As mentioned above, the standard bandwidth of 9.6 kbit/s (14.

4 kbit/s with some

providers) available for data transmission is not sufficient for the requirements of
today’s computers. When GSM was developed, not many people anticipated the
tremendous growth of data communication compared to voice communication.
At that time, 9.6 kbit/s was a lot, or at least enough for standard group 3 fax
machines. But with the requirements of, e.g., web browsing, file download, or
even intensive e-mail exchange with attachments, this is not enough.
To enhance the data transmission capabilities of GSM, two basic approaches
are possible. As the basic GSM is based on connection-oriented traffic channels,
e.g., with 9.6 kbit/s each, several channels could be combined to increase bandwidth.
This system is called HSCSD and is presented in the following section. A more progressive step is the introduction of
packet-oriented traffic in GSM, i.e.,
shifting the paradigm from connections/telephone thinking to packets/internet
thinking. The system, called GPRS, is presented in section 4.1.8.2.

4.1.8.1 HSCSD
A straightforward improvement of GSM’s data transmission capabilities is high
speed circuit switched data (HSCSD), which is available with some providers.
In this system, higher data rates are achieved by bundling several TCHs. An MS
requests one or more TCHs from the GSM network, i.e., it allocates several
TDMA slots within a TDMA frame. This allocation can be asymmetrical, i.e.,
more slots can be allocated on the downlink than on the uplink, which fits the
typical user behavior of downloading more data compared to uploading.
Basically, HSCSD only requires software upgrades in an MS and MSC (both have
to be able to split a traffic stream into several streams, using a separate TCH
each, and to combine these streams again).
In theory, an MS could use all eight slots within a TDMA frame to achieve
an air interface user rate (AIUR) of, e.g., 8 TCH/F14.4 channels or 115.2 kbit/s
(ETSI, 1998e). One problem of this configuration is that the MS is required to
send and receive at the same time. Standard GSM does not require this capability
– uplink and downlink slots are always shifted for three slots. ETSI (1997a)
specifies the AIUR available at 57.6 kbit/s (duplex) using four slots in the uplink
and downlink (Table 4.2 shows the permitted combinations of traffic channels
and allocated slots for non-transparent services).
Although it appears attractive at first glance, HSCSD exhibits some major
disadvantages. It still uses the connection-oriented mechanisms of GSM. These
are not at all efficient for computer data traffic, which is typically bursty and
asymmetrical. While downloading a larger file may require all channels
reserved, typical web browsing would leave the channels idle most of the time.
Allocating channels is reflected directly in the service costs, as, once the channels
have been reserved, other users cannot use them.

4.1.8.2 GPRS
The next step toward more flexible and powerful data transmission avoids the
problems of HSCSD by being fully packet-oriented. The general packet radio service
(GPRS) provides packet mode transfer for applications that exhibit traffic
patterns such as frequent transmission of small volumes (e.g., typical web
requests) or infrequent transmissions of small or medium volumes (e.g., typical
web responses) according to the requirement specification (ETSI, 1998a).
Compared to existing data transfer services, GPRS should use the existing network
resources more efficiently for packet mode applications, and should provide
a selection of QoS parameters for the service requesters. GPRS should also allow
for broadcast, multicast, and unicast service. The overall goal in this context is
the provision of a more efficient and, thus, cheaper packet transfer service for
typical internet applications that usually rely solely on packet transfer. Network
providers typically support this model by charging on volume and not on connection
time as is usual for traditional GSM data services and for HSCSD. The
main benefit for users of GPRS is the ‘always on’ characteristic – no connection
has to be set up prior to data transfer. Clearly, GPRS was driven by the tremendous
success of the packet-oriented internet, and by the new traffic models and
applications. However, GPRS, as shown in the following sections, needs additional
network elements, i.e., software and hardware. Unlike HSCSD, GPRS does
not only represent a software update to allow for the bundling of channels, it also
represents a big step towards UMTS as the main internal infrastructure needed for
UMTS (in its initial release) is exactly what GPRS uses (see section 4.4).
The main concepts of GPRS are as follows (ETSI, 1998b). For the new GPRS
radio channels, the GSM system can allocate between one and eight time slots
within a TDMA frame. Time slots are not allocated in a fixed, pre-determined
manner but on demand. All time slots can be shared by the active users; up- and
downlink are allocated separately. Allocation of the slots is based on current
load and operator preferences. Depending on the coding, a transfer rate of up to
170 kbit/s is possible. For GPRS, operators often reserve at least a time slot per
cell to guarantee a minimum data rate. The GPRS concept is independent of
channel characteristics and of the type of channel (traditional GSM traffic or
control channel), and does not limit the maximum data rate (only the GSM
transport system limits the rate). All GPRS services can be used in parallel to
conventional services. Table 4.3 shows the typical data rates available with GPRS
if it is used together with GSM (GPRS can also be used for other TDMA systems).
Users of GPRS can specify a QoS-profile. This determines the service precedence
(high, normal, low), reliability class and delay class of the transmission,
and user data throughput. GPRS should adaptively allocate radio resources to
fulfill these user specifications. Table 4.5 shows the three reliability classes
together with the maximum probabilities for a lost service data unit (SDU), a
duplicated SDU, an SDU out of the original sequence, and the probability of
delivering a corrupt SDU to the higher layer. Reliability class 1 could be used for
very error-sensitive applications that cannot perform error corrections themselves.
If applications exhibit greater error tolerance, class 2 could be
appropriate. Finally, class 3 is the choice for error-insensitive applications or
applications that can handle error corrections themselves.
Delay within a GPRS network is incurred by channel access delay, coding
for error correction, and transfer delays in the fixed and wireless part of the
GPRS network. The delay introduced by external fixed networks is out of scope.
However, GPRS does not produce additional delay by buffering packets as storeand-
forward networks do. If possible, GPRS tries to forward packets as fast as
possible. Table 4.6 shows the specified maximum mean and 95 percentile delay
values for packet sizes of 128 and 1,024 byte. As we can clearly see, no matter
which class, all delays are orders of magnitude higher than fixed network
delays. This is a very important characteristic that has to be taken into account
when implementing higher layer protocols such as TCP on top of GPRS networks
(see chapter 9). Typical round trip times (RTT) in fixed networks are in
the order of 10 to 100 ms. Using real unloaded GPRS networks round trip times
of well above 1 s for even small packets (128–512 byte) are common.
Additionally, GPRS exhibits a large jitter compared to fixed networks (several
100 ms are not uncommon). This characteristic has a strong impact on user
experience when, e.g., interactive Internet applications are used on top of GPRS.
Before sending any data over the GPRS network, an MS must attach to it,
following the procedures of the mobility management. The attachment procedure
includes assigning a temporal identifier, called a temporary logical link
identity (TLLI), and a ciphering key sequence number (CKSN) for data
encryption. For each MS, a GPRS context is set up and stored in the MS and inthe corresponding SGSN. This context
comprises the status of the MS (which
can be ready, idle, or standby; ETSI, 1998b), the CKSN, a flag indicating if compression
is used, and routing data (TLLI, the routing area RA, a cell identifier,
and a packet data channel, PDCH, identifier). Besides attaching and detaching,
mobility management also comprises functions for authentication, location
management, and ciphering (here, the scope of ciphering lies between MS and
SGSN, which is more than in standard GSM). In idle mode an MS is not reachable
and all context is deleted. In the standby state only movement across
routing areas is updated to the SGSN but not changes of the cell. Permanent
updating would waste battery power, no updating would require system-wide
paging. The update procedure in standby mode is a compromise. Only in the
ready state every movement of the MS is indicated to the SGSN.
Figure 4.17 shows the protocol architecture of the transmission plane for
GPRS. Architectures for the signaling planes can be found in ETSI (1998b). All
data within the GPRS backbone, i.e., between the GSNs, is transferred using the
GPRS tunnelling protocol (GTP). GTP can use two different transport protocols,
either the reliable TCP (needed for reliable transfer of X.25 packets) or the
non-reliable UDP (used for IP packets). The network protocol for the GPRS backbone
is IP (using any lower layers). To adapt to the different characteristics of
the underlying networks, the subnetwork dependent convergence protocol
(SNDCP) is used between an SGSN and the MS. On top of SNDCP and GTP, user
packet data is tunneled from the MS to the GGSN and vice versa. To achieve a
high reliability of packet transfer between SGSN and MS, a special LLC is used,
which comprises ARQ and FEC mechanisms for PTP (and later PTM) services.
4.2 DECT
Another fully digital cellular network is the digital enhanced cordless telecommunications
(DECT) system specified by ETSI (2002, 1998j, k), (DECT Forum, 2002).
Formerly also called digital European cordless telephone and digital European
cordless telecommunications, DECT replaces older analog cordless phone systems
such as CT1 and CT1+. These analog systems only ensured security to a limited
extent as they did not use encryption for data transmission and only offered a relatively
low capacity. DECT is also a more powerful alternative to the digital system
CT2, which is mainly used in the UK (the DECT standard works throughout
Europe), and has even been selected as one of the 3G candidates in the IMT-2000
family (see section 4.4). DECT is mainly used in offices, on campus, at trade shows,
or in the home. Furthermore, access points to the PSTN can be established within,
e.g., railway stations, large government buildings and hospitals, offering a much
cheaper telephone service compared to a GSM system. DECT could also be used to
bridge the last few hundred meters between a new network operator and customers.
Using this ‘small range’ local loop, new companies can offer their service without
having their own lines installed in the streets. DECT systems offer many different
interworking units, e.g., with GSM, ISDN, or data networks. Currently, over 100 million
DECT units are in use (DECT, 2002).
A big difference between DECT and GSM exists in terms of cell diameter
and cell capacity. While GSM is designed for outdoor use with a cell diameter of
up to 70 km, the range of DECT is limited to about 300 m from the base station
(only around 50 m are feasible inside buildings depending on the walls). Due to
this limited range and additional multiplexing techniques, DECT can offer its
service to some 10,000 people within one km2. This is a typical scenario within
a big city, where thousands of offices are located in skyscrapers close together.
DECT also uses base stations, but these base stations together with a mobile station
are in a price range of €100 compared to several €10,000 for a GSM base
station. GSM base stations can typically not be used by individuals for private
networks. One reason is licensing as all GSM frequencies have been licensed to
network operators. DECT can also handle handover, but it was not designed to
work at a higher speed (e.g., up to 250 km/h like GSM systems). Devices
handling GSM and DECT exist but have never been a commercial success.
DECT works at a frequency range of 1880–1990 MHz offering 120 full
duplex channels. Time division duplex (TDD) is applied using 10 ms frames.
The frequency range is subdivided into 10 carrier frequencies using FDMA, each
frame being divided into 24 slots using TDMA. For the TDD mechanism, 12 slots are used as uplink, 12 slots as downlink
(see Figure 3.4). The digital
modulation scheme is GMSK – each station has an average transmission power
of only 10 mW with a maximum of 250 mW.
4.3 TETRA
Trunked radio systems constitute another method of wireless data transmission.
These systems use many different radio carriers but only assign a specific carrier to
a certain user for a short period of time according to demand. While, for example,
taxi services, transport companies with fleet management systems and rescue
teams all have their own unique carrier frequency in traditional systems, they can
share a whole group of frequencies in trunked radio systems for better frequency
reuse via FDM and TDM techniques. These types of radio systems typically offerinterfaces to the fixed telephone network,
i.e., voice and data services, but are not
publicly accessible. These systems are not only simpler than most other networks,
they are also reliable and relatively cheap to set up and operate, as they only have
to cover the region where the local users operate, e.g., a city taxi service.
To allow a common system throughout Europe, ETSI standardized the
TETRA system (terrestrial trunked radio)9 in 1991 (ETSI, 2002), (TETRA MoU,
2002). This system should replace national systems, such as MODACOM, MOBITEX
and COGNITO in Europe that typically connect to an X.25 packet network.
(An example system from the US is ARDIS.) TETRA offers two standards: the
Voice+Data (V+D) service (ETSI, 1998l) and the packet data optimized (PDO)
service (ETSI, 1998m). While V+D offers circuit-switched voice and data transmission,
PDO only offers packet data transmission, either connection-oriented
to connect to X.25 or connectionless for the ISO CLNS (connectionless network
service). The latter service can be point-to-point or point-to-multipoint, the typical
delay for a short message (128 byte) being less than 100 ms. V+D
connection modes comprise unicast and broadcast connections, group communication
within a certain protected group, and a direct ad hoc mode without a
base station. However, delays for short messages can be up to 500 ms or higher
depending on the priority.
TETRA also offers bearer services of up to 28.8 kbit/s for unprotected data
transmission and 9.6 kbit/s for protected transmission. Examples for end-to-end
services are call forwarding, call barring, identification, call hold, call priorities,
emergency calls and group joins. The system architecture of TETRA is very similar
to GSM. Via the radio interface Um, the mobile station (MS) connects to the
switching and management infrastructure (SwMI), which contains the user
data bases (HDB, VDB), the base station, and interfaces to PSTN, ISDN, or PDN.
The system itself, however, is much simpler in real implementation compared
to GSM, as typically no handover is needed. Taxis usually remain within a certain
area which can be covered by one TETRA cell.
Several frequencies have been specified for TETRA which uses FDD (e.g.,
380–390 MHz uplink/390–400 MHz downlink, 410–420 MHz uplink/420–430
MHz downlink). Each channel has a bandwidth of 25 kHz and can carry
36 kbit/s. Modulation is DQPSK. While V+D uses up to four TDMA voice or data
channels per carrier, PDO performs statistical multiplexing. For accessing a
channel, slotted Aloha is used.
4.4 UMTS and IMT-2000
A lot has been written about third generation (or 3G) networks in the last few
years. After a lot of hype and frustration these networks are currently deployed
in many countries around the world. But how did it all start? First of all, the
International Telecommunication Union (ITU) made a request for proposals for
radio transmission technologies (RTT) for the international mobile telecommunications
(IMT) 2000 program (ITU, 2002), (Callendar, 1997), (Shafi, 1998).
IMT-2000, formerly called future public land mobile telecommunication system
(FPLMTS), tried to establish a common worldwide communication system that
allowed for terminal and user mobility, supporting the idea of universal personal
telecommunication (UPT). Within this context, ITU has created several
recommendations for FPLMTS systems, e.g., network architectures for FPLMTS
(M.817), Requirements for the Radio Interface(s) for FPLMTS (M.1034), or
Framework for Services Supported by FPLMTS (M.816). The number 2000 in
IMT-2000 should indicate the start of the system (year 2000+x) and the spec-trum used (around 2000 MHz). IMT-2000
includes different environments such
as indoor use, vehicles, satellites and pedestrians. The world radio conference
(WRC) 1992 identified 1885–2025 and 2110–2200 MHz as the frequency
bands that should be available worldwide for the new IMT-2000 systems
(Recommendation ITU-R M.1036). Within these bands, two times 30 MHz have
been reserved for mobile satellite services (MSS).
Figure 4.22 shows the ITU frequency allocation (from the world administrative
radio conference, 1992) together with examples from several regions that
already indicate the problem of worldwide common frequency bands. In Europe,
some parts of the ITU’s frequency bands for IMT-2000 are already allocated for
DECT (see section 4.2). The remaining frequencies have been split into bands for
UTRA-FDD (uplink: 1920–1980 MHz, downlink: 2110–2170 MHz) and UTRATDD
(1900–1920 MHz and 2010–2025 MHz). The technology behind UTRA-FDD
and –TDD will subsequently be explained in more detail as they form the basis of
UMTS. Currently, no other system is planned for IMT-2000 in Europe. More
bandwidth is available in China for the Chinese 3G system TD-SCDMA or possibly
other 3G technologies (such as W-CDMA or cdma2000 – it is still open
which system will dominate the Chinese market; Chen, 2002). Again slightly different
frequencies are used by the 3G services in Japan, which are based on
W-CDMA (like UTRA-FDD) or cdma2000. An open question is the future of 3G in
the US as the ITU’s frequency bands have already been allocated for 2G networks
or are reserved for other use. In addition to the original frequency allocations,
the world radio conference (WRC) allocated new terrestrial IMT-2000 bands in
the range of 800–1000 MHz, 1700–1900 MHz and 2500–2700 MHz in 2000.

Description
One innovation in HSCSD is to allow different error correction methods to be used for data
transfer. The original error correction used in GSM was designed to work at the limits of
coverage and in the worst case that GSM will handle. This means that a large part of the
GSM transmission capacity is taken up with error correction codes. HSCSD provides
different levels of possible error correction which can be used according to the quality of the
radio link. This means that in the best conditions 14.4 kbit/s can be put through a single time
slot that under CSD would only carry 9.6 kbit/s, for a 50% improvement in throughput.

The other innovation in HSCSD is the ability to use multiple time slots at the same time.
Using the maximum of four time slots, this can provide an increase in maximum transfer rate
of up to 57.6 kbit/s (i.e., 4 × 14.4 kbit/s) and, even in bad radio conditions where a higher
level of error correction needs to be used, can still provide a four times speed increase over
CSD (38.4 kbit/s versus 9.6 kbit/s). By combining up to eight GSM time slots the capacity
can be increased to 115 kbit/s.

HSCSD requires the time slots being used to be fully reserved to a single user. It is possible
that either at the beginning of the call, or at some point during a call, it will not be possible
for the user's full request to be satisfied since the network is often configured to allow normal
voice calls to take precedence over additional time slots for HSCSD users.

The user is typically charged for HSCSD at a rate higher than a normal phone call (e.g., by
the number of time slots allocated) for the total period of time that the user has a connection
active. This makes HSCSD relatively expensive in many GSM networks and is one of the
reasons that packet-switched general packet radio service (GPRS), which typically has lower
pricing (based on amount of data transferred rather than the duration of the connection), has
become more common than HSCSD.

Apart from the fact that the full allocated bandwidth of the connection is available to the
HSCSD user, HSCSD also has an advantage in GSM systems in terms of lower average radio
interface latency than GPRS. This is because the user of an HSCSD connection does not have
to wait for permission from the network to send a packet.

HSCSD is also an option in enhanced data rates for GSM evolution (EDGE) and universal
mobile telephone system (UMTS) systems where packet data transmission rates are much
higher. In the UMTS system, the advantages of HSCSD over packet data are even lower
since the UMTS radio interface has been specifically designed to support high bandwidth,
low latency packet connections. This means that the primary reason to use HSCSD in this
environment would be access to legacy dial up systems.