INTERNATIONAL JOURNAL OF SATELLITE COMMUNICATIONS, VOL. 14.

467-474 ( 1996)

FUTURE MULTIMEDIA COMMUNICATIONS VIA

SATELLITE
B. G . EVANS A N D R. TAFAZOLLI
Centre for Satellite Engineering Research, University of Surrey, Guildford, GU2 5 X H , U.K.

SUMMARY
We are now moving into an era when conventional speech service will expand to full multimedia
service offerings. This paper looks at ways in which satellite systems could provide such services in
both the mobile as well as the fixed bands and the integration and inter-operability of such systems.
Provision of multimedia services by ATM to integrate with an ATM core network is taken as the
basis of the future network.
Some of the constraints and difficulties of provision of ATM over satellite are discussed as well as
some potential solutions for the ATM wireless protocols and network architectures which potentially
could provide such services. Finally, some suggestions of how new orbit constellations could help meet
these requirements are given. 0 1996 by John Wiley & Sons, Ltd.

KEY WORDS: ATM over satellite; wireless ATM; integration; M E 0

1. INTRODUCTION early experimental results in Europe and the U.S.A.,

and the beginnings of some commercial offerings
Satellite communications for the past 5 years have for fixed point-to-point links which are promising,
been dominated by mobile and personal applications, although some problems still need to be addressed.
and we have seen major changes with the use of The interesting question is whether satellites can
low earth orbit (LEO) and medium earth orbit help extend the ATM bubble outwards from the
(MEO) orbits to join the geostationary earth orbit backbone network to the users. There are appli-
(CEO) in the first generation of satellite personal cations for lower rate access to the superhighway
communications networks. As mobile systems move from VSATs which need special networking con-
into their third generation we are witnessing the siderations. There is the whole question of mobility
integration of cellular and satellite into universal and whether ATM quality can be provided and
mobile telecommunication system (UMTS/ maintained, and for what classes of services and
FPLMTS), but are still being dominated by speech with what mobility constraints.
communications with the promise of multimedia to We clearly have two perspectives (Figure I ) on
come-but can this be delivered in frequency bands the situation of future radio multimedia delivery. On
(L,S) with much constrained spectrum and satellite the one hand we have the UMTSFPLMTS approach
systems technology? What do we mean by multime- which is moving to provide integrated access to
dia for mobile and personal communications and services currently available in different systems
where are the markets? This paper addresses these (cellular and cordless) and to incorporate an infra-
topics and shows that only a restricted class of structure which will allow both satellite and terres-
multimedia is possible for mobile systems, and pro- trial components. The aim is to provide global roam-
vides some pointers to future market application ing and add some high performance services to the
areas. speech services currently available. With the first
The world of multimedia communications is very generation UMTS coverage mobility is likely to be:
much driven by the backbone superhighway concept,
Home/office - pic0 cell 2 2 Mb/s
which it is assumed will largely be provided by
Local area - micro cell 2 385 kb/s
fibre optic cables. In many parts of the world,
Regional cellular - macro cell 2 1 4 4 kb/s
however, this is impossible, uneconomic or will take
Global satellite - mega cell 5 64 kb/s
too long to establish. It thus proposed that satellites
do have a niche role in extending or bringing into The UMTS specturm will be limited as follows:
being superhighways (GIIs) in the short-term- It is
0 1920-1980 MHz Terrestrial--6Qt60 MHz
very likely that asynchronous transfer mode ( ATM)
21 10-2170 MHz
will form the transmission standard for future super-
0 1980-2010 MHz Satellite-30+30 MHz
highways and this technique has largely been stan-
2 170-2200 MHz
dardized by terrestrial operators without recourse to
satellite operations. Can satellites thus provide the and even with estimated extensions the total spec-
quality of service expected from ATM? We have trum will not be much in excess of 200MHz which

CCC 0737-2884/96/060467-08 Received October I996

0 1996 by John Wiley & Sons, Ltd.
468 B. C. E V A N S A N D R. TAFAZOLLI

will limit the multimedia services to around 2 Mb/s of geostationary satellites to cover the North Amer-
and the mobility to pic0 cells. It cannot thus be ican continent and to deliver direct-to-home (DTH)
envisaged that satellite multimedia much above video and data services. The North American market
64 kb/s (MPEG4) can be conceived from the is ideally suited to this type of service as its urban
UMTUFPLMTS systems. However, UMTS provides centres are separated by large distances, total popu-
a valid standardized platform for interfaces and a lation density is low with a high percentage in non-
secure path for mobile migration towards a global urban environments, and satellite dish penetration is
environment. relatively high. In addition to these physical charac-
The second perspective is the use of satellites in teristics the GDP is high enough for service oper-
the context of the B-ISDN which comes from the ators to expect a high take-up of leisure services
fixed network ideas based upon ATM delivery or a (video-on-demand, computer games retrieval) for
‘Superhighway’. Standards introduced in these areas which the geostationary satellite is an ideal delivery
emanate from the ‘ATM Forum’ and the ITU and system (non-delay critical, asymmetric links). Other
have been led by terrestrial considerations. Only countries which have similar physical characteristics
now is the ATM Forum beginning to consider radio (not including satellite dish penetration) may not
access and the problem of extending the ATM bub- have the spending power with which to justify this
ble to include the radio channel. Satellites have regionally based type of system, although as can be
increased delays and channels which produce more seen with economic growth in China and south-east
errors than terrestrial. Note also that MPEG stan- Asia this may not always be the case.
dards for the source coding and organization of For business and computing applications, services
multimedia services assume a perfect channel. It is may have different characteristics and QoS require-
thus by no means clear that satellites can provide ments. Services such as video conferencing and
the quality of service (QoS) defined for ATM on high-speed computing are more delay critical and
terrestrial channels or the degree of mobility that thus are restricted in the way they can can be
will be achievable. From the preceding discussion delivered, in terms of satellite altitude. Users of
on spectrum availability it is clear that for full business and computing services may require a sys-
multimedia services via satellite one is forced to tem which can cope with both nomadic behaviour,
move out of L/S band and towards the new Ka- and would be flexible enough to respond to rapid
band allocations. Hence the recent Ka-band satellite changes in physical office location. In short, a user
filings. So we see a new positioning of satellite does not have to worry about service coverage if
multimedia coming from the two areas as shown he subscribes to a global system, and only a satellite
in Figure 2. network can provide a truly global service. One
system has been proposed which attempts to emulate
the terrestrial internet by establishing a large net-
2. MULTIMEDIA SERVICES AND
work of satellites as switching nodes within a sky-
APPLICATIONS
based computer network. In some quarters the large
number of satellites proposed for this system is
2.1. The market for satellite multimedia services
regarded as prohibitive but the concept is one that
Fibre-based systems will always be able to deliver is suited for solving the problem of nomadic com-
multimedia services more cheaply than satellite puting and communications.
systems-where it is economical to deploy the infra-
structure, the same can be said of terrestrial wireless
2.3. Technical aspects of satellite service provision
mobile services. Areas which do not have a high
enough traffic density, or favourable terrain, to jus- For all satellite systems, major concerns are the
tify the installation of fibre or terrestrial mobile satellites’ ability to deliver services with the required
trans receivers, will be more economically served QoS. Satellites are power limited (or sometimes
by satellite systems. Maritime and aeronautical users interference limited), and it is also desirable that
can also be served more easily by satellite systems, user terminals are made as small and as cheaply as
although for most multimedia applications these may possible, whether they be fixed, transportable or
be viewed as very small niche markets. The major mobile terminals. Bit error rate (BER), cell dropping
question is whether or not a large enough market rate and propagation delay become very important
for these services exists outside the coverage of the as measures of the QoS. In non-geostationary sys-
terrestrial infrastructure. tems the dynamics of the satellites (elevation, path
length, inter-satellite interference) will introduce a
new level of channel variation, and smaller, more
2.2. Current proposed satellite multimedia systems
personal user terminals will be less able to cope
Currently there have been many filings made to with increasing link margins. Any multimedia satel-
the ITU in which satellites have been proposed for lite system must address these problems, and firstly
the delivery of Ka band multimedia services to fixed should estimate the requirements for the provision
systems and VSAT type systems. These systems are of multimedia services to the relevant terminal types.
predominantly American and involve the deployment Satellite systems bring the benefits of point-to-
 FUTURE MULTIMEDIA COMMUNICATIONS 469

FlXED PERSPECI'IVE

Technology

1988 2005
MOBILE PERSPECTIVE
Figure 1 . The two satellite multimedia bubbles can either develop in parallel and be cemented via the core fixed ATM network or they
can be integrated into one system that supports a range of mobility

OR

Figure 2. Interconnection of satellite multimedia systems

point and point-to-multipoint service provision, rapid Table I illustrates typical QoS parameters for
deployment and universal access which cannot be multimedia services delivered through an ATM-
provided by other means. based satellite system.
Services can be subdivided into constant bit rate From the various trials within RACE 11, EUTEL-
(CBR) and variable bit rate (VBR). ATM provides SAT and INTELSAT indicators are that the CLR
real advantages for the latter in being able to accom- may be the critical parameter that determines the
modate the bandwidth on-demand features within link budget. On the satellite link the major differ-
the cell structure. Within the ATM definition QoS ences are thus:
is defined as a combination of BER, cell loss ratio
(CLR j cell delay tolerance (CDT). (i) The CLRIBER conditions will need advanced
The ATM Forum defines the QoS metrics between channel coding to overcome the bursty error
ATM layers with no recourse to the physical layer conditions on the channel.
(see Figure 3). However QoS parameters over satel- (ii) Availability conditions will need careful con-
lite are inextricably linked to the physical layer sideration as the satellite links, in general,
performance, and studies have shown that the fol- will not have the ability to re-route within
lowing is a reasonable reflection of the ATM Forum short periods.
requirements over a fixed satellite system. (iii) Long propagation delays will result in inef-
470 B. G. EVANS A N D R. TAFAZOLLI

Figure 3. Layered model of satellite.ATM

Table I.

Applications CLR CDT CDT BW BW BL BL

min max min max min max
(s) (s) (kb/s)(kbls) (kb) (kb)

Interlan 10-12 10-2 0.1 103 105 0.1 1

On-line transactions I 0-9 1 3 64 104 0.1 0.3
Electronic mail I 0-9 1 10 20 103 0.05 5
PC file transfer I 0-9 10 100 20
64 1 103
Group-4 facsimile 10-8 4 10 64
64 256 640
Data broadcasting 10-12 TBD TBD 9.6 64 nla n/a
Video conferencing I 0-9 I50 ms 350ms 128 104 1.6 40
Video telephony I 0-9 I50 ms 5 0 m s 64 2 2 10
Voice predictive coding I o-6 10 ms 150ms 16 n/a 2 3
Video mail 10-10 1 ms 5 m s 128 2 8 125
Multimedia conferencing I 0-9 TBD TBD TBD TBDTBD TBD

BL, Burst length; CDT, cell delay tolerance; BW, bandwidth; CLR, cell loss ratio.

ficient use of the satellite resource due to 0 Fixed MDR 52-100Mb/s

large numbers of retransmissions in adverse
conditions. Management of BOD is a prob-
lem. 3. ATM OVER SATELLITES

The ATM Forum has also developed specifi- ATM was designed for transmission over channels
cations which enable the use of multimedia services with very good error characteristics. For this reason
over ATM adaptation layers in conjunction with the protocols used in fixed B-ISDN networks were
MPEG. For MPEG2 systems over satellite with CBR simplified having less overhead and better through-
these are already in existence but for VBR only put. The broadband ISDN extension to mobile users
limited trials have been performed. MPEG2 is seen via satellite links can cause problems mainly due to
as the higher quality fixed service applications, the poor performance of the radio link and the extra
whereas the newer and less well defined MPEG4 at signalling needed to support the dynamic network
around 64 kb/s will form the basis of the mobile resource allocation (in a more general scenario as
multimedia specification. It is important to realize shown in Figure4, a satellite B-ISDN network can
that the MPEG/ATM VBR type services are still be dynamically reconfigured due to the move of the
very experimental and are not well validated. Satel- satellite nodes in addition to that of the mobile
lites need to provide variable bit and bandwidth on users).
demand services with data rates ramping from kb/s
to Mbls. Seamless integration of satellite and terres- 3.1. Advantages of using satellites for user access
trial systems is essential in order to provide global to the B-ISDN services
coverage and broadband access to the core network.
Study might indicate a range of user terminals as
0 ATM satellite links can offer bandwidth on
follows:
demand;
0 Mobile laptop I 6 4 kb/s 0 Allow higher flexibility concerning topology,
0 Transportable LDR I 1 28 kb/s reconfiguration and network expansion;
0 Transportable MDR I 2 Mb/s 0 Satellites offer broadcast and multipoint to
0 Fixed LDR I 2 M b / s multipoint capabilities;
 FUTURE MULTIMEDIA COMMUNICATIONS 471

m- LEO/MEO/GEO Satellite Constellation

Figure 4. Fixed/mobile ATM connections via satellite

0 Satellites can be attractive in rural and remote exists in an ATM header the entire cell is lost.
local loops. The affected QoS parameter in that case is
CLR, otherwise, an undetected error leads to
Advantages of using end-to-end ATM transmission the degradation of the cell mininsertion rate
over satellites for the user access to the B-ISDN ser- (CMR).
vices 0 At the ATM layer only the header is protected
such that errors in the information field are not
0 Same applications and protocols for call and
detected at this layer. In some AAL layers
bearer control can be used with the fixed net-
some error detection mechanism exists. The
work;
ATM QoS parameters that describe payload
0 Less interworking functionality.
errors are the CER and the severely errored
cell ratio (SECR).
3.2. Disadvantages of using end-to-end ATM
transmission over satellites for user access to the Media transmission considerations
B-ISDN services 0 Radio link constraints
0 BER performance issues
Delay jitter especially affects the voice/ multi- 0 Bit transmission and channel coding
media traffic;
Long transmission delays affect the throughput
Network considerations
of data protocols due to window sizelack con-
straints; 0 Dynamic network configuration
During the transmission wireless links may vary 0 Call admission control for mobile users
between good or very bad resulting in bursty 0 Satellite constellation
errors; 0 ATM user mobility
Retransmission of corrupted data could be 0 Handover
problematic due to:
0 the alteration of the strict sequence data
transmission; 3.3. Protocol considerations and solutions
0 protocol time-outs that are more likely to The operation of ATM via satellite can cause
be exceeded; serious problems to the existing protocols at differ-
In systems with very large bandwidth x delay ent layers of the B-ISDN protocol stack. The intro-
product interface equipment will require very duction of an ATM-link enhancement technique
large buffers to hold the transmitted frames; (ALE) developed by COMSAT,' gives a great
FEC can correct single bit errors and can detect improvement of the ATM cell loss rate on a satellite
multiple bit errors. If more than 1 bit error link. The ALE is inserted between the switch and
472 B. G. EVANS AND R. TAFAZOLLI

the satellite modem and incorporates a selective Management Plane

interleaving which does not impose any additional Control Plane
overheads. 7 /- User
The performance of AAL protocols on an ATM
satellite link depends on their error
correction/detection capability. For example AAL 1
(for user data with constant source bit rate and
employs a 3-bit redundancy check CRC and a parity
check on the SAR header) would most probably
cause a synchronization error at the destination when
Figure 5 . Modified B-ISDN wireless protocol reference model
undetected errors occur. The solution to this problem
is again to interleave the first byte of the AAL 1
payload.I
SDUs from the higher layer into packets that can
AAL 3/4 (user data with a variable source bit
be different in length from the standard ATM cell.
rate and a 10-bit CRC at the SAR level) is parti-
The length of the packets that are sent to the
cularly strong. The probability of undetected errors
underlying wireless packet transfer layer (similar to
can be greatly reduced if full-byte interleaving is
the ATM layer), can be optimized according to the
performed on the ATM cell payload that distributes
specific radio access technique needs.
the burst error into two AAL payloads.
The physical layer is responsible for the trans-
Finally, AAL 5 offers services similar to AAL 314
mission bit rates, modulation, logical-to-physical
but employs a 32-bit CRC which is a much stronger
channel mapping, and channel coding. In addition,
code for error detection. A length-check field is also
the physical layer is responsible for signal strength
present and undetected errors would rarely occur on
measurements, RF channel selection and power con-
a general FEC-coded satellite channel.
trol.
The SSCOP protocol that was originally designed
In another proposal for a personal communication
to provide error-free transmission of the signalling
network (PCN)4 a hierarchical ATM switching net-
messages can operate quite well in satellite net-
work is used for interconnection of PCN microcells,
works. SSCOP uses a very efficient selective retrans-
each of which is serviced by high speed shared
mission mechanism, a large enough (24-bit) modulus
access radio links based on ATM-compatible cell
sequence numbering mechanism for the data units,
relay principles.
a rate-based flow control mechanism and provides
network delay insensitivity. Because the use of the 0 In the presented approach, the protocol layering
current data link and transport layer protocols e.g. is harmonized with the ATM protocol stack
HDLC, TCP over satellite links results in inef- and new wireless, channel specific, physical
ficiencies, it is expected that SSCOP might be used medium access control and data link layers are
for applications with high-speed data transfer. added below the ATM layer (Figure 6). The
regular ATM network layer and control services
such as call set-up, VCI/VPI addressing, cell
3.4. Future wireless ATM protocol architectures
prioritization, and flow control indication will
Some proposals for the future wireless B-ISDN continue to be used for mobile services. Modi-
services, have already been presented in the fications to the existing signalling protocols
for the broadband wireless terrestrial are needed to support specific mobility related
networks. The general ideal behind those proposals function MAC layer considerations: a MAC
is to keep the same platform in the wireless protocol
architecture as is for the fixed networks, and to add
the necessary functionality to provide mobility and
improve the wireless link characteristics. The satel-
lite link characteristics and the related problems are
7 1
not addressed on those proposals. However, they
provide a reference point for a future satellite mobile
protocol architecture.
I ATM Adaptation Layer (AAL)
I
A modified B-ISDN protocol reference model that
is presented for future wireless networks uses a
packet transfer layer instead of the standard ATM
layer and on top of it the wireless adaptation layer
(Figure 5 ) . The wireless AAL layer depends on the
service and the radio access technique.' In this layer,
several different protocols might be fitted, each one
related to a particular service and radio access
scheme (TDMA, CDMA, PRMA etc.). As in the I Physical Layer
I
standard AAL layer, the wireless AAL converts the Figure 6. Wireless ATM protocol stack for PCN
 FUTURE MULTIMEDIA COMMUNICATIONS 473

layer design approach is needed for the wireless 4.2. ATM complications
multimedia applications. The two basic candi-
ATM is a connection-orientated protocol; after
dates for media access control CDMA, TDMA
call establishment it is assumed the data path will
may co-exist in future wireless ATM networks.
remain fixed for the duration of the call. Problems
Data link layer considerations: this layer is
will arise due to the dynamically changing satellite
needed to provide an adequate error recovery
channel between the mobile terminal and the fixed
mechanism from the wireless medium specific
earth station (FES). For example, due to the
problems (high bit error rates at the physical
dynamic motion of a MEO/LEO satellite, the chan-
layer) and from the delays or blocking at the
nel to the ATM switch may be lost. Other problems,
MAC layer. 'The ATM header could be com-
such as varying delay and handover need to
pressed into an abbreviated network layer
addressed as these features are inherent in a dynamic
header and then add a wireless medium-specific
satellite constellation.
data link header. The reduced VCI should be
acceptable for the mobile limited addressing
requirements for a single wireless channel, and 4.3. Possible solutions
can be expanded to standard ATM form at the
base station interfa~e'.~ As described above delivering high bit rate ser-
vices via MEOLEO satellites will require a large
number of satellites. With a MEOLEO satellite
constellation a large number of FES are required
4. SOME POSSIBLE APPROACHES due to the lower orbit altitude. This increase in the
number of satellites and FES could lead to a system
To deliver high bit rate ATM services to mobile that is not economically viable. To reduce the num-
terminals via satellite will require low delays, low ber of FES, inter-satellite links could be used. The
terminal power requirements and high minimum use of on-board ATM switches together with ISLs
elevation angles to maximize the probability. of a opens up the possibility of flexible routing and a
good channel. Use of GEO stationary satellites possible means of achieving the availability, BER
means lower elevation angles and large delays in and delay requirements in a similar way to that used
high latitude regions. These problems can be solved in terrestrial networks. The satellite constellation
by the use of satellites at much lower altitudes such described below addresses each of the issues raised
as ME0 and LEO orbit heights. Satellites at these above. The constellation is an optimized ME0 orbit,
lower altitudes have much smaller delays and lower offering lower delays and lower power requirements
terminal power requirements than satellites in a GEO than GEO satellites. The GEO satellite constellation
orbit. By using MEO/LEO satellites and selecting also requires fewer FES and offers higher elevation
an appropriate inclination angle these orbits can angles than a LEO constellation with a similar
offer much higher minimum elevation angles over numbers of satellites.
high latitude regions. Shown in Figure 7 are the mean and minimum
elevation angle statistics of the two highest satellites
over a range of latitudes. It is quite clear that
4.1. Trade-ofls
these high elevation angles will lead to a very low
probability of shadowing and therefore offer a very
To provide global coverage using ME0 and LEO high availability of service.
satellites with a high minimum elevation angle will
require a large number of satellites. For example
consider the two S-PCN LEO satellite constellation 90, . , . , . , . ,
I
proposals. First, Globalstar consists of 48 satellites
at a minimum elevation angle of lo", between f 7 0
latitude. Second, Indium with 66 satellites offers a
minimum elevation angle of 9.8" over the entire
surface of the earth. With both of these systems the
minimum elevation angle of the highest satellite
changes depending on the latitude position of the
mobile terminal. It is clear that the minimum elev- -
._
,4oL
-.
__:

ation angle in a future satellite-based ATM system

will have to be much higher than the proposed
530
w

20
-
-
-
- Highest Sot.
0. .0
Meon
Highest Sot. Minimum
Second Highest Sat. Meon
'_

minimum elevation angle of S-PCN constellation .... Second Highest Sot. Minimum

proposals, therefore requiring a larger number of

satellites. To calculate the number of satellites lo
01
t . ' I
required for an ATM constellation will require a 0 10 20 30 40 50 60 70 80 90
Lotitude (degrees)
definition of the service area and quality of service
to be delivered. Figure 7. Elevation angle statistics of MEO-32 constellation
474 B. G. EVANS A N D R. TAFAZOLLI

100 BOD Bandwidth on demand

90 CDMA Code division multiple access
80
CER Cell error ratio
CRC Cyclic redundancy check
70
E
FEC Forward error correction
t 60 FPLMTS Future public land mobile telecommuni-
cation systems
- 50
GDP Gross domestic product
? 40
a GI1 Global information infrastructure
30 HDLC High-level data link control
20 HEC Header error control
ITU International Telecommunications Union
10
LDR Low data rate
0 MAC Medium access control
Latitude (degrees) MDR Medium data rate
Figure 8. Diversity performance of M E 0 32 constellation
MPEG Motion Picture Experts Group
PRMA Packet reservation multiple access
SAR Segmentation and reassembly
As shown in Figure 8, should shadowing occur SDU Service data unit
the constellation offers dual and triple satellite visi- S-PCN Satellite personal communication network
bility (at a minimum elevation angle of 32”) over SSCOP Service specific connection orientated
a large range of latitudes, therefore minimizing the protocol
probability of complete loss of contact with the TCP Transmission control protocol
network. The final decision as to what type of VCI Virtual channel identifier
satellite constellation is suitable to carry ATM traffic VPI Virtual path identifier
will have to consider the issues raised above, and VSAT Very small aperture terminal
others such as the constellation capacity and satellite
payload complexity. REFERENCES
The constellation described above has properties I . D. M. Chitre, D. S. Gokhale, T. Henderson, J. L. Lunsford
such as a low number of required satellites, high and N. Mathews, ‘Asynchronous Transfer Mode (ATM) oper-
elevation angle statistics, very good satellite diver- ation via satellite: issues, challenges and resolutions’, Inter-
sity, and possible use of ISLs that make it suitable national Journal of Satellite Communications, 12. 2 1 1-222
( 1994).
for an ATM based satellite network, that can deliver 2. L. G. Cuthbert, J. C. Sapanel, ‘ATM the broadband telecom-
a high quality of service. munications solution’, IEE, London, 1990.
3. Y. Li and S. Andersen, ‘Multimedia traffic management
principles for guaranteed wireless network performance’,
APPENDIX: KEY FOR ACRONYMS Third Annual International Conference on Universal Personal
Communications, 1994.
4. D. Raychaudhuri and N. D. Wilson, ‘ATM-based transport
AAL ATM adaptation layer architecture for multiservices wireless personal communi-
B-ISDN Broadband integrated services digital net- cation networks’, IEEE Journal on Selected Areas in Com-
work munications 12, (8). October 1992.