched filters the highest absolute value is obtained at the
sampling instant. (The symbol alphabet comprises 32 near-orthogonal
symbols.) In addition, the polarity of the output signal from the selected filter
is measured.

23


Wideband TDMA (CD-900). Demodulator with channel equalization
Training sequence
(sync word)
Sync
correlator
Channel
impulse response h(t)
Sequence
R Memory inverter

h(T-t)

Matched Symbol Decision
Data word filter correlator circuit 6 bits
Suppresses (1 of 64) per word
time dispersion
Training sequence
(sync word) (Data words comprising 6 bits/32 chips

R: receiver front-end

Figure 2.15

WIDEBAND TDMA (CD-900). Modem based on digital signal processing

6 bits Code Generates
generator bipolar D/A
waveform
32 chip
sequence
± polarity
Sync.
word
Sync.
Corre- signal
Decision circuit lator
• Select input with h (t)
6 bits highest absolute
value. Polarity? Inversion C N
A/D
• Code conversion t =>-t
from 64 chips to
6 bits h (T-τ)

s(τ) h(T-τ)dτ Icoch
Convolution s (t) Time alignment of signals
Filter matched to
with different propagation
32 different code delays. Eliminates time
sequences Filter matched to dispersion and gives large
radio channel diversity gain against
impulse response Rayleigh fading

Figure 2.16

24


Examples of the output signal from the sync correlator (i.e. the measured
impulse response from the radio channel) have been published (see
figure 2.17).

AUTOTEL
Typical output signal from sync correlator
A Same as A

Different propagation paths

2µs

Figure 2.17

As mentioned above, the wide modulation bandwidth (B) in combination with
the advanced signal processing to handle the time dispersion due to multi-path
resulted in very efficient frequency diversity that suppressed most of the fast
fading. See figure 2.18.

25


Received signal power

f=900 MHz
(db) Sample separation: 30 cm
B=25 kHz
20 No suppression of Rayleigh fading

10

0
narrowband
-10

-20
0 100 200 300 400 Sample number
(db)
B=6 MHz
20 Frequency diversity nearly eliminates fast fading

10

0
wideband
-10

-20
0 100 200 300 400 Sample number

B: width of radio channel

Figure 2.18

This WTDMA system is of considerable general interest as it has several char-
acteristics of the G3 systems based on DSCDMA. Considerable bandwidth
expansion is used, which with CDMA terminology gives processing gain,
reducing the requirement on protection ratio. The protection ratio is further
lowered through the frequency diversity. The result is that a cluster size of
three is sufficient. (DS-CDMA develops these concepts further so that the
cluster size can be reduced to one.)
The overall performance of the testbeds based on NTDMA and WTDMA were
very similar. So why was NTDMA chosen instead of WTDMA?
A comparison between NTDMA and WTDMA based on the finding of the Pa-
ris evaluations indicates comparable spectrum efficiency and speech quality.
The choice therefore had to be made on the basis of other system and
implementation characteristics.
An advantage of WTDMA in areas with high traffic density is the large
number of speech channels per carrier. This reduces the cost of the base sta-
tion equipment for cells which must handle a large number of traffic channels.

26


Also all base-station sites can use the same radio channels. This facilitates
frequency planning of the individual networks.
However, the wide channel spacing imposes considerable limitations on the
gradual transfer of frequencies from the analog to the digital mobile telephone
system. Moreover, the cost per speech channel is high for small base station
sites that are located outside high-density urban areas and only have a traffic
volume for a small number of radio trunks.
Some additional drawbacks with WTDMA are also shown in figure 2.19. In its
assessment, the GSM group attached great importance to the apparently
greater technical risk inherent in WTDMA. Not only were highly linear trans-
mitter amplifiers an unproven technology in mobile radio applications but the
wide modulation bandwidth also imposed a heavy demand on high-speed di-
gital signal processing for channel equalization and detection.

Why was WTDMA rejected?

• Inflexible frequency planning
• Too complex base stations in rural locations

• High peak output power at portable terminals

• More complex, high-speed digital signal processing
(technical risk, increased power drain)

• Linear modulation (technical risk with linear transmitter amplifiers)

Figure 2.19

2.2 Overview of the radio subsystem
2.2.1 Introduction
Section 2.2 gives a first overview of the radio transmission after that a
connection has been established. Some of the features are covered in more
detail later, and also the signalling procedure to set up a call. A simplified
transmission block diagram is given in figure 2.20. The Viterbi procedure in
connection with convolutional channel coding is described in module DT12.

27


GMS Simplified Block Diagram (excl. ciphering)

Speech Channel Radio
Coding Coding

Speech Error Burst- Modulator
coder protection formatting Transmitter

data

Speech Viterbi Viterbi
decoder decoder equalizer Receiver

quality
inter-
leaving

Figure 2.20

The most important transmission specification items (“air interface”) are shown
in figure 2.21. The frequency band comprises 2 x 25 MHz in a duplex arr-
angement with 124 duplex channels with 200 kHz channel spacing. This
channel spacing allows a system data rate of about 270 kb/s with GMSK mo-
dulation and modest adjacent channel selectivity requirement (9 dB). 270 kb/s
corresponds to a symbol length of 3.7 ms. An overview of GMSK is given in
section 2.3.4. Speech coding is discussed in section 3.4.

Radio transmission specification for GSM
Frequency band: 890 - 915 MHz (uplink)
(frequency duplex) 935 - 960 MHz (downlink)
Channel spacing: 200 kHz
Modulation: GMSK
System data rate: 271 kb/s
TDMA Frame: 4.6 ms
Time slots: 8 x 0.58 ms
Data rate (full-rate traffic channel): 22 kb/s
Speech coder: Regular Pulse Exited LPC-LTP 13 kb/s
Diversity: Channel coding
Interleaving
Frequency hopping
Channel equalization

Figure 2.21

28


In the original design, full-rate speech coders having a data rate of 13 kb/s are
used. However, at a later stage it will be possible to use also half-rate speech
coders, by accomodating both full-rate and half-rate traffic channels. As well as
speech, data transmission at different speeds can also take place over full-rate
or half-rate traffic channels (see section 2.2.3).
During a call, the terminals and base need to exchange large amounts of infor-
mation, especially to prepare for the next hand-over to another cell. Sometimes
it may also be necessary to switch to a different radio channel in the same cell
for the connection, e.g. if the first channel is experiencing strong interference.
The Associated Control Channels are used for this signalling, either the Slow
Associated Control Channel (SACCH) or the Fast Associated Control Channel
(FACCH).
The FACCH is used during the actual channel-switching phase, during which a
large amount of information needs to be transferred. The SACCH is used, for
instance, for transferring from terminal to base measurements of the received
signal levels from nearby cells, needed for MAHO. In the outward direction, in-
formation is sent on the current transmitter power to be used by the terminal.
The SACCH has several additional functions as mentioned below. To indicate
that a 57 bit sequence is used for signalling (FACCH) an associated one bit
“stealing flag” is set.

2.2.2 TDMA structure for traffic channels
Each radio channel (carrier) is divided by a TDMA arrangement into 8
channels used for user traffic and system signalling. The Slow Associated
Control Channel (SACCH) is implemented by multiplexing each physical
channel (a certain time slot in a sequence of TDMA frames) between two
virtual channels within a multiframe of 26 basic TDMA frames. Most of the
time slots are used for traffic but, in some of the basic TDMA frames, the eight
time slots are used instead for the SACCH. Each SACCH frame are associated
with 8 traffic channels (one time slot per traffic channel). A multiframe has a
time of 120 ms (see figure 2.22).

29


Multiframe
Full rate 26 frames of which
120 ms 24 used for traffic
TF
0
TF
1
TF
2
TF
10
TF
11
SF TF
12
TF
13
... TF
22
TF
23
Idle

Half rate 26 frames of which
120 ms 2 x 12 used for traffic

TF
A0
TF
B0
TF
A1
TF
A5
TF
B5
SF TF
A6
TF
B6
... TF TF SF
A11 B11

TF = Traffic frame
Basic Frame SF = Signalling frame
1 frame = 120/26 = 4.615 ms GS = Guard slot
On = Transmitter switch on
0 1 2 3 4 5 6 7 Off = Transmitter switch off
T= Tail bits
Time slot Flag Flag GS
GS
T DATA Sync DATA T
On 3 57 bits 1 26 bits 1 57 bits 3 Off
0.031
0.546 ms ms
1 time slot = 4.615/8 = 0.577 ms

Gross data rate for full-rate physical channel: 114/0.004615 = 24.7 kb/s
Traffic channel:
24 x 114 = 2.736 bits are sent during 120 ms.
Therefore, data rate = 2.736/0.12 = 22.8 kb/s
SACCH: 114
114 bits are sent during 120 ms. Therefore, data rate = ≈ 1kb / s
0 .12

Figure 2.22

If the traffic consists of full-rate channels (8 traffic channels per carrier), 24
frames of the multiframe are used for user traffic and one frame for
signalling. One frame is not used. (This frame is used by the terminals to read
the base identity (“BSIC”) of carriers from other cells.)
If instead half-rate channels are used (16 traffic channels per carrier), each
time slot in a TDNA frame is time multiplexed between two traffic channels,
which occupy the time slot alternately each during 12 frames of a multi-frame.
The remaining two basic TDMA frames are used for two SACCH signalling
channels. Each SACCH is associated with 8 traffic channels.

2.2.3 Structure of data bursts in a TDMA time slot
A time slot of length 0.577 ms is used as follows (see figure 2.22).
Guard slot, timing advance
To prevent data bursts from different terminals overlapping in the input to the
base-station receivers, a guard slot with a time of 31µs has been introduced.

30


This is needed, above all, to cope with variations in the two-way propagation
time to terminals at different distances from the base. The guard slot
corresponds to a two-way propagation path of about 4.5 km.
This is considerably less than the maximum specified range (cell radius 35
km). Therefore, to prevent data bursts from different terminals overlapping in
the input to the base-station receivers, the base-station instructs the terminals
to insert a suitable delay between received and transmitted data bursts. See
figure 2.23.

Time alignment

MS1 MS2
BS

Base station transmits MS1 MS2

3 timesl.
MS1 receives and transmits RX TX

MS2 receives and transmits τ RX
2.5 timesl.
TX
τ

Base station receives
τ : propagation delay MS1 MS2

Figure 2.23

The delay is adjusted such that a transmitted burst from the terminal reaches
the base-station receiver at the right instant relative to the time-slot structure.
The closer a terminal is to the base station, the greater will be the delay
inserted. Thus, regardless of how far the terminal is away from the base, the
bursts arriving to the base receiver will always arrive roughly in the middle of
the intended time slot. The measurements made to determine the timing
advance can also be used to calculate the distance between the terminals and
the base. (During the first contact from a terminal to the base the terminal has
not yet been instructed about the suitable timing advance. Therefore signalling
bursts with much larger guard times must be used, see figure 2.43).
The timing advance values might be used as one of the parameters
determining when hand over between cells shall take place. It could also be
used for add-on position systems, which will be an important added value ser-
vice.
On/Off switching of transmitters. Tail bits.The transmitter pulse
corresponding to a data burst must have rounded start and end. If not,
additional widening of the spectrum relative to the basic modulation spectrum
of GMSK will occur. A small part of the guard slot is used for this rounding of
the transmitter pulse.
To facilitate channel equalization, each burst starts and ends with three bits
(0,0,0). The channel equalizer has to cope with time dispersion up to four
31


symbol intervals. The three bits in the beginning and end of the burst ensure
that channel equalization of the information bits can start and end in a known
state. The three bits at the end had the same function as the tail bits used at
convolution coding of finite length sequencies. See module DT12.

Synchronization and training sequence
The impulse response of the radio channel can change drastically during a
frame of 4.6 ms. This means that for each time slot, the receiver must carry
out bit synchronization and set the channel equalizer. The impulse response
can sometimes change even during a burst (if the terminal velocity is very
high and especially when moving up to 1800 MHz). Therefore, if the setting
of the channel equalizer was optimized with respect to the impulse response at
the beginning of the burst, the equalization may be sub optimum for the last
part of the burst. This results in increased ber. The degradation becomes larger
for large width of the doppler spectrum (depends on the terminal speed and
the radio frequency). See figure 2.24.

ber
GSM900
10-1
250 km/h
5

4

10-2

3

2

1 0 km/h
10-3
0 20 40 60 80 100
Bit position
(distance to training sequence)

Figure 2.24

To avoid the complication of having to adapt the channel equalizer to
variations in the impulse response of the propagation channel during a time
slot, short slots are used and, in addition, the training sequence is placed in the
middle of the burst. The setting of the channel equalizer is based on a known
bit sequence of 26 bits, which is also used for the bit synchronization. This
sequence can also be used to equalize the first part of the data burst, since the
received data burst is stored in a buffer before channel equalization and
detection are initiated.

32


The 16 bits in the middle of the sequence have good cyclic correlation char-
acteristics. To keep the good correlation characteristics up to the maximum
specified time dispersion, the first 5 bits of the basic 16 bit sequence are
repeated at the end, and the last 5 bits also placed at the beginning. See figure
2.25. The equalizer is further discussed in section 2.3.5.

GSM. Training (synchronization) sequence
Length of synchronization sequence is 26 bits
(several different sequencies used in system)

The good correlation characteristics apply to the mid 16 bits

This 16 bit word has perfect cyclic correlation characteristics

The 5 end bits on each side are taken from the other side of the 16 bit word.
The perfect correlation is preserved over ± 5 bits time shift
(which corresponds to the maximum width of the impulse response)

5 16 5

Figure 2.25

To reduce the risk for synchronizing to a distant strong cochannel carrier, 8
different training sequencies (“color codes”) are used see figure 7.18b.

Transmission of user information. Fast Associated Control Channel
Each data burst comprises two user sequences of 57 bits each. A flag bit is
associated with each 57-bit sequence and this denotes whether the sequence
contains normal speech information or if the sequence is instead being used
for system signalling (FACCH). A short break in the speech transmission will
hardly be noticeable, since the speech coder fills out the slot with information
taken from the previous speech frame.

2.2.4 Multiframe with SACCH
During a multiframe of 120 ms, 24 or 12 bursts will be allocated to a traffic
channel (corresponding, respectively, to full-rate and half-rate traffic
channels). In the former case, the average interval between bursts will be
5 ms, i.e. 200 bursts a second. The gross data rate (including channel coding)
for a full-rate channel will be 22.8 kb/s (200 x 114), and for a half-rate
channel 11.4 kb/s. In addition to the above, a Slow Associate Control
Channel (SACCH) is also included.
A full-rate traffic channel shall carry the signal from a 13 kb/s speech coder.
Channel coding increases the data rate from 13 to 22.8 kb/s. For each speech
frame of 20 ms, 260 bits will be output from the speech coder and 456 from
the channel coder. A 456 block is divided into 8 blocks of 57 bits, which are
interleaved over time slots within 8 consecutive traffic TDMA frames
(signalling frames and empty frames are skipped). Each burst carries informa-

33


tion from two adjacent 20 ms sequences from the channel coder. (See figure
2.45b.)

2.2.5 Duplex arrangement, MAHO, Frequency Hopping
The two time slots corresponding to a two-way traffic channel are mutually
displaced in time (see figure 2.26). The figure corresponds to the case in
which a base station not using frequency hopping has been allocated four
carriers, each of which carries eight physical channels in a TDMA frame. The
mutual displacement of the time slots for the outward and inward directions
corresponds to a quasi time-duplex arrangement (even if FDD is used by the
system, some TDD advantages is obtained). No duplex filter is required in the
terminals, therefore. Instead there is a fast T/R switch, which alternately
connects the transmitter and receiver to the antenna.
A terminal receiver also has time during a TDMA frame to measure the carrier
level of a signal from one of the six nearby cells (see figure 2.27). This is part
of the information needed for MAHO, see figure 2.28. This procedure is
further discussed in the next section on system signalling. To provide enough
time for receive, transmit and listening during each TDMA frame, the
terminal’s frequency synthesizer must be able to change frequency fast
enough. This has influenced the specification of the number of slots per
TDMA frame.
An option is also frequency hopping, in which the utilized duplex channel is
changed for each TDMA frame. To avoid collisions a coordinated hopping
pattern must be used within each cell. The base informs the terminals about
the hopping pattern on the Broadcast Control Channel and on the SACCH.
The frequency hopping arrangements are discussed in more detail in module
DM2 in connection with FH-CDMA.

Physical channels of a base station with four radio channels (carriers).
Total of 4 x 8 = 32 physical channels accessible
0 1 2 3
Inward 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 f1A
direction 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 f2 A
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
f3 A
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
f4 A

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 f1B
Outward direction 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 f2 B
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 f3 B
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 f4B

0 1 2 3

Displacement f1A + f1B form a duplex channel with 8 time slots
≈ 1/3 TDMA frame

Figure 2.26

34


TDMA, time duplex, frequency hopping and listening-in by terminal
Rx
c0 0 1 2 3 4 5 6 7
Inward
c1 Rx
direction 0 1 2 3 4 5 6 7
c2
(serving cell)

Tx
c0 I 0 1 2 3 4 5 6 7
Outward I
Tx
c1 d, e
direction I
0 1 2 3 4 5 6 7
c2
(serving cell)

Listening
d0
Outward Listening
direction e0
(adjacent cells)
c = Radio channels allocated to serving cell
c 0 + c 0 comprise a duplex pair d, e
I
= Radio channels allocated to adjacent cells
Tx = Transmitting interval
Rx = Receiving interval

Figure 2.27

Mobile Assisted Hand Over (MAHO)
The terminal measures:
Signal strength and BER on used channel

Signal strength from neighbour cells

Averaging over many field strenght measurements
to get local average over fast fading

Check that signal comes from neighbour cell
This information is transferred to base station

Figure 2.28

2.2.6 Diversity against fast fading
Instead of antenna diversity at the terminals, a combination of channel coding,
interleaving and coordinated frequency hopping is used to obtain large
diversity gain in respect to the multi-path fading. In addition, the modulation
bandwidth is so large that additional frequency diversity (or multi-path
diversity) is obtained in connection with the channel equalization. A necessary
condition is that the propagation channel has fairly large time dispersion.
Together, these give such high diversity and coding gains that the required
protection ratio (the local mean over the fast fading) will typically be 9 - 10 dB.
This is compatible with a cluster size of 3 x 3.
Interleaving a full-rate traffic channel, means that the 456 bits in a 20-ms
speech frame are split up into 57 bits sequencies, which are spread out over 8

35


TDMA frames, that is over 40 ms (see section 2.3.3). If the duration of a fa-
ding dip is not more than a few milliseconds, typically only one time slot (one
TDMA frame) is affected. The deinterleaver will then change the error burst to
a relatively random error sequence spread over 8 code words. Thus, 1/8 of the
bits in each code word will be subject to a ber of about 50%. (It is assumed
that the interval between fading dips generally is more than 8 frames.) The
FEC is so powerful that nearly perfect error correction is possible. Greater
interleaving depth cannot be used, as it would give rise to excessive transmis-
sion delay. Because the interleaving depth is only four for a half-rate channel,
the coding gain will be somewhat lower.
Fading dips that are longer than the channel coding with interleaving can cope
with, occur over quasi-stationary propagation paths – something that affects
portable terminals in particular. In this case, a fading dip could affect several
consecutive TDMA frames, which drastically reduces the effect of
interleaving. The situation can be much improved through frequency hopping,
whereby each physical channel is switched between different radio channels
which can be chosen, for instance, from a 4-group. For each TDMA frame,
the carrier frequency is changed. The size of a typical frequency hop is
usually large enough to give nearly uncorrelated fast fading in the different
frequency slots.
Another advantage of frequency hopping is that averaging occurs in respect
of co-channel interference from different cells.
Frequency hopping is specified only as an option. Without frequency
hopping, several dB higher protection ratio would be needed for portable
terminals.
A complication is that frequency hopping can not be used for the main
signalling radio channel - the Broadcast Carrier, see section 2.3.2. This must
be on a fixed frequency, known by the terminals. In some system
implementations, a larger cluster size is therefore used for the Broadcast
Carrier than for the frequency hopping traffic channels.

2.2.7 Background to the choice of primary radio parameters
The symbol rate over the radio channel and the primary TDMA structure are a
compromise between acceptable transmission performance over the worst
specified propagation channel and implementation/cost limitations. In gene-
ral, the factors summarized in figure 2.29a should be considered when
determining a suitable TDMA format. The specific air-interface requirements
that must be complied with at GSM are summarized in figure 2.29b (The
equalizing window is the width of the delay interval that the equalizer can
cope width. It is centered over the major part of the impulse response of the
propagation channel, see figure 2.31). This leads to the conclusions in figure
2.30a and b.

36


The TDMA structure is determined by:
Time for measurements on nearby cells
(Mobile Assisted Hand Over - MAHO)

Time duplex advantages

TIme for frequency switching
(Frequency synthesizer 1 ms)

Low TDMA overhead
(long bursts, narrow guard slots)

Low transmission delay due to TDMA formating
(small frame length - high system data rate)

Large range for given transmitter peak power
(portable terminal, moderate slots per frame)

Moderate equalizer complexity
(Equalizing window covering a small number of symbols)
(Short burst length - smaller than correlation time of channel)

Figure 2.29a

Overall system requirements for GSM
Micro diversity against fast fading

Equalizer window at least 16 µs

Fixed equalizer setting during data burst
- Reasonable equalizer complexity limits equalizing range
to 4 radio symbols, i.e. minimum symbol length 4 µs (250 kbaud).
Impulse response stationary over 0.25 ms (for high speed trains)
With training sequence in middle of burst, maximum burst length 0.5 ms.

Length of TDMA frame shall permit different frequencies for transmit,
receive and listen (MAHO).
- Switching time of frequency synthesizer assumed to 1 ms.
Minimum length of TDMA frame 4 ms.

Maximum transmission delay due to TDMA formatting and
interleaving 40 ms
- Total transmission delay should be less than 85 ms.
Interleaving depth must be less than 10.

Figure 2.29b

37


The TDMA structure is determined by:
Time for measurements on nearby cells
(Mobile Assisted Hand Over - MAHO)

Time duplex advantages

TIme for frequency switching
(Frequency synthesizer ≈ 1 ms)

Low TDMA overhead
(long bursts, narrow guard slots)

Low transmission delay due to TDMA formating
(small frame length - high system data rate)

Large range for given transmitter peak power
(portable terminal, moderate slots per frame)

Moderate equalizer complexity
(Equalizing window covering a small number of symbols)
(Short busrt length - smaller than correlation time of channel)

Figure 2.30a

Required TDMA frame lenght at GSM

0.5 ms 0.5 ms 0.5 ms 0.5 ms

Transmit Transmit

Receive

Measure
1 ms 1 ms 1 ms

4.5 ms
Switching time for frequency synthesizer: 1 ms
Burst length (TDMA slot): 0.5 ms

Figure 2.30b

The maximum system data rate (symbol rate) that could be used was
determined by the maximum acceptable equalizer complexity. The complexity
becomes less if no adaptation is necessary to the variations of the impulse re-
sponse during each TDMA burst. With the training sequence placed in the
middle of the burst, and considering the maximum terminal speed and the ra-
dio frequency, it was estimated that the maximum burst length, over which the
channel was nearly stationary, would be about 0.5 ms. (Long bursts are
desirable as the relative TDMA overhead would become less). Even with a
fixed equalizer setting during a burst, the maximum reasonable complexity
was at that time considered to correspond to an equalizer window (part of the
impulse response that must be handled, see figure 2.31) of 4 radio symbols.
As the nearly worst case time dispersion was considered to correspond to an

38


equalizer window of 16 µs, the minimum symbol length would be 4 µs, i.e. a
maximum symbol data rate of about 250 kbaud.

Received
Signal Power

Equalizer window
( 3 x delay spread)

Protection
ratio

Impulse response
Propagation
delay
Equalizer gives no suppression.
τ
Interference roughly equivalent
to cochannel interference

Figure 2.31

Another important parameter is the number of slots per TDMA frame. At the
time when GSM was specified, the frequency synthesizer of the terminal was a
major limitation with respect to switching time. Considering the need for
MAHO, each frame must contain slots for transmit, receive and listening. In
between must be room for the synthesizer to settle to a new frequency.
Therefore each frame should contain at least 8 slots. (Synthesizer chips are
now so cheap that terminals can be supplied with two synthesizers, so that
when one is active, the other can switched and settle to the next radio channel
to be used.) A disadvantage with a very large number of slots is the additional
transmission delay connected with the TDMA framing and interleaving.
Therefore, it was decided to use a TDMA frame with 8 slots.

2.3 Detailed systems description
2.3.1 Introduction
The previous section gave an overview of the use of the radio channels during
a call. The setting up of a call requires extensive system signalling for
synchronization of a terminal to the base station, for registration of the termi-
nal and for the allocation of a traffic channel. In preparation for hand over, the
terminals also measure the signal levels from adjacent-cell base stations and
transmit this information to the base (to be used for MAHO).
An alternative multiframe with a length of 51 TDMA frames is used for most
of the system signalling, instead of the multiframe having a length of 26
TDMA frames as in the case of traffic channels. The reason why two different

39


multiframe lengths have been chosen is explained in section 2.3.3. To
incorporate two multiframes of different lengths in the same overall structure,
the TDMA hierarchy has been extended with a superframe. In addition, the
requirement for secure encryption necessitated the introduction of a further
higher level in the TDMA structure – the hyperframe. See figure 2.32.

TDMA structure
Hyperframe = 2048 Superframes = 2 715 648 TDMA frames (≈ 3 h 29 m)
Superframe = 1326 TDMA frames = 6.12 s
= 51 Multiframes A (26 TDMA frames -120 ms)
= 26 Multiframes B (51 TDMA frames)

Multiframe A Multiframe B
0 1 2 3 24 25 0 1 2 3 4 48 49 50

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7

Basic TDMA frame

Figure 2.32

The hyperframe includes more than two million TDMA frames and it has a
duration of about 3.5 hours. The running number of a basic TDMA frame in a
hyperframe is one of the parameters determining the encryption key (see
section 2.5.2). An overview of the system signalling is given in section 2.3.2.
The achievement of considerable diversity gain for the traffic channels is
based on advanced channel coding. This is different for full-rate and half-rate
traffic channels and for speech and data transmissions with different data
rates. The signal channels are also protected by channel coding. An outline of
a few of the many channel-coding cases that occur is given in section 2.3.3,
where the interleaving structure for a full rate speech channel is also
described.
The GSM system uses GMSK modulation, which implies moderate filtering of
the modulation spectrum for MSK, on which a description of the modem can
be based. MSK is related to 4-QAM but it includes modifications that
complicate channel equalization (but results in an constant envelope signal).
This can be made simpler if the modulation of the signal to the equalizer
follows as closely as possible the basic 4-QAM arrangement. For this reason,
further modifications have been made to the MSK modulator and demodulator
arrangements. Modems and channel equalization are discussed in sections
2.3.4 and 2.3.5.
The radio subsystem of the GSM is part of a highly complex network structure
(core network), which includes mobile exchanges and the public telephone
transport network with common channel signalling. Advanced network
facilities are required to handle roamers (mobility management) and to protect
the network and its users from unauthorized usage and listening in. Before a
connection is established, a check is made to ensure that the terminal is
authorized to use the network for the requested service (Authentication) and
that the terminal has not been stolen (subscribers can report stolen units to the
Equipment Identity Register). To achieve a very high level of security

40


against listening-in, traffic channels and some sensitive system signalling sign-
als can be encrypted. A similar arrangement is used for authentication
(electronic signature). An outline of this is given in section 2.5.2.

2.3.2 Signalling, TDMA structure
2.3.2.1 Overview

As a preliminary to the analysis of the signalling procedures, a summary is gi-
ven below of the different traffic and signalling channels. The slow and fast
associated channels (SACCH and FACCH) have already been discussed in
section 2.2.2 and are therefore not dealt with here.
One of the duplex radio channels (carriers) allocated to a cell is given the task
of handling the system signalling (except the associated control channels). A
large proportion of the system signalling is sent via broadcast signalling
channels, i.e. channels accessible continuously to all the terminals. The carrier
that is used in a cell for signalling is therefore called the broadcast carrier. In
the outward direction, the base station transmits this carrier continuously at a
fixed frequency and constant power level. (If there is no information to be
sent, dummy bursts are inserted.)
Of the eight time slots in a TDMA frame (numbered 0 to 7), time slot 0 on the
broadcast carrier is used in the outward direction for broadcast or common
control channels, and in the inward direction for call requests from terminals
(Random Access Channel, RACH). Time slot 1 is used in both directions for
Stand-Alone Dedicated Control Channels (SDCCH). The SDCCH is assigned
exclusively (“dedicated”) to a terminal for signalling between base and termi-
nal during the setting up of a call or to exchange messages between the base
and the terminal, e.g. registering the terminal with a new cell or location area.
Time slots 2-7 on the broadcast carrier are used for traffic channels.

Traffic channels Signalling channels on Broadcast Carrier
• Full-rate: • Broadcast channels
Speech (TCH/FS) FCCH: Frequency Control Channel
9.6 kb/s (TCH/F9,6) SCH: Synchronization Control Channel
4.8 kb/s (TCH/F4,8) BCCH: Broadcast Control Channel
2.4 kb/s (TCH/F2,4)
• Half-rate: • Common control channels (CCCH)
Speech: (TCH/HS) Inward direction:
4.8 kb/s (TCH/H4,8) RACH: Random Access Channel
2.4 kb/s (TCH/H2,4) Outward direction:
PCH: Paging Channel
AGCH: Access Grant Channel
• SACCH and FACCH • SDCCH: Stand-Alone Dedicated CCH
(See section 2.2.2)

Figure 2.33

41


The first timeslot (slot 0) of the Broadcast Carrier is called the Broadcast
Channel. It is time multiplexed between several signalling channels. One set
of channels is the Broadcast Control Channels, which for instance gives the
terminals the initial information when they want to connect to a cell. See
figure 2.34a. The name “Broadcast” is used in connection with three different
types of channels, see figure 2.34b.

Broadcast Control Channel, BCCH
Time slot in TDMA frame, corresponding to slot 0 of the Broadcast Carrier
Only in the outward direction
Maximum power, continuous transmission
Used for
Information about the cell, that is, ID, available services
Measurement of signal level by terminal in connection with MAHO
Synchronization Channel, SCH
Used by terminal for frame synchronization (to TDMA structure of base)
Contains information about operator-ID and colour code (BSIC)
Frequency Correction Channel, FCC
Carrier information with a small frequency deplacement from carrier
Used by terminal to set channel oscillator frequency (system clock)

Figure 2.34a

Broadcast Radio Channel:
One designated radio channel (frequency slot) in each cell

Broadcast Signalling Channel:
Timeslot 0 of Broadcast Radio Channel

Broadcast Control Channel:
One of the signalling Channels multiplexed on the
Broadcast Signalling Channel

Figure 2.34b

Another set is the Common Control Channels, which are used when a trans-
mission channel shall be set up. See figure 2.35.

42


Common Control Channels

Inward direction: Random Access Channel, RACH
Slotted Aloha. Shortened burst (large guard time)
Page from terminal with request for allocation of SDCCH
Reply to page from fixed side

Outward direction: Access Grant Channel, AGCH
Reply to terminal page on RACH
Allocation of SDCCH for further signalling (call establishment)

Stand-alone Dedicated Control Channels, SDCCH
Inward and outward direction
Signalling in connection with registration and call establishment

Figure 2.35

Terminal synchronization to base station

When a terminal is switched on or moves into the coverage area of a new cell,
the first step is for the terminal to check and adjust itself to the local radio
environment in the cell. It has to scan over all the 124 radio channels allocated
for GSM and find the channel with the strongest broadcast carrier (It looks for
channels containing periods with sine wave bursts.) It then synchronizes to
this carrier and reads the data on the BCCH. This implies:
- Fine-tuning of the local oscillator frequency (master oscillator) to minimize
the frequency error between the terminal and the base station
- Setting the counters that determine the complex TDMA structure, so that the
terminal’s TDMA timing (TDMA hyper structure) coincides with that of the
base station
- Determination of the network’s ID code (several operators in a country may
be sharing the GSM band), the cell’s ID code and allocated radio channels
and the broadcast carriers in adjacent cells

The BCCH also informs the terminal about the maximum transmit power it
may use and the minimum input level to the receiver to be allowed to access
the base. See figure 2.36.

43


The Idle Mobile Monitors BCCH/CCCH
BCC: - fixed non - hopping carrier with
constant output power
- 2 parameters are set for each cell:
a) MIN - RX - LEV - A CCESS
b) MS - TX - PWR - MAX
Mobile must scan 125 RF carriers and search for PSW
(Pure Sine Wave burst). Then it shall synchronize on this carrier
and read BCCH data.

Mobile may include optional storage of BCCH carrier information
to reduce its carrier search time.

When mobile makes a call or answers a call, it shall use the CCCH uplink.
Then the mobile shall tune to a DCCH which is allocated by the network.

On DCCH there will be a call set up procedure and then the mobile
will be allocated a TCH.

Figure 2.36.
See also section 2.3.2.

Registering with the base station

After the successful locking on to the Broadcast Carrier of a suitable operator
and cell (step 1 in figure 2.37) the terminal has the right to start signalling. The
next step is to register.

GSM. Contacts between base and terminal before a call is set up
1. Selection of base station and ID check 4. Establishment of call from base
Select strongest broadcast carrier Base pages over cells within traffic area of terminal
Frequency synchronization Terminal replies on paging channel
Synchronization of TDMA frame Base assigns signalling channel to terminal
Check of base station ID (in cell where the terminal is situated)
(if unsuitable, another broadcast carrier selected) Information exchange between terminal and base
2. Registration Base assigns traffic channel
Paging of base by terminal Terminal switches over to traffic channel
The base assigns registration channel
5. Updating of location
Information exchange between terminal and base
Update of location (2 above) when terminal moves
(authorization, location registration)
to cell belonging to new traffic (location) area.
Terminal switches over outward paging channel
3. Establishment of call from terminal
Terminal page with request for signalling channel
Information exchange between terminal and base
Base assigns signalling channel
Terminal switches over to traffic channel

Figure 2.37

Through the registration, the terminal is accepted by the network so that it can
be reached by calls from the network. Part of the registration is that the
network stores information about the location area in which the terminal is
situated. Registration takes place both when the terminal is switched on or get

44


into radio coverage and when it enters a cell belonging to a new location area.
The terminal notices when it enters a new location area as this information is
transmitted continuously on the BCCH.
As part of the registration procedure, the GSM network checks that the termi-
nal is authorized for connection (Authentication) and may also check that the
unit is not listed as stolen (enquiry to the Equipment Identity Register, EIR).
Hereafter, the terminal monitors continuously the signalling channel (Paging
Channel, PCH) over which the base station pages terminals belonging to that
location area (that might comprise several cells).
If the terminal moves into another cell, synchronization to a new broadcast
carrier takes place. If the cell belongs to a different location area, re-
registration takes place so that the network can transfer paging signalling on
the PCH to the new location area. (The size of a location area is a trade-off
between heavy registration signalling in small location areas and heavy
paging signalling in large location areas containing many cells.)
Signalling for registration is initiated by the terminal which sends a paging
message over the RACH, whereupon the network assigns an SDCCH to the
terminal via a data message on the AGCH. Registration signalling includes
authentication.
As described in section 2.5.2, successful registration of a roamer with a base
station (new location area) results in the Visiting Location Register sending a
message to the subscriber’s Home Location Register with details of the
subscriber’s ID and where he should be paged for incoming calls. If necessary
a cancellation message is sent to the VLR where the roamer was registered
previously.

Call set up

Setting up calls to or from a terminal requires extensive signalling for transfer
of address information and allocation of a radio channel/time slot. Initially
signalling is via the CCCH (Common Control CH) and, subsequently, the
SDCCH (Special Dedicated CCH). The outcome is that the call is assigned to a
traffic channel. The procedure makes it possible for the network to determine
in which cell within the location area the terminal is situated.
The setting up of a call to a terminal is initiated by the network paging the
terminal over the PCH (Paging CH) in all cells belonging to the location area.
The terminal acknowledges the call on the RACH (slotted Aloha). The
procedure makes it possible to determine in which cell of the location area the
terminal is situated. Thereafter, the network sends a message to the terminal
via the Access Grant Channel (AGCH), instructing it to switch over to a given
SDCCH, which has an associated SACCH. The SDCCH is used for transmis-
sion of the calling and called-party numbers, for authentication, for sending
encryption keys, etc. Finally, a traffic channel is allocated to the terminal. The
SACCH is used for setting the transmission power.
For setting up a call from a terminal, the terminal sends a call request via the
RACH. The network sends back details of the allocated SDCCH on the
AGCH. Further signalling takes place as described above. See also figure
2.38.

45


Call set up on request from terminal
Base - terminal Terminal - base
1. Call request (RACH)
2. Allocation of CDCCH (AGCH)
3. Signalling dialog (SDCCH)
4. Traffic on traffic channel

Call set up by base
1. Paging (PCH) 2. Hear I am (RACH)
2. Allocation of CDCCH (AGCH)
3. Signalling dialog (SDCCH)
4. Traffic on traffic channel

Figure 2.38

Handover decisions are based on a host of different radio parameters
measured both by the terminal and the base-station equipment (MAHO). See
figure 2.39.

Handover and power control
The mobile measures:
- Signal strength from own base (TCH/BCCH)
- Quality from own base (TCH)
- Signal strength from the 6 strongest neighbour cells (BCCH)

The values are sent to BSS on SACCH
- The report takes 500 ms
- ARQ is not used since all values are updated continuously
- The transfer can be interrupted by other data messages

The Base measures:
- Signal strength and quality from mobile TCH
- Signal strength at unused channels (interference)
- BS MS Distance (For time alignment)
The BSS system decides handover and power control by evaluating the
measurements values.
Handover can also be initiated from the fixed network because of traffic aspects.

Figure 2.39

The measured data on local radio conditions at the terminal are transmitted to
base via the SACCH. The following information is used by the MTX when
deciding the best cell for handover:
- Carrier level (RXLEV) and connection quality (RXQUAL), i.e. the bit error
probability, for the connection between the terminal and the base.
Averaging is done over 12 seconds.

46


- Signal levels at the terminal receiver of broadcast carriers from nearby cells.
- Distance between base and terminal. This parameter is obtained from the
timing advance procedure.
- Interference level in the base receiver in idle time slots.

A variety of hand over algorithms based on these data can be used to
determine the handover instant. Some hysteresis is desirable so that repeated
hand overs back and forth can be avoided in overlap coverage areas.
Power control in the base and terminal transmitters reduces the average
interference level due to co-channel interference. This gives better average
speech quality, especially if frequency hopping is employed. Frequency
hopping implies averaging of the co-channel interference over several
channels and, consequently, the cell planning does not have to be based on
the worst case of interference. (However, the full advantage of interference
averaging is only obtained in systems based on DS-CDMA.) Control of the
terminal transmitter power also reduces the battery drain.
Power-control decisions are based on the carrier level and transmission quality
at the terminal and base. If these values are unnecessarily high, commands
will be sent to the opposite end of the link to reduce the transmit power. On
the other hand if the carrier level or connection quality is too low, the trans-
mitter power in the other end will be increased.
Discontinuous transmission and reception are used. Discontinuous transmis-
sion means that transmission is stopped during pauses in speech. This requires
reliable detection by a voice activity detector (VAD) of gaps in incoming
speech signal. The advantages of discontinuous transmission are lower
average levels of co-channel interference and lower power consumption by
the terminals. One drawback, however, is that totally silent intervals are
perceived as disturbing. It is therefore necessary to generate a rough approxi-
mation of the background noise level during pauses in speech (comfort noise).
Because the data rate required to describe the background noise is low, the
capacity of the SACCH is adequate for this.
With discontinuous reception, if the transmission quality is too low during a
20 ms speech frame, output of the detected signal is suspended. This is called
Frame Erasure and is controlled by the channel decoder, see section 2.3.3.2. If
only a single, isolated speech frame is disrupted, the previous frame will be
repeated. However, if several consecutive speech frames are disrupted, several
repetitions of a previous frame will cause strong quality degradation. For each
repetition after the first one, the output level is progressively reduced down to
0, and comfort noise is inserted instead.
The main signalling needed during a call is summarized in figure 2.40.

47


Signalling during a call
on SACCH
from base
power command
time advancement
frequency hop structure
frequencies used by adjacent channel

from terminal
BER on traffic channel
signal levels from adjacent cells
comfort noise

on FACCH
from base
command to switch channel (frequency and time slot)
Different training sequencies ("colour codes")
Mobile Assisted Hand-Over (MAHO)
Discontinuous transmission
Discontinuous reception

Figure 2.40

2.3.2.2 Adapting a terminal to the radio environment in a new cell

Time slot 0 of the Broadcast Carrier is used in time division multiplex for
several different signalling channels. The TDM structure comprises ten
consecutive number 0 time slots. The primary TDMA frames is in this case
part of the type B multiframe, which comprises 51 basic frames. One time slot
in the TDM structure is occupied by the FCCH, one by the SCH and the
others are divided between the BCCH and the CCCH (see figure 2.41).

48


Broadcast carrier. Signalling over time slot 0 (T0)

7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0

Logical T0 channel C F S B B B B C C C C F S B
(TDM channel)
TDM frame
(Ten x T0)
F: Frequency Correction Channel (FCCH)
S: Synchronization Control Channel (SCH)
B: Broadcast Control Channel (BCCH)
C: Common Control Channels (CCCH)
F and S occupy one time slot each per TDM frame
B and C together occupy 8 slots per TDM frame

Broadcast carrier. Signalling over time slot 1 (T1)

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0

Logical T1 channel S S S S

S: Stand-Alone Dedicated Control Channel (SDCCH)

Figure 2.41

When a terminal has no stored data on the radio environment at the time it is
switched on, the first step it takes is to scan all GSM channels and to record
the field strength. The terminal then returns to the strongest channel to check
if it is a broadcast carrier. If not, the terminal checks the next strongest
channel and so on. This procedure enables the terminal to find the strongest
broadcast carrier and, hence, it has preliminarily selected the best cell, to be
connected to.
The terminal then performs fine-tuning of the frequency and synchronization
to the base TDMA structure by first receiving the FCCH (Frequency
Correction Channel) and, thereafter, the SCH (Synchronization Channel). In
addition to synchronization data with respect to the TDMA structure, the SCH
also transmits the Base Station Identity Code (BSIC), which consists of the
network code and the base station’s colour code. (If the terminal were to seize
a carrier belonging to another operator, it would have to scan for a new
broadcast carrier.) The terminal then receives information on the BCCH. The
BCCH continuously transmits information on the identity of the network (ope-
rator) and cell, and on the channel allocations for the cell and the broadcast
carriers for the six adjacent cells.

49


2.3.2.3 Measuring the signal levels from adjacent cells

The terminal receives information over BCCH from the current base station
about the frequencies of the broadcast carriers from the adjacent cells. As
described in section 2.2.2, there is time during each TDMA frame for the ter-
minal to measure the level of a carrier from an adjacent cell. To ensure that
readings are reliable, averaging must be carried out over the fast fading.
Therefore, several measurements are made of each carrier before the mean
values are sent over the SACCH to the base. The terminal also needs to
identify the cell from which the carrier is being transmitted (during extreme
propagation conditions, the carrier from a remote cell may be stronger than
the carrier from the adjacent cell).
Identification of the measured carrier is done by noting the BSIC (Base Station
Identity Codec) transmitted on the SCH, which is placed in time slot 0 on the
broadcast carrier. The terminal measures the identity of a carrier during the
last, idle TDMA frame in the multiframe. (This applies to full-rate traffic
channels. On half-rate channels, one of the free time slots allocated to the
other half-rate channel is used.)
A complication here is that the base stations may not be mutually time
synchronized. This means that, to start with, the terminal has to listen to the
BCCH on the broadcast carrier from an adjacent cell for an entire TDMA
frame to be certain of capturing a 0 time slot. This alone is not enough,
however, since the 0 time slot is time multiplexed between several signalling
channels. The SCH uses only one in ten of the 0 time slots.
For this reason, two different multiframes are used. The inward and outward
traffic channels use multiframe A (26 basic TDMA frames), whereas the
broadcast channels use multiframe B (51 frames). This means that the idle
TDMA frame that is used for listening will slide over the TDMA frames in
multiframe B, thus ensuring that, after a number of A multiframes, the termi-
nal will have reached the correct 0 time slot in the broadcast channel of an
adjacent cell. See figure 2.42.

50


Sliding multiframes
51 frames = 235 s

FS B C FS C C F S C C FS C C FS C C

T T T A T T T T T T A T T T
T T T A T T T T T T A T T

TCH
T T T T A T T T T T T A T T
T T T T A T T T T T T A T T
T T T T A T T T T T T A T
T T T T A T T T T T T A T
T T T T T A T T T T T T A T
T T T T T A T T T T T T A T

1326 frames = 6.12 s
T T T T T A T T T T T T A
T T T T T A T T T T T T A
T T T T T T A T T T T T T A
T T T T T T A T T T T T T
A T T T T T T A T T T T T T
A T T T T T T A T T T T T
A T T T T T T A T T T T T
T A T T T T T T A T T T T T
T A T T T T T T A T T T T T
T A T T T T T T A T T T T
T A T T T T T T A T T T T
T T A T T T T T T A T T T T
T T A T T T T T T A T T T T
T T A T T T T T T A T T T
T T A T T T T T T A T T T

Figure 2.42

2.3.2.4 Structure of signalling frames

The structure of some of the signalling frames mentioned above is shown in
Figure 2.43. The frequency-correction frame (FCCH) consists of an
unmodulated carrier pulse with a small frequency shift relative to the nominal
carrier frequency. The synchronization burst (SCH) has a longer
synchronization sequence than the normal TDMA frame, to ensure that very
reliable time synchronization is achieved. The paging burst (RACH) from the
terminal to the base is shorter than the others, because the terminal may not
yet know the suitable timing advance which is needed to compensate for pre-
sent propagation delay. To prevent paging bursts from spilling over into
adjacent time slots, a much larger guard slot than normal is needed.

51


Signalling frames in the GSM.
Time slot 0.58 ms

Guard slot Guard slot

Normal burst (TCH)
Sync.

Tail Tail
Frequency correction (FCCH)

Tail Tail
Frame synchronization (SCH)
Extended sync.

Random access (RACH)
Sync.

Extended tail Tail

Figure 2.43

2.3.3 Channel coding and interleaving
2.3.3.1 Traffic channels

Traffic channels are either full-rate or half-rate, and both can be used for
speech or data channels with different data rates. It would take too long to
look at all the cases, but the channel coding of three types of traffic channels
are discussed below:
a) Full-rate traffic channel for speech (TCH/FS)
b) Full-rate traffic channel for 4.8 kb/s (TCH/F4.8)
c) Half-rate traffic channel for 4.8 kb/s (TCH/H4.8)
As mentioned above, the structure of a full-rate traffic channel corresponds to
the transmission of 26 TDMA frames in 120 ms. Twenty four of the frames
are used for traffic, which means that on average 200 traffic frames per
second are transmitted (one time slot every five milliseconds). The time for
one time slot in a TDMA frame is 0.577 ms, during which time 114 (2 x 57)
bits are sent. In the case of a half-rate traffic channel, 12 traffic frames are sent
in 120 ms, i.e. 100 time slots per second. Thus, a full-rate traffic channel
transmits 22.8 x 103 b/s (200 x 114), and a half-rate traffic channel
11.4 kb/s. See figure 2.44.

52


Channel coding GSM
From full rate speech encoder
260 bits every 20 ms ⇒ 50x260 = 13 kb/s

From channel encoder
Full rate channel
456 bits every 20 ms ⇒ 50x456 = 22.8 kb/s
Half rate channel
456 bits every 40 ms ⇒ 50x456 = 11.4 kb/s

4.8 kb/s data channel on full rate channel
+ 1.2 kb/s for network signalling functions
Channel coding:
240 input bits ⇒ 456 output bits

2.6 kb/s for data channel on full rate channel
Channel coding:
2.4 kb/s for data channel on half rate channel
Channel coding:

Figure 2.44

2.3.3.2 Speech transmission (TCH/FS)

The speech coder in the first-generation GSM outputs a binary signal in the
form of sequences of 260-bits at a rate of 50 blocks per second (20 ms per
block). The bits are classified according to their sensitivity to transmission
errors.
Class Ia includes bits for which transmission errors result in a strongly
disrupted output signal from the speech decoder. If transmission errors occur
in the class Ia group (despite FEC channel coding), the 20 ms frame will be
replaced by the preceding frame from the speech coder (“frame erasure”). To
enable error detection, three parity bits are inserted into the class Ia group
(Cyclic Redundancy Check, CRC).
Assigned to class Ib are bits for which transmission errors result in fairly large
degradation of the speech quality. Therefore FEC is used. Class Ia (including
the parity bits) and class Ib bits are then combined with four tail bits (since a
convolution code with constraint length of 5 is used). This gives a block
comprising a total of 189 bits, which is coded in a 1/2 rate convolution coder.
The coder outputs 378 bits (2 x 189).
The remaining bits from the speech coder are assigned to class II. Because
these are relatively non-critical as regards the impact of transmission errors on
speech quality, they are not given any protection against bit errors through
FEC. Thus, the channel coder arrangement outputs a total 456 bits every
20 ms, which matches to the capacity of a full-rate channel (see figure 2.45a
and b)

53


Channel coding and interleaving on TCH/FS

Ia Ib II
Block from 50 132 78 260 bits 50 times/s = 13 kb/s
speech coder

CRC
(error detection)

Block coding
25 25 3 66 66 78
for error
detection

tail

Resorting 25 3 66 25 4 189 bits
66 78
Adding of tails
Convolutional coding R = 1/2
(constraint length of 4)

Convolutional
coding 2 x 189 = 378 78
(class I bits)
456 bits 50 times/s = 22.8 kb/s

Split up into
8 subblocks 57 57 57 57 57 57 57 57
of 57 bits

Unaccessible
frame
TDMA frame
(Each time slot in TDMA accommodates 2 x 57 user bits)

Figure 2.45a

54


Fullrate speech channel GSM THC/FS, cont

456 bits interleaved over 8 half bursts
57 57 57 57 57 57 57 57

Burst format
1 57 1 26 1 57 1

148

8-burst block-diagonal interleaving

A B C D E

20 ms
AA AA BBBBCCCC D D D DE E E E
AA AA BBBB C C C C D D D DE E E E

ca 40 ms

Figure 2.45b

The reason for adding four tail bits is as follows. In a convolution code, which
is decoded by a Viterbi arrangement, the decoding is based on calculation of
the Euclidean distances between the received signal and different paths
through the trellis. Apart from the current radio symbol, several previous and
subsequent symbols contribute to the distance. Therefore, if there is a sudden
break in transmission after the last information symbol in the block, the
trailing symbols are given less decoding information and, consequently, have
a higher error rate than the other bits (see also module DT12).
The error rate for the last part of the burst can be substantially reduced if
enough tail bits are added to reset the memory cells in the channel coder to
zero at the end of each burst. The number of tail bits follows from the
constraint lenght. In this way, a known final state in the trellis is obtained in
the same way that there is a known starting state. Because the decoding can be
based on known start and end states, the number of paths through the trellis
will be lower at the start and end. This means that the bit error rate will be
lower for the first and last bits in the block input to the channel coder (see
figure 2.46). This is exploited by placing class Ia bits at the beginning and
end of the input sequence to channel coder.

55


Varying bit error rate over a TDMA burst

Bit error rate
from channel coder

Without tail

With tail

Bit number in
0 50 100 150
data sequence
to channel coder

Figure 2.46

A block from the channel coder comprising 456 bits is then divided into 8 sub
blocks of 57 bits each. These are inserted into the allocated time slot (physical
channel) in eight successive TDMA traffic frames by an interleaving arrange-
ment. (Besides 24 traffic frames, a multiframe also contains one signalling
frame and one idle frame. The non-traffic frames are skipped.) Since there are
114 user bits in each time slot, each time slot will therefore contain informa-
tion from two adjacent speech frames. The speech coder, the interleaver and
the TDMA frame arrangement create a time delay of about 40 ms in an ideal
case with unlimited processing speed, see section 2.3.3.5. (Additional delays
in the speech and channel coders bring the total transmission delay up to
about 70 ms.)

2.3.3.3 Transmission of 4.8-kb/s data on TCH/F

Apart from the data signal itself, 1.2 kb/s is needed for synchronization and
signalling in the fixed network. Accordingly, 6 kb/s need to be transmitted.
The incoming data signal is split up in blocks of 120 bits every 20 ms. Prior
to convolution coding at a rate of 1/3 and a constraint length of 5, waists of 4
bits is inserted between sub blocks comprising 15 bits (see figure 2.47). In ad-
dition, a 4 bit tail is added. The waist arrangement reduces the errors in nearby
bits in the same way as described for tails. The waist and tail extend the block
to 152 bits. The number of bits obtained after channel coding is 456, which
corresponds to a gross data rate on the traffic channel of 22.8 kb/s
(50 x 456).

56


4.8-kb/s data channel on a full-rate traffic channel (TCH/F4.8)

Block from 15 15 15 15 15 15 15 15 120 bits 50 times a second = 6 kb/s
data modem

Addition 15 4 15 4 15 4 15 4 15 4 15 4 15 4 15 4 152 bitar
of tail and
W W W W W W W T
waists
W: waist T: tail

3x152=456
Convolutional
coding: R = 1/3

Figure 2.47

An interleaving arrangement is introduced after the channel coding. The out-
put from the channel coder is 456 bits, 50 times a second. These are divided
into 19 words of 24 bits, and the words are distributed among the time slots
allocated to the traffic channel in 19 adjacent traffic frames (the signalling
frame and idle frames in each multiframe are skipped). This results in an
interleaving delay of more than 100 ms. (A longer delay is often permitted in
data transmission than in two-way speech.)

2.3.3.4 Transmission of 4.8 kb/s data on TCH/H

In the transmission of 6 kb/s over a half-rate traffic channel, the margin
available for channel coding is much smaller than in the last case. A half-rate
traffic channel accepts 456 bits every 40 ms, and 6 kb/s corresponds to 240
bits being transferred 25 times a second. Channel coding consists of the
following steps (see figure 2.48):
- Addition of 4 tail bits to the input block of 240 bits
(to fit a convolution code with a constraint length of 5)
- Convolution code with R = 1/2 generates 488 bits (2 x 244)
- 32 bits are removed by puncturing, leaving 456 bits (puncturing is
described in module DT12, section 3)

Interleaving involves dividing a block of 456 bits into 16 sub blocks, which
are placed in 16 consecutive traffic time slots in the half-rate traffic channel
(signalling frames are skipped).

57


4.8-kb/s data channel on a half-rate traffic channel (TCH/H4.8)

Block from 240 bits 25 times a second =
240
data modem 6 kb/s

tail

Addition 240 244 bits
4
of tail

2x244=488

Convolutional coding:
R = 1/2

Puncturing
of 32 bits 488-32=456
456 bits 25 times a second = 11.4 kb/s

Figure 2.48

2.3.3.5 Transmission Delay

An important parameter, affecting the subjective transmission performance of
the total, integrated network with fixed and mobile parts is the transmission
delay of the radio connection. See section 2.5.1. The one-way delay should be
kept below around 70 ms. It is influenced by the delay in the speech coder,
the TDMA framing and the interleaving. See figure 2.49.

Interleaving GSM
20 ms

Coded Speech frame n-1 Speech frame n Speech frame n+1
speech frame

Time slots

Speech frame n-1 Speech frame n
Detected speech frame

Total delay: ≈ 60 ms

Delay from interleaving: ≈ 20 ms

Fig 2.49

58


2.3.4 Radio modem
GSM uses GMSK (Gaussion filtered MSK) modulation with BTb = 0.3, where
B is the Gaussian filter’s 3-dB bandwidth and 1/Tb is the data rate of the input
signal to the modulator. (Tb = Ts, where Ts is the length of the radio symbols.)
This filtering complies with the specification that the modulation spectrum
shall be attenuated by 30 dB 200 kHz from the carrier frequency. Filtering of
the modulation spectrum in a filter having Gaussian characteristic and
BTb = 0.3 has only a fairly small effect on the main part of the spectrum (see
figure 2.50). The effects on the spectrum of filters with different values of
BTb is shown in figure 2.51.

Power spectrum S (f )
2

dB 4 QAM Nyquist filter
0
(Root Rized Cosine α = 0 , 35)

-20 MSK

-40

GMSK
-60
(B ⋅Tb = 0 ,3 )
-80 (f − fo )
0 1 / 2Tb 1 / Tb 3 / 2Tb 2 / Tb

Figure 2.50

59


0

BbT = (MSK)
1.0
-20
0.7
0.5

Spectral density -40

0.16
0.2
-60
0.25
0.3
-80 0.4

-100

-120
0 0.5 1.0 1.5 2.0 2.5
Normalized frequency: (f-fo)T

Figure 2.51

With increased filtering of the modulation spectrum, the waveform starts to get
distorted. The result is degraded receiver sensitivity relative to the theoretical
detector characteristics for MSK (which is the same as for 2PSK and 4QAM).
See figure 2.52. The figure also shows that there will be an additional
degradation due to limitations at the practical implementation.

60


Bit error rate as a function of normalized signal-to-noise ratio

10-1

5

2

10-2

5

2

10-3
Bit error rate

5

2

10-4

5

2

10-5

5

2

10-6
4 5 6 7 8 9 10 11 12 13 14 15 16
Eb/N0 (dB)

GMSK BbT = 0.20 (measured)
GMSK BbT = 0.25 (measured)
MSK BbT = 0.20 (measured)
Theoretical antipodal, 4QAM, MSK
BbT: normalized bandwidth of transmitting Gaussian low-pass filter
in GMSK modulator

Figure 2.52

The filter with BTb = 0.3 has fairly small effect on the waveform. Therefore,
apart from a considerably steeper spectrum flank, the characteristics of GMSK
is very close to MSK. We shall therefore study the simpler MSK type of mo-
dulation.
MSK can be considered either as a special variant of linear, 4-QAM or as
orthogonal 2-FSK with phase continuity in the transitions between successive
symbols.

61


Starting with standard Nyquist-filtered 4QAM (figure 2.53), the first step is to
go over to O-QAM (Offset QAM), see figure 2.54. This eliminates the paths in
the signal diagram through origo, which gives a considerable reduction in the
envelope variations.

Quadrature AM (4-QAM)

DSBS.C. modulator
db/2 (b/s) Baseband Baseband db/2 b/s
modem AM1 modem
(modulator) AM2 (modulator)
f0
db (b/s) Nyquist
Serial- filtering C1, C2, C3, C4 Serial-
parallel Antipodal parallel
converter converter
db = 1 signals 90 90
Tb
Baseband QM1 Baseband
modem QM2 modem
db/2 (b/s) (modulator) (modulator) db/2 b/s

Baseband Radio spectrum
spectrum C4 QM1 C1
PSK
f - f0 MSK
f CPM
0 d/4 -d/4 0 d/4
AM2 AM1

Signal
waveform
-4Tb 4Tb
C3 C2
t t QM2
-2Tb 0 2Tb
C1 C3 Deep dip in
Baseband impulse response signal envelope
C2 C4

Figure 2.53

62


Widening of the modulation spectrum in a class-C transmitter amplifier. OQAM (Offset QAM)
DSBS.C. modulator
db/2 (b/s) Baseband
modem AM1
(modulator) AM2
f0
A B db S P C1, C2, C3, C4
QAM modem Class-C
(modulator) transmitter Serial-parallel
b/s converter 90
Baseband QM1
Delay QM2
Tb modem
db/2 (b/s) (modulator)

1 QM1
QAM Tb = C4 C1
db PSK
MSK
CPM

OQAM AM2 AM1

f - f0

I A C3 C2
QM2
I B
C1 C3
Do not occur
C2 C4

Figure 2.54

The last step is to use another filtering than Nyquist filtering in order to obtain
a constant envelope signal, see figure 2.55. The drawback is considerably
larger spectrum width. The practical implementation of GMSK generally takes
advantage of the fact that also GMSK with good approximation can be con-
sidered as a form of linear 4QAM. See figure 2.56. (Even if GMSK strictly is
not a linear modulation, there are mathematical relations that can transform the
signal to a linear QAM modulation, "linearized GMSK").

63


MSK as linear modulation (QAM)
± cos ( π t )
2Tb
D/A S
±
2Tb 4Tb cosω0t
±
db/2 t t f0
2Tb
db b/s
Coder ± ± v
Tb 3Tb Tb
90
3Tb
t t sinω0t
db = 1
Tb db/2

D/A S
± sin ( π t )
Impulse response of transmitter filter S: 2Tb
0 < t < 2Tb

v = ± sin ( π t ) cosω0t ± sin ( π t ) sinω0t =
2Tb 2Tb
t π 1
2Tb = cos (ω0 ± ) t = ± cos 2π( ƒ0 ± )t
2Tb 4Tb
(Ts = 2Tb)
(The polarity of the exciting Dirac pulses is determined both by the preceding radio
symbols and by the value of the incoming baseband bit).

Figure 2.55

GMSK modulator (GMSK = Gaussian-filtered Minimum Shift Keying)
MSK interpreted as QAM (Complex signal representation)

10010110 Digital signal processing Radio monolithic circuit
in CMOS VLSI cosω0t
ƒ'i
Ib
ƒ'i φ' φ' cosω0t . cosφ
cos
cos(ω0t + φ)
∫ dt
Gaussian
(NRZ) filter
φ'
sin .
Qb -sinω0t sinφ
~
IsI = I
Complex envelope: -sinω0t
Q
φ ~
s = Ib +jQb = cosφ + jsinφ= ejφ ~
s
Qb = sinφ
~
φ' corresponds Complex signal: sejω0t = ej(ω0t+φ)
to MSK ~
Physical signal: Re[sejω0t] =cos(ω0t+φ) φ
~ I
φ corresponds (Normalize |s| =1 )
Ib = cosφ
to GMSK (cosω0tcosφ - sinω0tsinφ = cos(ω0t + φ)

Figure 2.56

If instead MSK is considered as a form of FSK, the starting point is to
introduce phase continuity between consecutive s 2-FSK symbols. This

64


reduces by half the minimum value of the frequency difference between the
two frequencies in the FSK pair, which gives orthogonality.
One of the symbols in the FSK symbol pair is displaced by + ∆f relative to the
suppressed carrier frequency, and the other by -∆f. ∆f = 1/4Ts is the lowest
value, giving an orthogonal symbol pair. See figure 2.57. In the signal plane,
the signal vector is rotated during each symbol either π/4 or –π/4
depending on the value of the corresponding input bit. As MSK is a type of
FSK modulation, it can be detected by a simple, non-coherent FSK detector.
However, this is non-optimum and the receiver sensitivity becomes about 3.5
dB less.

MSK (FFSK)

(1): + ∆f

(1) Binary
start point
ø (0) input signal

(0): - ∆f

f= f0 + fi = f0 ± ∆f
fi = 1 dø ∆f =
1
2 dt
∆f 4Ts
(1) Ts 2Ts
0 t
(0)
-∆f

ø
π/2

(1) Ts (0) 2Ts
t

−π/2 during one symbol:
π
linear phase change with ±
2
(Ts = Tb) (fi "instantanious" frequency relative to f0)
FFSK: Fast Frequency Shift Keying

Figure 2.57

MSK thus shows the following differences from normal Nyquist-filtered 4-
QAM:
- The transmitter filter does not have a Nyquist-related transfer function. To
obtain MSK with constant envelope, a transmitter filter is used whose
impulse response, ht(t), comprises a half sine period of duration 2Tb.

65


- The excitations of the transmitter filters in the I and Q channels are mutually
displaced by one bit period, Tb, which corresponds to incoming bits being
fed alternately to the I and Q channels. This gives O-QAM, which also
helps to give constant envelope.
- The polarity of the antipodal symbols in the I and Q channels are
determined by the value (polarity) both of the present input bit and of the
previous bit to the modulator. This correlation between the current and the
immediately preceding symbol complicates the channel equalization. It is
therefore eliminated through suitable precoding.

2.3.5. Channel equalization
The system data rate of the GSM is 271 kb/s, i.e. the duration of a symbol is
3.7 ms. The specification for GSM stipulates that the system must be able to
handle time dispersion up to 16 ms (16 µs equalization window), which
corresponds to four symbols. Time dispersion over 16 ms implies that the
impulse response of the channel can cover five sampling points (see figure
2.58). This is the width of the time discrete impulse response to be processed
in the channel equalizer (equalizer window, see figure 2.31).

Radio channel impulse response (time discrete)

16 µs

τ
Tb
3.7 µs

Figure 2.58

Because of the relatively short data burst in a TDMA time slot (0.55 ms), the
variation of the impulse response of the radio channel during this interval is
fairly small even for high speed terminals, especially as the training sequence
is placed in the middle of the burst. Thus, the setting of the channel equalizer
is constant during each burst and determined by the training sequence in the
middle.
The rapid changes in the channel’s impulse response are due to independent

66


fading of the signal components with different propagation delays. The
distance between two fading dips is roughly half the wavelength, i.e. approx.
0.15 m. For a terminal travelling at 90 km/h (25 m/s), there will therefore be
around 17 fading dips a second. One complication in going over from 900
MHz to 1800 MHz is doubled doppler frequency for a certain terminal
velocity. This results in a reduction in the correlation time, i.e. larger
variations of the impulse response during one data burst. (The time correlation
function is the Fourier transform of the doppler spectrum.)
Two types of channel equalizers have been evaluated:
a) Combination of linear equalization and decision feedback (see figure 2.59)
b) Maximum Likelihood Sequence Estimation (MLSE) based on a Viterbi
arrangement.

Suppression of intersymbol interference

Post-sampler impulse response
A a0
"1" t t n = τ + nTs
0 a -1 a1
T C R A a2
a-2 (t-τ)
(τ: delay of main propagation path)
-Ts 0 Ts 2Ts
Antipodal baseband transmission -2Ts
No
+ TR Digital
T C R K-1 K K+1 det
Digital output
input signal signal
••X
k-1 Xk Xk+1 • • + + + + + +
±a 1 ±a 2
"0" +1
"1" -1
Td Td Td Td
Linear equalizer for
suppression of a -1 and a -2 decision feedback
for suppression of a 1 and a 2
(Ts: Symbol duration) (Td: Time delay corresponding
to symbol length Ts)

Figure 2.59

Although the two options are fairly comparable in terms of performance and
complexity, the most widely used is the Viterbi arrangement.
The result of a simulation of the performance using a simple two-ray model of
the time-dispersive radio channel is shown in figure 2.60.

67


Channel equalizer performance (two-ray propagation model)
pb

10 -1 a) Linear equalizer (6 taps) & decision feedback
(4 taps)
b) Viterbi: 16 states (2 4)

10 -2 b a

T = Delay between rays
T =o
symbol time
T
10 -3
To
1 2 3 4 5

Figure 2.60

No diversity gain is obtained when there is a zero delay between the two
propagation paths. As the delay, T, increases, there is an initial fall in the bit
error rate, since the channel equalizer can distinguish between the two
propagation channels with independent Rayleigh fading and combine the
signals so that diversity is obtained. When the delay is more than four symbol
periods, the time dispersion window that the equalizer can handle is exceeded.
Any further increase in the time difference between the two rays will result in
a rapid deterioration in performance.
A suitable amount of time dispersion will therefore improve the detector char-
acteristics by means of frequency or multi-path diversity. See figure 2.61.

GSM - System performance
(early measurements)
BER
10-1
Without coding
With rate 2/3 conv. code
Flat Rayleigh fading (τ=0)
Time Dispersive Rayleigh fading (τ=3T)
10-2 (two-ray model)

10-3

10-4
8 12 16 20 24 28 Eb/No
dB

Figure 2.61

68


In MLSE, the demodulator determines the most probable input sequence
(maximum likelihood) considering the received sequence and the measured
impulse response of the propagation channel. The most likely transmitted
sequence is that having the shortest Euclidean distance to the received
sequence. The comparations are based on simulated sequences, which are
determined both by the transmitted sequences and the time dispersion
(impulse response).
The principle of this arrangement is shown in figure 2.62.

Viterbi arrangement

Simulated transmission channel
All conceivable with no noise or interference
sequences
? S ? MOD Propagation DEMOD
channel (time Simulated sequence
dispersion) (noise free)
A
Determine
Propagation channel the
impulse response Measure
Euclidean sequence
distance giving
between the shortest
Correlator A and B Euclidean
distance
Detected training Detected
sequence B
sequence
(with noise)
Data S Data

DEMOD

Receiver

Figure 2.62

In principle, the MLSE compares all possible transmitted sequences must be
generated in the receiver and the distance to the received sequence
determined. However, this is evidently not possible, as the number of
sequences is almost infinite (2114, since a data burst contains 114 user bits).
Nonetheless, the comparison can be made step by step using the Viterbi
procedure. Each step involves considering all possible sequences during 16
µs, which corresponds to the time-discrete impulse response, with a maximum
width of 5 samples. Besides the current symbol, the values of 4 other symbols
must be considered. This corre-sponds to 24 = 16 states. The procedure can be
interpreted as Viterbi coding of a convolution code.
The complexity of the equalizer increases sharply with an increase in the
number of states. During the original design of GSM, it was decided that the
equalizer would be too complex if more than 16 states were required. This,
therefore, had a decisive influence on the design specification. A trade-off had
to be made between the maximum time dispersion the system could handle
and the maximum system data rate, se section 2.2.7. In extreme cases (such as
in the Swiss Alps), time dispersion may exceed 16 µs, which means that
measures to reduce the time dispersion have to be adopted (such as suitable
antenna siting and use of directional antennas).

69


2.4 Radio performance
The GSM specification stipulates the highest permissible bit error rate for dif-
ferent traffic and signalling channels for different combinations of time disper-
sion, C/I and C/N. Thus, type approval testing involves a multitude of
measurements. Only some of these are dealt with here.
The measurements are made through laboratory tests, in many cases using a
fading and time dispersion simulator. The simulator generates a transfer
function (impulse response) for the propagation channel with several fading
signal components with varying delays. The parameters that describe the rapid
fading (Rayleigh or Rice, maximum Doppler frequency) can be set for each
subsignal. The channel simulator can be set to a number of standardized time
dispersion models (see figure 2.63). It is assumed in the majority of test cases
that the subsignals with varying propa-gation times are subject to independent
Rayleigh fading. The specification stipulates that the channel equalizer shall
be able to cope with time dispersion within a 16-µs window.

Test-channel power profiles (type-approval testing of GSM)

dB Rural (RA) dB Typical Urban (TU)
0 0

-10 -10

-20 -20

-30 t -30 t
µs µs
0 5 10 15 20 0 5 10 15 20

dB Bad Urban (BU) dB Hilly Terrain (HT)
0 0

-10 -10

-20 -20

-30 t -30 t
µs 0 5 10 15 20 µs
0 5 10 15 20

Figure 2.63

Type testing includes measurement of the noise-limited sensitivity (transmis-
sion performance for Eb/N0 = 8 dB) and the performance with co-channel
interference (C/Ico = 9 dB) and with adjacent channel interference
(C/Iadj = -9 dB). The type approval specification gives the permissible
degradation for a number of traffic and signalling channels for these three test
cases. A few of the test specifications are shown in figure 2.65. Other test
cases are the bit error rate at the demodulator output (i.e. without the aid of
channel coding) at a high C/N (C = -115 dBW), with no interference and
with the TU propagation channel.

70


Type approval test criteria for the GSM
Noise-limited Co-channel interference C/Ico =9dB
sensitivity
C=-132dBW Eb/No =8dB Adjacent channel interference C/Iadj=–9dB

TU50 TU3 TU50
Frequency hopping: Frequency hopping: Frequency hopping:

Without With Without With Without With

TCH/FS FER 3% 2% 21% 3% 6% 3%
RBER Ib 0.4% 0.3% 2% 0.2% 0.4% 0.2%
RBER II 8% 8% 4% 8% 8% 8%

TCH/F 4.8 BER 10-4 10 -4 3 10 -4 10-4 10 -4
TCH/F 2.4 BER 2•10-4 10 -5 3 10 -5 3•10 -5 10 -5

SDCCH FER 10% 4% 22% 4% 10% 4%

TU50 = Typical Urban, 50 km/h
FER = Frame Erasure Rate
RBER = Residual Bit Error Rate (Ib and II: see section 2.3.3.2).

Figure 2.64

Several parameters are related to the speech transmission quality: the residual
bit error rate (RBER) is specified for the speech blocks during which error
correction for class 1a bits works satisfactory, i.e. the CRC decoder has
verified the block. The frame erasure rate is the number of blocks in which
class 1a bits from the convolution coder contain errors (detected by the CRC).
The performance requirements shown in figure 2.64 corresponds to the
Typical Urban channel model (TU) with two terminal speeds: 3 and 50 km/h.
The require-ments are also given for the cases with and without frequency
hopping. As is clearly apparent, frequency hopping gives a considerable
improvement in performance at the low terminal speed (3 km/h).

2.5 The fixed network
2.5.1 The speech path

One of the major problems in connection with the fixed telephone network is
delayed echoes. The main cause is the hybrid between the 2-wire and 4-wire
part of the network, see figure 2.65 and 2.66.

71


Terminals Transit station
(Telephone sets)
Local Trans- Local
station mission station
channel
(Communi-
cations link)

Local network Trunk network (4-wire) 2-wire
(2-wire) FDM/TDM
Trunk connection

4-wire
Multi-
Hybrid Balance plexing Multi-
load equip- Amplifiers
plexing
ment equip-
(FDM ment
2-wire or 2-wire
TDM)

Terminal Terminal

Figure 2.65

Telephony: two-wire and four-wire connections

Hybrid 4-wire Hybrid

Potential
2-wire instability 2-wire

(singing)
Balancing (echo effects)
circuit

Repeaters

Figure 2.66

For economic reasons, the final part of the access network, connecting the
telephones consists of two wires, i.e. the telephone line is used for both trans-
mission directions. On the other hand the long distance transport network
must have different transmission equipment and lines for the two directions.
This is called 4-wire. The hybrids forming the interface between the 2-wire
and 4-wire parts of the network give fairly low isolation between incoming
and outgoing lines, i.e. signals from the distant subscriber arriving to the hy-

72


brid is reflected back to him. This could lead to instability if the gain in the 4-
wire network becomes too large.
Even if this does not happen echoes could be noticeable. Fairly strong echoes,
delayed more than about 30 ms, become disturbing and the annoyance
becomes worse with increasing delays and increasing echo level. In the nor-
mal fixed network problems with delayed echoes occur only for very long
transmission lines, especially when geostationary satellites are used. See figure
2.67.

Interconnection of DLMR and Satellite repeater in
public telephone network geostationary orbit

Transmission delay
50-100 ms 12 kb/s 64 kb/s One-way delay

km
2x36000 km 1/4s EX
ES EC

0
00
R T TDM

36
EX
EC TM MXT EC ES
TDM
T R

DLMR 64 kb/s speech codec
EC: Echo control
TM: Transmultiplexer
64 kb/s 12 kb/s
ES: Earth station
EX: Telephone exchange (It is assumed that a 12-kb/s speech codec is
MTX: Mobile telephone exchange used for the digital land mobile radio system)

Figure 2.67

To solve the problem with echoes in extreme situations, echo control is
introduced, either echo suppression or more advanced echo cancellation, see
figure 2.68. An echo suppressor breaks the incoming line to a hybrid when
the outgoing signal dominates. An echo cancellator includes an adaptive
network that is set to the same characteristics as 2-wire input to the hybrid.
Thus a compensating signal is generated that is added in opposite polarity to
the outgoing echo from the hybrid.

73


Echo control

Level
detector
2-wire
Echo Control 4-wire
unit

Level
detector

Echo suppressor

Error
signal
2-wire
Echo Echo 4-wire
simulator

Echo canceller

Figure 2.68

Echoes become more of a concern when digital cellular systems are included
in the fixed network. The reason is the considerable additional delay caused
by the combination of advanced speech coding (much more than for 64 kb/s
PCM), TDMA formatting and especially interleaving. This means that echo
control is needed also in local and regional networks. Another complication is
additional echo effects due to less isolation between speaker and microphone
at hands-free operation of terminals.
Cellular systems use other types of speech codecs than the 64 kb/s PCM
codecs used in the fixed network. This makes it necessary to introduce
transmultiplexers in the interface between the mobile and the fixed network.
See figure 2.67. It has been a design problem to avoid degradation in the
speech quality, when passing through the transmultiplexers.

2.5.2 Switching and control
2.5.2.1 Structure

The Public Land Mobile Networks (PLMN) run by different operators
constitute islands in the Public Switched Telephone Network (PSTN). When
the PSTN initiates a call to a mobile terminal belonging to a PLMN, the call
request is fed to the interface between the PSTN and the PLMN. The interface
consists of the operator’s Gateway Mobile Switching Centre (G-MSC). Details
on all the sub-scribers belonging to the PLMN are contained in the Home
Location Register (HLR) database.
In the simplest case, there will only be one HLR in the mobile network that is
connected to the G-MSC. An incoming call is therefore routed direct to the
HLR, which contains a host of subscriber data, e.g. the subscriber’s mobile

74


network number, which mobile services shall be provided to each subscriber,
security codes to prevent unauthorized access to the network, and the location
of roamers registered in another MSC with its associated Visiting Location Re-
gister (VLR).

2.5.2.2 Calls to mobile terminals

The first step when a call request is made to a mobile subscriber is to send a
data message to the mobile network (operator) serving the subscriber. This ini-
tial request follows the standard calling procedure on the PSTN (or ISDN)
number. Every mobile subscriber is thus allocated a normal phone number
(Mobile station ISDN number) comprising a country code and a network ope-
rator code NDC (National Destination Code) plus a subscriber number issued
by the operator. By means of the country code and the NDC, the fixed
network (e.g. the PSTN or ISDN) establishes a connection with the operator’s
gateway (G-MSC) (see figure 2.69).

GSM. Connection to the public telephone network
Public Switched PLMN
Telephone
Network On registration of mobile
(PSTN or ISDN) terminal in V-MSC Base
MSRN
G-MSC V-MSC
Mobile MSRN
subscriber (Routing of call
to V-MSC) Base
ISDN
(Normal
telephone HLR VLR TMSI
number)
IMSI
MSISDN IMSI

MSC : Mobile Switching Centre
PLMN : Public Land Mobile Network Terminal
G-MSC : Gateway MSC
HLR : Home Location Register
: Visiting MSC
SIM (IMSI)
V-MSC
VLR : Visiting Location Register
SIM : Subscriber Identity Module
IMSI : International Mobile Subscriber Identity
MSRN : Mobile Subscriber Routing Number
TMSI : Temporary Mobile Subscriber Identity

Figure 2.69

The MSC transfers the call to the HLR for the called subscriber. Here, the
number is replaced by the subscriber’s International Mobile Subscriber
Identity (IMSI), which is also stored in the Subscriber Identity Module (SIM)
or, if not used, permanently coded into the terminal. The IMSI has the same
form as a normal telephone number, i.e. a country code, mobile network code
and the subscriber number.
In the case of a non-roaming subscriber, the terminal will be within a location
area belonging to the G-MSC (assuming that the operator uses only one HLR),
in which case paging can be initiated immediately. If the terminal is roaming

75


within the area covered by the operator, the call request must be processed
further within the operator’s PLMN. The HLR associated with the G-MSC will
hold information on the V-MSC with which the terminal is currently
registered. This information will have been sent to the HLR by the VLR with
which the terminal has registered (through the IMSI) and which has allocated
an MSRN (Mobile Subscriber Routing Number) to the terminal. This
temporary number comprises a country code, a trunk code for the VLR and a
temporary subscriber number. Using the MSRN, the G-MSC is able to connect
the call to the V-MSC, which then pages the terminal in accordance with the
procedure described in section 2.3.2.

2.5.2.3 Security against unauthorized access

Because a mobile radio link is more vulnerable to listening-in and
unauthorized usage than is the public telephone network, the following
subsystems features are available:
Encryption of messages and some system signalling associated with a given
subscriber. Temporary subscriber numbers are also used for the identification
of subscriber during the main part signalling procedure, to prevent an
unauthorized party from being able to locate a subscriber by intercepting the
IMSI. The IMSI is replaced by the TMSI (Temporary Mobile Subscriber
Identity) as soon as possible in different signalling procedures over radio
links.
Authentication. This involves checking that a terminal, wishing to make a
connection to the network, is authorized and has subscribed to the requested
service. Encryption procedures are used for transmission of electronic signa-
ture (signed response) to prevent unauthorized parties from sending false
authorization data.
Checking the (manufacturer’s) serial number of a terminal unit. The network
can ask a terminal to provide this number so that it can be checked against a
database (Mobile Equipment Identity Center) containing details of equipment
reported as stolen or as not functioning properly.
Encryption and authentication are based on matching keys (KI) stored both in
the subscriber SIM and in the authentication database (Authentication Center,
AUC) connected to the subscriber’s home switching center (H-MSC) and
HLR. In addition to the IMSI, the SIM (Subscriber Identity Module) contains
the encryption and authentication keys, and the algorithms for computing the
enciphering sequence and the response to the authentication request from the
network. The corresponding information on all the subscribers belonging to
the HLR are held in the AUC database. The algorithms are the same for all
subscribers but the encryption KI is individual. KI is the secret key used by the
encryption and authentication system (see also the session-key procedure
described in Module S1).
In the terminal the signed response (SRES) sequence to be transmitted for
authentication is formed in the terminal. The authentication algorithm
combines the key (K1) with a random sequence (RAND), which is different
for each call. The RAND is transmitted to the terminal, which has access via
the SIM to K1 and the authentication algorithm. The terminal can therefore
generate the signed response (SRES) and send it to the network. The terminal
will only be accepted by the network if the SRES sent as reply by the terminal

76


is identical to the sequence that has been computed by the AUC and stored in
the VLR. The VLR can request the SRES from the HLR (see figure 2.70).

GSM. Security arrangements RAND

Base station Terminal
H-MSC V-MSC Encryption SIM
unit, K c KI , A3 , A 5, A8
SRES

RAND = Random sequence
Triplet Kc = Session key
AUC VLR SRES = Signed response
RAND KI = Secret key
Kc AUC = Authentication Centre
SRES H-MSC = Home Mobile Switching Centre
Terminal procedures VLR = Visiting Location Register
Authentication TDMA
RAND K I RAND K I Frame number

Algorithm Algorithm Algorithm
A3 A8 A5
Kc Encryption sequence
SRES
Plain text
+ Cipher text
Cipher text
+ Plain text

Figure 2.70

The final encryption sequence that is added to the plain text to produce the
cipher text is obtained in two steps. First, the session key, Kc, is obtained by
combining the random sequence as above with the encryption key, K1 (the
secret key) using the encryption key algorithm (A8). The final encryption
sequence is then obtained by combining Kc with the current frame number (in
the hyperframe) using the encryption algorithm (A5).
On the fixed network, the authentication and encryption processes require ac-
cess to a triplet of parameters: SRES, RAND and Kc. Since the RAND must be
different every time it is used, The HLR (with support from the AUC)
computes several triplets with different RANDs for each terminal. On receipt
of a request from the VLR, the HLR transmits a number of triplets for a
roaming terminal. For encryption/decryption of outward and inward traffic,
the VLR (or HLR if the terminal is not a roamer) sends the current encryption
keys, Kc, to the relevant base stations. The arrangement with different Kc keys
for each call means that the only units that know the secrete key K1 is the
AUC and the terminal.
Only the actual data sequence is encrypted – not the training sequence.
Encryption and decryption is implemented through by Modulo-2 addition of
the identical encryption sequences to the transmitted and received signals.
To prevent stolen or defective equipment being connected to the network, a

77


special database (Mobile Equipment Identity Center) is set up. The
manufacturer gives every terminal unit a serial number (Mobile Equipment
Identity), which is stored in a memory that is difficult for unauthorized per-
sons to access and modify. The terminal will transmit this number when
requested to do so by the network.

78


3 D-AMPS, original system
3.1 Overview
The development of digital mobile telephone systems has differed considera-
bly in the USA as compared to Western Europe. This explains the many
essential differences between the Digital AMPS system (D-AMPS), originally
called American Digital Cellular (ADC) and GSM. The analog mobile
telephone system, AMPS (Advanced Mobile Phone System), covered almost
the entire USA around 1985 and served then the needs of the majority of its
users. The overriding problem was that AMPS started to become saturated
1990 in the major urban areas owing to the shortage of frequencies. There was
short term only limited possibilities to increase the traffic capacity by reducing
the cell sizes. The principal reason for adding digital speech transmission to
AMPS was to obtain a substantial improvement in spectrum efficiency.
Work on the specification of a digital mobile telephone system started in
1988, when the Federal Communications Commission (FCC) initiated a study
phase with the stipulation that a new digital system must offer the long-term
potential to provide spectrum efficiency ten times higher than that of AMPS.
The study work was co-ordinated by the TR-45.3 working group set up by
the Telecommunications Industry Association (TIA). The work was based on
a demand specification issued by the Cellular Telecommunications Industry
Association (CTIA) and the result was the IS-54 standard – the Digital-AMPS.
Two system solutions were presented to the working group: narrowband
FDMA (N-FDMA) and narrowband TDMA (N-TDMA). The overall system
requirement was that a threefold improvement in spectrum efficiency should
be obtained to start with by accommodating three speech channels in a
30-kHz radio channel. Two variants of N-FDMA were submitted: one by
ATT/Bell and the other by Motorola. The N-TDMA system proposal was
submitted by Ericsson. Pilot systems were developed and demonstrated in
field tests. From the results of these, the working group drew up a
recommendation for N-TDMA, followed by a detailed standard specification:
EIA/TIA-IS-54. The considerations taken info account when selecting TDMA
are summarized in figure 3.2.
The D-AMPS system was to start with an add-on to the analog AMPS, the two
systems forming an integrated system. The digital system should be gradually
introduced in the main urban areas. System capacity was increased by
replacing an analog 30 kHz base-station unit for one speech channel by a di-
gital unit for a 30 kHz TDMA radio channel for three speech channels. Each
time a unit is exchanged in a cell, the total number of speech channels in the
system was increased by two.
The analog system alone would continue to serve the rural areas for a long
time yet, which meant that mobile terminals would have to switch between the
the analog and the digital system to make full use of the integrated system
both inside and outside the main urban areas. The original D-AMPS system
also depended on the analog system for most of the signalling, e.g. for setting-
up calls and allocating traffic channels. The signalling complexity of the origi-
nal D-AMPS system was therefore much lower than in the GSM system.
Initially, the D-AMPS network constituted isolated islands within an otherwise
analog system. It must be possible to exchange an analog base station in an
individual cell for a digital one, without the need to change the overall cell-
frequency structure. This would hardly be possible if the system were based

79


on FDMA, whereby a 30 kHz analog channel would be replaced by three di-
gital 10 kHz channels. Because these would end up in the same cell, adjacent
channel selectivity approaching 70 dB would be required. This is
incompatible with the adjacent channel performance of existing analog
systems, which must use cell structures based on an interleaved channel plan,
with a guard band of several channels between each utilized channel in a cell.
This permits adjacent channel selectivity of only 30-40 dB. The unrealistic
high requirement on adjacent channel selectivity using adjacent channels in
the same cell was avoided with the use of TDMA. This was the principal rea-
son why N-TDMA was chosen instead of N-FDMA. See figure 3.1.

FDMA
30 kHz channels

10 kHz channels

Co-existence in same cell puts high require-
ments on Adjacent Channel Attenuation

Figure 3.1a

TDMA fits into the frequency reuse plan for AMPS

A A/D
A A A A
A A/D
A A A A/D
A A/D
Replace one 30 kHz single voice path A: old analog system
channel with 3 TDMA voice paths D: digital/TDMA system

Figure 3.1b

The TDMA arrangement thus accommodated three speech channels per 30
kHz channel. Continued advances in speech coding could eventually result in
a further doubling of the capacity (six speech channels in a 30-kHz frequency
slot). As in the GSM, the system solution for the D-AMPS system therefore
allows each carrier to carry twice the number of half-rate traffic channels. An-
other way in which the spectrum efficiency can be enhanced in the long term
is through tighter geographical packing (smaller cell clusters). Thus, the long-
term potential of the D-AMPS system complies with the FCC objective of
more than ten times improved spectrum utilization as compared to that of the
Analog AMPS. To start with, however, the digital system will have to be
adapted to the same cell structure (cluster size) as used for the analog system.

80


Development work continued after 1990, resulting in a new specification
IS-136. The original D-AMPS was augmented by a complete arrangement of
signalling channels plus control and monitoring equipment, making it an
autonomous radio system, see section 7.2. It became one of several system
options for the new PCS band around 1900 MHz. The name was also
changed to TDMA 136. (Other PCS standards are IS-95 based on DS-CDMA
and GSM.)
More recently there have been preparations to adapt D-AMPS to the
increasing market for data transmission. It has been decided to use EDGE
(with 200 kHz channels) also for D-AMPS (TDMA 136). See section 7.3.

3.2 System background
System considerations

At the over-all system discussions, the points indicated in figure 3.2 were ta-
ken into account. Several of these have been referred to above. Some of the
key requirements on the system are summarized in figure 3.3.

D-AMPS. System considerations
No new frequency band available
Gradual introduction into A-AMPS cell structure:
- In existing 30 kHz channel raster (cluster site)
- In existing cells (sites)

Dual-mode terminals
A-AMPS only system in rural areas for many years
Mix of A-AMPS and D-AMPS in high density areas

Two system alternatives evaluated:
Advanced 10 kHz FDMA (ATT/Bell, Motorola)
Extremely narrow-band TDMA (ERA)
(3 time slots in 30 kHz radio channel)

Limitations of FDMA alternative:
3 adjacent radio channels to be used at one site result in
extreme requirements on low out-of-channel radiation

Figure 3.2

81


D-AMPS. Technical considerations
Channel spacing 30 kHz
At least 3 time slots per frame
(to meet capacity requirements)
(to make time duplex possible)
(to make MAHO possible)
Bandwith per speech channel 10 kHz
With low adj. channel requirements ≈ 50 kb/s system
data rate is possible over a 30 kHz radio channel.
With 30% TDMA overhead, data rate from channel
encoder could be ≈ 50(3x1.3) ≈13 kb/s

Equalizer
With 50 kb/s system data rate and using 4QAM,
the symbol length becomes 40 µs.
Two-tap equalizer gives 40µs equalizing window.
Equalizer must adapt to variations over a data burst
(burst length ≈ 6.5 ms)

Interleaving depth
Length of TDMA frame (3 slots) ≈ 20 ms
(long burst length decreases relative TDMA overhead)
Maximum TDMA formating + interleaving delay 40 ms
Maximum interleaving depth 2

New technology available
Improved low-rate speech coders
Linear transmitters
Adaptive DSP

Figure 3.3

The main difficulty was to reduce the bandwidth requirements, so that three
speech channels could be fitted into one 30 kHz radio channel. This was
accomplished by the trade-offs given in figure 3.4. These are further discussed
in section 3.3.

82


System trade-offs
Requirement:
3 speech channels to fit into 30 kHz radio channel
Complied with through:
Low rate speech coder (8kb/s)
Moderate amount of channel coding
Linear modulation (4QAM)
Low TDMA overhead, i.e. long bursts
Low requirements on adjacent channel selectivity
Corresponding system parameters:
System data rate: 25 kbaud = 50 kb/s
Frame length: 20 ms (or 2x20ms = 40ms)
Slot length: 6.67 ms
Maximum interleaving depth: 2

Figure 3.4

As the frequency planning should be closely aligned with that of the analog
system, frequency hopping could not be used. Due to the narrow modulation
no appreciable gain could be obtained from multi-path diversity in the
equalizer. The scope for introducing interleaving was also limited because of
the relatively long duration of a TDMA frame. This meant that if reasonable
diversity gain against fading due to multi path propagation was to be achieved
for quasi stationary connections (handheld terminals), antenna diversity would
be necessary also at the terminals. However, this was not introduced as it did
not give any improvement in spectrum efficiency as the cluster size was
determined by the analog system. (In the corresponding Japanese system,
which is not integrated with an analog system, the terminals use antenna
diversity.)

3.3. Radio specification
The requirement that three speech channels should be time multiplexed on a
30-KHz radio channel imposed heavy demands on the subsystems involved –
speech coders, channel coders, radio modem and TDMA formatting.
Advances made since the GSM specification was finalized enabled the data
rate from the speech coder to be reduced from 13 kb/s in GSM to around
8 kb/s in the D-AMPS system. The coding rate is also higher in the D-AMPS
system than in GSM.
The bandwidth expansion in the modulator is considerably less than in the
GSM, thanks to the use of linear Nyquist-filtered 4QAM modulation. The
drawback is that the transmitter output stage must have good linearity char-
acteristics.
To reduce the relative overhead in TDMA formatting, longer time slots are
used than in GSM. This means a smaller relative increase in the system data
rate in order to accommodate guard slots and synchronization sequences. The
drawback with long time slots is that the channel equalization must follow the
changes in the impulse response of the propagation channel during a data

83


burst (rendering channel equalization more complex) and that the possible
interleaving depth will be insignificant (otherwise the transmission delay
would become too long).
A summary of the principal radio parameters is given in figure 3.5. Since lin-
ear modulation is used and the requirement for adjacent channel selectivity is
only moderate, a 48 kb/s system data rate can be used over a 30 kHz channel.
The π/4 arrangement reduces variations in the signal envelope which, in turn,
reduces the requirement for linearity in the transmitter amplifier. Nonetheless,
a fairly complex transmitter amplifier with special arrangements to improve
linearity is required.

Transmission specification for the USA
Digital MTS (D-AMPS/IS-54)
Frequency band: 825–850 MHz
(frequency duplex) 870–895 MHz
Channel spacing 30 kHz
Modulation π / 4-DQPSK (4QAM)
System data rate 48.6 kb/s
TDMA Frame 40 ms
Time slots 6 x 6.67 ms
Full-rate channel data rate 13 kb/s
Speech coder Code-excited LPC (CELP) also called
Vector-sum excited LPC (VSELP)
7.95 kb/s
Diversity Channel coding
Interleaving

Fig 3.5

The TDMA structure is shown in figure 3.6. As in GSM, the timing of the
inward and outward structures differs to make time duplex possible (the
received and transmitted time slots in the terminal occur at different times).
In addition, the base instructs the terminals of suitable delays between
incoming and outgoing data bursts, in order to offset different delays in
propagation times between terminals close to and remote from the base sta-
tion. The procedure is basically the same as in GSM (timing advance).
In some situations, a terminal will need to transmit before it has received an
instruction from the base station about what timing advance it should use. To
eliminate the risk of the transmitted data burst covering two time slots at the
base-station receiver, a reduced length data burst is used. This is also similar
to the procedure in the GSM.

84


D-AMPS:TDMA structure

TDMA frame
40 ms

1 2 3 4 5 6

6.67
ms

P Terminal Base
6
6 Data Sync Data SACCH carrier&
Base
ID Data
16 28 122 12 12 122
3 22
3 Reduced length data burst

Data Base & Data
Sync SACCH carrier ID Spare
28 12 130 130 12
12

Sync = Known sequence, used by receiver for frame synchronization
SACCH = Slow Associated Control Channel
CDVVC = Coded Digital Verification Code
RSVD = Reserved for future use
System data rate: 324/0.0067 = 48.6 kbits/s
Net data rate per traffic channel: 260/(0.0067x3) = 13 kb/s

Fig 3.6

To facilitate a subsequent transition to half-rate channels, a TDMA frame is
divided into six time slots instead of three. Initially, two of these are used for
each full-rate channel, which means that three traffic channels are available.
The time slots are handled somewhat differently in the two directions. In the
inward direction, as in the GSM, guard slots are interposed between data
bursts in adjacent time slots, and short intervals are used for ramp-up and
ramp-down of the transmission pulse. These intervals are not required in the
outward direction (TDM instead of TDMA), since the transmitter will not be
pulsed and the data blocks will be packed close together without any
intervening guard slots.
Each time slot carries an information sequence, system signalling (Slow
Associate Control Channel, SACCH) and an ID code giving information about
the base station/operator and carrier.
A synchronization sequence is also needed, which is also used as training
sequence for the channel equalizer. Each time slot contains 260 traffic bits;
since 25 frames per second are transmitted and two time slots per frame are
used for a full-rate channel, the gross data rate for a traffic channel will be
13 kb/s (260 x 25 x 2).
A Fast Associate Control Channel (FACCH), which is needed for fast
signalling during handover, is obtained by “stealing” time frames from the
speech signal (compare GSM).

85


The channel equalization requirements differ greatly between the D-AMPS
system and GSM. A larger equalizer window is specified for the D-AMPS sys-
tem (40 ms), which corresponds to the length of only one symbol period. On
the other hand, the channel equalizer has to be adapted to variations in the
impulse response of the propagation channel during a data burst. If a Viterbi-
type channel equalizer is used, the number of states will be four, since the mo-
dulation type employed is 4-QAM. The complexity of implementation is
roughly the same as in GSM.
Linear Nyquist-filtered modulation is used to minimize the modulation
bandwidth. This complicates the transmitter output amplifier as the signal
envelope is not constant. To reduce the linearity requirement for the amplifier,
a phase shift of π/4 is introduced between each symbol (see figure 3.7a).

α = 0.5
Nyquist characteristic f < 1
(frequency response) |N(f)| = 1 för
fN 2
| N(f)| f > 3
|N(f)| = 0 för
α=0 fN 2
Nyquist
impulse response
|N(f)| = cos2 π f 1 för 1 < f < 3
2fN 4 2 fN 2
hN(t)
α=1
0.5 α = 0.5
α = 0.3 α=0
α=1

f t
fN ts
0 0.5 1 1.5 2 -2 -1 -1 -2

1 1 cos2 (......) called raised cosine
ds = fN = d
ts 2 s ds: symbol data rate

Fig 3.7b

Thus, the chance of the signal envelope reaching the zero level in some sym-
bol transitions is averted. This implies a less stringent requirement for linearity
on the transmitter amplifier, since class B or B/C output stages have strong
nonlinearity at low power level (see figure 2.4). For the Nyquist
characteristic, N(f), a Rised Cosine has been selected, with the value of the α
parameter being 0.35. To ensure matched receiver conditions, the same
selectivity of the transmitter filter and the receiver filter has been introduced:
|N(f)| = |N(f)R(f)|, that is |N(f)| = √ |N(f)|.
The modulation spectrum is thus of the Root Rised Cosine type, with
α = 0.35. In the case of 4-QAM with a bit data rate ≈ 48 kb/s, the Nyquist
frequency, fn≈12 kHz, which corresponds to α 3-dB bandwidth of
2fn ≈ 24 kHz. In theory, there is infinite attenuation of the modulation
spectrum outside a bandwidth of 2(1 + α)fn ≈ 32 kHz. The spectrum
therefore has steep flanks with little power outside the 30 kHz channel slot.

86


Amplifier stage. Relation between input and output levels
Class B/C
Pout

Class A

Pin

Figure 3.8

D-AMPS: p / 4-DQPSK modulation

Modulator for standard 4-QAM
~
s = Re [ s4-QAM - ejkπ/4 . ejω0t ]

D/A T(f)
db
/2 ejω0t

Serial- s Linear
db
transmitter
parallel ~
s4-QAM ejkπ/4 π/2
converter amplifier

db
/2
D/A T(f)

Tb = 1
db

Ts = 2Tb R(f)
ejω0t Receiver
kTs frontend

Channel
equalizer e-jkπ/4 π/2
~
s4-QAM

R(f)
kTs

Nyquist characteristic: N(f) = T(f) . R(f)
(Differential coding omitted)

T(f): Transmitter filter response R(f): Receiver filter response

Fig 3.9

87


The phase shift of π/4 between each symbol is introduced in the output from
the 4-QAM modulator. A compensating phase shift of -π/4 per symbol is
introduced before the channel equalizer, so that the input signal to the channel
equalizer is a standard 4-QAM signal (see figure 3.8).
Differential modulation has been specified so that the receiver in principle can
be realized without the need for a phase-locked local oscillator. This means
that the 4-QAM modulator is equipped with a precoder in the I and Q
channels, the output signals from which correspond to the difference between
the current bit and the preceding bit. For detection, a decoder has been
introduced that gives the inverse function to that of the precoders. However,
because the D-4QAM signal format is unsuitable for the channel equalizer,
transition to D-4QAM takes place after it.
The difference in the data rate from the speech coder (7.95 kb/s) and 13 kb/s
is exploited for channel coding. The principle is the same as in the GSM (see
figure 3.10). In D-AMPS, the bits are also grouped into three classes
according to their sensitivity to transmission errors.

88


Channel coding and interleaving for a full-rate traffic channel

Ia Ib II
Block from
12 65 82 159 bits 50 times/s = 7.95 kb/sIa
speech coder

CRC

12 7 65 82
Block coding for
error detection

Tail

Regrouping 4 6 65 6 3 5 89 bits 82
Adding of tail

Convolutional
coding, R = 1/2

Convolutional
260 bits
coding 2 x 89 = 178 82 (13 kb/s)

Division into two
subblocks of
130 130
130 bits

Interleaving

20
ms TDMA frame (CRC: Cyclic Redundancy Check)

Figure 3.10

The output signal from the speech coder occurs in data bursts of 159 bits
every 20 ms. (Speech coding is discussed in section 3.4.) When the signal is
passed through the channel coder, the number of bits is increased to 260. The
channel coder is adapted to the characteristics of the speech coder such that
only certain sensitive bits (77) are protected by FEC channel coding
(convolution code with R = 1/2 and constraint length 5.) Twelve of these 77
bits are very critical and are therefore also protected by error detection coding

89


(seven parity bits). If this group of 12 bits should be subject to transmission
errors even after FEC, the same measures will be taken as in GSM
(discontinuous reception, i.e. replacing the speech frame with the previous
one). 5 tail bits are added, so that the shift register in the encoder is reset to
zero at the end of each burst.
After channel coding, interleaving takes place over two time slots 20 ms apart.
See figure 3.11. The figure indicates that the total one-way transmission delay
is about 47 ms, which is about the same value as for GSM. The practical value
is around 70 ms, as the digital signal processing will cause some additional
delay. See also figure 3.12.

D-AMPS. Mapping from speech frames to TDMA burst

Block diagonal interleaving
Speech frames
A2 B2 C2 D2 E2 F2
to transmitter
A1 B1 C1 D1 E1 F1

TDMA bursts A2 B2 C2 D2 E2
A1 B1 C1 D1 E1 F1

Speech frames 46.7 A B C D
from recceiver

Fig 3.11

D-AMPS Transmission delay
20 ms

Speech frame n Speech frame n+1

Transmitted n-2 n-1 n n+1
time slots n-1 n n+1 n+2
6.7 ms
Detected speech frame n-1 Detected speech frame n

Transmission delay:
In speech coder 20 ms
Due to interleaving 20 ms
Due to TDMA formatting 6,7 ms
Total delay 46,7 ms
(Delay in speech decoder neglected)

Fig 3.12

The diversity gain from channel coding is much lower than in the GSM at low
terminal speeds, as the D-AMPS system has few support facilities for the
channel coding. Interleaving is insignificant and no frequency hopping is
utilized to break up stationary fading structures. See figure 3.13.

90


Channel coding for D-AMPS. Performance
ber ber
10-1 10-1

Solid line: class II Solid line: class II
Dashed line: class I Dashed line: class I

10-2 10-2

10-3 10-3

10-4 10-4
12 14 16 18 20 22 24 12 14 16 18 20 22 24
Eb/No dB Eb/No dB

Speed: 50 mph Speed: 3 mph

Small coding diversity gain for hand-held terminals Class I and class II refer to figure 3.10
- No effective interleaving
- No frequency hopping

Fig 3.13

During the initial system studies, antenna diversity at the terminals were con-
sidered, but not included. The proposed antenna diversity arrangement at the
terminals was based on the possibility offered by TDMA to listen to the
received signal from both antennas during free time slots occurring before the
time slot to be received. Thus the best antenna could be connected well before
this time slot and therefore the switching transients disappear before the start
of the receive time slot. Performance close to normal selection diversity will
be obtained with only one receiver channel. A small degradation relative to
this case occurs at high terminal speed, as the fading structure could change
somewhat between the channel measurements and the end of the active
receive slot.
It was estimated that the required protection ratio for hand-held terminals was
about 16 dB without, and 11 to 12 dB with, antenna diversity. Even without
it, the protection ratio for the D-AMPS system is slightly better than that of the
existing analog system. There was therefore no obvious gain from employing
antenna diversity so long as the D-AMPS system is forced to use a cell
structure that was entirely determined by the analog system.

3.4 Speech coding
3.4.1 64 kb/s log PCM
Digitization of a analog signal is in principle achieved in two steps: sampling,
in which a time-discrete signal is obtained, and quantization which produces
the wanted digital representation of the input signal. See figure 3.14 and 3.15.
The resolution of the quantization is usually given as the number of bits

91


required to describe each sample of the digital signal. The more bits, the lower
the quantization noise.

PCM speech coder

Encoder Anti-aliasing filter Sampling Quantizer
8 kHz Time-discrete
Analog 64 kb/s
fg=3.5 kHz to digital
speech (8-bit resolution)

Decoder Reconstruction

64 kb/s Digital to time- Time-discrete Analog
discrete fg=3.5 kHz
signal speech

Figure 3.14

Conversion from a time-discrete signal to a digital signal
Time-discrete
signal Digital signal
Quantizer Coder
(High sampling (Code-word sequence)
frequency)

Output signal (quantized)
Binary code Input signal
5 Volt 5 0101
Quantized
4 Linear quantization 4 0100 signal
3 3 0011
Quantization steps
2 2 0010
1 1 0001
Quantization noise
0 0 0000

-1 -1 1001
-2 -2 1011
-3 -3 1011 Example:
Input signal 2.7 V ⇒
-4 -4 1100
output signal 3 V ⇒
-5 -5 1101 Code word 0011
-5 -4 -3 -2 -1 0 1 2 3 4 5
Time discrete input signal Volt

Figure 3.15

According to the sampling theorem, to avoid aliasing distortion it is necessary
to sample a signal that occupies a band of W Hz at least 2W times per second.
See figure 3.16.

92


Sampling theorem
A baseband signal s(t) limited to band W, without loss of information can be
represented by a time-discrete signal s'(t), obtained by sampling s(t) more
than 2W times a second.
If this condition is met, no aliasing distortion will occur. The original signal can be
reconstructed by filtering s'(t) through a rectangular low-pass filter with a cutoff
frequency W. Sampling at
ts = m . To
A s(t) s'(t) B
Time-discrete
W W
signal processing
To < 1
Low-pass filter 2W Reconstruction
(anti-aliasing) filter
A and B
1 W

f

Figure 3.16

Due to practical limitations, somewhat higher sampling rate is used in practice.
A speech signal can, with moderate degradation, be cut off around 3 kHz, and
it is sampled 8000 times a second.
If a typical speech signal (with fairly high peak factor) is processed by perfect,
linear sampler a signal-to-noise ratio S/N is obtained after a psophometric filter
in its output:
S/N = 6n - 4 dB
n is the number of bits per sample, S is the power of the speech signal (of
constant mean power) and N is the power of the generated quantization noise.
When measuring the S/N of a telephone line, a psophometric filter is generally
used to get a better estimate of the subjective quality degradation. This is
discussed in module DT13.
It is assumed above that the range of the sampler matches the top-to-top
variations of the speech signal. This is a critical adjustment. If the sampler
range is too small, overload distortion occurs, if the sampling range is too
large, all the sampling steps are not used and thus the effective resolution
corresponds to a less number of bits.
A speech signal in the telephone network has a short term mean power that
typically varies over a range of 30 to 40 dB. It is necessary that the sampler
adapts to these variations. This is called companding. The standard speech
codec for telephony is adapted by replacing the linear sampling algorithm
with a log function, i.e. very small sampling steps are used for low input
signals, and the steps gradually increased for larger input amplitudes.
Adequate speech quality in the fixed telephone network roughly corresponds
to S/N = 30 dB. That means that somewhat higher quality is required for the
speech codec. With a speech signal of constant power and using a linear
sampler (linear PCM), 6 bits resolution is necessary. This corresponds to a
data rate of 8000x6 = 48 kb/s.
With a real speech signal with varying power and using a log type of sampler
it is found that 8 bits resolution is necessary (log PCM), i.e. the data rate
becomes 64 kb/s, see figure 3.17.

93


Signal-to-noise ratio for PCM speech codec
(S/N)psoph

60

M
C
rP
50

ea
in
tl
i
-b
11
8-bit log-PCM
40

CCITT

30

M
C
rP
ea
in
tl

20
bi
8-

Speech
level, dBm0
10

-50 -40 -30 -20 -10 0
(definition of dBm0 in module S2, section 2.1)
psoph: measured with psophometric filter, see module DT13, figure 1.7

Figure 3.17

The figure also indicates the minimum requirements for the S/N for speech
codecs to be used in the fixed telephone network, set up by CCITT. The
requirements are as fairly high, as there might be several conversions between
analog and digital format in a long distance connection.
The procedure of sampling and quantization is is from historic reasons called
with the peculiar name PCM: Pulse Code Modulation. The standard speech
coder used in the fixed telephone network is therefore called 64 kb/s (log)
PCM.
There is also a secondary standard intended for band-width critical transmis-
sion situations (i.e. ocean cables) called 32 kb/s ADPCM (Adaptive Differen-
tial PCM). This codec standard has been used for cordless telephone (section
5). The speech quality is marginally less than for 64kb/s PCM, and the codec
is somewhat more complicated. An advantage of these two types of codecs is
very low delay.

94


3.4.2 Speech codecs for GSM and D-AMPS
Background
Below is given a short review of basic concepts and relations which are
directly related to the speech codecs used at the original GSM and D-AMPS
systems. Due to the stiff requirements on good frequency economy for
cellular systems, speech codecs with much lower data rate than 64 kb/s must
be used, even at the expence of somewhat lower speech quality than the 64
kb/s PCM codec, increased delay and much higher complexity.
These advanced speech codecs belong to the class of hybrid codecs, which
use a model for the speech generation process, is based on an excitation
which is filtered in a vocal tract filter. See figure 3.18. The function of the
speech encoder is to estimate (analyze) the excitation and the transfer function
of the vocal tract filter. This information is quantized and transmitted to the
speech decoder, which can then regenerate (synthesize) the speech signal. In
most cases, the parameters of the vocal tract filter are estimated 50 times per
second, and the excitation 200 times per second. The information from the
speech encoder is delivered as packets to the channel encoder every 20
millisecond. (The value 20 ms corresponds to the time over which the vocal
tract filter has a more or less constant setting).

Speech generation model (speaker or speech decoder)
excitation
Excitation e(t) Vocal tract Speech spectrum: S(f)= E(f) . H(f)
generator (H(f))
E(f)

a) Voiced sounds; quasi-periodic excitation (line spectrum)
b) Un-voiced sounds; noise-like excitation (continuous spectrum)

a) IE(f)I2 IS(f)I2 formants
line
spectrum

f f
4kHz fundamental 4kHz
frequency
b)

IE(f)I2 IS(f)I2

f f
4kHz 4kHz

Figure 3.18

95


For voiced sounds, the excitation is determined by the vocal cord. The transfer
function of the vocal tract filter is determined by the settings mainly of the
mouth and tongue. There is a considerable redundancy in the speech signal.
Advanced speech encodecs eliminate a large part of the redundancy at the
processing and put it back at the decoder. The result is that the data rate can
be reduced without excessive degradation of the speech quality.
The vocal tract filter is generally characterized by 10 - 12 LPC coefficients
(LPC: Linear Predictive Coding). These coefficients are determined by the
mutual correlation between 10 - 20 consecutive speech samples (sampling rate
8000 /sec). See figure 3.19. The transfer functions of the vocal tract filter and
the inverse filter can be computed from the LPC coefficients.

LPC-analysis determines the transfer function of the
speech tube filter (and the inverse filter)

Due to the filtering of the excitation in the vocal tract, the speech samples
are correlated over 1-2 ms

The inverse vocal tract filter eliminates this redundancy

The corrrelation between near-by samples determines the
correlation function (the LPC coefficients)

The vocal tract filter and the inverse filter are defined by the LPC coefficients

speech Inverse vocal excitation Excitation
tract filter analysis

LPC excitation
analysis
(coded)
speech encoder
speech decoder

speech excitation
Vocal Excitation
tract filter synthesis
(formants)

Figure 3.19

The amplitude function of the vocal tract filter comprises a number of pass
bands, called formants. Their mid frequencies and bandwidths must be
accurately reproduced by the decoder using the LPC coefficients. This is
neccessary for good understanding of the speech content. At a hybrid
encoder, the redundancy introduced by the vocal tract filter is eliminated by
the inverse vocal tract filter, which means that the original excitation of the
speaker is reconstituted, if the input speech is fed through this filter. The
encoder must then estimate the parameters of a suitable model of the excita-
tion (very difficult) and sends over these parameters and the LPC coefficients
to the decoder.
The subjective speech quality and the possibility to recognize the speaker
mostly depends on an accurate estimate of the vocal cord information at

96


voiced signals (vocals and voiced consonants). This is the major part of the in-
formation in the excitation. For voiced signals, the signal from the vocal cord
is nearly periodic. This means that the excitation is very redundant. A large
part of this redundancy can be eliminated by filtering of the excitation in in
the Long Term Prediction (LTP) filter. The output signal from LTP-filter is
called the innovation. The LTP is performed in principle by taking the
difference between the excitation waveform, corresponding to one vocal cord
period, and the previous one. Therefore, the vocal cord frequency must be
estimated and used to set the parameters of both the LTP-filter in the encoder
and the inverse LTP-filter in the decoder.
In figure 3.20 there are shown typical wave forms related to the signal proces-
sing of a vocal:
a. the input speech signal to the encoder
b. the excitation after the STP (inverse) filter
c. the innovation after the LTP filter
The figure b (corresponding to the original excitation formed by the speaker)
has a high degree of periodicity, which largely is eliminated in figure c.

97


Wave forms for vocal, before and after filtering

Inverse vocal
tract filter
Excitation Long-Term Innovation
Speech signal LPC-filter
(Short-Term predictor
predictor) (LTP)

LPC-coefficients Vocal cord frequency

1. Speech
signal

2. Excitation

3. Innovation

Figure 3.20

The GSM speech codec
At the first generation of GSM speech codec the Regular Pulse model is used
to estimate the innovation. This means that the innovation waveform during
each 5 ms interval is represented by a series of equidistant pulses ("regular").
Part of the matching of the model to the waveform consists of a choice
between three possible starting times (”phases”) of the first pulse. Then the
amplitude and polarity of the pulses are selected for optimum match. This sig-
nal processing of the speech signal lies behind the full name of the speech
codec: ”Regular Pulse Exited LPC with LTP”.
The signal processing of the excitation also comprises an automatic gain
control, which compensates for the large level differences between different
speech sounds (phonemes). Vocals are much stronger than consonants and in
addition these is a large difference in average power between speech signals.
In the decoder, an inverse gain control is introduced, so that the original level
differences are restored. (This corresponds to the compander function at log
PCM).

98


Speech codec with STP (LPC) and LTP (GSM)

Speech
signal STP excitation innovation
vocal LTP Innovation
tract filter analysis

Power
measure-
ments STP (LPC) LTP-
analysis analys innovation
(coded)
Speech Speech LPC vocal cord
encoder power coefficients information 9.4kb/s
Speech
decoder
STP excitation Innovation
Vocal LTP analysis
tract filter

Voiced sounds have quasi-periodic excitation
This redundancy is eliminated by LTP (Long Term Prediction)

Speech codec generation 1 for GSM: 260 bits 50 times/s 13 kb/s

Figure 3.21

Figure 3.21 contains a simplified block diagram of the speech codec. Of the
total data rate 13 kb/s from the encoder, 9.4 kb/s is used for the transmission
of the innovation.
In order to reduce the quality degradation due to error bursts, the following
arrangements have been added to the speech decoder:
1. The maximum rate of change of the volume control at the decoder has been
limited. Without this facility, transmission errors on this side channel, could
result in subjectively very disturbing effects.
2. Transmission errors, which hit the low order LPC coefficients also result in
a strong degradation of the speech quality. A few additional bit positions in
a 20 ms speech frame are also very critical. If a 20 ms frame contains so
many errors that error correction is not possible of these bits (class lA bits in
figure 2.45a), the current frame is replaced by the previous one (frame
erasurc). This is called discontinuous reception.

The D-AMPS speech codec
The speech codec for D-AMPS is about three times more complex than the
speech coder for GSM. The speech quality is more or less the same for the 8
kb/s D-AMPS codec and the 13 kb/s GSM codec. The main design difference
between the two encoders is the estimation of the excitation. The other parts of
the encoder: the automatic gain control and the calculatian of the LPC and
LTP coefficients are in principle the same.
Also the D-AMPS speech encoder contains in principle a LTP block for gene-
ration of the innovation. (The practical implementation combines the excita-
tion and the LTP functions through a more complicated code book arrange-
ment, which directly defines excitation vectors with optimum periodicity. The
arrangement is called ”adaptive code book”, as the periodicity of the vectors
is determined by the estimated pitch frequency.) A much more complicated
arrangement than at GSM is used to estimate the innovation.

99


Different names for the procedure are: Vector Quantizing, Code Excited Lin-
ear Prediction (CELP) and analysis through synthesis.
The model for the innovation comprises a very large number of digital wave
forms (called codes or vectors) stored in a code book. Identical code books
are used at the encoder and the decoder. The dominating problem in the
encoder is to select the wave form that gives the best representation of the
innovation waveform over a 5 ms interval. The procedure for this selection is
called analysis through synthesis.
Analysis through synthesis has certain similarities with the human speech ge-
neration process. See figure 3.22. An important part of the detailed speech ge-
neration, especially the intonation, is based on a feed back loop: the speaker
listens to the generated speech and makes corrections until he is satisfied with
the quality of the generated speech. The procedure can be called synthesis
through analysis.

Synthesis through analysis; Speech generation process

Brain Ear

Speech analysis

Speech synthesis

Excitation Speech
generator tube
speech

Figure 3.22

Analysis through synthesis is based on a codec arrangement in which the
encoder contains an exact copy of the decoder. In principle, the estimation of
the excitation (or innovation) is based on a procedure according to which all
codes in the code book are tried and the resulting errors between the incoming
speech segments and the corresponding output from the decoder is stored.
The best estimate corresponds to the code that gives the smallest error. The
number of this code is transmitted to the encoder on the receive side.
The estimation is improved by measuring the error after an adaptive weighting
filter controlled by the estimated LPC coefficients. The weighting filter brings
out those parts of the spectrum for which the level of the short term speech
spectrum is low. The reason is the masking effects of the ear, that is, a higher
noise level can be tolerated in frequency slots with strong speech components.

100


Analysis through synthesis; Speech encoder
Adjust excitation and LTP parameters for best match (minimum error)

Speech encoder

Optimizer e weighting STP
(for minimum e)
error signal filter analysis

synthesized
speech signal Speech
Excitation
generator STP input
LTP
(code filter
generator) excita- inno-
tion vation
Power
closed measure-
50 times/s ment
code vocal cord LPC-
number information coeff. Speech
power
Excitation STP Speech
generator LTP Speech
(code
generator) tube filter

Speech decoder
Example:
Speech codec for D-AMPS (gen 1): 159 bits 50 times/s ⇒ Data rate 7.95 kb/s

Figure 3.23

To simplify and speed up the comparison process, the code vectors are for-
med by adding together sub vectors contained in two code books. Therefore,
the name of the speech coder is: Vector sum exited LPC (with LTP). Another
name of the codec is CELP (Code Exited LPC).
A similar type of codec is used at the first generation of the PDC system.

3.5 Comparison of GSM and D-AMPS
The GSM and the D-AMPS system need to satisfy roughly the same communi-
cation needs. They are both cellular mobile telephone systems having
extensive coverage, serving both rural areas and major population centers.
Because of the need to coexist with the established analogue mobile telephone
system in the same or adjacent bands, both systems must employ frequency
duplex. Both the GSM and the D-AMPS system also have to meet the same
requirements in terms of transmission delay in speech transmission. Speech is
the dominating application, although the need to interface with ISDN was
more of a consideration at the design of the GSM.
However, the background conditions for the systems are otherwise quite diffe-
rent, as a result of which there are essential differences in numerous system
parameters. The main differences are given in figure 3.24.

101


Comparison of the GSM and D-AMPS system, A

GSM
New frequency band allows wide freedom in selection of channel spacing
Total spectrum efficiency essential (combination of bandwidth requirement
and geographical packing density)

D-AMPS
Channel spacing is the same as in the analog system
Need to minimize bandwidth requirement per speech channel
(geographical packing density determined by the existing analog system)
Basic specification finalized two years later than for GSM
(more advanced technological solutions)

Figure 3.24

The different requirements and design criteria have also resulted in essential
differences between the systems in the TDMA structure and signal processing
(see figures 3.25-3.27).

Comparison of the GSM and D-AMPS system, B

a. GSM
Maximum system data rate and length of time slots determined by need
for reasonable complexity of channel equalizers.
• No adaptation during a TDMA time slot, i.e. short data bursts required
(0.58 ms).
• Equalizer able to cope with time dispersion having a maximum length
of four symbol periods (16 ms).
Comprehensive signal processing justified to reduce cluster size
(channel coding, interleaving depth of 8, frequency hopping).

b. D-AMPS
Maximum system data rate determined by channel spacing
(linear modulation and 30-kHz channels give approx. 50 kb/s).
Necessary with small reduction in capacity due to guard slots and
synchronization sequence. Therefore need for long time slots (6.7 ms).
Cluster size determined by existing cell structure. Acceptable with protection
ratio only moderately better than in the analog system
(reduced channel coding, small interleaving depth, no frequency hopping).

Figure 3.25

102


Comparison of the GSM and D-AMPS system, C

TDMA structure: GSM TDMA structure: D-AMPS

1 2 3 4 5 6 7 8 1 2 3 4 5 6
0.6 ms 6.7 ms
4.6 ms 40 ms
= 150 bits per time-slot = 320 bits per time-slot

Frame time determined by terminals’ Long time-slot period owing to:
needing to be tuned to three frequencies: • long time per radio symbol
transmit, receive and listening (= 5 ms). • low relative TDMA overhead requires
The time-slot shall be short enough many bits per time-slot.
(= 0.5 ms) to permit fixed equalizer setting.
2 x 3 = 6 time-slots to allow subsequent
transition to half-rate speech coders.

Time-slot are so long that channel equalizers
must be adapted during each time-slot,
although equalizer window of only one
symbol period is needed.

Figure 3.26

Comparison of the GSM and D-AMPS system D

Speech quality

Analog
(64 kb/s PCM)

Acceptable
quality
GSM

PS
D-AM

9 16 18
10 20 30 40 C dB
I
Carrier to interference ratio (dB)

Figure 3.27

103


Comparison of the GSM and D-AMPS system, E

GSM D-AMPS
Channel spacing 200 kHz 30 kHz
Modulation GMSK QAM (π/4-DQPSK)
System data rate 271 kb/s 48.6 kb/s
TDMA frame 4.6 ms 40 ms
Time slots 8 x 0.57 ms 6 x 6.67 ms
Bit rate, full-rate speech channel, net 13 kb/s 7.95 kb/s
(with channel coding), gross 22 kb/s 13 kb/s
Rate (channel coding) 0.57 0.68
Bandwidth expansion* 200/(8 x 13)=1.92 30/(3 x 8)=1.25
Interleaving depth (full-rate speech codec) 8 2
Frequency hopping Possible No

* Channel spacing per speech channel divided by the data rate from the speech coder

Figure 3.28

A summary of the most important radio transmission parameters for the GSM
and D-AMPS system is given in Figure 3.28.

Comparison of the GSM and D-AMPS system, F
Analog FM GSM D-AMPS
(NMT 450) without with
antenna diversity
Number of
speech channels/25 MHz 1000 1000 2500 2500
Protection ratio 18 dB 9 dB 16 dB 11dB
Cluster size 3x7 3x3 3 x7 3x4
Speech channels/cell 47 111 119 208
Capacity improvement 1 2,4 2,5 4,4
factor

Figure 3.29

A summary of the principal radio network parameters is given in figure 3.29.
The D-AMPS system requires a much smaller frequency slot per speech
channel. On the other hand, the basic version, without antenna diversity at the
terminals, requires a considerably higher protection ratio than GSM. This in
itself is not significant, so long as the cluster size is determined by the
requirement for integration with the AMPS. The total spectrum efficiency
(speech channels per cell per MHz) is much the same as for GSM. However, a
stand-alone D-AMPS system with antenna diversity (cluster size of 3 x 4) has
significantly better spectrum efficiency than the GSM. (For the corresponding
Japanese system, antenna diversity has been specified also at the terminals).
A general comparison of the background conditions for the GSM and the
D-AMPS system is shown in figure 3.30.

104


Comparison of the GSM and D-AMPS system, summary
GSM D-AMPS
• New frequency band allocated (although • The D-AMPS system is introduced gradually in a
a part of the band is for a limited time frequency band that is already in use and within the
used by analog MTS) existing cell structure for analog MTS (AMPS).
Channel width therefore predetermined at 30 kHz (or
• No strict requirements on narrow submultiple).
bandwidth per speech or traffic channel
• The analog AMPS is to be retained for a long time
• Width of radio channel not critical in rural areas. Principal reason for the D-AMPS:
(although inappropriate with very wide improved spectrum efficiency in urban areas
radio channels)
• TDMA not really justified with fewer than three
• Technical risk too high for introduction of speech channels per radio channel (carrier)
modulation with varying signal envelope
• FDMA cannot meet requirement for adjacent
channel selectivity

• Principal problem: How to accommodate TDMA with
three time slots in a 30-kHz radio channel.

The GSM achieves excellent The D-AMPS system achieves excellent spectrum
geographical spectrum efficiency efficiency in the frequency domain but poor
(small cluster size) geographical spectrum efficiency (unless antenna
diversity introduced).

Figure 3.30

105


4 PDC, generation 1
The Personal Digital Cellular system (PDC) system (previously called Japan
Digital Cellular, JDC or Pacific Digital Network, PDN) was specified by the
Japanese telecommunications authority after extensive technological
development work and studies. The largest operator is NTT Docomo, with
57% of the market 1999. The total number of PDC subscribers in Japan was
45 milj. in middle 1999.
PDC is similar to the D-AMPS system but in some respects more advanced
and, consequently, provides higher spectrum efficiency. The main radio char-
acteristics are summarized in fig 4.1

PDC: Personal Digital cellular
a. Key radio transmission parameters
Outward 810 - 830 MHz Duplex separation
Inward 940 - 960 MHz 130 MHz
Channel width 25 kHz
Width per one-way speech channel 8.3 kHz
Modulation 4 QAM (π/4 -DPSK)
System/data rate 42 kb/s
Time per radio symbol ≈ 50 µs
Time slots inTDMA-frame 3 x 6.67 ms
Data rate for traffic channel, incl. 11.2 kb/s
channel coding
(C/I)coch. for Pb=10-2 to speech decoder 13 dB
Diversity against fast fading channel coding
antenna diversity
b.- TDMA frame structure (traffic channel)
6.67 ms
280 bits
Inward
R P Traffic channel Sync. SACCH Traffic channel guard
4 2 (FACCH) 112 20 24 (FACCH) 112 slot

Outward
R P Traffic channel Sync. SACCH Traffic channel
4 2 (FACCH) 112 20 30 (FACCH) 112

R = Ramp time P = Preamble

c. Traffic and signaling channels
TCH
Traffic (SACCH)
Channels (FACCH)

DCCH (Dedicated)
Control UPCH (User Packet Channel)
Channels BCCH
CCCH PCH (paging)
SCCH (signalling)

Figure 4.1
106


The PDC system has much the same TDMA structure as D-AMPS, with a
TDMA arrangement allowing three speech channels per carrier. As in
D-AMPS and GSM, free time between transmit and receive time slots is used
by the terminals to listen to carriers from adjacent cells (Mobile Assisted
Handover). In principle, the same arrangements for FACCH and SACCH are
used as in the D-AMPS system.
The system also includes provision for half-rate traffic channels on and half-
rate speech codecs are used. Decomo has recently introduced a slow packet
data service (i-Mode) to which around 250 000 subscribers were connected in
middle 1999. One time slot is used, which gives 9.6 kb/s data rate. Future
plans are to add a 20 kb/s packet service to PDC. (On the other hand more
advanced data transmission similar to GPRS/EDGE are not planned. The next
step is a full G3 system based on W-CDMA)
An important difference between the PDC and the D-AMPS systems is that
advanced antenna diversity (Post Detection Diversity) also at the terminals
constitutes an integral part of the system solution for PDC. This has influenced
the design of the system in two respects:
a) Antenna diversity gives moderate suppression of the time dispersion over
the radio channel. This together with suitable antenna arrangements on the
base-station side has proved to be adequate to cope with the effect of time
dispersion without the need to introduce channel equalization. It is also
likely that the propagation conditions are less extreme in Japan than in the
USA.
b) Antenna diversity is as effective as channel coding with the supporting
arrangements used at GSM to bridge over the fading dips caused by multi
path propagation. The required protection ratio is therefore considerably
lower than for D-AMPS.

Channel coding in the PDC uses a lower rate than the D-AMPS. This
compensates for the need to employ a lower system data rate, owing to the
narrower channel spacing (25 kHz in Japan as against 30 kHz in the USA).
Otherwise, the speech coders, the speech coder data rate and the modulation
type are largely equivalent to those in the D-AMPS. The data rate from the
channel coder is 11.2 kb/s in PDC as against 13 kb/s in D-AMPS.
The spectrum efficiency of the PDC is better than in the D-AMPS (without
antenna diversity) partly because of the narrower channel spacing and partly
because of the lower protection ratio (≈ 13 dB). This allows a cluster size of
3 x 4. NTT has also developed an advanced base-station antenna, which
allows the tilt of the antenna lobe below the horizontal to be adjusted
individually for each base-station site. This reduces the average co-channel
interference and can also reduce strong reflections from remote objects, i.e.
reduce the delay spread. A small drawback is stronger reflexes from near-by
objects, which could increase the effective side-lobe level.

107


5. Cordless Telephone
5.1 Overview
Cellular systems with wide area coverage have been discussed in sections 2 to
4. Another type of system covers only local areas, and at least in the
beginning were designed for very small handheld terminals and micro/pico
cells only. There were no requirements for operation with large cells (macro
cells). This meant:
- reduced requirements on transmit power,
- very small time dispersion of the propagation channel, which allowed fairly
high data rate without any need for equalization and
- very good frequency economy due to extensive frequency reuse.

The basic system cannot handle roamers between local areas, even if several
local coverage areas can be interconnected.
These characteristics and limitations result in reduced costs both for the
terminals and the core network, which together with much improved
frequency economy would make these systems an attractive alternative for
local applications, such as wireless PABX and radio local loop. The original
application for cordless telephone was wireless extension of the normal
telephone line (radio local loop) for domestic users and small enterprises.
However, the extensive penetration and reduced cost for wide-area cellular
systems have made the market for cordless systems less than originally fore-
seen. Additional applications are as complement to cellular systems, using
double-mode terminals, and for point to multi-point applications (fixed radio
systems) with moderate requirements on data rates. These additional
applications for fixed radio access mainly apply to DECT. See figur 5.1 to 5.3.
Fixed applications with directive antennas (nearly free sight) several meters
above ground at both ends can have a range up to several kilometers, both
due to improved link budget and reduced time dispersion.

WLL - For Urban and Suburban areas

PSTN

Base
Station Base
Controller Stations

Fixed
Access
Unit

WLL: Wireless Local Loop

Fig 5.1

108


Evolution to cordless
in the home/neighbourhood To fixed
public
network
RNC
DAN

Coverage
Enhancement
Unit
DAN: Dect Access Node
RNC: Radio Node Controller

Fig 5.2

Evolution to cordless in the home
and in small office To fixed
public
network
RNC
DAN
WRSs

WRS
with
Intercom

WRS: Wireless Relay Station

Fig 5.3

The development of cordless telephone has gone through several steps. To
start with, several types of extremely simple analog systems were introduced
as a wireless extension of fixed domestic telephones (CT-1).
The first system based on digital speech transmission was CT-2, which
originated in England. The market for CT2 was fairly small, and it found little
use outside of UK. It was based on FDMA/TDD and operates in the 900 MHz
band. In addition to beeing a replacement for CT-1 systems, typical
applications were to connect cordless terminals to Tele Points and as Wireless
PABX. Telepoints are small base stations with a range of 30 to 50 m which are
located on walls or on lamp posts in business centers, railway stations, airports
and the like. The terminals can access the public telephone network through

109


these telepoints, but generally only for calls made by the terminals. A Wireless
PABX is an office exchange to which cordless terminals are connected.
The radio access for CT-2 was based on time duplex and 100 kHz radio
channels for one two-way speech channel per carrier. See Figure 5.4, which
summarizes the radio parameters. The same arrangement for antenna diversity
is used as for DECT, see section 5.2.
Figure 5.5 describes the time division duplex (TDD) structure. In the figure
are shown two types of frames: one is used mainly for time synchronization in
conjunction with the initial setting-up of calls from terminals. The other type
of frame is used during calls. 64 bits is transmitted in each direction during
each 2 ms frame. This corresponds to a 32 kb/s traffic channel.

Digital Cordless Telephone CT-2

Frequency range: 864 - 868 MHz
FDMA/TDD
Channel spacing: 100 kHz
Speech coder: 32 kb/s ADPCM
Base station: Antenna diversity for both directions
No need for echo control

Figure 5.4

CT-2
Signalling frame
base terminal terminal base
G D Sync. D G D Sync. D G
Bits 4.5 16 34 1 5.5 1 34 16 4.5

Time duplex frame
(2 ms)
Traffic frame
G D User traffic D G D User traffic D G
4.5 2 64 3.5 64 4.5

G: Guard slot
D: System signalling
User traffic: 64 bits/2 ms ⇒ 32 kb/s

Figure 5.5
A more advanced system is DECT (Digital Enhanced Cordless
Telecommunications - original name Digital European Cordless Telephone).
The air interface is based on FDMA/TDMA/TDD. The overall system concept
was verified in 1988 through a 900 MHz test system, which was set up in
Sweden. Based on the results from this test system, ETSI developed the

110


specification for a Pan-European system operating in the 1800 MHz band. The
specification was adopted as an official EC standard in 1992. A wide range of
different interface specifications have been developed for DECT, such as for
ISDN services, emulation of GSM signalling and speech processing, different
types af data application, incl. point-to-multi point fixed networks.
The basic DECT standard covers 1880-1937 MHz, but frequency allocations
in different regions are typically 20 MHz. In Europe, the band 1880-1900
MHz is used. In 20 MHz, 10 radio channels can be placed. The terminals can
in principle scan over these10 radio channels, and each channel comprises 12
two-way speech channels. Seemless handover is possible, between time slots
and between frequency slots.
The key radio parameters are listed in figure 5.6. Figure 5.7 describes the
TDMA/ TDD structure used for traffic. The TDMA frame can be structured so
that several time slots can be allocated to users which need wide-bandwidth
(“multi slot”). Asymmetric traffic can be handled, as a connection can be
allocated different number of slots in the inward and the outward directions.
With multi-slot, the maximum user data rate is 552 kb/s.
A recent addition to the set of specification is similar to EDGE, i.e. there is an
option to use to use high-level modulation (8PSK) during the information part
of a TDMA slot. In combination with multi-slot this makes it possible to offer
user data rates up to 2 Mb/s.

DECT Digital Enhanced Cordless Telephone (DECT)
European (ETSI) standard 1991
Frequency band: 1,880 - 1,900 MHz
Channel spacing: 1.7 MHz
System data rate: 1.15 Mb/s
Modulation: GFSK (BT=0.5)
TDMA-time duplex: TDMA frame 10 ms
2x12 time slots
Transmitter power: mean 10 mW
peak 250 mW
No FEC channel coding of user traffic
No channel equalization
Speech coder: 32 kb/s ADPCM
Antenna diversity at base station for
both inward and outward directions.
Echo suppression required in some network configurations
ISDN compatible

Figure 5.6

111


TDMA/TDD frame
(10 ms)


1 2 3 4 5 6 7 8 10 11 12 1 2 3 4 5 6 7 8 10 11 12

Sync. I Guard
C slot 64
bits 1 6 320
C channel: system signalling (including CRC error detection)
I channel: user information (320 bits/10 ms 32 kb/s)

Figure 5.7

DECT and GSM networks can be combined by means of double-mode termin-
als and interface arrangements (gateways) between the corresponding core
networks. GSM gives wide-area coverage. However, when a subscriber is
within range of one of the local DECT networks, a switch-over to DECT
results in lower call charges and improved frequency economy. Also special
subscription conditions often apply to radio-PABX networks. A dual mode
terminal will be only little larger than a GSM terminal. It has been discussed to
introduce a similar double-mode arrangement for domestic applications,
giving the user the advantage that there will be no radio charges when using
the radio terminal in his home.
(Similar cost advantages might be possible using GSM only, if suitable pico-
base stations for PABX applications become available. Large organizations
could make an agreement with a GSM operator to install their own pico cell
GSM network at the premises.)
DECT is a bearer service for radio access. It can cooperate with many diffe-
rent fixed networks, through different gateways at the interface. It can service
many types of terminals. The maximum user rate has been extended by multi-
slot techniques in combination with more bandwidth efficient modulation.
One offered service is (2B+D) ISDN (B: 64 kb/s traffic channel, D: 16 kb/s
signalling), another is packet transmission (DPRS, DECT Packet Radio Ser-
vice) for wireless local area networks (W-LAN) with moderate user rates and
Internet access. The number of DECT terminals was 45 milj. at the end of
1999.
Another digital cordless system is the Personal Handy Phone System (PHS)
in Japan. See section 5.3. In 1997 the PHS networks in Japan had about 7
milj. subscribers. PHS has some similarities to DECT, but more limited
capabilities, as the goal was to establish a low-cost system with extremely
small portable phones and with speech as the dominating service.

5.2 DECT
Speech coding, echo control

The ADPCM 32 kb/s (G.726) speech coding standard is used, as the change-
over to micro and pico cells has resulted in such a large improvement in
frequency economy that the resulting larger bandwidth per speech channel is

112


acceptable. The advantages with this speech codec are improved speech
quality, lower cost and lower power drain in comparison with the low-rate
codecs for generation-two cellular systems.
Compared to CT2, the larger transmission delay at DECT due to the TDMA
formationg is a drawback, when DECT is connected to the public telephone
network. In this case echo control is generally needed to suppress echos.

Network structure

A typical DECT-network, connected to a PABX, consists of a number of base
stations, which together form a pico cell structure with fairly large overlap
regions between nearby cells. The placement of the base stations is not critical
besides the requirement for complete coverage of the service area. See figure
5.8.

Cordless Telephone: Radio - PABX och Telepoints

Handover
PS PS PS
m
00m -30
5-1 15
CH 1-12
CH 1-12 CH 1-12 CH 1-12

Telepoint CH 1-12
FSn
FS1 FS FS
FSn FS
FS1
FS
FS RE
RE
Small office

Radio link exchange
Residental

Wireles

PABX Local Network

FS
FS

Public telephone network

FS: Fixed Station PS: Portable Station RE: Radio link Exchange

Figure 5.8

113


Diversity, dynamic channel allocation

The radio part of the DECT system differs from the GSM in two essential
respects: the diversity arrangement against multipath fading, and the
decentralized dynamic channel allocation.
The DECT system combines TDMA with TDD, i.e. the same radio channel is
used for transmission in both directions. This means that the multi-path fading
has precisely the same time and spatial structure in the up and down
directions. The antenna diversity at the base station receivers is therefore
effective also in the outward direction. The same diversity antenna is used for
the two time slots in a frame, which constitute a two-way traffic channel. The
prerequisite is that the time between the two time slots is short enough so that
fading conditions are unchanged. (Time difference much smaller than the
correlation time of the propagation channel.)
The antenna diversity in the DECT system provides roughly the same
diversity gain as what is achieved through channel coding, interleaving and
frequency hopping in the GSM system. Therefore, the speech channels are not
protected by FEC. The main reason, why antenna diversity is better than FEC,
is that channel coding in this case gives only a small improvement, even if
supported by interleaving. The reason is that frequency hopping cannot be
utilized. As most terminals are quasi-stationary, the error bursts during fading
dips are therefore too long to be handled by interleaving only.
Another advantage of antenna diversity is that the time dispersion is reduced
to some extent. The reason is that the time dispersion is most prominent when
the dominant propagation path fades down during fading dips. Then weak
secondary paths with larger delays will be of larger relative importance, and
the delay spread will be increased.
The combination of the antenna diversity and the small time dispersion in
micro and pico cells allows modulation bandwidths of about 2 MHz without
any need for equalization. DECT can tolerate delay spread up to 200 nS.
However, the high system data rate and lack of equalization make DECT
unsuitable for macro cell networks, with mobile terminals.
The other important feature of DECT is decentralized dynamic channel
allocation. Each base station determines together with the involved terminal a
suitable channel (time slot) for the call. The decision is based on quality data
for the available time slots, which have been measured and stored by both the
base station and the terminal A duplex channel is chosen that provides an ac-
ceptable C/I (and C/N) for both transmission directions. It might happen that a
call in progress suffers strong cochannel interference. If that should happen,
the call is switched to another more suitable time slot, using the same
procedure as for the original channel selection.
This dynamic channel allocation gives a large improvement of the frequency
efficiency. The explanation is that with traditional fixed frequency planning it
is necessary that the reuse distance is large enough to handle the worst
propagation conditions with respect to C and I, considering antenna
separations and shadowing. In contrast, when dynamic channel allocation is
employed, it is possible to utilize the fact that the propagation geometries for
C and I and also the shadow effects are considerably more favourable for the
average case than for the worst case. The result is a considerable reduction of
the average reuse distance. It might even happen that the same time slot can
be used for connections in adjacent cells. See figure 5.9.

114


Dynamic channel allocation in the DECT system

T Ch1 T
B B B Ch1
Ch2
T
T Ch2
Ch1: Channel 1 B B B Ch1
Ch2: Channel 2
Ch1 T
T

The reuse distance can be shorter when the distance between the base (B) and the
terminal (T) is short and shadow conditions are favourable.
T and B each measure the C/I ratio of available channels

Figure 5.9

A further important advantage of dynamic adaptive channel allocation is the
much simplified installation planning, since there is no need for traditional
cell-frequency planning with fixed allocation of channels to different cells.
Several studies have been made of the possibilities to extend the range or
coverage of DECT, which is seriously limited by time dispersion due to high
symbol data rate and lack or equalization. The range extension mainly applies
to applications with fixed terminals. The range can be extended through
means which gives better tolerance against time dispersion (diversity also at
the terminal side) and reduced time dispersion (better propagation conditions
through better placed directive antennas).
As is shown in figure 5.10, extreme values of the normalized rms value of the
time dispersion only occur during deep fading dips. As mentioned above,
these can be suppressed by diversity, which bridge over the dips. See figure
5.11 which shows the improvements with selection diversity (requires two re-
ceiver channels) and with preamble diversity (the best antenna is selected by
channel measurements on a preamble just before the start of the information
part of the data burst). Preamble diversity gives somewhat reduced
performance as the propagation channel could vary during the duration of a
TDMA slot.

115


DECT: Relation between time structure of fading and delay spread
Fast fading average C/N0 = 83.7 dB.Hz

C/N0 [dB.Hz] 90 fd = 5Hz
80

70

60
Time (a)
Normalized delay spread τrmsRs =0.15
100
τrmsRs

101

102
Time (b)
τrms: Delay spread
Rs : System data rate

Figure 5.10

Diversity performance for different delay spreads

40
Required Eb/No to obtain 1%

35
failure rate (dB)

30

25

No diversity
20
Preamble diversity
Selection diversity
15
0 50 100 150 200 250 300 ns
τrms

Figure 5.11

Typical values of the delay spread for different situations is indicated in
figure 5.12.

116


Delay spread characteristics for DECT

Typical rms delay 90th percentile rms
Environment type
spread (ns) delay spread (ns)

Sports halls, exhibition centres 15-50 40-150

Open environments, railway 40-140 105-400
stations, airports

Underground, underground 80 120-140
streets, corridors

Parking garage 18-35 35-50

Urban street LOS 20-70 30-105

Urban street NLOS 30-60 60-275

Town squares 50-100 105-150

Suburban 80-160 150-330

Figure 5.12

5.3 PHS

PHS uses a TDMA frame with only 4 time slots, three for traffic and one for
signalling. As a result, the system data rate is three times lower than for DECT,
which makes it more tolerant to time dispersion, but only allows low rate data
services, as the multi-slot capabilities are limited. A more advanced (coherent)
detector is used than in GSM, which gives somewhat better receiver sensitiv-
ity, but requires very high carrier stability. The small number of time slots in a
frame makes the system unsuitable for dynamic channel allocation. A
comparison of the basic system parameters for DECT and PHS is shown in
figure 5.13.

117
M


PHS DECT

Access technique FDMA/TDMA/TDD FDMA/TDMA/TDD
Carrier spacing 300 kHz 1728 kHz
LO stability 3 ppm 25 ppm
Modulation π/4 DQPSK GFSK
RF power (peak/average) 80/10 mW 250/10 mW
Sensitivity for 0.1% BER -95 dBm -89 dBm
Speech codec 32 kb/s ADPCM 32 kb/s ADPCM
Tolerance to time dispersion 500 ns 200 ns
Frame duration 5 ms 10 ms
Time slots/frame 4+4 12+12
Speech channels/carrier 3 12
Packet data 24 kb/s 552 kb/s
Seamless handover no yes

Figure 5.13

118


6 Further development of NMT
6.1 Shut-down of NMT 900 networks
NMT 900 shares frequency band with GSM, and already from the beginning it
was clear that the whole of this band eventually should be used by GSM (EC
directives). It also a good reason why the spectrum for NMT 900 should be
reduced, as the penetration was gradually reduced, as more and more
subscribers preferred GSM. The situation in Sweden is typical for what is hap-
pening in other countries, having analog system in the same 900 MHz band as
GSM.
The Swedish NMT 900 network had at its peak in 1995 close to one million
subscribers. The number dropped to 400 000 in the fall av 98.
The spectrum reduction, according to a PTS directive, started March 96, when
part of the NMT 900 band was redistributed to the three GSM operators, see
figure 6.1.

Mobile Telephone, Sweden
Mid 95: Penetration > 20% (>2 .10 6 terminals)
Estimated increase 1995: > 60%
40% GSM 60% NMT (number of subscribers)
New subscribers: > 80 % GSM
March 96: NMT900 frequencies transferred to GSM
(net 3 . 0,6 MHz)

NMT GSM
935 941,7 944,1 948,9 949,1 953,9 954,1 958,9 B T
A B C A B C D
890,1 896,7 899 903,9 904,1 908,9 909,1 913,9 T B

A: Telia
B: Comviq
C: Europolitan
(D: Cordless) (Penetration 99 ≈ 60% - close to saturation)

Figure 6.1.

Further reductions followed early 98 and early 99. The remaining NMT band
in 99 was 2x1.9 MHz. The NMT operator Telia Mobitel found in 1998 that the
operation would soon become uneconomic due to unsufficient subscriber base
and the costs connected with fitted the network into the reduced spectrum.
The subscribers were therefore giving notice that the network would be
shutdown at the end of year 2000.

6.2 Modernization of NMT 450
The situation is quite different with respect to NMT 450, which has large pe-
netration also outside the Nordic countries, especially in Eastern Europe. In

119


Sweden the only operator is Telia with more than 250 000 subscribers. Their
license for the present (analog) NMT450 lasts until 2004. In many less
developed countries the speech service complemented with a limited range of
data services will for a long time be the dominant service for many
subscribers. It is therefore not considered to be a major disadvantage that the
fairly small system bandwidth (total allocation 2x4.5 MHz) would make it
impossible to offer similar, wideband data services as UMTS. However many
systems were heavily overloaded, and therefore, there was an urgent need for
new technology that could give substantial capacity improvement. The situa-
tion was similar as in the US when D-AMPS was introduced. Also, it was evi-
dent that similar data transmission capability would be needed as was offered
by the GSM enhancements, i.e. GPRS, (section 7.3).
The 450 MHz band has an important advantage in much better propagation
conditions than in the higher frequency bands used by GSM and UMTS. (The
global propagation loss for the same antenna gains is about 8 dB higher at 900
MHz than at 450 MHz and there is at least a further 8 dB increase going from
900 to 2000 MHz. In addition, the shadow effects becomes more pronounced
at higher frequencies.) This could give an economic advantage to NMT 450,
especially in less costs for the fixed radio network outside of areas with high
traffic density.
Therefore, the GSM group, consisting of operators from the Nordic countries,
Holland and Switzerland, decided in May 98 to initiate a two-step
modernization process.
The first step was to enhance the analog system, above all making it possible
to introduce advanced phase-controlled base station antennas. Also, power
control and improved means to handle portable terminals should be included.
Decisions about such enhanced standards were taken 1998, and systems
started to become available commercially in 1999. The advantages with the
advanced antenna are both improved link budget due to higher antenna gain
(increased cell sizes and/or reduced transmit power) and improved frequency
reuse due to very narrow antenna lobes, i.e. increased system capacity. It was
estimated that the capacity per cell site could be increased more than 5 times.
The second step was to introduce digital speech together with data services.
Several proposals have been presented by system organizations. One
possibility is to keep the present channel width (25 kHz) and for instance use
a system solution similar to TETRA. Another possibility is to use 200 kHz
channels and base the system on GSM technology (GSM400). It has also been
suggested to base the system on cdmaOne and use around 1 MHz wide radio
channels (CDMA450). Qualcomm and Lucount are interested to market such
a system in Eastern Europe. In October 99, the NMT user group (NMT-MoU)
with operators from 30 countries recommended that GSM400 and CDMA450
are further developed to commercial systems. Most operators preferred
GSM400 which was specified jointly by Ericsson and Nokia.
The total capacity improvement by a combination of step 1 and 2 is estimated
to about 15 times.
The limited frequency band for NMT 450 would probably not allow more
than one operator. PTS is preparing a change in the law, which would force
dominating cellular operators to open their networks for other operators on
reasonable terms. Telia hopes that with this arrangements they could continue
their NMT450/GSM400 service after 2004.

120


7 Further developments of GSM
and D-AMPS
7.1 Improved speech codecs

The first versions of GSM, D-AMPS and PDC were optimized for speech,
using full-rate speech codecs. One line of further development is the
introduction of half-rate speech codecs. The main reason is that the existing
operators in the 900 MHz band need increased traffic capacity. (In several
countries, the present 900 MHz operators are not allowed to apply for
frequency allocations in the 1800 - 1900 MHz band.) Half-rate coders have
been introduced in the Japanese mobile telephone system PDC. In Europe,
ETSI has developed a corresponding standard for GSM, but the interest from
operators has been fairly small.
The speech quality of the half-rate coders is similar to the previous full-rate
speech coders during good conditions. However, they are less robust, i.e. the
speech quality is considerably degraded by acoustic background noise (car
and office operation, several speakers) and the coders also distort other signals
than speech (such as back-ground music while waiting for connection in a
telephone exchange).
Another line of development is to take advantage of the continuous
developments in the areas of speech coding algorithms and VLSI to improve
the speech quality of full-rate codecs. The overall objective is that the speech
quality in all respects shall be at least equal to the 32 kb/s ADPCM speech
coder, which for a long time has been a secondary standard (G.726) within the
public telephone network. Besides the requirement on good speech quality
under good conditions, there are requirements on low transmission delay and
acceptable speech quality, even with relative high bit error rates and acoustic
background noise.
A new full-rate speech coder with improved performance has been
standarized for GSM (EFR: Enhanced Full Rate). The first practical application
in an Ericsson system was in 97.
Studies and standardization work are also going on in ITU study groups.
Several new speech coder standards have been issued. One of them (G.729)
gives roughly the same performance as the 32 kb/s ADPCM standard G.726
but at 8 kb/s. The block diagram for G.729 has similarities to the D-AMPS
codec but the design is more advanced, see figure 7.1. The main performance
requirements are summarized in figure 7.2. The algorithmic delay is the
theoretic minimum delay assuming unlimited speed of the digital signal
processing.

121


Coding principle of the "FT/USH" 8 kbit/s ACELP coder

ITU Standard G. 729 8 kb/s "Algebraic CELP" Input
Speech

LPC Info LPC anal
quant
Past & interp.
exitation To Open loop Perceptual
PTTCH weighting
analysis
LPC
T Sp Info
Adaptive
code book
Synthesis
filter -
K Algebraic Sc
code book

MSE Perceptual
search weighting

Gains (7+7)
Gain VQ Pitch lag (8+5+1) M
Index (16+16) T
P
LPC info (18+1) X
Digital
Output

10 ms frame
80 bits/frame
Alg.code index (16+16) Digital
Input
D
Pitch lag (8+5+1) M
Gain Gain 7+7 T
Past prediction P
exitation and VQ X
LPC
T Adaptive Sc Info
code book
Synthesis Postfilter
filter
K Algebraic Sp
code book
Output
speech

Figure 7.1

122


Main terms of reference for the ITU-T 8Kb/s speech coder

Parameter Requirement Objective
Speech quality in Not worse than that
error free conditions of G.726 at 32 kbit/s

Ber 10-3 Not worse than that of Equivalent
random errors G.726 under similar conditions to G.728

Randomly distributed 3% missing frames As small as
missing frame rate < ∆0.5 MOS possible

Algorithmic delay < 16 ms < 5 ms
Total codec delay < 32 ms <10 ms

Speech quality dependancy Not worse than that As low as
on the input signal level of G.726 at 32 kbit/s possible

Ability to transmit DTMF, CCITT No. 5, 6, & 7
signalling/information tones CCITT R2, Q.23, Q.35, V.25

Tandeming capability 2 asynchronous with 3 asynchronous
for the speech a total distortion <4 asynchronous
< 4 asynchronous G.726 G.726 at 32 bit/s

Implementation fixed-point implementation

Capability to operate needed
at different bit rates (9.6 kbit/s to 6.4 kbit/s)

Performance in the presence of Not worse than
background noise (car noise, 32 kbit/s G.726
bubble noise, multiple talker)

G.726: 32kb/s ADPCM standard
(also other rates possible)

Fig 7.2

In figure 7.3 the trade-off is summarized between speech quality and data rate
for a few speech coders of interest in connection with cellular systems.

123


Quality characteristics of a perfect tranmission link
(no background noise, no bit errors, no tandeming)

MOS:
Low delay ADPCM PCM
5 Excellent

G.729 G.711
G.729A G.728 G.726
EFR
4 Good
PDC
HR FR FR
+ +
GSM GSM
HR IS-54
3 Fair
2 4 8 16 32 64
kb/s
IS-54: Used for US DS-CDMA HR: Half Rate
MOS: Mean Option Score FR: Full Rate

Fig 7.3

7.2 D-AMPS with Digital Control Channel
7.2.1 Introduction
The original specification for D-AMPS (IS-54B) was based on the concept of
system integration with the previous analog AMPS. Analog AMPS was used
for all the signalling before a call was set up on a digital traffic channel within
the D-AMPS part of the system. To make a fully digital system possible which
is completely independent of the analog AMPS, a new specification IS-54C
was established. This was the basis for the new D-AMPS standard IS-136 of
1994. Recently D-AMPS has been given the designation TDMA-136.
This section gives a short summary of the most important characteristics.
A more detailed description can be found in Ericsson Review No 2, 1994:
A New Standard for North American Digital Cellular.
The main new feature of IS-54C is the introduction of a digital control
channel, DCC. On the same time additional requests from the operators (User
Performance Requirements for a Digital control Channel issued by Cellular
Telecommunication Industry Association) have been satisfied. The main new
operational features are:
• Support for microcell operation, incl. incorporation of private networks
(R-PABX)
• Sleep mode provision for idle terminals to enhance battery life time
• Increased control channel capacity and flexibility, incl. support of new data
services, i.e. asynchronous data, group 3 fax and especially short message
service, SMS.

124


The DCC is based on the same transmission structure as used by the traffic
channels within D-AMPS. The same modulation, system data rate and basic
TDMA frame are used. One of the radio channels in each cell accomodates
the DCC, which replaces one of the digital traffic channels (DTC). See figure
7.4a.

Incorporation of DCC in basic TDMA structure
a. One radio channel per cell contains DCC time slots
DCC channel uses time slots A

DCC DCC DCC
A B C A B C A B

TDMA frame
40 ms

b. Outward direction
32 "A" time slots form a super frame (time multiplex)

F-BCCH F-BCCH E-BCCH E-BCCH .... S-BCCH .... SPACH SPACH ....

Super frame
32 x 20 = 640 ms

F-BCCH: Fast BCCH (Broadcast control channel)
E- BCCH: Extended BCCH S- BCCH: SMS broadcast channel
SPACH: PCH + ARCH + SMSCH PCH: Paging channel
ARCH: Access Response Channel SMSCH: SMS point-to point channel

C. Inward direction
No hierarchical TDMA structure
AII "A" time slots used by
RACH: Random Access Channel

Figure 7.4

As in GSM, the DCC shall perform the following basic functions:
• Guide the terminals to lock on to the radio channel comprising the DCC
• Synchronize to the DCC, incl. frame syncronization
• Broadcast messages about the network structure
• Registration of terminals
• Location updating
• Paging terminals to initiate the setting up of a traffic channel
• Handling requests from terminals of channel allocation for system
signalling or for traffic.

125


7.2.2 Description of the DCC
7.2.2.1 Burst structure

The structure of a DCC burst in a TDMA time slot is fairly similar to a traffic
burst, see figure 7.5. Additional data fields in the outward direction are SCF
and CSFP. The SCF is the return channel in ARQ arrangements, which
improves the performance of the inward paging channel. The CSFP field ma-
kes it possible for the mobiles to synchronize to the base station timing of the
TDMA hierarchy (basic TDMA frame and super frame). In the inward
direction a PREAM field is included. This field contains no information. Its
purpose is to give the base receiver time to adjust the AGC to the level of the
incoming burst.

Burst structure of DCC

Downlink (single access)

Sync SCF Data CSFP Data SCF RSVD
28 12 130 12 130 10 2

Uplink (multiple access)

G R PREAM Sync Data Sync+ Data
6 6 16 28 122 24 122
20 .48,6=324 bits
6 2/3 ms =
3

G. Guard time SCF: Shared Channel Feed back
R: Ramp time CSFP: Coded Superframe Phase
Pream: No information
(AGC settling time)

Figure 7.5

7.2.2.2 TDMA/TDM structure

In the inward direction, all the DCC slots are used as a universal paging
channel (PCH), see figure1c. It can operate either as a contention channel
(slotted Aloha) or on reservation basis. Immediate acknowledgement is given
by the base on the SCF channel, so that immediate repetition of the page can
be made, if the first page was not received successfully.
In the outward direction, many different types of logical channels must be set
up. Therefore the DCC slots in 32 basic TDMA frames form a super frame,
which can be considered as a TDM arrangement with 32 channels, see figure
1b. In the first two slots in each super frame is placed the F-BCCH channel,
which contains the most essential broadcast information, which must be
repeated often. The channel also informs the mobiles how the other slots in
the TDM arrangement are utilized. Additional broadcast data is transmitted
over the EBCCH, which is organized in a way that permits several repetition
rates. The number of slots allocated for E-BCCH can vary. Next follows a
number of slots for the S-BCCH, which contains SMS messages of the
broadcast type. The remaining slots in the TDM arrangements are used for a

126


combined channel “SPACH“, which is used for the outward paging channel
(PCH), the access response channel, ARCH for point-to-point outward
signalling and the SMSCH channel (point-to point SMS).

7.2.2.3 Hyperframe, paging classes, SMS frames

In order to increase the chance of success of an outward page over the fading
radio channel, each page is sent in two successive super frames, which form a
hyperframe, see figure 7. 6. After two paging attempts, there might be a delay
of several hyperframes before there is a new possibility (hyperframe) to page
a certain mobile. There are 8 paging classes corresponding to different time
intervals between the hyperframes assigned for paging of a certain group of
mobile. Class 1 gives the possibility to send a page every hyperframe, class 2
allows pages to a certain group of mobiles every second hyperframe and class
8 allows for paging every 96 hyperframe. The terminals know in which
hyperframes they might receive pages, so that they can go to sleep during the
other hyperframes. If too many page requests arrive at a paging slot, the
overflow is handled by a later paging slot.

Hyperframe
1,28 s

Primary superframe Secondary superframe
...... PCH ...... PCH ...... ...... PCH ...... PCH ......

overflow
repetition of pages

SMS frame = 12 hyperframes

...... ...... ......

Hyperframe

4 subchannels for SMS:
subchannel 1: timeslot every hyperframe
subchannel 4: 1 timeslot per SMS frame

Figure 7.6

The SMS-frame applies to the S-BCCH (used for sending broadcast SMS
messages). Different repetition rates are suitable for different types of SMS
messages. The repetition time can be chosen by allocating the messages to 4
subchannels with different repetition rates. A SMS frame consists of 24 super
frames.

127


7.2.3 New system features
7.2.3.1 SMS (Special Message Service) to terminals

SMS of the broadcast type has been discussed above. SMS messages of the
point-to-point type are sent either over the traffic channels or over the DCC on
the SMSCH. The messages can be up to 239 characters long. The longer
messages require several time slots, and an ARQ arrangement might be used.

7.2.3.2 Hierarchical cell structures, MACA

Mobile-assisted channel allocation (MACA)
When a connection is set up on a digital traffic channel, the MAHO is based
on measurements by the mobile terminal of the signal levels from surrounding
cells. This has already been implemented in the original D-AM PS according
to IS-54B. Similar arrangements are implemented for the DCC, i.e. when the
mobiles are in the idle mode. The base station sends a neighbour list on the
BCCH, which informs the mobiles where to look for potential cell reselection.
At system access (call origination) the mobile sends over the measured signal
levels on the channels indicated in the neighbour list.

Hierarchical cell structure
The hierarchical cell structure is based on dedicated frequency bands for diffe-
rent cell types, e.g. macro cells, public micro cells and private micro cells.
Due to frequency economy considerations, it is generally desirable to allocate
traffic to the smallest cell, whenever there is a choice. IS-54C provides two
mechanisms for forcing down the traffic to the micro cell layer. See figure.
7.7. One mode selects a preferred micro cell, whenever the signal level from
the micro cell exceeds a specified minimum level. The other mode selects a
micro cell if the difference between the measured signal levels for the macro
cell and the micro cell is less than a specified off-set value (biased cell
selection).

128


Hierarchical cell structure

Signal strength from macrocell 1

Signal strength
Signal strength from microcell 3
from microcell 2 offset

offset
SS-SUFF

Select 1 Select 2 Select 1 Select 3 Select 1

Microcell 2 selected when Microcell 3 selected when
signal strength from microcell the difference in signal level between
exceeds SS-SUFF macrocell and microcell is less
than specified offset

The parameters "SS-SUFF" and "offset" are transmitted in the neighbour list

Figure 7.7

7.2.3.3 Virtual Mobile Location Area (VMLA)

Instead of fixed location areas used at GSM and the original AMPS, IS-54C
utilizes a more flexible concept VMLA. When a mobile registers, the network
control sends over a list of cell numbers, which defines the current location
area for the mobile. When the mobile moves outside of the current location
area and informs the network control about this, it is given a new list of cell
numbers, defining a new location area. The advantages with this arrangement
are:
• different classes of mobiles (i.e. with different speed characteristics) can be
assigned location areas of different sizes
• the location area can be centered around the mobile.

7.3 Adaptation to Data Transmission
7.3.1 Introduction
The original versions of GSM , D-AMPS and PDC were optimized for speech
transmission. Due to the rapidly increasing interest for data services, a gradual
evolution has taken place. GSM has led the way in this respect. Therefore, this
section concentrates on the GSM evolution. However, the last step EDGE
(section 7.3.4) is also added as an option to D-AMPS (TDMA-136) using 200
kHz channels.
With the introduction of EDGE, GSM and D-AMPS comply in principle with
the lower data rate 384 kb/s, which has been specified for G3 systems, such as
UMTS.

129


The advantages are:
a. Present G2 operators might offer G3 type of services using most of the
existing equipment for G2 systems and their available frequency bands.
b. The wideband CDMA systems, which offer user data rates up to 2 Mb/s,
will at least in the beginning be limited to areas with very high traffic dens-
ity, as the maximum cell sizes will go down for wideband systems. (Also
the propagation conditions are more difficult at 2 GHz than at 900 MHz.)
This motivates integration of G2 systems using EDGE and the more
advanced G3 systems based on WCDMA.

The original GSM specification (phase 1) included a framework that made
possible moderate-speed circuit-switched data channels (up to 9.6 kb/s net
user data + 2.4 kb/s network signalling) and short-message service (SMS)
based on packet transmission. The detailed specification of these add-on
services (phase 2) was completed 1996. Of these, only SMS has been used to
any extent.
The GSM standardization is now in phase 2+. This phase includes above all
advanced data transmission services: high-speed circuit-switched data service
(HSCSD) and the general packet radio service (GPRS). These will be extended
to include the EDGE concept.
HSCDS allows higher user rates by allocating a user several time-slots per
TDMA frame.
GPRS is a connection-less packet transmission service. The ETSI
standardization activities were completed late 97 and sent out for review
during 1998. Commercial service will start 2000.
A further improvement is EDGE (Enhanced Data Rate for GSM Evolution).
EDGE allows higher throughput when the quality (C/I and C/I) of the radio
channel is good enough. It can be combined both with HSCSD (ECSD:
EDGE-based Circuit Switched Data) and GPRS (EGPRS). The work by ETSI
to standardize EDGE started early 1998 and the standard was issued early year
2000. It will be introduced 2001.
The increase in the system data rate is obtained by switching over to 8PSK
(which however requires several dB higher C/I and C/N than the original
GMSK modulation). See section 7.3.4 about EDGE.
HSCSD and GPRS are based on the original GSM specification, i.e. the origi-
nal structure of signalling and hand-over arrangements is used as far as
possible. The normal GSM is used during the call set-up procedure. During a
call it is also possible to switch between the original GSM services and the
new data services. Some modifications are necessary in connection with
HSCDS and GPRS. These changes in the detailed arrangements are not taken
up here. In this section, mainly the new facilities, needed for GPRS
(connection-less packet transmission), are discussed.
As mentioned above, similar developments have taken place for D-AMPS.
Data transmission of packet type could to a limited extent be accomodated on
the digital signalling channel. By allowing the use of 200 kHz channels,
GPRS/HSCSD/EDGE can be added as an option. Similar, but more limited
capabilities are planned for the normal 30 kHz channels. To allow D-AMPS
operators in the future offer to wideband services of the same type as within
UMTS, it has been discussed to standardize a system for 5 MHz radio
channels based on one of the FRAMES proposals for UMTS (Wideband
TDMA).

130


7.3.2 HSCSD
The maximum user data rate can be increased by allocating several (n) times
slots to one user. At the BSC (Base Station Controller), the incoming data
stream is divided up into n independent channels (full rate traffic channels,
called HSCSD channels), which are combined in the terminal after channel
decoding and ARQ for each channel. The same applies to the other direction.
The maximum user data rate is determined by the 64 kb/s links between BSC
and MSC. (Presently only one such link is used for each connection.)
The net data rate is influenced by the rate of the FEC channel coding. If fairly
large and varying delays can be accepted, ARQ (BEC: backward error
control) can be introduced to further decrease the error rate (non-transparent
service). In principle, each time slot is used for one of the data transmissions
modes (TCH/F9.6 or TCH/F4.8) according to the original GSM specification.
In addition a raw rate of 14.4 kb/s (net rate - in addition some end-to-end
signalling overhead must be included) is introduced by reducing the amount
of channel coding (data rate from the channel coder 22 kb/s). The maximum
rate per user over the air interface is nx14.4 kb/s, if one user is allocated n
slots.
The normal GSM terminals cannot receive and transmit simultaniously. That
means that the maximum number of traffic slots that can be allocated in either
the inward or outward direction, is limited by the need to:
- include slots to transmit control signals in the other direction
- measure the level of signals from near-by cells (for MAHO)
- include guard times in connection with change of radio channel

The maximum number of slots that can be used in either direction is therefore
4, but the total number of traffic timeslots (for both directions) per basic
TDMA frame should not be more than 5.
These restrictions can be eliminated, if full frequency-duplex terminals are
introduced and no frequency-hopping is used. In that case one user can be
allocated up to 8 time slots (but the total user data rate should not exceed 64
kb/s due to limitations in core network).

A drawback with circuit-switched operation for data transmission over cellular
networks is the risk that the connection is interrupted due to dropout of the ra-
dio connection. Then a new call must be set up and the full data message
transmitted again. This will probably limit the usage of HSCSD, i.e. GPRS will
become the dominating data sevice.

7.3.3 GPRS
7.3.3.1 Overview

GPRS can be seen as an overlay on the normal circuit-switched GSM system.
Many GSM features can be used also for GPRS, but some features must be
implemented by packet technology, i.e. in connection with authentication and
assignement of temporary ID:s. Also the billing system will be changed. With
GPRS the user charge will typically be based on how many packets that have
been transmitted, not on how long time a terminal is connected to the network.

131


Independent packet routing within a packet based network is supported by a
new logical network node SGSN (Serving GPRS Support Node). The Gateway
GSN (GGSN) is a logical interface to external packet data networks. The
SGSN is responsible for the delivery of packets to the terminals within its ser-
vice area. A much simplified block diagram of the complete network is shown
in figure 7.8.

BTS

MS BTS BSC HLR

BTS
MSC/ ISP
SGSN GGSN
VLR network
BTS Backbone network

MS BTS BSC
BTS: Base Transceiver Station
BTS BSC: Base Station Controller
SGSN: Serving GSN
GGSN: Gateway GSN
Base station subsystem GPRS network GSN: GPRS Support Node

Figure 7.8

An hierarchical protocol structure according to the ISO/OSI reference model is
used to handle the data transmission. See figure 7.9 and in more detail in
figure 7.10. The ISO/OSI model is described in module DM2.
Large information blocks from the SNDCP (Subnetwork Dependent
Convergence Protocol) are segmented and placed in LLC frames (LLC:
Logical Link Control). Different frame lengths are possible, the maximum
permitted length is 1600 octets. Each frame contains parity bits for ARQ on
the LLC level (FCS: Frame Check Sequence) and a frame header (FH) with
routing information.

132


Appl.

IP
IP: Internet Protocol
SNDCP SNDCP
SNDCP: Subnetwork Dependent
Convergence Protocol
LLC LLC
LLC Relay LLC: Logical Link Control
RLC RLC
RLC: Radio Link Control
MAC MAC
Physical Physical MAC: Medium Access Control
Layer Layer

MS BSS SGSN

Figure 7.9

GPRS transmission plane

Network Network
layer layer

SNDCP SNDCP GTP GTP

LLC LLC TCP/UDP TCP/UDP
LLC relay
RLC RLC BSSGP BSSGP IP IP
MAC MAC Frame delay Frame delay L2 L2
PLL PLL Physical Physical Physical Physical
layer layer layer layer
RFL RFL

Mobile station Base station subsystem Serving GPRS support node Gateway GPRS
(MS) (BSS) (SGSN) support node
(GGSN)

Figure 7.10

The next layer is the RLC layer (Radio Link Control). See figure 7.11. A LLC
frame is broken up in a number of radio blocks, which also comprises a
header (BH:Block header) and parity bits for selective ARQ (BCS: Block
Check Sequence). Two types of radio blocks are used: data blocks and
signalling blocks. Both type of blocks start with a MAC header, comprising
USF, T and PC fields. The USF (Uplink State Flag) is used in connection with
the reservation of radio blocks on the inward traffic channel. The T field is a
flag, which tells if a block is used for data transfer or for signalling. The PC
field is used in connection with the power control.

133


Radio Link Layer. Block structure

User Data USF T PC RLC header RLC data BCS

MAC header RLC data block Block
check
sequence

Control USF T PC RLC /MAC signaling information BCS

MAC header RLC/MAC control block Block
check
sequence

Figure 7.11

LLC
FH Information field FCS

RLC
MAC
BH Information field BCS BH BCS BH BCS

interleaving + (FH) + FEC + tail
Physical
layer

TDMA slots FH: Frame Header
(4x114 = 456 bits) FCS: Frame Check Sequence
BH: Block Header
BCS: Block Check Sequence

Figure 7.12

The relations between the data sequences in the three layers are shown in
figure 7.12 and in more detail in figure 7.13

134


GPRS transformation data flow

Packet (N-PDU) PH User data Network layer
SNDCP layer
Segment Segment
SNDCP layer
LLC frame FH Info FHC LLC layer

LLC layer
Segment Segment Segment RLC/MAC layer

RLC block BH Info BCS Tail
456 RLC/MAC layer
Convolutional encoding Physical layer
114 114 114 114
Normal burst Burst Burst Burst Burst

PH: Packet Header FCH: Frame check sequence
FH: Frame Header BSC: Block check sequence
BH: Block Header

Figure 7.13

The RLC blocks are numbered (TFI: Temporary Flow Identifier) and the
receive side can request retransmission of erroneus blocks. Besides ARQ
based on the BCS, FEC coding can be applied. See section 7.3.3.3.
The radio blocks are fed to the Physical layer, which is based on slots in the
normal 8-slot TDMA frame. Each of the 8 slot positions constitutes one Packet
Data Channel, which is multiplexed between traffic channels and different
signalling channels.
One radio block is allocated 4 slots in successive frames, i.e. one block
comprises 4x114 = 456 bits. Most of the slots in a Packet Traffic Channel
(PTCH) are used for data transfer (PDTCH: Packet Data Traffic Channel) but a
small percentage of the slots is used for signalling (PACCH: Packet Associated
Control Channel). The arrangements correspond in principle to the signalling
structure of the original GSM.
The PTCH is shared between several simultanious connections (sessions) un-
der Media Access Control (MAC), see section 7.3.3.2. If multi-slot is used,
one connection is served by more than one PTCH.
In principle, the same signalling facilities as in the orignal GSM are used for
timing advance, power control and MAHO. Examples of extended facilities
are:
- possibilities for several hand-over protocols (one alternative is that the
terminal by itself determines suitable handover)
- paging groups to permit Discontinuous Reception DRX (the terminal has to
listen to pages only during certain time intervals and can go to sleep in
between).

135


The network control is decentralized with the RLC and MAC performed by the
BSC units in the BSS. The hand over is also placed in this layer. Radio blocks
can be destroyed during hando ver, but that is handled by ARQ at the LLC
layer. Some of the functions of the LCC layer are placed at the BSC (LCC
Relay).

7.3.3.2 Media Access Control (MAC)
To permit efficient sharing of a common channel resource between several
sessions, the principle is that reservation of time slots (groups of four) for radio
blocks is only given as long as there are information stored in the buffer
memory on the transmit side. As soon as all the buffered data bits have been
successfully transferred (incl. possible ARQ retransmissions) the channel
allocation is released and can be used by other connections. When a new burst
of data arrives to the buffer, a new reservation of time slots must be made. The
network might even interrupt long data sequencies, in order to reduce waiting
times and to obtain a more fair access to the medium, when operating close to
the capacity limit.
The MAC is more complicated for uplink data transfer (multiple access). In the
downlink direction the base controller has full information about requests for
transmission capacity and can store them in a common queu (single access).

Uplink transmission
The BSC controls the inward data traffic by means of Uplink State Flags (USF),
which have 3 information bits. Of the 8 available flags, one is used to mark the
slots which make up the inward access channel (Aloha) (USF = Free). The
other 7 flags (R1 - R7) are used to reserve 4-group of time slots for terminals.
An example of the allocation procedure is given in figure 7.14. To simplify the
description of the functionary, the TDMA frames are designated by running
numbers. Groups of 4 time slots placed in 4 TDMA frames is the basic trans-
mission unit.

Mobile-originated traffic. MAC
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 ......

Downlink ChRes A-B
TS 0 F F F F F F F F F F F F F
Uplink PA PA

Downlink ACK-B ACK-A
TS 1 R1 R1 R1 R2 R2 R3 R1 R1 R1 R2 R1 R1 R1
Uplink A1 A3 B2 A3

Downlink NACK-A
TS 2 R1 R1 R1 R2 R3 R2 R1 R1 R1 R1 R1 R1 R1
Uplink A2 B1 A4

Downlink
TS 3 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1 R1
Uplink

PA: Paging packet from terminal
F: Flagcode "free" (TS0 uplink available for paging from terminals)
R2, R3: Flagcodes which point to assigned packet slots uplink

Figure 7.14

136


Time slot 0 (uplink) in the figure is used as Packet Random Access Channel
PRACH). Two terminals make request (PA) during frame 1 and 3 respectively.
The PRACH is a contention channel, with a risk of collision of simultanious
requests from several terminals. To minimize the risk for repeated collisions, a
more refined procedure is used than in original GSM. It is a protocol of the
“persistance“ type, see module S5, section 3.3.
If channel capacity is available, time slots are allocated to the terminal
(immediate assignment), including the maximum number of radio blocks that
could be transmitted. This information is sent on the PAGCH (Packet Access
Grant Channel). Also a two-step reservation process can be used. In the second
step the terminal comes back with a Packet Resource Request, and the base
replies with a Resource Assignment. This signalling is performed on the two-
way Packet Associated Control Channel (PACCH). Compare the signalling in
the original GSM. If no channel capacity is available for the moment, the termi-
nal is informed by a Packet Queuing Notification.
In the example, the BCS makes an immediate assignment by sending back a re-
servation for the two calls (A and B) on times slot 0 (downlink) as a radio block
during frames 8 to 11.
USF = R2 is used to mark reservations for call A, and USF = R3 is used for call
B. The reservation message also includes information about which frequency
slot and timeslot within a frame shall be used. In this case, both calls will be
serviced by timeslot 1 and 2. The terminals therefore will monitor the signalling
sent on these slots. The reservation for radio blocks is made by transmitting the
corresponding USF flags over the down link channel during the block position
before the allocated position.
The first two allocations for call A are during frames 16 to 19. Therefore during
frames 12 to 15 USF = R2 is transmitted on the downlink channels TS1 and
TS2. Call A is also allocated frames 20 to 23 on TS1 and 24 to 27 on TS 2.
Therefore USF = R2 is also transmitted downlink during frames 16 to 19 on
TS1 and during frames 20 to 23 on TS 2:. In the same way call B is allocated
two timeslots on TS1 and TS2.
The two B blocks were transmitted successfully, and therefore a positive ack-
nowledgement (ACK-B) is transmitted back during frames 32 to 35 on a
PACCH (Packet Associated Control Channel). On the other hand, block A3 was
not transmitted successfully. Therefore a negative acknowledgement (NACK-
A) is transmitted back during frames 32 to 35. The NACK-A message indicates
that blocks A3 must be retransmitted (selective ARQ) and also gives additional
reservation for the retransmission. A3 is retransmitted during frames 40 to 43.
The transmission was succesful this time and ACK-A is transmitted during fra-
mes 48 to 51.

Down link transmission
For down link data transfer, the first step is a Packet Paging Request on a
paging channel (PPCH). The terminal replies with a Packet Channel Request.
The rest of the set-up procedure corresponds to what has been described above
for inward data transfer.

Point-to-multipoint
Data can also be transmitted from the network to several terminals (point-to-
multipoint or broadcast). To set up this connection a special channel is used

137


(PNCH: Packet Notification Channel) to inform the terminals concerned about
the resource assignment for the packet transfer.
Multiframe
The basic TDMA frames are grouped in multiframes comprising 52 basic fra-
mes. The main reason for the multiframes is for discontinous reception, i.e. the
terminals have to listen to pages from the BSC only during a specified part of
each multi slot.

7.3.3.3 Channel coding for the PDTCH

An adaptive channel coding arrangement is used with the possibility to
dynamically select one of four channel coding procedures. These are called
CS-1 to CS-4. In addition to the coding of the information sequence also the
USF is protected by a block code.
CS-1 is a combination of a rate 1/2 convolution code and a long BCS (Block
Check Sequence), resulting in a net data rate of 9.05 kb/s. CS-4, which only
contains a short BCS and no FEC, has a net data rate of 21.4 kb/s. The
convolution code that is used for CS-1, is punctured to get the higher rate
codes CS-2 and CS-3, which give net data rates of 13.4 and 15.6 kb/s
respectively. To reduce the degradation due to error bursts, interleaving over
four time slots (TDMA frames) is used. In addition frequency hopping may be
introduced. This is especially motivated with quasi-stationary terminals.
The set of bursts that results from a single user data package is marked by a
Temporary Flow Identifier (TFI), which is used at the receiving side to
reassemble the user data package.
When there are block errors, the radio blocks, which contain errors, are
identified by their TFI numbers (Temporary Flow Identifier). The TFI:s are
sent back to the transmitter side, so that corresponding blocks can be
retransmitted. The numbering is modulo 128, and 64 blocks can be handled
by the ARQ protocol. Positive acknowledgement is sent back for successfully
received blocks. As gradually the block errors are cleared, the error control
window can be moved forward. Radio blocks without errors are fed to the
LLC layer, so that LLC frames can be generated on the receive side.
The optimum code arrangement, i.e. the code which gives the highest
throughput, depends on the ber from the data demodulator at the receive side.
This quality information must be transmitted back to the transmitter side,
which gives a certain delay. Adaptation is therefore not possible to the
variations due to the fast fading, The measured ber values are therefore
averaged over the fast fading.
With extremely low ber, the number of retransmissions becomes very low
even without FEC. Therefore CS-4 gives the highest throughput. When the
quality of the radiochannel degrades more and more, it is optimum to
introduce successively more FEC. The reduction of the basic data rate
(without the influence of ARQ) is more than compensated for by increased
probability of successful block transmission. In principle, we have the relation
given in figure 7.15. Already CS-3 gives most of these advantage - CS-2 and
CS-1 give small additional gains at very marginal radio channels. Typical rela-
tions between channel quality and throughput for the different coding
alternatives are indicated in figure 7.16. The figure corresponds to a
propagation channel with short fading dips (fairly high terminal speed or low
speed in combination with frequency hopping).

138


Data transmission over fading radio channel.
Throughput characteristics with and without FEC

Throughput ARQ Radio channel
Channel Modulator
retrans-
mission coding baud
di b/s dn b/s db b/s ds

GMSK db = ds
Throughput 8PSK db = 3ds
(16QAM: db = 4ds)
db
At GSM: ds = 270.8 kbaud
dn1
(also using 8 PSK)
loss due to (1)
dn2 retransmissions
(2)

C Eb
I No
(1) Only error detection coding
(2) Both FEC and error detection coding

Figure 7.15

The performance of this adaption scene is degraded by inaccurate estimates of
the channel quality at the receive end of the link and measurements delays.
Therefore, at EDGE it is complemented by an additional adaption based on
advanced ARQ.

Adaptation of code rate to channel quality for maximum throughput

Throughput
kb/s

20
CS4

16
CS3

CS2
12

CS1
8

4

C/I dB
0
3 7 11 15 19 23 27

Figure 7.16
139


7.3.4 EDGE
7.3.4.1 Introduction and background

Edge stands for Enhanced Data rate for GSM Evolution. It is a further evolu-
tion of HSCSD and GPRS, giving the option of increased system data rate.
However it will mainly be used in connection with GPRS, and we limit the
discussion to this application. The combination of GRS and EDGE is usually
referred to as EGPRS.
GPRS used originally a dynamic link adaptation to the channel quality based
mainly on varying the code rate. In addition simple ARQ arrangement with re-
petition of radio blocks gave a further link adaptation of the throughput.
EGPRS extends the adaptation of the code rate by allowing a choice between
GMSK and 8PSK (a dynamic selection of suitable combinations of channel
coding and modulation is used depending on the quality C/I and E/N0 of the
radio channel). The net transmission data rate (including the effect of channel
coding) will be reduced, when the quality of the radio channel goes down.
The throughput will in this case be further reduced as more packets must be
retransmitted (ARQ).
The suitable choice of modulation and coding is based on measurement of the
channel quality (average over the fast fading) which is transmitted to the other
end. Practical evaluations indicated that this link adaptation did not perform
fully satisfactory. Therefore, a more refined link adaptation procedure is used
at EDGE than at the original GPRS (and HSCSD). A large part of the link
adaptation is obtained by an advanced ARQ procedure, called type II hybrid
ARQ. The link adaptation used at the original GPRS, which was based on dif-
ferent modulation and coding schemes (MCS) in combination with simple
ARQ is called type 1 hybrid ARQ.
EGPRS obtains the best throughput only when the radio channel has very
good quality. This has two consequences:
a. An effective dynamic link adaptation is necessary that dynamically selects
modulation and code rate to the local average of C/I and C/N.
b. EGPRS for GSM is typically introduced in existing cell structures, i.e. 9 cell
clusters. As much higher protection ratios would be needed for the highest
throughputs, the highest user data rates could only be used in part of each
cell. (The dynamic link adaptation makes it possible to go down to 3 cell
clusters, which is often preferable in the US, where available bands for
D-AMPS are more restricted.)

Background
Several alternative modulation arrangements were proposed and studied
before 8PSK was finally selected. One possibility was Offset 4QAM (similar to
what is used for DAMPS), another Offset 16QAM, which gives a further
increase in data rate for a given modulation bandwidth. 16QAM was
discarded mainly due to very high linearity requirements on the transmitter
output stage. This made it necessary to introduce considerable back-off to
comply with the GSM requirements on out-of-channel radiation. The large
back-off reduced the transmitter efficiency considerably and therefore 8PSK
gave nearly the same throughput as 16QAM for the same battery drain. An

140


additional disadvantage with 16QAM is that the radio symbol do not all have
the same energy, which gives a certain complication at the receiver (AGC).
The final proposal for an international standard, sent to ITU in May 98, was
based on an agreement between study groups in Europe and USA, i.e. the
standard should be used both for GSM and DAMPS (IS-136). 200 kHz
channel spacing will be used also for evolved DAMPS by combining several
30 kHz channels to form the necessary 200 kHz channel. EGPRS can in
principle (very good radio channel, all 8 time slots allocated to a user) comply
with the lower of the two user data rates (384 kb/s) specified for UMTS.

7.3.4.2 Modulation, burst structure

The modulation will be either GMSK or linear 8PSK. See figure 7.17. With
8PSK the system data rate becomes 69.2 kb/s per time slot (three times higher
than with GMSK). Linearized 8PSK is used, i.e. the modulation can be con-
sidered to be of the QAM type. Also the pulse forming is the same as for
GMSK, resulting in the same spectrum envelope. In the same way as the D-
AMPS modulation (figure 3.7a), the signal plane is rotated a suitable amount
between each transmitted symbol in order to reduce the linearity requirements
on the transmitter output stage. A rotation of 3π/8 is used instead of π/8, as
this simplifies the blind detection at the receiver if GMSK ot 8PSK was used
for a transmitted packet.

141


EDGE 8PSK modulation
( Linearized Graycoded 3π 8PSK )
8
S P 8 PSK conv.
Pulse
Gray- modulator binary
forming
coding to δ-pulse
j3π
j( )
e 8

GMSK
0,0,0 0,1,1 symbol rate = bitrate = 270.8 kbaud
0,1,0 kb/s

0,0,1 1,1,1 8 PSK
symbol rate = 270.8 kbaud
bit rate = 812.5 kb/s
1,0,1 1,0,0 1,1,0
(channel spacing: 200 kHz)

x Si+1, Si+3 ....
Si, Si+2 ....
x x Pmin 2
= 0.26 = 0.07 =11.5dB
Pmax

x x

0.26 Put
1
1 (Put)max

x π
8 x

x x Pin
3π (Pin)
0.07 1
8 max

Fig 7.17

The burst structures (“normal burst”) during a time slot is shown in figure
7.18.a. As no flags are needed the data sequencies contain 58 bits. The bursts
for the GMSK and the 8PSK modes are compared in figure 7.18b.

Edge TDMA burst
2x58x3 =348 data bits per burst
4x348 = 1392 data bits per 20 ms interleaving block symbols carrying
Gross bitrate 69.6 kb/s/timeslot 3 bits each with 8 PSK

3 58 26 58 2
Training
Tail Data Data Tail
sequence

0.55 ms
Multiframe 24 traffic frames in 120 ms i.e. 20 time slots per sec.
Data rate per time slot 348x20 = 69.6 kb/s
(including bits for channel coding and headers)

Fig 7.18a

142


EDGE. TDMA burst structure
Normal burst for GMSK
Bit Number (BN) Length of field Contents ot field

0-2 3 tail bits
3 - 60 58 encrypted bits(e0, e57)
61 - 86 26 training sequence bits
87 - 144 58 encrypted bits (e58, e115)
145 - 147 3 tail bits
(148 - 156 8.25 guard period (bits)

- where the "tail bits" are defined as modulating bits with states as follows:
(BNO, BN1, BN2) = (0, 0, 0) and
(BN145, BN146, BN147) = (0, 0, 0)

Training Training sequence bits
Sequence (BN61, BN62 BN86)
Code (TSC)

Normal burst for 8PSK
Bit Number (BN) Length of field Contents of field
(bits)
0-8 9 tail bits
183 - 260 78 training sequence bits
435 - 443 9 tail bits
444 - 468 24.75 guard period

- where "tail bits" are defined as moduladng bits with states as follows
(BN0, BNI BN8) = (0,1,0;1,1,1;1,1,0) and
(BN43S, BN436 BN443) = (0,1,0;1,1,1;1,1,0)

Training Training sequence symbol
Sequence (BN183, BN184 BN260)
Code (TSC)

Figure 7.18b

143


7.3.4.3 Link adaptation

As mentioned above the link adaptation is based on type 2 hybrid ARQ. The
first step is the selection of a suitable MCS, i.e. combination of modulation
type and coding rate. The original GPRS arrangement is extended by the
choice between GMSK and 8PSK, giving 8 different combinations, see figure
7.19.

EGPRS Link Quality Control by adaptation of modulation and code rate to C/I
≈ throughput
70
Scheme Modula- Maximum Code rate
tion rate FEC
60
(kbps) Packet bit rate (per time slot)(kbps)

MCS-8 8PSK 59.2 1.0 50

MCS-7 44.8 0.78
40
MCS-6 29.6 0.50
30
MCS-5 22.4 0.38

MCS-4 GMSK 16.8 1.0 20

MCS-3 14.8 0.89 10

MCS-2 11.2 0.69
0
0.53 0 5 10 15 20 25 30 35 40
MCS-1 8.4
C/I (dB)

Figure 7.19

The next step in the link adaptation is a more advanced ARQ arrangement
than in GPRS. It is called incremental redundance (IR). The information part
of a packet is to start with coded with a low rate code, i.e. a R=1/3 convolution
code. From this sequence two or three sub blocks are generated by heavy
puncturing by different structures, which should give sequences as disjunct as
possible. Each sub block is detectable by its own. To start the transmission of
a packet, one of the sub blocks is transmitted. If ARQ is necessary, one of the
other sub blocks is transmitted. (If further ARQ is necessary, the third sub
block could be used.)
The normal mode of operation is that the detector, if needed, combines the in-
formation in the two sub blocks. This gives a large increase in the probability
of successful detection, both due to increased redundancy and to time
diversity. (It might happen that the memory at the receiver is not sufficient to
store the information about the first packet. In this case, the less effective type
1 hybrid ARQ is used.) The procedure is described in principle with 3
subblocks in figure 7.20, and in more detail (with two sub blocks) i figure
7.21. The final step is interleaving over 4 TDMA slots as for GPRS. This could
also be combined with frequency hopping.

144


Encoding and sub block puncturing for incremental redundance

Data FCS Tail
R=1/3 encoding

Codeword
Puncturing

S1 S2 S3

FCS: parity bite for error detection

Fig 7.20

EGPRS Link Quality Control
Principle of operation
3 bits 32 bits 612 bits

USF RLC/MAC CRC FBI E Data = 74 octets = 592 bits BCS TB

Rate 1/3 convolutional coding

36 bits 96 bits 1836 bits

1392 bits puncturing

SB = 4 36 bits 96 bits 1256 bits 1256 bits
P1 P2

1392 bits
(4 x 348 = 4 x 3 x 3 x 58)

Fig 7.21

7.3.4.4 Performance

EGPRS will be used both in GSM and D-AMPS (TDMA-136) networks. In a
typical GSM application, cluster size 9 and frequency hopping will be used.
Figure 7.22 gives the CDF for different user data rates for a total normalized
spectrum utilization of to 0.33 bps/Hz/site. The figure shows that 30 % of the
users can obtain at least 384 kb/s for 8 time slots.

145


3/9 reuse, frequency hopping

Average packet bitrates for offered load that achieves 0.33 b/s/Hz/site
30% of users: > 384 kb/s, 97%: >144 kb/s
100

90

80

70
C.D.F%
60
Standard GSM EDGE
50

40

30

20

10

0
0 10(80) 20(160) 30(240) 40(320) 50(400) 60(480) 70(560)
kb/s
Average packet bit rate per user per timeslot
(bitrate per 8 timeslots in brackets)

Fig 7.22

The available system bandwidth is more restricted at D-AMPS. A minimum
EGPRPS system corresponds to an allocation of only three 200 kHz radio
channels (cluster size 3). No frequency hopping is possible. The performance
in this case is shown in figure 7.23. For a system loading that gives a total
normalized spectrum utilization of 0.50 bps/Hz/site, 40 % av the users
achieves at least 384 kb/s.

146


1/3 reuse, no frequency hopping

Average packet bitrates for offered load that achieves 0.50 b/s/Hz/site
40% achieves: > 384 kb/s, 90% >144 kb/s
100

90

80

70

60

50

40

30

20

10

0
0 10(80) 20(160) 30(240) 40(320) 50(400) 60(480) 70(560)
Average packet bitrate per user per timeslot kb/s
(bitrate per 8 timeslots in brackets)

Fig 7.23

147


8 Cell structures
8.1 Additional frequency allocations to digital
mobile telephone systems
The rapid increase of the number of mobile telephone subscribers has led to
insufficient traffic capacity in several countries in the allocated frequency
bands around 900 MHz. Most of the new subscribers join one of the digital
mobile telephone networks. In the Nordic countries, the number of subscribers
in the NMT system does not increase further. According to EC directives, the
frequency bands for the present analog 900 MHz systems shall gradually be
given over to GSM. See section 6.1.
On the other hand there are good reasons to continue and improve the
NMT 450 system due to the better propagation characteristics, which allow
larger maximum cell sizes to be used. See section 6.2.
It is doubtful if it is economic viable to extend the 900 MHz GSM networks to
give full coverage of the northern part of Sweden. The cell sizes will be even
more restricted at 2000 MHz due to difficult propagation conditions. (The
propagation loss is roughly 16 dB larger at 2000 MHz than at 500 MHz.) That
means that the 2000 MHz UMTS networks will be limited to areas with high
traffic density where small cells can be used. In scarcely populated areas,
UMTS must therefore be complemented by GSM or by satellite networks
requiring near free-sight conditions.
As different types of cellular systems often have different coverage characte-
ristics and offer different user data rates, there will be increasing need to give
the user the option to access several cellular systems be means of multi-mode
multi-band terminals. This is one aspect of hierarchical cell structure. An other
aspect is a mixture of different cell sites within the same system, see section
8.2.

8.2 Hierarchical cell structures
One of the requirements on the futurecellular systems is that they shall handle
both densely and scarsely populated areas in an optimum way, considering
both infra structure costs and frequency economy. This requirement also
applies to the digital mobile telephone systems of today. Also on metropolitan
areas there is often a mixture of hotspots and areas with fairly low traffic dens-
ity. Therefore hierarchical cell structures have already been introduced to
some extent in G2 systems.

148


HCS - Hierarchical Cell Structures

PICO

MACRO

MICRO

Figure 8.1

A hierarchical cell structure comprises cells of different sizes. See figure 8.1.
Areas with considerably different traffic densities are covered by structures
with different cell sizes, i.e. the larger the traffic density, the smaller cells are
used. Without micro and pico cells, the traffic capacity will be unsufficient in
areas with high traffic, and the cost for the fixed radio network excessive in
areas with low traffic, if very small cells are used.
The highest traffic density can be found indoors, for example in large offices
and conference centra, assuming that most of the wire-connected telephones
will be replaced by radio-connected ones. In this case, pico cells are used with
dimensions of about 30 m. A pico cell might be fed by a base station placed in
a corridor and with a coverage area consisting of the corridor and the
adjoining rooms. Outdoors or mixed indoors-outdoors, traffic concentrations
exist in shopping streets, sports and business centra and airport. These are
covered by micro cells with typical dimensions of 100 meters. A microcell can
have a base station on a lamp post or house wall and cover a length of street.
Even in major cities there are areas with so low traffic density that it is not
economic viable to use micro cells. Therefore, a complement is macro cells
with approximately 1 km size. They are often called umbrella cells, as an
important function is to fill in gaps between the high traffic areas, serviced by
micro and picocells. As before, larger cells with sizes up to around 30 km
(large macro cells) are used in rural areas.
The highest level in hierarchy will be satellite cells, serviced by landmobile
satellite systems.
The main part of the traffic will come from traffic concentrations (hot spots) at
densely populated areas, outdoors and indoors. The major part of the traffic in
a cellular system will come from these traffic concentrations, which should be

149


covered by micro and pico cells. This is a precondition for enough total traffic
capacity in the future cellular systems.
The introduction of hierarchical cell structures gives certain complications on
the system level. Even if the frequency economy is improved, by the
introduction of micro and pico cells different allocation of radio channels
should be made to the different levels of the hierarchy. This could be a pro-
blem due to unsufficient spectrum allocation to an operator. One reason for
the use of different frequencies for different layers is that the system control
shall be able to force the terminals to use the lowest hierarchical layer, that gi-
ves an adequate connection (see figure 7.7). However, a limitation is that
micro cells cannot serve fast moving terminals, as this would result in such a
high handover rate that the system signalling would be overloaded, or the
transmission quality degraded.
There are also additional requirements on signalling and control logic in order
to handle handover between cells on different levels, incl. MAHO (Mobile
Assisted Hand Over). The principle of MAHO, as used at GSM, works also in
hierarchical cell structures, see figure 8.2.

Mobile Assisted Handoff (MAHO)
Extended to hierachical cell structures
t

MS receives Rx Rx Rx Rx

MS transmits Tx Tx Tx Tx

MS measures cell 1
M
MS measures cell 2 M

MS measures cell 3
M
MS measures macro cell M

Figure 8.2

8.3 Land mobile satellite communication
Worldwide satellite communication has been used for many years for traffic to
ships, using frequencies close to 1 GHz. The operator is Inmarsat, which is
jointly owned by a large number of telecom administrations. Geostationary
satellites are used. These are placed in an equatorial orbit 36000 km above the
earth, i.e. a typical propagation distance would be 40 000 km. The difficult
link budget has been a limitation, which means that high performance termin-
als must be used with fairly high EIRP (Effective Isotropic Radiated Power)
and G/T. G: gain of receiver antenna, T: noise temperature of receiver sys-
tem). See figure 8.3.
150


Link budget
Gt G
Pt Pr
Lo+ M

EIRP = P + Gt dBW L : propagation loss (isotropic antennas) System noise
t o
M : fading margin temperature: T
Pr = EIRP - Lo - M + G k: Boltzmann's constant

C = Pm ⇒ P = C/N + k + T
No kT r o

Lo + M = EIRP + G - k - T - C/No
Lo + M = EIRP + G - C/No - k (C/No = I . Eb/ )
T db No
G: antenna gain for receiver system
Gt: antenna gain for transmitter system

Figure 8.3

As the satellites must be placed over the equator, a further drawback is that the
polar areas are not covered. If the service is extended to land mobile termin-
als, there would also be problems with strong shadowing in hilly terrain and
even outdoors in metropolitan areas, especially if the elevation angle to the
satellite is small.
The same basic satellite system has recently been used also for slow data
services, e.g. fax and short data messages. The combination of low data rate
(db) and low required Eb/No means that terminals with much reduced EIRP and
G/T can be used. The dimensions and cost of the terminals could therefore be
reduced considerably. A large number of terminals have been installed on
fairly small ships and even on vehicles. A portable version has about the same
size as a large briefcase. Free sight is necessary, and the directive antenna
must be pointed towards the satellite.
A new generation of geostationary Inmarsat satellites with high gain, multi-
lobe antennas started operation in 1997. The satellite EIRP and G/T are con-
siderably improved, which makes it possible to use fairly small mobile termin-
als for digital speech, even if their radio performance (EIRP and G/T) still
must be considerably better than for terminals for terrestrial networks.
The fast expansion of terrestrial mobile telephone systems has caused a large
interest to establish also land mobile satellite networks. Even if the cost for
terminals and calls will be several times larger (estimate 3 times) than for the
corresponding terrestrial services, there have been optimistic (perhaps
unrealistic) estimates that there would be a potential satellite market
amounting to a few percent of the total mobile telephone traffic. This has been
enough to motivate the large investments needed to establish such satellite
systems. Foreseen applications are given in figure 4.10. The main application
is probably as gap filler, i.e. the satellite network is used outside of the
coverage area of the terrestrial network, which is normally used by the
subscriber. A limitation that could be quite serious is the very limited indoor
coverage. Due to the difficult link budget, only small margins for shadowing
are possible.

151


Land mobile satellite networks, which can be rapidly installed (if the satellites
with earth support are already in place), could also complement the primitive
public telephone networks in many developing countries.

Use of Mobile Satellite Communication
• Extend coverage to areas not covered by any terrestrial radio
network
• Extend coverage to areas serviced by other, non-compatible
terrestrial networks
Problems:
Difficult link budget
small shadow margin (indoor coverage very limited)
high cost per call
larger, more expensive terminals
Inferior frequency economy
large cells
Advantages:
Large (worldwide) coverage by one flexible system

Figure 8.4

A satellite system, that shall be an extension of the terrestrial mobile network,
must be designed so that the size of the dual-mode terminal (terrestrial +
satellite service) is only moderately larger than a standard terminal for the ter-
restrial network. This requirement could be complied with by using advanced
geostationary satellites with very large multi-lobe antennas. If a lower orbit is
used, the size of the satellite antenna can be considerably reduced and some-
what larger link margins might be possible. (If the orbit height is changed, the
link budget remains the same for the same area of the footprint of the satellite
antenna on earth.)
That is one reason why most of the land mobile satellite systems with
worldwide coverage will be based on satellites in lower orbits than the
geostationary, either low flying satellites in orbits with around 1000 km height
(LEO: Low Earth Orbit) or in intermediate orbits (ICO: Intermediate Circular
Orbit) about 10 000 km above the earth. See Figure 8 .5 and 8.6. The height
interval between the LEO and ICO orbits is unsuitable due to a belt of
intensive radiation (Van Allen belt), which would rapidly damage the satellite
electronics.
Other advantages using the LEO and ICO orbits are considerably lower
propagation delay than for the geostationary orbit and that good coverage is
obtained also of regions close to the poles (less shadowing if the satellites are
well above the horizon).

152


Mobile satellite communication - portable terminals

Highly
Elliptical HEO
Orbit

Intermediate
ICO Circular Low
Orbit Earth LEO
Orbit

Geostationary
Orbit GEO

Figure 8.5

Land mobile satellite communication

Iridium Globalstar Odyssey ICO G.C.
Start of operation 1998 2000 2000 2000
Type of orbit LEO LEO ICO ICO
Orbit period (hours) 1.7 2 6 6
Satellite height (km) 800 1 400 10 000 10 000
Frequency band 1.7 GHz 1.9/2.2 GHz
(terminals)
Number of satellites 66 48 12 10
One-way delay 100ms 200 ms
Cost 109 $ (excl. terminals) 3.4 1.8 1.7 2.4

Figure 8.6

Examples of systems using LEO and ICO respectively are the Iridium and the
ICO Global Communication. There are also satellite systems with regional
coverage, using geostationary satellites.

153


Iridium

The first worldwideland mobile satellite service is the Iridium system
originally proposed by Motorola. The system comprises 66 satellites, which
are placed in 6 polar orbits, each with 11 satellites. The satellites have
advanced antenna arrangements, which generate a large number of lobes. The
whole earth surface is covered with about 4000 cells. See figure 4.13. 4.8 kb/s
speech coders are used. The required terminal EIRP is about 2 W (1W trans-
mitter power and 3 dB antenna gain).
The LEO-satellites move fast relative to a fixed point on earth (orbit period
around 100 minutes) and the moving cells are fairly small. A terminal is
therefore within the coverage area of a cell only for a few minutes. Thus hand-
over between cells is needed, even if the terminals are not moving.

Figure 8.7

A further problem is that each of the earth stations, which connect the
satellites with the fixed public telephone network, can see only a few of the 66
satellites at a certain time.
A large number of earth stations would therefore be needed, if it would have
been necessary that all satellites at all times would be within sight of an earth
station. However, only a small number of earth stations will be needed in the
system, as the satellites operate both as an access network and a transport
network. Independent of the position of the terminal to be connected, an earth
station only needs to establish a connection with any suitable satellite above

154


the horizon. From this satellite, the call can be relayed via several satellite-to-
satellite links to the satellite, which is best situated to service the terminal.
These connections (the transport network) consist of microwave links between
nearby satellites. The ar-rangement evidently requires switches in the satellites
and that the radio signals are detected in the satellites. The fixed terrestrial
infrastructure is based on GSM technology.
Commercial Iridium services started late 1998, that with certain technical
problems, such as unsatisfactory handover. The Iridium consortium has made
roaming agreements with many terrestrial operators (i.e. using GSM and
D-AMPS). Dual-mode terminals will be used.

ICO Global Communication (Inmarsat P)

The complications mentioned above with the LEO orbit are avoided, if the
satellite network is based on the ICO orbit. The satellites are in a considerably
higher orbit, and therefore move much slower (orbit period around 6 hours),
also the cell sizes are somewhat larger. Handover between cells would not be
necessary. Due to the higher satellite height, each earth station can service
satellites over a fairly large part of the earth. Therefore, there is no need for a
satellite transport network. INMARSAT (strictly speaking the affiliary “ICO
Global Communications“) will use 12 Satellite Access Nodes, each with 5
antennas to establish the contacts between the satellites and the fixed part of
their network.
10 satellites will be placed in two orbits with 45 degrees inclination to the
equator. No signal processing is performed in the satellites (“bent pipe“). (An
advantage with this arrangements is that the transmission parameters might be
changed by modifications of the terrestrial part only)

Networks compatible with UMTS

The land mobile satellite networks, mentioned above, are originally designed
for speech as the dominating service. They belong to the same generation as
GSM and D-AMPS. Only low-speed data services can be offered. A new ge-
neration of satellites is planned, which will give much improved data service
to both mobile and fixed terminals. They could complement UMTS/IMT-
2000. One example is Teledesic.

155


Appendix. Follow-up questions
1. Channel coding gives improved frequency economy over a
propagation channel with fast fading. Explain.
2. List advantages of TDMA
3. List draw-backs of TDMA
4. Several of the experimental systems, which formed the background
for the GSM specification, gave the same frequency economy in spite
of different radio bandwidth per speech channel. Explain.
5. Which signal processing measures against multipath propagation
were included in the experimental system DMS-90?
6. The relation between C/I and speech quality is different for GSM and
for an analog mobile telephone system. Explain.
7. Describe the principle for the channel equalization at the
experimental system based on wide band TDMA.
8. Which considerations determined the system data rate and the length
of a TDMA frame at GSM.
9. Which were main reasons why wideband TDMA was rejected for
GSM.
10. Explain the meaning of SACCH and FACCH. Give a few examples
of their use.
11. List a few advantages and disadvantages of GMSK.
12. Explain the multiplexing together of the traffic channel and the
SACCH at GSM.
13. How is the TDMA structure arranged at GSM to permit both half-rate
and full-rate traffic channels?
14. Why are the data bursts used for the
RACCH (Random Access CCH) shortened?
15. Which maximum interleaving depth can be used at GSM for the
full-rate and half-rate traffic channels?
What limits the interleaving depth?
16. Why is frequency duplex used at GSM?
17. The transmit and receive time slots at GSM are separated by 1/3
frame time. Why?
18. Give two advantages of the frequency hopping procedure used at
GSM.
19. The TDMA structure at GSM comprises hyper, super and multi
frames in addition to the basic frame. Explain why.
20. Explain the concepts broadcast carrier and broadcast (signalling)
channel.
21. How is the initial synchronization of a terminal to the base station
accomplished?
22. Describe how a terminal registers at the base station.
23. Which information is used for the hand-over procedure?
24. Why is power regulation used at GSM?
25. What is meant with discontinuous transmission?

156


26. What is meant with discontinuous reception?
27. Why are two types of multiframes used at GSM?
28. Describe how a terminal measures the signal levels in the frequency
slots assigned to the adjacent cells, and determines from which cell
the signals originates.
29. Describe the channel coding procedure for class 1A bits from the
speech coder.
30. Why are 4 tail bits added to the sequence from the speech coder
before the convolution coding? Why is block coding used in addition
to the convolution coding?
31. Why are start and end bits added to the data bursts in the time slots in
the TDMA frame?
32. Why is the training sequence placed in the middle of each data burst?
33. Describe the contacts between HLR and VLR during call setup to a
roaming terminal.
34. Describe the principle for authentication.
35. Describe the main features of the enciphering procedure at GSM.
36. Which are the main differences between QAM (Nyquist filtering) and
MSK.
37. Describe the principle of channel equalization based on MLSE.
38. At a TDMA system, time overlaps at the base receiver input must be
avoided between data bursts coming from distant and from nearby
terminals. How is this solved at GSM?
39 AT GSM frequency hopping can be used as an option and it
necessary to use equalization against time dispersion. Discuss how
the performance of these system tools might be affected by:
a. by moving the frequency band up to 2 GHz
b. operation in micro and pico cells
40. AT GSM, which information is transmitted on:
a. the Stand-alone Dedicated Control Channel (SDCCH)
b. the Broadcast Control Channel (BCCH)
c. the Common Control Channel (CCCH)?
41. What is the main motive for using TDMA instead of FDMA at
D-AMPS?
42. At D-AMPS much longer TDMA frames and time slots are used than
at GSM. Why?
43. At D-AMPS, very small interleaving depth is used and no frequency
hopping. Why? Which are the consequences?
44. It has been discussed to introduce antenna diversity also at D-AMPS
terminals.
Which are the arguments for and against?
45. Antenna diversity is used at the terminals at PDC. Why?
46. Antenna diversity gives a certain reduction of the time dispersion.
Explain.
47. Why is QAM modulation with Nyquist filtering used at D-AMPS and
PDC (instead of GMSK used at GSM).
48. The TDMA structure at D-AMPS can be used both for half rate and
full rate traffic channels. Explain.

157


49. The signalling possibilities are more limited at D-AMPS (IS-54) than
at GSM. What is the reason?
50. At D-AMPS is used π/4-DQPSK.
a. Explain the meaning
b. Which are the advantages?
51. The requirements on the channel equalizer are different for GSM and
ADC. Explain.
52. a. The principle for the full rate speech coder at GSM is called
Regular Pulse Exited LPC with LTP. Explain.
b. The principle for the full rate speech coder at D-AMPS is called
“CELP” or “Vector-sum Exited LPC with LTP” and is based on
“analysis-through-synthesis”. Explain the meaning of these concepts.
53. Estimate the total width of the modulation spectrum at D-AMPS
based on the system data rate 48 kb/s and that the modulation is
4QAM with root rized square characteristic with α = 0,35.
54. Explain the concepts Radio LAN, Radio PABX, Tele Points and
Radio Local Loop.
55. Which are the main differences between the radio specifications for
DECT and CT2?
56. Explain how the frequency economy is improved at DECT through
dynamic channel allocation based on C/I measurements.
57. Base stations diversity can be used at DECT for both transmission
directions. Why is this possible?
58. Why is no frequency-cell planning needed at DECT?
59. The motives for error correction channel coding for speech
transmission are less at DECT than at GSM. Why?
60. Explain the MAHO features in connection with the digital signal
channel for D-AMPS
61. a. How is the adaptation to different user data rate obtained at HSCSD?
b. ARQ can be used at HSCSD. What are the advantages and
disadvantages?
62. How is the dynamic adaptation of the transmission parameters to the
quality of the radio channel obtained at GPRS?
63. Describe the combined FEC and BEC at EDGE to obtain dynamic
adaptation to the quality of the radio channel. What is the main
advantage of this arrangement compared with what is used at GPRS?
64. At EDGE, the maximum user data rate for terminals close to the cell
boundary is considerably less than 384 kb/s. Explain.
65. EDGE can be used for different cluster sizes. Explain.
66. Describe the reservation protocol used at GPRS for inward traffic.
67. AT EDGE can be used the modulation “linear, Gray-coded 3π/8-
8PSK”.
a. Why is the signal plane rotated between consecutive symbols.
Why is a rotation of 3π/8 used instead of π/8.
b. Originally also 16QAM was proposed. List a few pros and cons for
this modulation in comparation with 8PSK
c. 8PSK results in worse link budget than GMSK for two reasons (for
the same DC-power to the transmitter). Explain. Which are the
system consequences?

158


68. Which are the advantages and disadvantages to use:
a. satellites in lower orbits compared to geostationary satellites
b. satellites in a LEO orbit in corporation to an ICO orbit ?
69. Which are the motives to introduce land mobile satellite networks?
70. Which are the motives to introduce micro and pico cells?
71. Which are the motives to use hierarchical cell structures with
umbrella cells?

159


Author professor Sven-Olof Öhrvik
in cooperation with Ericsson Radio Systems AB
unit ERA/T, Core Unit Radio System and Technology
Publisher Ericsson Radio Systems AB
T/Z Ragnar Lodén

Ericsson Radio Systems AB
Torshamnsgatan 23, Kista
S-164 80 Stockholm, Sweden
Telephone: +46 8 757 00 00 EN/LZT 123 1246/1 R10
Telefax: +46 8 757 36 00 © Ericsson Radio Systems AB, 2000

160

Dm1 e

More Related Content

What's hot

Similar to Dm1 e

Dm1 e