null null

Voice over IP in Mobile Networks
Ludde Algell
Department of Communication Systems
Lund Institute of Technology
1
2
Executive Summary
Title:
Voice over IP in mobile networks.
Author:
Ludde Algell
Tutor:
Ulf Körner
Objective:
Voice over IP is rapidly transforming fixed line networks. The new technology
slashes prices on telephony in an unprecedented manner. However it is not
clear if Voice over IP will have the same effect on mobile telephony. The
objective of this thesis is to determine if there are essential differences between
fixed and wireless voice over IP and in which ways IP will affect mobile
operators.
Analysis:
In order to determine how VoIP will affect mobile operators the following
areas were investigated:
•
A background description of VoIP
•
The current market for fixed VoIP
•
Radio access technologies and VoIP
•
VoIP and new services in the mobile network
Conclusions: Mobile operators are unlikely to, in the coming 3 years, experience the same
type of competition from VoIP as fixed line operators are experiencing. The
main reason is the bandwidth-limited radio interface that requires services, such
as voice calls, to be highly optimized. At present VoIP is an inefficient way of
delivering voice service. There will also be a lack of attractive VoIP handsets.
Currently, VoIP calls made with VoIP applications are not substituting mobile
calls. VoIP applications are mainly used for international calling. By offering
subscribers the possibility to use VoIP applications over mobile broadband it
should therefore be possible for a mobile operator to increase revenue, without
cannibalizing on existing voice revenue.
The threat from alternative access technologies like WiMAX and WiFi is not
imminent for mobile operators’ voice revenue. Mobile operators with 3G
networks could however see negative impact on mobile data revenue. WiFi
network can be launched fast and at low cost. It is therefore recommendable
for operator’s to implement a WiFi strategy.
3
4
Foreword
This report is the result of a Masters thesis project at Millicom International Cellular and the
Department of Communication Systems at Lund Institute of Technology.
Thanks to Won-Suck Song at Millicom and Ulf Körner at the Department of Communication
Systems. Also thanks for great support, ideas, and corrections from Thomas Beijar, Philip
Henriksson and Anjna Mehta
Ludde Algell
Stockholm, December 18, 2005
5
6
1
Introduction............................................................................................................... 11
2
Introduction to IP Telephony .................................................................................... 13
2.1
Circuit switched networks .....................................................................................................13
2.2
Packet-switched networks......................................................................................................14
2.2.1
TCP/IP............................................................................................................................................... 15
2.2.2
RTP.................................................................................................................................................... 16
2.3
IP telephony protocols ............................................................................................................16
2.3.1
Session Initiation Protocol ............................................................................................................... 17
2.3.2
H.323 ................................................................................................................................................. 17
2.4
Categories of VoIP ..................................................................................................................17
2.5
Speech Coding and Codecs ....................................................................................................19
2.5.1
Codec negotiation in VoIP............................................................................................................... 19
2.5.2
Wideband codecs.............................................................................................................................. 20
2.5.3
Efficiency and Bandwidth................................................................................................................ 20
2.6
Header compression................................................................................................................22
2.6.1
Slow start and congestion control ................................................................................................... 22
2.6.2
Conext ............................................................................................................................................... 23
2.6.3
Robust Header Compression ........................................................................................................... 23
2.6.4
Performance improvements ............................................................................................................. 23
2.7
Measuring voice quality .........................................................................................................24
2.8
VoIP and Quality of Service ..................................................................................................25
2.8.1
Bandwidth ......................................................................................................................................... 25
2.8.2
Latency .............................................................................................................................................. 25
2.8.3
Jitter ................................................................................................................................................... 26
2.8.4
Reliability.......................................................................................................................................... 26
2.8.5
Reliability.......................................................................................................................................... 27
2.9
Provisioning QoS .....................................................................................................................27
2.9.1
IntServ or Flow Based Control........................................................................................................ 28
2.9.2
DiffServ............................................................................................................................................. 28
2.9.3
Smart routing .................................................................................................................................... 28
2.9.4
Blocking VoIP traffic by providing sub par quality....................................................................... 28
2.9.5
Comments ......................................................................................................................................... 29
7
3
4
Current market for fixed Voice over IP ..................................................................... 30
3.1
Business fixed line VoIP .........................................................................................................30
3.2
Residential fixed line replacement VoIP .............................................................................31
3.3
Instant Messaging and Softphones .......................................................................................32
3.4
Major VoIP providers ............................................................................................................36
3.5
Future development ................................................................................................................37
3.6
Market Value ...........................................................................................................................38
3.7
Comments .................................................................................................................................38
Voice over IP over Wireless Channels....................................................................... 40
4.1
Basics of Radio Communication ...........................................................................................40
4.1.1
Frequency, Power and Coverage:.................................................................................................... 40
4.1.2
Radio Resources and Cost ............................................................................................................... 42
4.1.3
Non Line of Sight Communication ................................................................................................. 43
4.1.4
System Capacity Improvements ...................................................................................................... 43
4.1.5
Bandwidth Limits ............................................................................................................................. 45
4.1.6
Evolution Radio ................................................................................................................................ 45
4.1.7
Spectrum Licenses............................................................................................................................ 46
4.1.8
Spectrum Availability ...................................................................................................................... 46
4.1.9
Comments ......................................................................................................................................... 46
4.2
Cost structure of cellular network........................................................................................48
4.2.1
4.3
Comments ......................................................................................................................................... 50
Voice over IP and WCDMA ..................................................................................................51
4.3.1
Softphone’s impact on capacity....................................................................................................... 51
4.3.2
Latency .............................................................................................................................................. 53
4.3.3
HSDPA.............................................................................................................................................. 53
4.3.4
Comments: ........................................................................................................................................ 56
4.4
Voice over IP and CDMA-2000 ............................................................................................57
4.4.1
EV-DO .............................................................................................................................................. 58
4.4.2
Evolution........................................................................................................................................... 59
4.4.3
Comments ......................................................................................................................................... 59
4.5
Voice over IP and WiMAX ....................................................................................................60
4.5.1
Performance:..................................................................................................................................... 62
4.5.2
Deployment: ..................................................................................................................................... 63
8
4.5.3
4.6
VoIP and WiFi .........................................................................................................................65
4.6.1
Voice over WiFi ............................................................................................................................... 66
4.6.2
Deployment....................................................................................................................................... 67
4.6.3
Issues ................................................................................................................................................. 69
4.6.4
Comments ......................................................................................................................................... 70
4.7
5
Comments ......................................................................................................................................... 64
VoIP and FLASH-OFDM ......................................................................................................72
4.7.1
Technology ....................................................................................................................................... 72
4.7.2
Deployment....................................................................................................................................... 72
4.7.3
Comments ......................................................................................................................................... 73
Mobile Operator implemented Voice over IP............................................................. 74
5.1
IMS ............................................................................................................................................74
5.1.1
IMS – Technical overview............................................................................................................... 75
5.1.2
Implementation ................................................................................................................................. 77
5.1.3
Convergence ..................................................................................................................................... 78
5.1.4
New Services .................................................................................................................................... 78
5.1.5
Comments ......................................................................................................................................... 80
5.2
Push-to-talk ..............................................................................................................................81
5.2.1
Services offered through PoC.......................................................................................................... 81
5.2.2
Technology Performance ................................................................................................................. 82
5.2.3
Deployments ..................................................................................................................................... 83
5.2.4
Comments ......................................................................................................................................... 84
5.3
Unlicensed Mobile Access ......................................................................................................84
5.3.1
Equipment:........................................................................................................................................ 84
5.3.2
How does it work?............................................................................................................................ 85
5.3.3
Deployments ..................................................................................................................................... 85
5.3.4
Network load .................................................................................................................................... 87
5.3.5
Growth of broadband access............................................................................................................ 87
5.3.6
IMS and UMA .................................................................................................................................. 88
5.3.7
Comments: ........................................................................................................................................ 88
5.4
Application based VoIP in 3G networks .............................................................................89
5.4.1
Cost of mobile data........................................................................................................................... 89
5.4.2
Price on telephony ............................................................................................................................ 90
5.4.3
Interconnect fees............................................................................................................................... 91
5.4.4
International traffic is initiated from fixed line network................................................................ 91
5.4.5
Alternative methods for making international calls ....................................................................... 92
9
5.4.6
Using softphone over a 3G network................................................................................................ 93
5.4.7
Comments: ........................................................................................................................................ 94
6
Conclusion................................................................................................................. 95
7
Resources ................................................................................................................ 101
8
Acronyms................................................................................................................. 106
9
Appendix.................................................................................................................. 107
9.1
Link budget ............................................................................................................................107
9.2
OFDM .....................................................................................................................................108
9.3
CDMA (Code Division Multiple Access) ...........................................................................109
9.4
List of Codecs.........................................................................................................................113
9.5
Frequency bands ...................................................................................................................114
10
1 Introduction
The initiative for this thesis comes from mobile operator Millicom International Cellular. The
company sees it as important to be well informed in what way VoIP will affect mobile
operators.
The key difference between VoIP and traditional telephony is that for the first time the
applications and the transport will be separated. This is a major shift in the infrastructure of
telephony. Traditionally the network and the application have been one and the same, the
phone network, PSTN, was build for one purpose i.e. delivering voice communication. Now
with the use of one IP-infrastructure for communication voice will become just one, albeit
important, application. Combined with the significantly lower cost structure for IP-networks
VoIP will dramatically change the telecommunication industry.
Mobile voice over IP refers to a system in where the last part of the communication is carried
over a wireless link. This could either be a cellular system where current technology is replaced
with voice over IP or it could mean using VoIP capable handset used in a wireless local area
network.
Skype is mentioned a few times in this thesis. This is because the brand is well known and
easily related to VoIP. It should be noted however that when Skype is used as an example
there are many other solutions that could provide the same service.
Problem discussion
Voice over IP is quickly transforming the fixed line networks in a rapid pace. The new
technology slashes prices on telephony in an unprecedented manner. However it is not clear if
IP will have the same effect on mobile telephony. The objective is therefore to investigate if
there are essential differences between fixed and wireless voice over IP and in what ways IP
will affect mobile operators.
Areas of inquiry
The questions that are aimed to be answered are as follows.
•
Why are fixed line networks leading the way in voice over IP rollout?
11
•
When is voice over IP expected to be commercially available for mobile operators?
•
Which operators are in the lead for a voice over IP rollout?
•
How will VoIP be introduced in mobile networks?
•
Are applications such as Skype a threat to operator’s revenue?
•
Alternative access technology and VoIP.
Delimitations
The thesis focuses on the technical issues of Voice over IP, leaving opportunity for continued
research with a more market oriented approach.
Thesis structure
To give a broad understanding of VoIP, chapter 2 gives a broad background of VoIP, how it
differs from traditional telephony and why quality and reliability have yet to make significant
progress.
Chapter 3 describes the market for fixed VoIP. It outlines the major providers of IP-telephony
that provide replacement service to traditional telephony and those which provide services that
complement traditional telephony.
Chapter 4 investigates why VoIP as seen in fixed networks will not successfully penetrate the
wireless world without considerable adoptions. This is done by explaining the fundamentals of
mobile communication and the challenges that exists for providing bandwidth and coverage
with radio. The chapter then discusses different third generation network access technologies
and presents a time frame when VoIP is expected to replace existing technology.
Mobile operators not only see VoIP as a way to lower the cost per delivered bit. They are more
interested in the new services that IP can bring. The most promising services, as described in
chapter 5, include Unlicensed Mobile Access, Push-to-talk. The chapter then explains how
these services relate to a new echo system called IP Multimedia Subsystem. The chapter
further discusses how VoIP applications may be deployed in a WCDMA network without
necessarily negatively affecting the operator’s revenue.
The thesis is summarized in chapter 6.
12
2 Introduction to IP Telephony
IP telephony or Voice over IP, VoIP, uses packet-switched networks for call transportation whereas traditional
telephony uses circuit-switched networks. This chapter will first explain the two types of networks and then
continues with a more in-depth description of packet switched networks. Knowledge of the IP-packet structure
and the technique of compressing IP headers is fundamental for understanding the problems of efficiently
introducing voice services using IP in cellular networks. The chapter also explains why good quality is harder to
achieve over VoIP as compared to circuit-switched networks.
2.1 Circuit switched networks
Circuit switched technology has been used in the traditional telephone network, PSTN for
more than one hundred years. Over circuit-switched networks, when a call is made, the
connection between the parties is maintained throughout the call. Because the points are
connected in both directions there exists something similar to a circuit, hence the circuit
switched name.
In the early days of telephony the path was truly dedicated. This means that for a call between
New York and Los Angeles a physical line of copper would be established over the entire
distance. Now various calls are digitalized and trunked together in high capacity fiber optic
cables, this means that each call does not have its own physical wire. Instead it is called a
virtual circuit since all the information follows a unique path throughout the call.
Circuit switched technology is well suited for real-time application such as voice. By routing
the information over a unique path, for the duration of the call, the information is received in
the order it was sent and in an even flow see Figure 1.
13
Figure 1. The information is sent in a single path throughout the call.
A major drawback with circuit switched technology is that it is a fairly inefficient way of
sending data. Whether or not information is sent the connection is occupied, thus consuming
128kbps throughout the call even if no information is transmitted.
2.2 Packet-switched networks
In a packet-switched network there is no reserved or dedicated capacity for each user. The
principle is more of a best effort sort; the capacity will be shared among the users but no
guarantees regarding throughput or other parameters can be given. The packets that contain
the voice may traverse the network independently of one another, see Figure 2. The packets are
then assembled in the right order at the receiving end.
Figure 2. The information packets may traverse the network independently. At the
end the packets are put in the right order.
This strategy works very well for bursty traffic rather than continuous; when browsing a
webpage no information must be sent after a page is downloaded. Instead, the capacity
becomes available for other users. Several users can also share one connection and the
14
available resources are distributed evenly. In the circuit switched case, new users will be
blocked if system has reached its maximum capacity.1
Packet switching allows each packet to be routed through the fastest and least congested path
to its destination. If a connection is lost along the path, packets will find an alternative route,
thus making packet switching very flexible.
Introducing telephony on packet switched networks is initially not a perfect match. Because
the packet switched networks were built with a focus on serving as much data with best-effort
quality, issues like latency [2.8.2] and jitter [2.8.3] were not a focus point.
2.2.1
TCP/IP
TCP/IP is the family name for a set of communication
protocols that are used to provide a means of communication
through data networks. The family members include, among
others, Internet Protocol (IP), Transmission Control Protocol
(TCP) and User Datagram Protocol (UDP). TCP/IP, is like
other protocols divided into layers with each layer handling
different tasks.
In the protocol stack (see Figure 3) IP resides in the network
layer. IP provides two services to the upper layers, addressing
and fragmentation and reassembly of long and short TCP
protocol data units. IP gives no guarantees on flow control,
reliability or error recovery to the underlying layers. Instead higher
Figure 3.
The IP stack
levels handle these functions. IP is thus a best effort delivery
service.
The TCP and UDP protocols are found in the transport layer. They provide a flow of data for
the upper layers. Common for both TCP and UDP is that they partition the data into
appropriate sizes before handing them over to the underlying layers. TCP is considered reliable
1
http://computer.howstuffworks.com/ip-telephony6.htm
15
as it provides guarantees of delivery where as UDP has no such built in functions and is
therefore considered less reliable. When sending a packet a number, or checksum, will be
placed in the UDP field. By looking at this checksum at the endpoint it is possible for a router
to determine if a packet has been damaged or not. If damage is detected, that packet will be
discarded.
2.2.2
RTP
RTP is used to provide end-to-end delivery service for data that has real time characteristics
like voice or video. The RTP protocol gives the possibility of adding a notion of time to packet
communications. While encoding at the source a time stamp is set on the packet before it
enters the network. At the destination this time stamp is used to regenerate the content at the
same rate it was encoded. A receiver can be used at the destination for setting a delivery pace
independent of network induced jitter.2
2.3 IP telephony protocols
An IP telephony protocol, or session protocol, is needed to provide the same services and
capabilities that the PSTN can offer. When two endpoints are communicating, or in any form
exchanging information, a session is said to be established. The session protocol’s function is
to handle the signaling between two endpoints. More in specific it has to:
•
Locate an endpoint
•
Contact the endpoint
•
Find out the capabilities of the two endpoints, e.g. what codecs [2.5] that are supported,
what sampling rate can be used etc.
•
Modify existing session
•
Terminate session
There are currently two major standardized protocols to handle voice over IP. The Session
Initiation Protocol and H.323.
2
Delivering Voice over IP Networks, D. Minoli, E. Minoli. Wiley
16
2.3.1
Session Initiation Protocol
The Session Initiation Protocol, or SIP has emerged as the predominant protocol for handling
multimedia sessions. Its origins come from the Internet organization Internet Engineering
Task Force, IETF. The relationship with the Internet world can clearly be seen since SIP is
based on two other widely used protocols, HTTP and SMTP.
2.3.2
H.323
SIP and H.323 can be viewed as two competing standards. There is however a clear shift
towards SIP. H.323 is said to be rather complicated in comparison and the specification is
almost six times as long as the SIP specification. The complexity of H.323 comes from its use
of a multitude of other protocols. Developers also find SIP easier to work with since it is, like
its HTTP foundation, text based. H.323 is on the other hand is binary based and will thus be
rather complicated to debug or extend.3 That the two standard comes from groups with
different view on network strategies is clear. H.323, from the telephony world, is said to be
network centric while SIP puts the intelligence at the endpoints and runs over a “dumb”
network.
2.4 Categories of VoIP
VoIP can be divided into five major categories: IP-to-IP, IP-to-PSTN, PSTN-to-IP-to-PSTN,
PSTN-to-IP. IP-to-PSTN-to-IP. With the interconnection of various networks users may use a
mixture of the categories depending on whom they call. It is thus not necessary to have
different terminals for the different categories.
•
In IP-to-IP two users are connected to each other through the Internet, without the
involvement of the PSTN, see Figure 4. The consumer premises equipment, CPE could be
a standard PC or mobile phone equipped with a softphone client such as Skype, AIM or
MSN Messenger [3.3.2].
3
Comparison of H.323 and SIP for IP Telephony signaling, Ismail Dalgic,
Hanling Fang
17
Figure 4. IP-to-IP. Softphone on a computer
•
In IP-to-PSTN the call IP-call is routed to a gateway that connects the call the PSTN, see
Figure 5. Users can thus bypass long distance charges by the use of local call termination.
Such call is associated with termination fees and will thus not generally be free too the
users.
Figure 5. IP-to-PSTN. A user can call regular phones from his computer
•
PSTN-to-IP, in this case a user can receive a call anywhere in the world using the same
local number, see Figure 6. The caller dials a local number that connects the call through a
gateway to the Internet. The call is then routed to user’s destination.
Figure 6. PSTN-to-IP. A user can receive calls from the PSTN
•
PSTN-to-IP-to-PSTN. When both endpoints of a call are using the PSTN, see Figure 7.
The call is the connected to a gateway and then transported over an IP-network. This kind
of service is becoming increasingly common and major players such as British Telecom are
upgrading their network with IP technology. PSTN-to-IP-to-PSTN is also commonly
found when using calling cards for long distance calls.
18
Figure 7 PSTN-to-IP-to-PSTN. Calling cars can use VoIP in parts of the transmission
•
IP-to-PSTN-to-IP. At present two users that use a wire line replacement service and want
to call each other will be routed over the PSTN, see Figure 8. Eventually it will be possible
that the initiating SIP-server look up the desired phone number and sees that this is an IPconnection and routes the call over IP the entire way
Figure 8. Subscribers to two different VoIP operators can be connected over the
PSTN.
2.5 Speech Coding and Codecs
In its most fundamental form, speech coders are analogue-to-digital converters. The analogue
speech waveform is sampled periodically and for each input voltage level a digital value is
assigned. To achieve a high quality output, either a complex codec or high bandwidth is
necessary. Speech codecs can also have a built in error correction feature to be used over
unreliable connections. The drawback is that the speech quality degrades as more bits are used
to build up redundancy.
2.5.1
Codec negotiation in VoIP
VoIP makes it easier to introduce new codecs due to the fact that the endpoints are generally
more intelligent. During the call set-up, two endpoints can negotiate on which codec to use,
based on available resources and network conditions. Therefore, it is possible to use wide-band
codecs that produce a better sounding voice stream.
19
2.5.2
Wideband codecs
The traditional PSTN has a sampling rate of 8 kHz. With that rate it is possible to distinguish
frequencies up to 4 kHz. Fricative sounds such as “f” or “s” contains energy at higher
frequencies than 4 kHz. As a result, “f” and “s” become indistinguishable in a traditional
phone call. Figure 9 plots an “e” and an “s”. One can clearly see that the “s” has a peak of
energy that lies past 4 kHz. By implementing a wide band codec, the perceived quality is under
good network conditions, superior to that of normal telephony.
110
100
90
Energy [dB]
80
e-sound
70
s-sound
60
50
40
30
20
10
0
5000
10000
15000
20000
Frequency [Hz]
Figure 9 “s” and “f” contains energy at frequencies above 4000 Hz making them
indistinguishable.4
2.5.3
Efficiency and Bandwidth
If low bandwidth usage is a requirement, complex codecs has to be used. Complex codecs
bring long packetization delays and the need for high computing power. In IPv4 the
RTP/UDP/IP header is 40 bytes. A payload (here voice) of 40 bytes would result in 50%
efficiency. The required bandwidth BW can be derived from the output rate R (in bits/s),
header size H (in bits) and payload sample size S (in milliseconds), using
BW = R +
4
H
S
Global IP Sound
20
At a bit rate of 64kbps it will take 5 ms to accumulate 40 bytes but the required bandwidth
would be quite high. With 8 kbps it will take 40 ms to accumulate 40 bytes5, and the required
bandwidth will be lower at the expense of an increased delay.
There is an engineering trade-off between achieving an acceptable packetization delay and low
bandwidth. Many systems use 20ms sample sizes and a bit rate of close to 10 kbps to achieve
a compromise between bandwidth, efficiency and introduced delay.
A commonly used codec is G.711. It is found in fixed telephone replacement services like
Vonage. G711 uses a bit rate of 64 kbps with 20 ms sample duration. If deployed over an
Ethernet link it will use an overhead of 58 bytes consisting of
•
IP, 20 bytes
•
UDP, 8 bytes
•
RTP, 12 bytes
•
Ethernet 18 bytes
Resulting in a bandwidth of
BW = 64kbps +
5
58 8bits
= 87.2kbps
20ms
Bur Goode Senior Member IEEE, Voice Over Internet Protocol (VoIP)
21
2.6 Header compression
For applications such as voice where the payload
often ranges around 20 bytes the overhead
becomes a considerable part of each packet, see
Figure 10. With header compression techniques
the header can significantly be reduced.
Figure 10. The IP header can be large
compared to the payload
The information in the header is what enables
communication over networks, large and small.
The information consists of the destination
address, source address, ports, protocol
identifiers, error checks etc, see Figure 11. This
information is fairly consistent between two
consecutive packets. For examples it is highly
unlikely that the Source IP Address would be
different during a session. Other fields in the IP
packet may change more frequently, but if does
it often happens in a non-random manner. All
this implies that it is possible to represent the IP
information in fewer bits.
2.6.1
Figure 11. An IP-packet with different
information fields
Slow start and congestion control
TCP/IP relies on slow start and congestion avoidance. This means that the transfer speed is
kept low initially to determine network congestion. As the first chunks of data are delivered
successfully, speed is increased. If packets are lost the TCP protocol interprets this as
congestion in the network and lowers the sending data rate and the process starts over again.
This works fine in a wire line environment where the data pipe does not suffer significant
interference and packets are rarely lost due to the physical link. In a radio environment the
situation is quite different. The radio channel is as far as one can get from a stable and reliable
data link. Packets are frequently lost and speed decreases and therefore reaching high
throughput becomes a problem. Essentially, this means that TCP/IP is not a perfect match for
22
air link protocol. Some of the issues mentioned above can be solved using features such as
forward error correction (FEC). 6
2.6.2
Conext
When initiating a session in a system that uses header compression, the first packets are sent
uncompressed to build up a context. This context is formed on both the compressor side and
the de-compressor side. The number of packets required to build a context after a data flow
has been initiated is closely related to the quality of the transfer link. With high, Bit Error
Rate, BER it will take more time. When the context is established on both sides the
compressor will reduce the headers to the extent possible. If an error occurs resulting in
several lost packages the context may have to be rebuilt. For this build-up a number of
uncompressed headers must be sent.7
2.6.3
Robust Header Compression
Current implementation of header compression such as RObust Header Compression, ROHC,
are good at handling packet loss without losing the context, hence the word robust. By sending
fewer bits over the air interface the risk for a corrupted package decreases. Because TCP/IP
relies on slow start and congestion control, packet loss is seriously affecting network
throughput. By implementing a header compression technology the gain is twofold, first by
sending less overhead and second by decreasing the risk for packet loss.
The header can be reduced from 40 bytes down to a minimum of 1 byte under ideal
conditions. Typical average values should however around 2-4 bytes for IPv4 and 3-5 bytes for
IPv68 Within 3GPP the Robust Header Compression or ROHC was chosen as the standard
header compression technique. It will be implemented in Release 6 of UMTS at the end of
2006 or 2007.
2.6.4
Performance improvements
Compared to circuit switched solutions the overhead caused by IP leads to a lower system
capacity for handling voice calls. Simulations in a UMTS shows that with VoIP, capacity can
6
Northstream White Paper, Operator Options Beyond 3G
7
Interview Joakim Enerstam, Effnet
8
Interview Mats Nordström, Ericsson Research
23
reach only 50% of a circuit switched implementation. With header compression is possible to
achieve 90%.9
Header compression techniques must be implemented on both ends of the communication
link. In UMTS network it can be found both in the terminals and the RNC’s. The standard is
to be implemented in 3GPP Release. 6. It is then left to infrastructure providers, handset
manufacturers and operators to implement these standards in their products and services. An
estimate is that header compression using ROHC will be operational in 2007.10
2.7 Measuring voice quality
Voice quality is measured using subjective techniques. A number of listeners will under
controlled conditions give a score to a specific call or codec. 5 is excellent, 4 is good, 3 is fair, 2
is poor, 1 is unacceptable. An average is calculated to determine the Mean Opinion Score or
MOS. Several conditions may be simulated such as packet loss environmental noise and
tandem encoding/decoding to measure their effects on perceived quality. Tandem
encoding/decoding occurs when the connection crosses networks using different types of
codecs11.
9
10
Voice-over-IP-over-Wireless, K. Svanbro, J. Wiorek, B. Olin
Interview, Mats Nordström, Ericsson Research.
11
http://www.cisco.com/univercd/cc/td/doc/product/access/acs_mod/1700/1750/
1750voip/intro.htm
24
2.8 VoIP and Quality of Service
Real time applications, such as voice, are very sensitive to variations in network conditions. A
standard file download is not. However, these applications share the same infrastructure
making it difficult to provide capacity and reliability parameters when designing a network. In
this chapter the different parameters that determine quality of service are considered and how
quality can be increased or guaranteed.
The main parameters that affect quality of VoIP communication are:
•
Bandwidth
•
Latency
•
Jitter
•
Reliability
2.8.1
Bandwidth
If a communication link offers insufficient bandwidth, packets will be dropped and lost.
Therefore, it is important to foresee that the bandwidth requirements are met through out the
whole link.
2.8.2
Latency
Latency is the time it takes for one packet to traverse the network from one node to another,
see Figure 12. The ITU has set the limit for latency to 150ms. If latency is above 50ms there are
echo effects that must be handled.
25
Figure 1212 Latency is introduced in various parts of the network
2.8.3
Jitter
Jitter is variation in latency. Because packets traverse the network independently, they may
require different time to reach their destination. For data services this has a negligible effect.
However, during a voice conversation it is important that each voice packet is delivered at the
same rate it was sent. To avoid jitter, buffers must be used in the receiving end. The buffer
will also fix packets that arrive in the wrong order. Such buffers increase latency, thus making
a trade-off between low latency and jitter.
2.8.4
Reliability
Reliability can be measured in many different ways. The most common method is looking at
packet loss or bit loss. Since RTP data is sent over the unreliable UDP protocol, most codecs
have built in error correction. Such error correction can handle error correction of up to
roughly 5 percent before quality degrades.13
12
12
http://www.protocols.com/papers/voip2.htm
26
2.8.5
Reliability
Reliability can be measured in different ways, the most common is however to look at packet
loss or bit loss. Since RTP data is sent over the unreliable UDP protocol most codecs have
built in error correction. Such error correction can handle error correction of up roughly 5
percent before quality degrades.14
Reliability can also be measured in terms of availability of the network (do you mean
“unavailability”?). The traditional PSTN has a very high reliability, close to 99.999%, or a few
minutes per year. In order for VoIP to gain widespread usage, temporary outages must be
minimized. Broadband service for enterprises promises service availability of 99.8%15. This
means that the service can be unavailable more for than one hour per month. Additionally, the
risk of power outage and planned maintenance must be included. Some argue that these lower
requirements make IP-infrastructure less expensive, as compared to traditional
telecommunication systems.16
When a VoIP service fails, software failure is the most common cause, accounting for more
than 47% of the cases17. This indicates that VoIP has some maturity issues that must be
addressed.
2.9 Provisioning QoS
The common way to provision QoS over IP has been over-dimensioning the network. By
utilizing a bigger data pipe network related problems will diminish. If all the links in a network
route traffic at 30% of peak capacity, there should not be any problem with congestion or lost
packets.18 Others argue that over-provisioning will not hold when more and more of the traffic
consists of real-time voice or video.19 The underlying debate is whether it would be less costly
to overprovision than to introduce control functions to constantly monitor the network.
15
http://www.bredbandsbolaget.se/portal/FORETAG_INTERNETACCESS
16
Ulf Olsson - Ericsson White Paper – Combinational Services.
17
www.lighreading.com/document.asp?doc:_id=53864
18
Interview Stefan Hagbard TeliaSonera
19
Ericsson White Paper, Softswitch in mobile networks
27
Other than increasing bandwidth there are few alternatives that can be used to handle high
priority traffic. Traffic could be routed around congested links or packets can be given priority
in router queues. Three key concepts are IntServ, DiffServ and smart routing, as described
below.
2.9.1
IntServ or Flow Based Control
IntServ is based on reserving resources through the network and is sometimes called a “flow
based” solution. When an application requests a certain QoS level, resources are reserved on
each of the routers in the network. IntServ, however, is not seen as a solution that would scale
well in a large IP network with thousands of users. The problem is that the traffic generated
from controlling the network is quite large compared to the small amount of traffic that VoIP
generates.20
2.9.2
DiffServ
In DiffServ each IP packet is flagged with a QoS priority. Throughout the network these
packets will be given priority in routers thus avoiding congestion. The main benefit of DiffSer
is that it is simple compared to IntServ, but will still offer significant improvements on current
best-effort networks.21
2.9.3
Smart routing
IP is relies on destination-based routing, sending all packets over the shortest path without
regards to the load on different links. By using smart routing, packets will take the best route,
which may not be the shortest, based on information on congestion and other parameters in
the network.
2.9.4
Blocking VoIP traffic by providing sub par quality
Broadband operators are interested in taking a share in the expanding market. Currently, many
of them are only providing a fat data pipe and there have been attempts to interfere with VoIP
traffic. In the US, Federal Communications Commission (FCC) reached a $15.000 consent
decree with Madison River Communication after it was found to have interfered with VoIP
traffic. The company pledged not to block VoIP traffic in the future.22 However, it is still
20
Bur Goode, Voice over Internet Protocol
21
Bur Goode, Voice over Internet Protocol
22
http://informationweek.smallbizpipeline.com/60405214
28
possible for operators to indirectly block traffic. By giving priority to traffic from their own
services such as telephony or TV a network owner could effectively set a lower priority to
other types of traffic. It will be clearly significantly more complicated to determine if such
operation can be deemed illegal.
2.9.5
Comments
Modern codecs produce a data stream of down to 5 kbps. Lower bit rate codecs give inferior
quality. It is not credible that the codec will gain any significant improvements in the coming
years. It has also been shown that the overhead produced by the Internet Protocol has a
significant impact on the resulting bit rate.
Considering that circuit switched telephony has been around for more than 100 years VoIP is
still in its infancy. There is thus a challenge for VoIP providers to implement systems for
maintaining end-to-end quality of service.
29
3 Current market for fixed Voice over IP
The Voice over IP market is in a turbulent phase. According to Sandvine there are some 1100 VoIP providers
operating today.23 These operators range from multinational long distance carriers to small local operators and
applications that run on standard PCs. This chapter discusses the different segments and describe how each is
deployed. The market for VoIP can be divided in to 3 major categories:
•
Business fixed line VoIP
•
Residential fixed line replacement VoIP
•
Softphone VoIP
3.1 Business fixed line VoIP
A survey made by Heavy Reading among companies in the telecommunication industry
showed that three quarters have already deployed VoIP in parts of their network. Still, the
survey found that the vast majority of the traffic was carried over traditional circuit switched
technology. The respondents expected that CAPEX savings would be non-existent or low
while OPEX is considered to have a greater potential to lower costs.
24
23
http://www.lightreading.com/document.asp?doc_id=72854
24
www.tns-infratest.com
30
3.2 Residential fixed line replacement VoIP
The fixed line replacement segment is dominated by companies that are competing heads up
with the wire line service providers. The goal with their service is to get a look a feel that is not
different from traditional telephony. A typical provider like the US based company Vonage
provides the client with a handset with the same functions as a normal telephone, differing
only in that it plugs into the broadband connection instead of the telephone network. Vonage
uses the term “Bring your own broadband” meaning that Vonage piggybacks on infrastructure
provided by broadband network operators. A user plugs his phone in to an adapter and can
then use Vonage’s services. While Vonage is considered the market leader in this segment,
there are several options available.
The traditional fixed line telephone companies are overseeing their current networks to
determine, which upgrades are needed to maintain competitiveness. British Telecom of the
U.K has announced its plans to make a £10bn investment in its network. BT sees it as
necessary to increase capacity in order to offer new services such as VoIP, broadband and TV.
25
25
www.tns-infratest.com
31
3.3 Instant Messaging and Softphones
3.3.1
Instant Messaging
Instant Messaging, or IM, is an electronic communication tool through which users can send
text messages and files. It could be viewed as a cross between email and telephone; more
personal and direct than an email and less intrusive than a call. It provides an effective medium
for communication that requires swift but not immediate answers like “can we meet after
lunch?”
A survey conducted by Opinion Research Corporation showed that 66 percent of 13-to-21year olds send more instant messages than email. A year ago that number was 49 percent. 20
percent of the respondents said they were interested in making phone calls from their IMapplication. According to AOL, quoting data from ComScore Media Matrix, more than 80
million people regularly use IM.26
3.3.2
Softphone
The softphone is an evolved version of Instant
Messaging enabling users to place phone calls to other
IM-clients or the PSTN, see Figure 13. By searching the
Internet one can, in a few minutes find more than 90
applications that can be downloaded to a PC and
provide PC-to-PC and PC-to-PSTN calls27. These
applications vary from those that use proprietary
closed protocols like Skype to open source projects
that use the SIP protocol like Wengo.
Softphones are not targeted as being fixed line
replacements. A strong incentive for not being a
Figure 13. Softphone Skype
26
http://news.zdnet.com/Study%3A+teenagers+favor+IM+to+e-mail/2100-
9588_22-5944265.html?part=rss&tag=feed&subj=zdnn
32
telephone service is that the providers do not have to comply with several laws and regulations
to which traditional telephone companies must comply. These include the ability to call
emergency services or for federal agencies to legally intercept if necessary. Countries like China
have plans to ban the use of unregulated VoIP services.28
There are a great number of VoIP-capable IM/Softphones available. A search at
www.download.com gives over 100 hits of IM-application. Below is a selection of the most
popular applications listed.
Yahoo / MSN
Yahoo has built up a large user base in Japan topping 4.7 million subscribers in first quarter of
200529. In June 2005 Yahoo acquired DialPad to have access to DialPad's Pc-to-phone calling
capabilities. Since then Yahoo has teemed up with British Telecom to provide BT customers
with Pc-to-phone calling through Yahoo messenger30. Yahoo is believed to have 21.9 Million
instant messaging clients in the US31.
Microsoft has offered its VoIP-capable instant messenger service MSN Messenger for a
number of years. In 2005 Microsoft acquired Teleo, a company that provides PC-to-phone
access. Goldman Sachs analysts see the move as a way for Microsoft to better integrate
services currently found in its portfolio such as MSN, Outlook or even Xbox games.
According to Nielsen/NetRating MSN had 27.3 million unique users in September.32
Yahoo and MSN have now decided that their respective services should be compatible with
each other, which further manifests the interest in IM-services.
28
http://www.forbes.com/technology/feeds/afx/2005/09/08/afx2214918.html
29
Source: Point Topic
30
http://www.bt.com/btcommunicator/index.jsp
31
,31,31http://customwire.ap.org/dynamic/stories/M/MICROSOFT_YAHOO?SITE=CAS
DT&SECTION=HOME
33
Google
Google launched its Instant Messenger service Google Talk in 2005. Google uses the open
Instant Messaging protocol XMPP. The idea is to connect people to advertisers that buy
services from Google.
AOL
AOL provides one of the most widely deployed Instant Messenger services today, most
noticeably in the US market with 51.5 million users33. It has long had PC-to-PC VoIP
capabilities but has recently launched a service that enables clients to call regular phones. The
service will not be softphone based however. Instead users will use an adapter to which they
connect their regular landline phone. AOL’s strategy is thus not to directly make their existing
IM clients to start making calls to the regular phones. Instead the company is targeting its base
of broadband subscribers to move over from traditional telephony to VoIP. 34
Trillian
Trillian is an application developed by Cerulean Studios that makes it possible for users to
connect to multiple IM platforms, using only one application. A user can be connected to
MSN Messenger and AOL simultaneously without having to launch both applications.
Skype
During 2005 Skype has created a lot of buzz. It differs from the traditional IM client because it
is based on distributed peer-to-peer technology thereby not relying on central servers. Skype
reports that their software has been downloaded over 200 million times and that it has over 60
million accounts registered. Over 2 million people have opted to use its PC-to-PSTN
SkypeOut service35. Skype was acquired in September 2005 by eBay, which is interested both in
Skype’s user base and the possibility of integrating its auctioning service with live
communication. Skype will have an estimated revenue of $60 million 2005 and an estimated
revenue of $200 million 2006.36
34
http://www.lightreading.com/document.asp?doc_id=80587
35
www.skype.com
36
http://www.centerformarketintelligence.com/analystviews/20050927-
WeeklyReport.htm
34
The Switchboard
The Switchboard is a different kind of Softphone because it is a program that is runs online.
One could compare it with having an email-client like Outlook versus an online email account
like Hotmail. Switchboard uses Java to enable people to make free calls across the world.
Switchboard has, in spite of their interesting technology, not attracted any significant number
of subscribers.
35
3.4 Major VoIP providers
Table 1 Major VoIP service providers
Company
8x8
Product
Packet 8
AOL
Total Talk
AT&T
Call Vantage
Misc
Wire line replacement, offers video
telephony to both business and residential.
Unlimited monthly US plan for $19.95
Available both as integrated with AOL’s IM
client AIM and in an adapter version.
Monthly unlimited $34.95 to call landline in
the US and Canada
53.000 subscribers years end 2004. $29.95
unlimited monthly plan.
BroadVoice BroadVoice
$19.95 unlimited international plan.
BT
BT
Communicator
Cablevision
Optimum
Voice
Venture with Yahoo Messenger. Softphone
with international rate starting at 0.5p
/minute. Requires fixed line subscription
with BT
Unlimited national $34.95
601.208 subscribers37
Comcast
Comcast
Cox
Cox
Google
Google Talk
Microsoft
MSN
Messenger
Skype
Skype
37
38
Service Type
IP-to-IP
IP-to-PSTN
PSTN-to-IP
IP-to-IP
IP-to-PSTN
IP-to-IP
IP-to-PSTN
PSTN-to-IP
IP-to-IP
IP-to-PSTN
PSTN-to-IP
IP-to-IP
IP-to-PSTN
IP-to-IP
IP-to-PSTN
PSTN-to-IP
Bundled services with TV and Internet
IP-to-IP
connection from 39.95. 83.000
IP-to-PSTN
subscribers38
PSTN-to-IP
Bundled services, TV and Internet. 130.000 IP-to-IP
subscriber39
IP-to-PSTN
PSTN-to-IP
User open standard protocol Jabber. Service IP-to-IP
connected with Goggle’s Gmail service.
IM client integrated in Windows. MS claims IP-to-IP, soon
180M active user accounts / month. Has
integration
acquired Teleo for PSTN capabilities
with PSTN
Announced deal with Yahoo to let users
communicate using respective IM client.
Softphone. 60M+ registered users. The
service. Accounted for 46.2 percent of
VoIP minutes used in North America.
IP-to-IP
IP-to-PSTN
PSTN-to-IP
http://www.lightreading.com/document.asp?doc_id=84312
,38 http://www.cabledatacomnews.com/dec05/dec05-1.html
36
Time
Warner
Vonage
Vonage
Yahoo
Yahoo
Messenger
Skype users accounts for 35.8 percent of
individual VoIP users.40
Unlimited local calling, unlimited instate
calling, unlimited long distance $39.95
854.000 subscribers41
Several national and international plans.
Unlimited national from $24.95. 1M+
Users. Market leader in replacement
category
Has recently acquired DialPad which will
give PSTN capabilities to Yahoo Messenger
IP-IP
IP-PSTN
PSTN-IP
IP-to-IP
IP-to-PSTN
PSTN-to-IP
IP-to-IP
Soon
integration
with PSTN
3.5 Future development
The way forward is all but clear. Telecom operators will not stand by and let software players
set the terms of the market. However, the proprietary nature of these applications, could
hinder the development of the market. A step forward has definitely been the announcement
of Yahoo and MSN to make their services compatible with one another. Google entrance in
the market and their choice of and open standard protocol is seen as a big thing. From the
cellular world companies like Verizon and Cingular already offer their subscribers access to
IM-clients like MSN and Yahoo. Typically they charge a flat fee of $2-$3 per month that allows
subscribers to send and receive an unlimited amount of instant messages.42
For the operator offering a softphone could be a way to more tightly integrate their
subscribers. A subscriber would have one contact number and calls would be routed to
different points of access. When away from the computer, calls would be routed the
subscriber’s mobile phone. During a business trip to Japan, for example, calls made could be
answered on his laptop PC at a lower cost compared to is anwered on his mobile phone.
Operators have an opportunity to expand their businesses into areas where they have so far
been quite unsuccessful. Because the system, ideally, would be based on open standards the
subscribers would have access to their favorite IM-applications and at the same time the
40
http://www.lightreading.com/document.asp?doc_id=75833&site=lightreading&
WT.svl=news1_3
41
http://www.cabledatacomnews.com/dec05/dec05-1.html
42
http://news.zdnet.com/2100-3513_22-5298633.html
37
enhanced features provided by the operators. The operator would need to create software that
lies on top of other IM-application and aggregate their functions and features much like
Trillian [0]. This is not seen as a technically challenging operation.43
3.6 Market Value
The market for VoIP products and services is expected to grow dramatically over the next few
years. Forecasts from Frost and Sullivan predict that the global market will be worth 8.15
billion in 2007.44
45
3.7 Comments
The growth of broadband connections seems to be the single most important cause for the
uptake of VoIP.46 Many of the VoIP providers are also providers of broadband connection.
Examples of such are: Comcast, AOL, Cablevision.
43
44
Tele2
,45 http://www.tns-infratest.com/06_BI/bmwa_english
/Faktenbericht_8/Abbildungen/Folie404.JPG
46
Arthur D. Little, Broadband Report 2004
38
Instant Messaging and softphones are markets that are growing exponentially. Operators
should develop solutions in order to increase their market footprint. Such applications should
be made using open standards.
The proliferation of VoIP over fixed lines will indirectly have an effect on wireless VoIP in
that it speeds up the evolution to an all-IP telecommunication scenario. With an all-IP
infrastructure on the fixed line side, upgrading mobile network will become more economical.
39
4 Voice over IP over Wireless Channels
The mobile sector is lagging the fixed sector with a couple of years. This chapter will explain why there is a
difference between fixed and mobile VoIP. We first use a general approach describing the fundamentals of
radio communication. This is of use when discussing future development of new access technologies. Next we look
at the cost structure of a mobile network. This will give some incitements for deploying VoIP but also some
arguments against it. We then look at access technologies such as WCDMA, CDMA2000, WiMAX and
WiFi and see if or how VoIP will be introduced.
4.1 Basics of Radio Communication
A spectrum is a group of frequencies in which a radio service can be deployed. Spectrum is a
limited resource and has proven to be a valuable asset. A number of interests must be
considered when distributing these resources: military communication must not be hindered;
television broadcasts should be received with good quality, etc. Every country has a federal
agency that handles the distribution of this resource.
4.1.1
Frequency, Power and Coverage:
Frequency and range are related in that a higher frequency yields a shorter range. In the
frequency band used for mobile communication, 500-2500 MHz, a rule of thumb is that
doubling the frequency will halves the range. This is because higher frequency is subject to
greater attenuation47 and that it does not benefit from diffraction (to bend around buildings) as
much as lower frequencies48. Deploying a network in the 950 MHz band would therefore
require about as many base stations as in the 1900 MHz band. Lower frequency bands are
thus highly interesting for telecommunications operators. The drawback of lower frequency
bands is that they do not provide large amounts of spectrum. In Sweden for an example, the
available spectrum in the 450 MHz band is only 2 x 4.5 MHz + 2 x 1.8 MHz. This can be
compared with UMTS, which has been allocated 2x60 MHz in the 2 GHz band. Because the
area covered by one base station is very large at low frequency the throughput-per-area will be
moderate making it unsuitable for densely populated areas.49 There is no difference, however,
47
Fredrik Tufvesson, Department of Electrical Science, LTH
48
Wireless Communication
49
www.PTS.se
40
in throughput depending on at what frequency the band is located, 5 MHz bandwidth will
provide the same throughput at 1 GHz, as it will at 10 GHz.
Power
Coverage is dependent on power. The stronger the signal the farther the signal will reach.
Although it is possible to increase output power from a base station almost indefinitely, doing
the same from a battery driven terminal is simply impossible. Using high output power would
also cause interference on other base stations.
Gugliemo Marconi conducted the first radio transmissions in 1895, occupying a great amount
of spectrum but capable of sending only a small amount of information. In 1901 the first
transatlantic transmission with a high power amplifier blanked a great part of the globe. In fact,
using that technology only 50 simultaneous conversations could take place in the whole world.
Bandwidth, Frequency and Coverage
The relationship between bandwidth and coverage can further be explained by the following
equation, leaving modulation and coding unchanged
B
R = 10 log
Br
f + log fr in which R is the required performance improvement in dB, B is the bit rate to achieve at
frequency f. Br and fr are the references used. is a propagation factor for frequency. The
value of is 33.4 according to the COST 231 – Hata-model and 20 for free space propagation.
The graph below plots R as a function of frequency, assuming the two aforementioned values
of .
Looking at the plot for =35, to achieve 100mbps in the 5 GHz band, one must improve the
link budget (see [9.1]) with 30dB, corresponding to reduced coverage. If one instead uses 1
GHz, R needs to be only 6.5dB higher to achieve the same coverage50, see Figure 14.
50
Masahiro Umehira – Research and Development of Broadband Wireless
Access Technologies
41
Figure 14 Increased bandwidth or frequency leads to less coverage
The equation states that an increase in bandwidth has a negative effect on coverage. With less
coverage there will be a need for building new and expensive base stations. It is thus unlikely
that there will be any significant substitution from circuit switched to VoIP before the latter
can achieve similar performance over the radio link.
4.1.2
Radio Resources and Cost51
The cost of building a network can be described by the following equation:
C system N user Buser Aservice f (Q)
Csystem = the cost of the system
Nuser = the number of users
Buser = Bandwidth of user
Aservice = the service area covered (volume indoors)
f(Q) = function of the required Quality of Service
51
J. Zander – Economics of Broadband Wireless Access Systems
42
This equation shows that it is not possible to scale bandwidth and maintain coverage and cost.
In order to offer a higher bandwidth Buser to the users and maintain system cost Csystem, one or
more of the other parameters would have to be lowered. This means either reducing coverage
(Aservice), fewer users (Nuser) or a reduced quality of service (f(Q))
The cost of providing ubiquitous coverage increases with transfer speed. A likely scenario is
thus that high data speeds are concentrated to areas where people actually are e.g. airports,
train stations and cafes. In other areas, users will have access to moderate access speeds. Big
cellular technologies like WiMAX and UMTS will thus be complemented with hotspot
solutions like WiFi. The equation that governs how cost relates to transfer rates tells us the
importance of using available resources efficiently. It also gives an indication that VoIP over
the radio interface will first be introduced when there are solutions available that are as
effective, in a radio resource perspective, as circuit switched traffic.
4.1.3
Non Line of Sight Communication
An essential characteristic in mobile radio communication is that there is no need for line of
sight between the user and the base station. At frequencies above 2.5 GHz non line-of-sight
propagation is very limited. In fact there are hardly any mobile networks deployed in
frequencies above 2.5 GHz.52
Frequencies above the line of sight requirement can be used for services where line sight is not
an issue e.g. backhaul transmission from base station or wireless broadband access. At higher
frequencies the signal becomes more susceptible to atmospheric conditions like fog, rain or
snow. This drawback has to be weighted against the great amount of available bandwidth.
4.1.4
System Capacity Improvements
The number of simultaneous conversation has, obviously, increased dramatically since
Marconi’s first broadcasts [0]. In the last 45 years it has increased by a million times. To
increase capacity there are four main options:
•
Frequency division
•
Modulation techniques
•
Spatial division
52
Patrik Wikström, Netcom Consultants
43
•
More spectrum
Of the million-fold improvement, a 45x improvement comes from the use of more spectrum.
A 4x improvement from the ability to divide the spectrum into ever narrowly defined slices, or
frequency division. Frequency division leads to less interference outside the intended
frequency and remove the approximations that are needed to “tune in”. A 5x improvement
comes from techniques such as AM, FM, time division and approaches to spread the
spectrum. The rest, or 1600-fold improvement comes from confining radio areas with the use
of cells.
It could be interesting to view these improvements in terms of how many conversations, voice
or data, a system can handle. It turns out that this number has doubled every 30 months for
the past 104 years. This relation is sometimes referred to as Coopers law.53 Cellular telephony
is therefore based on creating small islands of non-interfering zones thus increasing the system
throughput.
53
http://www.arraycomm.com/serve.php?page=cellCooper
44
4.1.5
Bandwidth Limits
There is an upper bound for how much data that can be transferred in a given link. Claude
Shannon described this limit in 1948.
C = BW log 2 (1 + S / N )
C= Error free data
BW = Bandwidth
S=Signal level
N=Noise level
Signal power inherently decreases with distance. At best, noise is kept constant. Today’s radio
technologies are approaching the Shannon limit by the use of effective modulation and coding.
The noise level N is the thermal noise that always is present and could be seen as a constant.
There are then two parameters left, Signal level S and Bandwidth BW. As described [0] it is not
possible to increase output power without causing increased interference. Thus it is not likely
that there will be any significant improvement on system capacity through improved radio
bearers.54
Current radio technologies are approaching the Shannon limit and there is not much room for
radical improvements in transfer speed in the coming years. At a given frequency and
bandwidth the number of base stations required to give coverage and adequate throughput is
roughly the same for competing access technologies. Today’s 3G technologies are increasingly
gaining benefits from economics of scale making the introduction of a completely new radio
system to deliver the same type of service unlikely.
4.1.6
Evolution Radio
A study conducted by Rysavy research found that the difference in performance between radio
access technologies is becoming less noticeable. In a live network many factors will ultimately
limit throughput and therefore, the introduction of a new radio bearer will have less impact.
Instead, the focus should be on improving the signal to noise ratio, SNR, by interference
54
Rysavy Research – From GPRS to HSDPA and beyond
45
coordination between sectors and the use of intelligent antennas or MIMO technology.55 Such
improvements can be made on any type of access channel, thus making the choice of
technology less and less important.
4.1.7
Spectrum Licenses
Some frequencies have been made license-exempt, e.g., the 2.4 GHz band in Europe.
Microwave ovens, DECT-telephones, WiFi and Bluetooth are example of technologies that
use this frequency band. A communication service deployed in the license-exempt band would
thus be subject to severe interference. This is why all major communication networks operate
in licensed frequency bands. In these bands the license holder is exclusively allowed to
operate, thus keeping the radio environment free from interference
In 2000 various European countries held auctions for licenses to the 3G networks. In the UK
the 5 licenses generated an income of 37.8 billion and in Germany it generated 54 billion.56
In retrospect some argue that this price was too high and that it has hurt the ongoing
deployment of networks. It does however give an indication of how valuable this resource is
and at the same time the difficulties associated with free unlicensed spectrum
The agency that is responsible for licenses can set terms under which a technology may be
used. As an example in Europe, spectrum in the 3.5 GHz band only allows fixed or nomadic
use, thus making it useless for an operator interested in offering mobile service.
4.1.8
Spectrum Availability
As mentioned above spectrum is a limited resource. Available frequencies vary from country
to country making it hard to deploy a standard that can be adopted through out the world and
thus benefit from economics of scale. See table [9.5],
4.1.9
Comments
From a VoIP perspective there is no “preferable” frequency, VoIP is from a radio perspective
not different from any other service; low frequencies are bandwidth limited and cover a large
area, high frequencies are bandwidth abundant but with limited coverage.
55
Rysavy Research – From GPRS to HSDPA and beyond
56
http://www.gsmworld.com/gsmeurope/faq/3g.shtml
46
Since available spectrum is limited the most likely scenario is that existing networks move over
to VoIP when it provides the same capabilities and efficiency as circuit switched technology.
Due to interference in the license exempt bands it will be hard to deploy reliable services as the
number users increases. The idea of deploying a city- or nation wide network in the license
exempt band to offer a VoIP service is therefore greeted with skepticism.
New access technologies has only had a marginal effect on system capacity compared to the
gains derived from creating non-interfering cells. The creation of non-interfering cells has,
without question, been the most important factor for fulfilling Cooper’s law. This evolution is
likely to continue, with the use of more small high capacity cells where user can access
advanced network services.
Different access technologies only show marginal differences in capacity when deployed in the
same frequency band. Rolling out a new network like WiMAX would thus be more of a
political decision about freeing up spectrum resources. In Europe where the 2 GHz band is
reserved for UMTS technology a WiMAX solution would have to be deployed in the 3.5 GHz
band. To achieve acceptable coverage a large number of sites would have to be built. This
makes it plausible that new access technologies will not compete heads up with existing
technologies but more find a niche market such a fixed wireless broadband. From a mobile
VoIP perspective it is thus not likely that new technologies could bring any significant
improvement over already deployed infrastructure using circuit switched technology.
VoIP can bring significant improvements for network operator in form of a network that is
easier to maintain. This benefit has to be weighed against the increase in bandwidth over the
radio interface that a VoIP solution implies. This also shows the fundamental difference
between fixed line VoIP and wireless VoIP. In the fixed world the last mile access bandwidth
will always be abundant from a voice perspective. Using 10 kbps or 100 kbps will make no
difference when the available bandwidth will be 10-100 Mbps. The radio spectrum will as
shown always be a scarce resource that will require efficient use.
It is thus likely that the core network will start the VoIP-transition leaving the radio interface
unchanged until VoIP is as effective as circuit switched solutions. Current networks are highly
47
optimized for carrying voice traffic and new technology using VoIP in the radio interface will
only be introduced when it can provide equivalent as effective.
4.2 Cost structure of cellular network57
As seen in the graph below, the major expenses for a mobile operator are associated with
marketing, billing and administration, constituting almost 55% of total cost. Equipment costs
such as base stations and switching equipment constitute only a minor part, roughly 15% taken
over system lifetime.
58
Operator costs are divided into two main groups, capital expenditure (CAPEX) and operating
expenditure (OPEX). CAPEX comes from investments made in infrastructure, such as base
stations, base station sites, switching equipment or other expenses related to building the
network. OPEX originates from three different kinds of costs:
Customer driven: Includes costs to get new subscribers to the network, terminal subsidies
and dealer commissions.
57
,57 J. Zander- Low Cost Broadband Wireless Access – Key Research Problems
and Business Scenarios
48
Revenue driven: Associated with the cost to get people to use the services in the network or
the cost that the traffic is generating. Marketing and sales staff belongs to revenue driven
costs.
Network driven: Cost of running the network, transmission, site rental and maintenance.
The estimates below show that CAPEX costs are dominated by the cost of sites. Site costs
include acquisition and civil engineering and will vary considerably depending on site location.
The costs of base station and network equipment are both approximately half that of the cost
of the sites.
49
One can summarize the cost structure as:
•
The cost of equipment is not the dominating expenditure of overall network CAPEX or
OPEX
•
Equipment cost is likely to be reduced over time as prices fall, and the fraction between
equipment cost and overall cost will be reduced even more
•
Site construction and deployment are a major expenditure.
•
Maintaining the network is costly
4.2.1
Comments
Since a large part of the cost of running the network comes from administration and
maintenance there could incentives for introducing an IP-based infrastructure. Legacy
telephone network requires a variety of experts on different parts of the system. With a more
IP-centric system there will be easier to find skilled people at a lower cost.59
The cost of sites is dominating the capital expenditure. New access technologies like WiFi that
is easy to install could be a viable path for extending network coverage and capacity especially
if an investment could be shared between the location owner and the operator. The cost for
the equipment constitutes a minor part and is likely to fall as more base stations are deployed
59
Interview Stefan Hagbard TeliaSonera
50
and more manufacturers enters the marker. This will make it hard to provide a solid business
case for a technology like WiMAX to compete for the same type of clients as current mobile
operators. It is thus likely that new access technology will find its niche market as a provider of
broadband access in rural areas with underdeveloped infrastructure.
4.3 Voice over IP and WCDMA
WCDMA is the leading 3G-radio access technology and is sometimes said to be the European
3G technology referring to CDMA2000 as the American version. WCDMA stands for
Wideband Code Division Multiple Access. WCDMA has similarities with the GSM system
since it uses the same packet core network found in GPRS. In fact, a network that has been
upgraded to support GPRS needs only a new RAN, Radio Access Network. However, the new
RAN is substantially different and requires large investments to achieve reasonable coverage.
There were two main focuses when developing UMTS/ third generation network, first and
foremost it was seen as necessary to increase the capacity of handling voice calls. The second
reason was the possibility of offering new and advanced services. The developers found it
most suited to give the standard both circuit- and packet switched capabilities. The higher bit
rates provided by the network would enable services such as video telephony and interactive
gaming. The uptake on these services has however been slow and operators are consistently
seeking ways to spur network traffic.
4.3.1
Softphone’s impact on capacity.
A user equipped with a laptop PC and a 3G data card, could
use a softphone [3.3] to place phone calls Existing softphones are
developed for fixed broadband connection with little or no
concern is taken for bandwidth consumption. A WCDMA system
is, like other CDMA based systems, interference limited. This
means that when a user transmits data he adds to the interference
in the system. It is therefore essential to use resources as
efficiently as possible. With a higher bit rate overall system
capacity falls dramatically.
Figure 15. PC-Data card for
3G connectivity
Harry Holma provides in WCDMA for UMTS a simplified
example how packet overhead affects system capacity. A 12.2 kbps voice carries 244 bits per
51
20 ms. This will be our reference. In tests I have conducted Skype has consumed a bandwidth
of roughly 30 kbps. This means that each packet contains 600 bits per 20 ms packet. The
reduction in capacity would then be:
Skype _ bit _ rate
10 log
circuit _ switched _ bit _ rate 600 _ bits 10 log
= 3.9dB
244 _ bits Resulting in a reduced system capacity of
110(3.9 /10) = 59%
These calculations will only give rough estimates of the impact of VoIP applications in the
radio network. But they do, however, give and indication on the importance of managing
available resources in an efficient way.
The major penalty that is associated with the use of Skype is produced by the overhead from
the IP-protocol as described earlier. The fact that Skype uses a 13.3 kbps codec has only minor
effects. Header compression is however to be implemented in coming releases of both
CDMA2000 and WCDMA.
With the use of header compression it will be possible to reduce overhead on to approximately
3.5 bytes. With a 12.2 kbps codec this would give a penalty of
244 + 3.5 8 10 log
= 0.5dB
244
Resulting in a capacity reduction of only
110(0.5 /10) = 11%
Many networks have both GSM and WCDMA. An operator can move voice users between
the different access technologies depending on network load. This will however not be
52
possible with a softphone solution. Legacy packet based network like GPRS cannot provide
the radio link that a VoIP call requires. It is both bandwidth limited and suffers from high
latencies, possibly exceeding 800 ms. Handoff between different base stations is not as smooth
as in WCDMA.
4.3.2
Latency
Latency could potentially be a problem for VoIP communications in current releases of
WCDMA. A typical value today is close to the 150 ms limit that ITU has set for achieving
acceptable quality [2.8.2]. Latency issues become even more problematic in loaded networks.
4.3.3
HSDPA
Many WCDMA network operators are in the process of planning or implementing 3GPP
Release 5. The key element of release 5’s higher data rates capabilities is the introduction of
HSDPA, High Speed Downlink Packet Access. In HSDPA a new transport channel is
introduced, HS-DSCH or High Speed Downlink Shared Channel. Two of the most prominent
features of WCDMA, fast power control and variable spreading factor are not found in
HSDPA. Instead the upgrade introduces:
•
Adaptive Modulation and Coding, AMC,
•
Fast Packet Scheduling, FPS,
•
Short Transmission Time Interval, TTI.
•
Hybrid ARQ.60
Power control and AMC
HSDPA introduces the Fast Link Adaptation function and higher order modulation (16
QAM). Instead of controlling the output power the system focus on adapting the modulation
to the radio environment. Higher order modulation requires improved EC/Ior ratio. High ratio
is available to users close to node B or users that have good radio environment in general. This
feature, commonly known as rate adaptation or link adoption, is very efficient for services that
can tolerate variations in the data rate.61 In a live system the scheduler evaluates different users
radio condition, how much data that are in node B’s buffer, time passed since last transmission
60
UMTS Networks: Architecture, Mobility and Services, by Heikki Kaaranen,
Ari Ahtiainen, Lauri Laitinen, Siamäk Naghian, Valtteri Niemi
61
Ericsson White Paper, WCDMA Evolved
53
and other parameters. The scheduler uses this information to determine which user to send to
and what modulation and number of codes that shall be used.
Short TTI
The short TTI makes it possible for the scheduler to take advantage instantaneous favorable
radio conditions. Thus with a large number of users in a cell it will be possible to transmit on
high data rates a large part of the time by switching between users with a short time interval,
see Figure 16.62
Figure 16. Short TTI makes it possible to benefit from instantaneous favorable radio
conditions. 63
Latency
In HSDPA intelligence is moved from the RNC to Node B. This way retransmission can be
controlled directly by Node B, which means shorter delay and faster retransmissions. HSDPA
will also use Hybrid-ARQ, which will decrease the number of retransmissions in the radio
access Latency will also be reduced, field test show latency numbers down to 40ms in the radio
network compared to 150 ms in current releases.64
62
Rysavy Research, Data capabilities GPRS to HSDPA and beyond.
63
Ericsson White Paper – Evolution of WCDMA
64
Interview Johan Sköld, Ericsson Research
54
Performance:
Using 15 codes it will be possible to achieve data throughput of 14.4 Mbps. 15 codes will
however not be available in initial releases due to the complexity of memory handling and the
processing power required. System capacity will also increase compared to WCDMA due to
the use of higher order modulation and an improved scheduling process, optimized for
increasing system throughput. In WCDMA the scheduling is optimized to achieve data rates
with little variation through out the cell.
Improved scheduler benefits packet switched traffic
The scheduler in HSDPA is highly optimized for packet switched traffic. Today WCDMA
only uses dedicated channels for data. The dedicated channels are relatively inflexible and it
takes time to adjust to changing requirements. When using a circuit switched voice service, a
16 kbps dedicated channel will be able to support it. With VoIP, there will be a need to
sometime update the header compression context and a higher bandwidth is thus required.
This is not done easily with current WCDMA.
With the scheduler used in HSDPA it will be possible to intelligently hold packets in the base
station depending on what type of service that is being transmitted. Since speech is produced
with an interval of 20 ms where each packet contains only small amount of information a few
packets could be sent in one of the high capacity time slots if the service parameters allows.
Packets in line to be sent over the air interface can also be stripped of their IP-header and be
sent as one big packet. By trading between delay and capacity HSDPA will be able to offer
higher data using VoIP than existing circuit switched solution.
HSDPA is as mentioned an improvement only to the downlink. Tests show however that
system voice capacity would benefit from an introduction of IP-based telephony even if it
were only initially applied to the downlink.65
65
Interview Mats Nordberg, Ericsson Research
55
Evolution
HSDPA will be succeeded by HSUPA sometimes in 2007-2008, see Figure 17. HSUPA, High
Speed Uplink Packet Access will bring improved performance to the uplink. HSUPA will use
some of the characteristics of HSDPA such as short TTI and HARQ. Due to physical
characteristics of the uplink it will not be possible to introduce higher order modulation in the
same way as with HSDPA. The increment in uplink capacity is believed to enhance overall cell
throughput with as much as 85.66
The advantage of the UMTS system is that it allows operators to follow a highly standardized
path. UMTS integrates operators’ legacy GSM network with the data solutions of tomorrow.
The massive industry support renders infrastructure and end user terminals at competitive
prices.
Figure 17. Evolution of WCDMA.
4.3.4
Comments:
UMTS/WCDMA is the dominant 3G technology today and is likely to become even more so
in the near future. Its close connection to GSM/GPRS makes it the natural step for such
operators to upgrade their networks. In Europe the choice of WCDMA is very much a
political one and the future steps are likely to involve several regulatory agencies and other
governmental institutions.
66
Rysavy Research, Data capabilities GPRS to HSDPA and beyond.
56
There is also a significant reduction in system capacity when using a non-optimized VoIP
application like Skype. The reduction is due to the fact that the Skype uses a high bit rate codec
and that the protocol produces a lot of overhead. From an operator’s perspective replacing a
highly efficient circuit switched solution with a wasteful packed solution seems highly
unrealistic. It is thus not in the operator’s interest to move their voice service users over to
VoIP before there are some significant improvements done.
Although technically possible to use VoIP in a WCDMA network today it is not likely to have
any major impact before the aforementioned upgrades are implemented.
According to industry representative a fully functional end-to-end duplex VoIP solution is
likely to be available on the WCDMA market in 5 years time.67 The reason for implementing
this will be that it provides a more efficient solution than existing circuit switched technology.
The GSM/GPRS networks that have already been deployed are expected to remain in service
for a long period of time, some say past 2015. These network provide efficient voice service
using circuit switched technology meaning that VoIP will coexist with legacy equipment.
4.4 Voice over IP and CDMA-2000
Qualcomm developed the first commercially available CDMA standard in the 1990s. It has
since then been deployed in various countries in the US, Asian and South American markets.
CDMA2000 is an enhancement to that original technology and provides in particular higher
data rates. Today there are some 270 million subscribers using CDMA technology. 68
Although considered part of the IMT2000 standard, which is a “3G stamp”, CDMA2000 1X is
not thought of to be a genuine 3G standard due to only moderate data rates. In a comparison
with the GSM/UMTS standard CDMA2000 1X is more of an equivalent to the GPRS
overlay.69 It is instead the EV-DO that provides the enhanced data capabilities.
67
Interview Ericsson
68
http://www.cdg.org/worldwide/index.asp
69
3G Standards: The Battle Between WCDMA and cdma2000 – paper presented
at Nordic ICTR Workshop, August 2004, Helsinki
57
4.4.1
EV-DO
CDMA2000 1xEV-DO stands for Evolution-Data Only and is an enhancement to the data
capabilities of CDMA2000. It offers peak downlink data rates of 2.4Mbps. The data
capabilities in DO are handled in a separate channel thus requiring additional spectrum. This
overlay lacks circuit switched capabilities making it effective on providing high data rates.
In a CDMA2000 typically 1.25MHz is used for voice and with EV-DO an additional 1.25MHz
is occupied for providing enhanced data capabilities. EV-DO uses much of the same
techniques found in HSDPA to improve data rates. These include higher order modulation,
efficient scheduling adaptive modulation and coding. The key difference is however that EVDO does not have circuit switched capabilities. There are currently some 17 Million
subscribers using EV-DO 3G technology.
58
4.4.2
Evolution
Rev A is mainly an enhancement on the uplink. Downlink throughput could achieve 3.1 Mbps
and uplink would be boosted up to 1.8 Mbps.70 Rev A also brings some important
improvements for maintaining quality of service using flow based control.
Initially DO was to be followed by DV, or Data-Voice. DV would have circuit switched
capabilities. Qualcomm has said they are not going to push a technology shift to EV-DV and
the future for it is a bit unclear.71 With Rev B, EV-DO will be able to provide scalable
bandwidth solutions. This will make it possible for operators to deploy EV-DO in bandwidths
from 1.25 MHz up to 20 MHz providing a maximum 46 Mbps in the downlink and 27 Mbps
in the uplink.72
Figure 18. Evolution of CDMA2000
4.4.3
Comments
A problem for operators that decide to go with EV-DO is that they cannot allocate voice
services in the EV-DO frequency band since it has no circuit switched support. Today, data
only accounts for a small part of network utilization and in even lesser extent to revenue. Since
an operator cannot use the EV-DO spectrum for voice they will lose efficiencies gained from
70
http://www.qualcomm.com/technology/1xev-do/revA.html
71
Northstream White Paper, Operator options beyond 3G.
72
http://www.qualcomm.com/technology/1xev-do/solution.html
59
load balancing and trunking efficiencies. Another issue with EV-DO is that a terminal cannot
maintain voice and data connections simultaneously much like GSM today.73 It is therefore
necessary for operators with EV-DO to look at VoIP in order to use their specttrum
efficiently. Verizon has already shown some interest for such solutions.74
When WCDMA has been upgraded with HSDPA/HSUPA there is likely commercial systems
available that uses EV-DO and VoIP. This means that for the time being UMTS operators do
not have to put major focus on full-duplex VoIP alternatives.
4.5 Voice over IP and WiMAX
WiMAX stands for World-wide Interoperability for Microwave AXess. It has emerged from
the IEEE 802.16 WirelessMAN (Metropolitan Area Network) group of standards. The
standard has been around a few years, first as 802.16 in 10-66GHz and later in 2003 as 802.16a
in 2-11GHz.75
To promote the spread of the standard WiMAX-Forum was founded in 2001. IEEE is only
involved in the MAC and physical layer of a standard, so the forum fills an important part in
the process of obtaining interoperability. The WiMAX forum has evolved from being
compromised by a few tech companies to a massive force supported by over 300 industry wide
enterprises, ranging from chip manufacturers to mobile operators.
Today there are a few twists to the 802.16 standard, the original, 802.16a has been more or less
overrun by newer revisions. Currently work is being focused on 802.16-2004 or 802.16d which
was ratified in October 2004. Next step will be 802.16e, a mobile flavor that follows the
802.16-2004 standard yet incompatible with it. The mobility of 802.16e is likely to come in
incremental steps starting with fixed/nomadic use where users have to turn off their
equipment during transport.76
73
Rysavy Research, Data capabilities. GPRS to HSDPA and beyond
74
http://www.unstrung.com/document.asp?doc_id=71150
75
Signals Research Group, WiMAX Opportunities and challenges in a
Wireless World
76
Rysavy Research, Data capabilities. GPRS to HSDPA and beyond
60
Usage:
The 802.16-2004 is a fixed wireless technology. A fixed technology is one that is not easily
movable, e.g. they have large outdoor antennas. This means that it is thought of to be a
replacement technology for DSL or broadband cable providers in such areas where
infrastructure is non-existent or inadequate for high speed Internet. It is likely to be deployed
in scarcely populated rural areas where distance makes it difficult to use DSL technology. It
can also serve as a high-speed data link for corporate subscribers.
Initially CPE will be an outdoor fixed antenna and an indoor modem. The trend is however to
move towards indoor self-installable equipment, see Figure 19.
Figure 1977. WiMAX deployment models
Technology:
Although the final word has not been said in the standardization process there are some key
elements that will not be changed. WiMAX uses OFDM multiplexing technology with variable
77
http://www.lightreading.com/document.asp?doc_id=82739&page_number=1&image
_number=1
61
levels of QAM modulation78 WiMAX uses many of the same technologies that are found in
HSDPA and EV-DO to increase throughput and spectrum efficiency, these include HARQ,
adaptive modulation and efficient scheduling.
Spectrum:
Initially 802.16-2004 was a 2-11GHz-band technology. There is also potential to deploy
WiMAX in the cellular bands and 700MHz. The most likely are however, 2,1GHz, 2.3GHz,
2.5GHz, and 3.5GHz.79 This poses some threats to achieving cheap mass market CPE since
these would have to comply with a number of frequencies. Initial focus is however likely
however to be on 3.5GHz. The 3.5 GHz band is in many countries restricted to fixed or
nomadic use making thus making it a less plausible competitor to existing mobile networks
such as UMTS.
4.5.1
Performance:
It is a bit difficult to assess real network performance of access technologies. The actual
throughput that a user may experience depends on a variety of factors. Conducting
measurements on a single isolated base station may lead to numbers that are very favorable
compared to a live system that receives interference from a multitude of surrounding base
stations and active users. Since there are no major deployments of WiMAX systems there are
no “real world” data available. The standard, however, supports up to 70 Mbps in 20 MHz
spectrum.80 Such high rates will not be available over large distances since it requires 64QAM
modulation, which requires excellent radio environment. Figure 20 shows the capacity of a 3.5
MHz carrier deployed at 3.5 GHz. As shown the data rates are more down to earth than
numbers often seen in the press where 70 mbps over 50 km seems highly realistic.
78
Intel White Paper, Understanding WiMAX and 3G for Portable/Mobile
Broadband Wireless
79
Michael Thelander: WiMAX, Opportunities and Challenges in a Wireless
World
80
http://www.wi-fiplanet.com/tutorials/article.php/3550476
62
Figure 20. WiMAX performance over 3.5 MHz bandwidth deployed at 3.5 GHz81
When used in smaller frequency bands like 5 or 10 MHz it is not clear if it has a direct
advantage to HSDPA.82 The technology will however be more efficient in the uplink due
better orthogonality than WCDMA systems. An advantage of OFDM based technology is that
it is easy to scale up if more bandwidth is available something that cannot be done with in a
WCDMA. This makes it attractive for point-to-point operation at high frequencies where
bandwidth is abundant.
QoS:
To be able to blend a variety of services, including VoIP, there needs to be way to prioritize
between traffic. The WiMAX standard has support for QoS in its MAC layer.83 This is
probably an even more important factor than achieving higher and higher data rates.
4.5.2
Deployment:
Interoperability testing for 802.16d has recently been initiated. There are thus no “real”
WiMAX solutions available today. What exists are a number of “WiMAX ready” or “preWiMAX” solution. Companies providing such products states that only small software
81
http://www.lightreading.com/document.asp?doc_id=82739&page_number=1&image
_number=2
82
Rysavy Research, Data capabilities. GPRS to HSDPA and beyond
83
http://www.wimaxforum.org/technology
63
upgrades will be necessary to achieve compliance and interoperability, however some research
believes that new ASICS solutions will be necessary, especially for the CPE.84
In July, Sprint and Nextel stated that the combined entity is under obligation to
“fulfill its voluntary commitment to meet certain milestones for offering service in 2.5GHz
band, unless circumstances beyond its control prevent the merged entity from reaching those
milestones.” The merged company is to provide 15 million Americans with wireless broadband
within 4 years and then additionally 15 millions within 6 years. This could lead to a significant
boost to WiMAX access technologies. Sprint is cooperating with Intel to provide chip sets and
with Motorola for WiMAX trials starting late 2005 through 2006.85
4.5.3
Comments
WiMAX like EV-DO does not have circuit switched capabilities and will thus need to support
VoIP, the revenue generated from voice services is too large to ignore. It is not clear however
how the mobile flavor of WiMAX will be marketed or even built. Going heads up with
operators of traditional UMTS or CDMA2000 technology is far too challenging for probably
any company. In some aspects WiMAX offers higher throughput than UMTS. But the success
for a new technology will probably not lie in download speed alone. This fact has been proven
for 3G where only recently the data traffic in the networks has started to increase due to the
introduction of new services.
WiMAX should have potential as a fixed wireless broadband solution has potential to be
economically successful. The mobile flavor 802.16e, see Figure 19, will meet fierce competition
from both CDMA2000 and WCDMA: As said, the actual throughput is likely to be fairly
similar regardless of what technology is to be used. The high throughputs that are often
mentioned with WiMAX are mainly because of use of more spectrum (20 MHz) and favorable
radio conditions and less about the physical characteristics of the radio channel. The mobile
flavor will also have a hard time to explain what the actual difference between different 3G
technologies actually is, selling only on higher data bit rate will probably fail. As voice
84
Michael Thelander: WiMAX, Opportunities and Challenges in a Wireless
World
85
Unstrung,
http://www.unstrung.com/document.asp?doc_id=78580&WT.svl=news2_2
64
communication is still what constitutes the lion share of mobile operator’s revenue the supply
of attractive handsets is crucial. Considering the time it has taken since the initial launch of
UMTS to meet market demand of slick and functional terminals the business case for mobile
WiMAX looks even grimmer.
4.6 VoIP and WiFi
WiFi is the acronym for Wireless Fidelity. WiFi has been developed by the IEEE, under the
working name 802.11, to provide medium range wireless connectivity. The different flavors of
the 802.11 family are all characterized by their use of unlicensed spectrum in either the 2.4GHz
(802.11b, 802.11g) or the 5.8GHz (802.11a) band. Coverage can be up to 100 meters under
ideal conditions. The 2.4 GHz band is widely used by different electronic devices ranging
from microwave ovens to DECT-cordless phones. Interference from such devices makes it
hard to deploy reliable communication services. 802.11a uses frequencies in the 5 GHz band
over 12 non-overlapping channels with OFDM technology. For a list of WiFi standards, see
Table 2.
Table 2. Different flavors of WiFi
Standards under consideration
802.11e
Enhances the 802.11 Media Access Control layer for quality-of-service features,
such as prioritizing voice or video traffic.
802.11k
Creates a way for access points for pass specific radio frequency and health and
management data to higher-level management applications
802.11n
Designed to boost throughput, not raw data rate, to 100M bit/sec. The idea is to
make WLANs feel like 100M bit/sec Ethernet LANs
802.11r
Allows fast, secure, seamless handoff of a VoWiFi connection among access points.
Standard established
802.11a
WLANs for the 5-GHz band, with a data rate of 54M bit/sec. Is becoming less
deployed
802.11b
WLANs in the 2.4-GHz band, 11M bit/sec data rate
802.11d
Enables 802.11 hardware to work in various countries where it cannot today.
802.11g
Also in the 2.4-GHz-band, but used OFDM modulation to reach 54M bit/sec
802.11h
Supports measuring and managing the 5-GHz radio signals in 802.11a WLANs
802.11i
Repairs weaknesses in the Wired Equivalent Privacy encryption scheme
Source: OECD based on data from Network World Fusion
65
Figure 21. WiFi evolution of the most common standards
User equipment
The main driver for hotspot deployment is the proliferation of WiFi equipped laptops. The
market share for laptops is increasing on the expense of desktop shipments and in some
markets laptop share exceeds desktop share. Strategic Analytics expects that the number of
WiFi equipped laptops will top 140 million by 2008.86 According to research firm Dell’Oro the
business segment of WiFi equipment will be worth 1.1 billion dollar and by 2009 it will be
worth 3.5 billion dollars.
4.6.1
Voice over WiFi
Voice over WiFi, or VoWiFi, is the name for providing VoIP over WiFi technology. It has
been seen as specially appealing to enterprise users that can combine their wireless network
with VoIP.
WiFi is seen as a cost effective access technology that can cover small areas at low expense.
Due to its simplicity it does not require any special knowledge to set up a WiFi site.
86
http://www.gsmw orld.com/news/media_2004/
66
Handsets
The pioneers in the VoWiFi handset market are UTStarcom and Symbol Technologies. These
handsets have been available for the corporate market since 2002.
Shipments have been kept at low numbers mainly due to a hefty price tag. New handsets are
however entering the market. Vonage announced in December 2005 that it would bring its
first WiFi handset to the market. At $80 Vonage subscribers will be offered a UTStarcom
F1000. With the device user will be able to place and receive calls at any public hotspot using
their existing Vonage subscription. The relative immaturity of the technology is reflected in
that it will only be possible to connect to public hotspots, i.e. networks that are open. This will
effectively block Vonage users from using the service at closed networks like those provided
found at Starbucks.87
Handsets capable of handling WiFi connectivity in addition to cellular technologies like GSM
or CDMA are referred to as dual mode handsets. Research firm ABI Research predicts that 50
million dual mode handsets will be shipped in 2009.88 More on this in [5.3]
Amongst other, Nokia have announced that they are releasing several new models with WiFi
in 2006.89
With the emergence of Softphones like Skype that offer PocketPC versions new players like
Dell are likely to provide a compelling alternative for providing voice service over WiFi
networks.
4.6.2
Deployment
The deployment of WiFi access points is accelerating as broadband penetration increases and
equipment prices are falling. Today there are about 100.000 public hotspots according to
Informa Telecom and Media90. Research firm ON World expects that by 2009 there will be
174.000 hotspots in Europe alone.91 The number of people connecting to hotspots is growing
87
http://news.zdnet.com/2100-1035_22-5992450.html
88
http://www.3g.co.uk/PR/April2004/6982.htm
89
www.nokia.com
90
http://news.zdnet.co.uk/
communications/wireless/0,39020348,39222683,00.htm
91
www.onworld.com
67
and by the end of 2004 they were about 20 Million.92 Analyst firm Ovum expects however that
the uptake of WLAN will be a slow process in many corporations due to lack of security
guarantees and the availability of existing fixed Ethernet connections.93
Roaming agreements are being established and today it is possible for a user to connect
through WiFi at almost any place in the world. Boingo is a WiFi hotspot aggregator that offers
revenue sharing agreements to hotspot providers so that these do not have to systems to
handle access and billing. Boingo has currently signed up some 20.000 hotspots worldwide.94
T-Mobile has a WiFi network built up by over 7000 hotspots in the US. These are located in
places such as FedEx stores, Kinko’s Office and Print Centers and airports. Subscribers of TMobile mobile phone services can include WiFi connectivity in bundles plans.
There are plans in cities like Philadelphia and San Francisco to rollout citywide WiFi networks
in the next couple of years.95 The idea is to offer affordable broadband access to people in
unprivileged groups.
The main driver for VoWiFi deployment is from leveraging the investments in WiFi networks
that are being deployed in corporate offices. With a VoWiFi solution workers could potentially
increase the functionality of PDAs and laptops. The Radicati Group estimates that the share
of corporate phone lines using VoIP will grow to 44% in 200896.
In a corporate environment a VoWiFi could be integrated into existing PBX by using a
VoWiFi gateway making it possible to connect to the PSTN
In July 2005 Skype announced a deal with Boingo that will enable users of Skype to access
Boingo’s WiFi network.97 The deal lets Skype user get access to Boingo’s nationwide WiFi
network for a discount price of $7.95 per month for unlimited use. For Boingo the deal could
92
http://www3.gartner.com/5_about/press_release/pr17feb2004a.jsp
93
Ovum, Wireless Voice Over IP
94
www.boingo.com
95
http://www.phila.gov/wireless/briefing.html
96
OECD, Development of voice over WiFi by integrating mobile networks
97
http://www.skype.com/company/news/2005/boingo.html
68
boost network traffic and with that revenue. Skype has also shown interest for launching a
Skype capable handset.
4.6.3
Issues
There are a number of issues associated with the emergence of VoWiFi. These are mainly
related to the nature of the underlying protocol stack of the 802.11 family. WiFi was developed
for data usage and delivering high quality voice has not been a priority. The IEEE is in the
process of setting the 802.11e standard. This new member is expected to bring additional QoS
guarantees to WiFi. Since a standardization process is frequently subject to delays the major
player has developed proprietary solutions in order so insure a satisfying level of reliability.
Network Capacity
Offering VoIP services in WiFi networks will put extra stress on network management. The
network has to consist of enough access points to ensure low packet loss and sufficient
capacity. The network has to be monitored so that no access point is receiving an unproportional amount of traffic. Ideally the network should be carefully planned so that
effective penetration through walls, floors and ceilings can be assured. The challenge of
roaming between different access points and sub network has to be addressed. The goal is that
the switchover can be made in less than 50ms.
Bandwidth is also a potential problem in WiFi networks. Both the air interface and backhaul
need careful planning. The bandwidth offered by a base station is most adequate in an
unloaded network and could support up to 5-8 calls simultaneously using 802.11b.98 But with a
number of users running time critical real time applications like voice, packet loss will become
a concern.
Due to the limited coverage and capacity of an WiFi-access point it will be necessary to use
quite a large number of them to achieve acceptable quality. This in turn means that switching
between different access points will become an even greater problem since it will happen
frequently, see Figure 22.
98
OECD, Development of Voice Over WiFi
69
Figure 22. Multiple access points will be needed to achieve good quality in a WiFi
network. This requires that handoffs between access points can be done effectively.
Security
The security concerns of WiFi are not unjustified. The most commonly used technology to
increase security is called WEP. WEP is considered fairly easy to crack requiring some 10-15
million packets, equivalent of 2 weeks of network traffic in a small network. Since there are
programs available to automatically do this it poses a serious threat.99 There are however new
standards like WPA that provides enhances security. With upcoming standards in the 802.11
family security will further be addressed.
Using security standards such WPA a user has to be re-authenticated on every switch of access
point, a process that takes some 500ms. To solve this issue the IEEE have developed the
802.11r standards.100
4.6.4
Comments
There has been considerate debate over the business case of deploying WiFi networks. Some
believes WiFi will become a commodity that will be taken for granted, much like hot water in a
hotel room or a newspaper at a café. For a mobile operator offering WiFi-service to their
subscribers could be a way to increase customer loyalty. A study conducted by Pyramid
research found that the WiFi venture of T-Mobile would more be justified if the churn from
99
100
http://www.wi-fiplanet.com/tutorials/article.php/2106281
Mobile VoIP Competitive Landscape, On World
70
T-Mobile cellular business was decreased by only 0.2%.101 On the side is Verizon, which in
August 2005 pulled the plug on its WiFi rollout in New York City. The company stated that
they are instead pursuing with its strategy to deploy its EV-DO technology.102
Creating a nationwide WiFi network will require and significant amount of roaming
agreements. These agreements will not only involve comparable WiFi providers but most likely
also various wire line operators, mobile operators and ISP’s. As shown in previous chapters it
will not be cheaper to use WiFi compared to other access technologies for creating a
nationwide footprint. For mobile operators VoWiFi should not be seen as a major threat.
Voice communication is not only about placing calls; it is also, perhaps even more, about being
able to receive calls. It is likely that people will continue to pay a premium for this feature
exclusively offered by mobile operators.103 With bundled offers mobile operators are likely to
be able to retain minutes placed and received in their network. However, mobile operators are
aggressively promoting broadband connections using 3G networks. These may encounter
fierce competition from citywide rollout of WiFi networks. The advantage of building a
citywide network is that its coverage is clearly manageable. As described in [4.1.2] providing
high bandwidth over a densely populated and limited area makes economic sense. A mobile
operator will not only be judged on its peak performance but also if it can provide reasonable
coverage. It will therefore be necessary for operators to also service less profitable regions.
Voice over WiFi is likely to gain considerable traction in the near future. The key point is that
it the technology is bandwidth abundant considering the small coverage area. This makes
tackling the radio limitations described above less problematic. It is also a very cheap
technology meaning that operators or business owners will not stand or fall with the choice of
technical solutions.
101
Pyramid Research, Is WiFi wagging the 3G dog?
102
http://www.devxnews.com/article.php/3501611
103
Interview Jens Zander
71
4.7 VoIP and FLASH-OFDM
Flarion is the company behind FLASH-OFDM, Fast Low-latency with Seamless Handoffs
Orthogonal Frequency Division Multiplexing. Flarion is currently in the process of setting its
solution as the IEEE 802.20 standard for mobile broadband. However, the process has slowed
down with the announcement of the mobile version of 802.16e. IEEE does not allow
overlapping standards and the similarities between the two technologies might make it hard to
getting an IEEE stamp on 802.20.
4.7.1
Technology
A key feature of the Flash-OFDM system is that it can handle prioritization. When a user
enters the network that users profile is determined. The system can set priority according to
the user’s profile; a high paying customer can “skip the line” in a congested network. The
system can also support different priority classes. Where for example voice is prioritized over
simple web browsing. This type of differentiation makes Flarion’s system suitable for
emergency communication. In the event of a large-scale emergency, rescue workers will be
given priority over any other type of traffic104. Accordingly a Flash-OFDM system has been
implemented in Washington D.C to support their testing of public safety network.105
The Flarion technology is available today in a wide range of frequencies from 400Mhz up to
3.5Ghz. It uses 1.25MHz of paired spectrum or 5MHz paired if deployed with the
abovementioned Flexband technology. Due to restrictions in the issued 3G licenses Flarion
technology may find it hard to move into the 1.9-2,1 GHz territories in Europe. Moreover
Flarion offers a ready to ship consumer products such as PC-Cards although the company
expects third party manufacturers to produce the devices under license.106
4.7.2
Deployment
The weakest side of Flarion solution is its current deployment. At the moment operators are
conducting trials at several different location although a fully commercial deployment has yet
104
Signals Ahead, No.1 April 19, 2004
105
Government of District of Columbia.
106
Northstream White Paper, Operator options beyond 3G
72
to be seen. This might change with the decision of the finish government to go for the Flarion
option in the deployment of a 450MHz nationwide data network.
Nextel has since 2004 had a test system up and running in Raleigh, North Carolina. With the
merger with Spring Nextel has dropped its plans for deploying a nationwide FLASH network
since Sprint has followed the ED-DO path. Operator implemented VoIP
4.7.3
Comments
What makes FLASH-OFDM interesting is that is based on OFDM. A seems to be well
understood that the next generation radio access technology, Super 3G or 4G, will be based on
this technique. Also interesting is the highly data centric approach that FLASH-OFDM
provides where low latency has been a priority during development. In 2005 CDMA giant
Qualcomm acquired Flarion. This move is seen as a recite on the potential of the Flarion and
the OFDM technique.
73
5 Mobile Operator implemented Voice over IP
When speaking with mobile operators and equipment providers, their view of Voice over IP in the mobile
networks is not about implementing a full duplex VoIP solution to replace existing circuit switched technology.
Operators are seeking ways to introduce new services and additional content in a cost efficient manner. For this
IP has showed to be very promising. In this chapter we take a look at how the industry views the ongoing
transition towards IP and VoIP.
5.1 IMS
IMS stands for IP Multimedia Subsystem. It is a concept developed by 3GPP to enhance
GSM-operators’ capabilities in deploying IP-based services in 3G cellular networks. With IMS
an application can be developed and launched and instantly be accessible through various
platforms and different network access technologies, see Figure 23.
Figure 23, IMS will enable fast and efficient content distribution over multiple
platforms.
74
The promising deliverables of IMS have piqued the interest of the fixed network world in its
path to upgrading to next generation networks. These fixed network operators are currently
among the strongest supporters of IMS because they have a more urgent need to seek
additional revenue; telephony revenue is falling and broadband prices keep plunging so the
operators need a quick fix to create revenue generating applications.107
The Internet can be seen as one big cloud that contains every possible format and application.
The majority of these applications run independently of one another making it difficult to
develop services that can be accessible without downloading extra plug-ins or applications.
The strength of the telecommunication industry has traditionally been its ability to create
standardized services e.g. SMS, and MMS. Consider the example of a cellular subscriber who
whishes to look at a video clip: if it is done from a mobile phone the subscriber will most likely
not have to download a new application to view the clip. However, if the same clip were
accessed from a standard PC over the Internet, the recipient may need to download additional
applications (QuickTime, Window Media Player, Flash, Real Video, etc.). Subscribers of phone
services thus have higher expectations that things will work. The telecommunication industry
sees this as an asset justifying their premium prices.108 IMS will be able to make the real time
adaptation, depending on what type of access the subscriber is on. This is referred to as
network awareness. If the subscriber’s use a laptop pc connected to a broadband access a
richer and fuller content can be delivered as opposed to if the user access the same content
from a mobile phone. The idea is to deliver content with minimal adjustment over different
access technologies.
5.1.1
IMS – Technical overview
IMS will function as a bearer of signaling and traffic over an IP-layer and operate as a “routing
engine” or “session control” application that can match different user’s profiles with the right
call/session-handling servers.
IMS relies on SIP as its basic signaling mechanism. Simply put, SIP is a way of finding and
routing control signals between endpoints. The IMS architecture extends this function and
puts it into a framework of networks elements to fulfill various roles such as service triggering,
107
Light Reading; Telecom’s Technology Hotspots
108
Interview Vodafone
75
authentication (is the person who he claims to be), authorization (is this person allowed to
access such data), and network interconnection. It is thus an architecture that provides IP
connectivity between endpoints. IMS is also thought to provide greater support for QoS
functions currently not found in SIP.109
CSCF and HSS
With IMS a user will access content through a dynamically associated standardized access
point, the Call Session Control Function, CSCF. The CSCF routes different kinds of
applications to endpoints in the network
The HSS, or Home Subscriber Server, creates a database for subscriber data with identity
profile and billing permissions for all the services and devices in the account. The HSS can also
interact with the Home Location Register HLR in mobile networks and keep information
consistent over multiple networks. What this means is that a user will be able to access the
same personalized services regardless of what type of access technology he is using. A user can
start a session from his home PC watching a video clip and maintaining a conversation with a
friend talking about the video clip. When the user leaves his home he can maintain the session
on his mobile phone. The transition between the platforms is handled by IMS.
As shown in Figure 24, IMS operates outside the packet core of the network making it access
and even backbone agnostic.
109
Webinar Light Reading,
http://www.lightreading.com/webinar_archive_home.asp?webinar_id=27493
76
Figure 24, IMS system architecture
5.1.2
Implementation
With Release 5, IMS is introduced in the core network. The IMS framework does not
specifically introduce a VoIP service. It is more a way to simplify the establishment of IP
sessions between terminals. It is however left to the Open Mobile Alliance to specifically
design the applications and to ensure interoperability of handsets from different
manufacturers. This is where the ball currently lies in the rollout of PoC.110
In an industry survey presented second quarter of 2005, ABI Research asked 125 people within
the telecommunication industry when they expected that IMS would be available,
110
Interview Anna Svensson, Ericsson
77
When will IMS be universally available?
30
25
20
(%) 15
10
5
0
2005
2006
2007
2008
2009
2010
2011
2012
2013
Source: ABI Reasearch
5.1.3
Convergence
Convergence means eliminating barriers that exists between different platforms and networks.
The idea of being able to offer services that will run on all existing platforms is attractive to
both operators and specific service providers like sports network ESPN. It will ideally make it
possible for them to develop a single service that can be broadcasted to a variety of platforms
and networks without specific adaptations. IMS enable operators to rapidly develop or buy
and launch new services into their network.111
The possibility of offering blended services and faster service introduction are the main
benefits for service providers. For equipment vendors, network convergence and faster
service introduction are seen as the key benefits of IMS.
5.1.4
New Services
In a GPRS network it is possible to establish a connection between a terminal and a location
on the Internet. What IMS does is to improve this by making it a lot easier to establish IPconnections between two terminals. It is not a specific service that will be offered to customers
but instead it is seen as a service enabler. Higher-level application will use IMS as an efficient
way to enhance communication between people. It is possible to achieve the same bells and
111
Interview Vodafone
78
whistles in other ways without using IMS, but it gets increasingly difficult to provide seamless
integration between services and networks as the number of application grow. IMS will reduce
the need for “inventing the wheel” every time a new application is to be launched.112
Presence and Group List Management
Presence and group list management will be two of the top
features in the IMS system. With presence it will be
possible to allow a set of users to obtain information
about the availability status of other users, see Figure 25.
IMS can distinguish between different media types, users
and user preferences. The system will also be aware of
what devices a user is available on, e.g. laptop, PDA,
mobile phone, the type of connection he is using and
adjust the content to give the user a better experience.
When accessing a movie clip from a laptop using high
speed broadband connection the user receives high quality
images and sound whereas when accessing from a mobile
phone the content has to be adapted to moderate
bandwidth and small screen size.
Figure 25, Buddy list
With group list management users will be able to create network-based group definitions that
can be used by any service within the IMS system. A user can set up buddy lists, block lists,
public and private groups and different chat groups. In short, the user will be able to control
all different forms of communication easily from one place.113
These services provided by IMS can then be integrated into applications such as “Push-toTalk”. IMS is attractive to the industry because it will be significantly easier to develop services
at a lower cost, as compared to developing stand-alone applications.114
112
Interview Vodafone
113
Ericsson White Paper, IMS – IP Multimedia Subsystem, Ulf Olsson
114
Interview Ericsson
79
Combinational services
IMS will also make it possible to offer combinational services. An example of a combinational
service is one in which two persons are playing a game while at the same time maintaining a
voice conversation. With IMS it will be possible to do this over one radio bearer using VoIP.
Another scenario is that the use of the contact list will accelerate the demand for sharing
photos and other files while at the same time having a conversation. This can be done today
but it requires two separate radio bearers thus consuming resources inefficiently.115
5.1.5
Comments
IMS is more seen as an evolution than a revolution. Services could be developed and deployed
in the network without the use of IMS. In the long run however it is seen as necessary to be
able to fast and over multiple platforms launch new services. The mobile centric view focus
has now moved towards the fixed line industry since this sector is seriously lacking attractive
services to compete with both fixed VoIP services and mobile telephony.
IMS could be seen as a way for network equipment providers to push a new technology to the
market. For content providers it is seen as an effective way of delivering their product.
However, considering the low market uptake on services launched in 3G networks, some
operators are not looking for a way to more easily launch new services but to actually find new
services to launch.116
It is clear that third party applications that run outside IMS could provide equal functionality as
IMS. The goal for the industry is however to create an eco-system in which they can keep
control over network services and charge accordingly.117
From a VoIP perspective it is IMS’s capabilities to manage point-to-point IP connection and
the interaction between different networks is the most interesting. IMS will however make it
into the mobile networks long before VoIP does. Many of the features found in IMS are
dependent on VoIP for successful implementation. These can be the different push-services
115
Interview Nokia
116
Tele2
117
Interview Nokia
80
and services based on presence. VoIP is necessary for applications provided by IMS to
seamlessly interact.
5.2 Push-to-talk
Push-to-talk will become a standardized feature in coming releases of UMTS. This can be seen
as the first network backed technology to use voice over IP in the UMTS world.
Push-to-talk is a service that introduces one-to-one and one-to many instant voice
communication. The principle is straight forward, just push and talk, the answering party does
not even have to activate its terminal to answer the call, making it similar to a walkie-talkie. 118
Nextel has offered the service in its iDen network for almost 10 years.
Push-to-talk can be divided into three categories:
•
Proprietary packet switched
•
Proprietary circuit switched
•
Standardized packet switched, or Push-to-talk over cellular (PoC)
The third option is the one drawing most attention lately. PoC refers to the Push-to-talk
standard backed by the Open Mobile Alliance. OMA seeks to make PoC available across a
variety of terminals and platforms thus increasing interoperability and service attractiveness.
PoC will be first widely deployed application using Voice over IP.
Ericsson, Nokia, Siemens and Motorola drafted the PoC specification in the fall of 2003. The
specification uses functionality found in IMS.
5.2.1
Services offered through PoC
The first service likely to hit the market is one to one communication, see Figure 26. This can
either operate in auto answering mode or in a fashion similar to today’s phone call where the
caller must accept the incoming call in order to initiate a conversation.
118
http://www.nokia.com/nokia/0,8764,46740,00.html
81
Figure 26. Push-to-Talk, One-to-one.
Additional service will be PoC “one-to-many”, see Figure 27, where the participants are either
pre-defined groups or instantly created groups, the initiating user push on button and will
instantly be connected to multiple other users.
Figure 27. Push-to-Talk. One-to-many
Since PoC will be developed in close cooperation with the IMS standard PoC will add typical
IMS features like presence management, instant messaging and the possibility to share photos
during a PoC conversation119. As PoC is a half-duplex service it is not likely to have a negative
impact on traditional voice revenue. It fills a different need, like a blend between a text
message and a normal phone call.
5.2.2
Technology Performance
In a Push-to-talk over circuit switched technology, timeslots will be allocated for each user
through out the whole session. During periods of silence no data is transmitted but the timeslots are still allocated but no information is sent affecting negatively system capacity. With a
circuit switched solution, transport latency between the endpoints will never be a problem and
typical values are around 150ms. Setup latency would be approximately 3-5 s
119
http://www.3g.co.uk/PR/Feb2004/6603.htm
82
GRPS/EGPRS also uses time-slots as radio resource but available time-slots can be shared by
a group of user and are not constantly allocated. With a packet based approach the choice of
codec can be done quite freely. Codec negotiation is done by the end the terminals and there is
no need for transcoding equipment in the network. Thus to increase network capacity it will be
possible to use low bandwidth consuming codecs like half-rate AMR if the loss of quality is
considered acceptable.
With a packet based Push-to-talk solution transport latency will be higher compared to circuit
switched push-to-talk. This fact is due to some network state transition timers resulting in that
the end-to-end transport latency could be as much as 3 seconds. It should be noted however
that since push-to-talk only offers half duplex communication latency is not as big of a
concern as with full duple duplex communications.
Estimates shows that PoC will be between 5.5 – 13 times more efficient from e network
perspective as circuit switched push-to-talk. This efficiency gain is associated with the fact that
using packet switched technology it is possible for several users to share one time slot in the
GPRS radio channel. In the core-network the packet switched alternative does not show any
great differences in terms of efficiency120.
5.2.3
Deployments
The standardized version of PoC has yet to gain full traction. There are a number of
proprietary solutions readily available on the market. Nokia offers one solution where it can
provide the necessary PoC servers and handsets to operators. Vodafone Sweden offers Nokia’s
solution and subscribers gets unlimited usage of the service for $9.8 per month. Users can also
sign up for a one-day subscription for $0.98.121
The standardized architectural framework for PoC is as mentioned above already done. The
process is now in the hands of OMA to solve interoperability between terminals and different
vendor’s solutions. 122
120
Northstream, Overview and comparison of Push-to-talk solutions
121
www.vodafone.se
122
Interview Anna Kristoffersson, Ericsson
83
5.2.4
Comments
Push-to-talk over cellular is a good example of how VoIP will make it in to the networks. It
has gained attention because it offers a new service deployed over existing infrastructure and
with considerable efficiency gains over circuit switched technology
Being an operator and manufacturers backed service it will become integrated with other
services in IMS such as presence thus creating a functional and, for the user, easy service.
5.3 Unlicensed Mobile Access
Unlicensed Mobile Access, UMA, is an initiative from 3GPP to provide access to the same
services found in GSM/GPRS/UMTS services over unlicensed spectrum technologies such as
Bluetooth or WiFi. The UMA protocol will provide both roaming and handover between
cellular networks and public or private WLANs using dual mode handsets and VoIP.123
Development of UMA began in late 2003 and was finished in 2004124. Participants in the
forum are Alcatel, British Telecom, Cingular, Ericsson, Kineto Wireless, Motorola, Nokia,
Nortel Networks, O2, Research in Motion, Rogers Wireless, Siemens, Sony Ericsson, TMobile US.
Research shows that 80% of mobile calls are being made from either the office or from the
home.125 The industry is enthusiastic about using broadband connection to deliver the same
services found in mobile networks: Mobile operators see a chance of lowering cost per
delivered data bit and increasing indoor coverage, fixed line operators are seeing a new source
of revenue as calls are being connected through their data network.
5.3.1
•
Equipment:
Terminals must to be equipped with both a WiFi or Bluetooth radio chip and a cellular
radio chip (GSM, WCDMA, CDMA2000 etc), this is called a dual-mode handset.
•
WiFi/Bluetooth access point to connect the terminal to the broadband connection
•
Broadband connection to connect to the network controller
•
A network controller to handle the handover between cellular access and broadband
access.
123
http://www.umatechnology.org/
124
Ovum Report – Mobile VoIP – Technical Overview
125
Northstream Operators Options Beyond 3G
84
5.3.2
•
How does it work?
When within range, a user with an UMA-enabled device connects to the local network. For
voice services it will use VoIP, see Figure 28.
•
Through the existing local Internet connection the device connects to the UNC to be
authenticated and authorized
•
If approved, the subscribers’ current location is stored in the core network is updated and
all traffic will be routed through the Unlicensed Mobil Access Network, UMAN instead of
the cellular network
•
When losing connection with the unlicensed wireless network it connects to the licensed
cellular network
•
During a GSM/GPRS session on a cellular network a subscriber can seamlessly be handed
over to the UMA network with no interruption of the service. Much in the same way as a
handover within the cellular network.
Figure 28. Unlicensed Mobile Access126
5.3.3
Deployments
Residential
BT in Britain is currently offering its BT Fusion based on the UMA standard127. BT offers
it’s subscribers 100 anytime, any network minutes for £9.99 per month and 200 minutes for
126
http://www.umatechnology.org/overview/index.htm
85
£14.99 a month. When not connected to its home access point via Bluetooth a subscriber
uses Vodafone’s cellular network in the table referred to as BT Mobile Network. At home
the subscribers call at a rate equal to that of a fixed line. To the right are the prices from
two Swedish operators, which will serve as reference.
Table 3. Price comparison between BT Bluephone and Swedish operators. Prices in
Sterling Pounds.128
Call type
Time
National
landline
In mobile
network
Other mobile
Text message
Picture
message
Voicemail
International
BT Mobile
Network
Connected to BT Fusion
Home Hub
Vodafone
Sweden
Tele2
Sweden
Anytime
Daytime
8.5p
Evenings &
Weekends
5.5 p
10p
Anytime
5p
Anytime
3.5p
10p
10p
10p
5p
3.5p
30p
10p
30p
30p
10p
N/A
30p
10p
N/A
5p
3.5p
5p
10p
from 15p
10p
From 15p4
10p
from 15p
From 5p
There is currently only one phone model available with BT’s offer, a Motorola V560. Users
interested in the service must be subscribers to BT landline phone and BT broadband.
Business
Potential business clients would be retailers, hospitals and other sectors with large “on
premises mobile work force”. In a business case presented by Motorola with a 1500 people
work force on a 90.000 m2 facility, investments in infrastructure and system integration would
be $337.500 resulting in cost savings of $40 monthly per user129. According to the company
the investment would be recovered within two years.
127
http://news.zdnet.co.uk/communications/wireless/0,39020348,39203738,00.ht
m
128
www.bt.com, www.tele2.se, www.vodafone.se
129
Analysis Research, 2005
86
5.3.4
Network load
An operator’s system load over time can typically be as seen in Figure 29. Users are at home the
network load falls considerably leaving the system with excess capacity. This situation may very
well vary; in suburbs and other residential areas the curve might appear to be different. The
benefits from using UMA to offload cellular networks need to be individually analyzed fro
each site.
Figure 29. System load varies during the day. Using alternative access technologies
will thus not always make sense since there could be excess capacity. (Figure not
based on real data)
5.3.5
Growth of broadband access
Research firms expect the number of dual-mode handsets sold in 2009 to be around 100
millions mainly targeted on the business market. Other research firms predict that the majority
of phones will be dual-mode. Nokia will launch a number of WiFi equipped models during
2006.
In combination with the rapid growth of broadband connections, which estimated to reach
370 million by 2010130 up from a current 175 million131, there are some arguments that the
130
Arthur D. Little Broadband Report 2004
131
http://www.dslforum.org/dslnews/pdfs/Q22005briefingsummary.pdf
87
UMA technology could get a foothold on the market. A study by Disruptive Technologies
predicts that there will be 5.5 million homes using UMA handsets over WiFi.132
5.3.6
IMS and UMA
Although UMA is an independent service it can be integrated into IMS. UMA is an important
part in IMS since it provides the physical access to between two types of networks. With UMA
it will be possible to deliver bandwidth-consuming media content within the IMS framework
in a cost efficient manner when such connections are available.
5.3.7
Comments:
There are a number of issues that must be solved before any large-scale deployment will be
seen. At present there are only proprietary solutions that can offer Quality of Service
Guarantees [2.7] in the WiFi-access. This means that an of-the-shelves access-point might
offer unacceptable quality. There is however standards under development that will be able to
prioritize voice traffic over data traffic. For large-scale deployment vertical roaming and billing
agreements have to be established.
UMA is a technology initiative that has yet to prove its sustainability as a business model. The
adoption by the market will be highly dependent on the meeting the expectations of promised
cost savings.133
UMA and its use of VoIP technology has similarities with Push-to-talk in that it is an industry
initiative that has been brought up in order to offer a new attractive services at a lower cost.
The success of the technology depends on how the industry will convince their customers,
who many times already enjoy plans with unlimited voice minutes, the need for yet another
technology.
As seen in the tariff table, prices for BT Fusion are lower than for a mobile call. The user gets
access to their phonebooks and the solution will, if everything works out, provide an
interesting service. There are however some issues with BT’s solution. The major concern
being the limited range of Bluetooth wireless technology with a maximum reach of some 10
132
http://www.wirelessweek.com/article/CA626287.html?spacedesc=Departments
133
Interview Nokia
88
meters, which could result in a very limited indoor coverage. This is very much in contrast to
the original idea of providing improved indoor coverage. As shown in the table prices are
exceeding by far those in other European countries. One also has to consider that the
subscriber have to be a subscriber to both BT landline and BT broadband. UMA is more likely
to attract corporations where many of the calls are intra-corporate/intra-building. Operators
that have deployed WCDMA have more problems providing adequate indoor coverage due to
less penetration capabilities in the higher frequency bands. Operators are always trying to meet
demand with the lowest possible resources. From a UMA perspective this means that not all
areas would potentially benefit from moving users from one access technology to another as
described by Figure 29.
As discussed in previous chapter the cost of providing ubiquitous coverage increases with
bandwidth. With such model it is of interest to investigate if deploying voice service over high
capacity small range access technologies does make economical sense. The chances are greater
that bandwidth consuming application like video or file downloads will have to meet the
economical rationale for using hotspot access. But as presented in the business case by
Motorola, there probably are situations where an investment in WiFi access will be
economical.
5.4 Application based VoIP in 3G networks
With presence function, file sharing, short messages together with voice and video, Instant
Messaging applications create an attractive bundle of applications. In combination with an
attractive user interface these IM-clients has become a very usable tool for communication.
The next step for these applications is to be made available on mobile phones. This move can
soon happen as mobile phones start using more advanced operating systems capable of
running a breed of applications.
5.4.1
Cost of mobile data
Prices on mobile data services have fallen considerably in Sweden. From a $2 price tag per
megabyte it is now possible to subscribe to data bundles where the cost per megabyte is below
3 cents.
89
Table 4. Price of mobile data offered by Swedish operators
Cost for data
Operator
Monthly
Price ($)
Price / MB
allowance (GB)
TeliaSonera
3
109
0.036
Vodafone
1
63
0.063
Tele2
1
25
0.025
The price war is likely to continue as operators seek ways to increase network traffic. Since all
major operators in Sweden offer their subscribers unlimited access to TV-broadcast over
cellular networks, network capacity does not seem to be the bigger issue. The vast majority of
calls are still being made over the GSM network leaving operators with both WCDMA and
GSM with plenty of available capacity.
5.4.2
Price on telephony
In 2004 the average subscriber in Sweden used 69 minutes of airtime. 134 Prices are falling
rapidly and potential clients are offered considerable subsidies on handsets
The operators are bundling more and more minutes and the plans are basically flat-fee. The
ability to receive phone can be viewed as a flat-fee in most counties since the subscriber is not
charged for incoming calls.135
Table 5. Price Comparison between Swedish mobile operators
Cost for voice
Operator
Offer
Price ($)
Price / min
Other
TeliaSonera
Unlimited innetwork and fixed
line calling. 136
51
N/A
All calls are
associated with a
connect fee of
$0.076
Vodafone
5999 anytime any
network minutes137
94
0.016
Tele2
3000 anytime any
network minutes138
63
0.021
134
www.PTS.se
135
Interview Jens Zander
136
www.teliasonera.se
137
www.vodafone.se
138
www.tele2.se
90
5.4.3
Interconnect fees
When a phone call is terminated in another operator’s network the initiating party will have to
pay a interconnect fee. In Sweden this rate is regulated and should mirror the cost of
production of the call plus reasonable return on the investment in infrastructure Currently
Swedish operators pay more for terminating a call than it charges from their subscribers.139 If
the operator only provides the data connection it does not have to pay anything to other
operators. Since interconnect fees are a major source of income it should be further analyzed
how it affect individual operators. A reasonable idea is that small operators pay more than they
receive.
5.4.4
International traffic is initiated from fixed line network
Subscribers interested in long distance calling have traditionally not been using their mobile
phone. They have either called from their landline and/or used a calling card.140 In Sweden,
international traffic from mobile phones only accounts for 18 percent of that traffic initiated in
the fixed network.141 Those who already have Skype are likely to make the great majority of
their international calls from Skype. What they are interested in is thus a broadband
connection. It should thus be possible to make part of these make their calls using a 3G
connection. Initially with a laptop PC and when available using attractive handsets
139
http://www.telekomidag.com/FMPro?-DB=artiklar.fp5&-
lay=cgi_tc4&id=15329&-Format=/ti/nyheter/artikel.html&-find
140
Marcus Nylén, Alpha Telecom
141
www.pts.se
91
5.4.5
Alternative methods for making international calls
People interested in making cheap international calls have for more than 40 years been able to
do so by using calling cards. There are three main groups that make international calls, see
Figure 30. They are very much ethnical. Newly immigrated persons do not have access to a
fixed line phone. Therefore calls are either made from a phone booth or a friend’s home using
a pre-paid calling card. The second group is the established immigrant. They have a fixed line
at home and can use post-paid calling plans. The third group is the ones that do not make that
many calls internationally. These do not in general care so much about the cost of the call
when they every one in a while make one. This group is also the one that has the strongest
financial power.
92
Figure 30. The ethnical triangle of international calling
Calling card companies believe that the main impact from VoIP-applications like Skype is on
the top level of the pyramid. With time it does however seem to also impact the other groups.
Since Skype requires both a PC and a broadband connection the financially weakest group
might use “Skype-centers” for communicating. In such the user gets access to a Skype
equipped PC with broadband connection.
5.4.6
Using softphone over a 3G network
It was previously shown that it is possible to run a VoIP-application like Skype in a WCDMA
network [4.3.1]. Using such solution, with the prices listed above, the per minute price would
be competitive but still be undercut by the operator’s own monthly minutes, see Table 6.
Table 6. Price comparison between VoIP application and operator’s implemented
solution
Operator
TeliaSonera
Vodafone
Tele2
Monthly
allowance
(GB)
3
1
1
Price
($)
109
63
25
Price / Monthly Price/minute
MB
minutes using
Skype142
0.036
6000
0.0183
0.063
2000
0.0315
0.025
2000
0.0125
Handsets
Handsets that will be able to run VoIP application are likely to be found in the upper segment
due to the relative complexness of running an external application. By time this is likely to
142
Assuming 0.5 MB / min
93
change and more and more terminals will be able to do this. Considering that low end mobiles
make up the bulk
5.4.7
Comments:
As shown above, the per-minute price for using Skype vs. traditional telephony is not very
different when looking at domestic calls. To be able to use “free” mobile calling with Skype
both users have to have a monthly data account thus increasing revenue even more.
That is why, in my opinion, Skype does not pose a direct threat to mobile operators. A
competitive war on monthly minutes between operators is probably worse than users starting
to use additional data services.
Current Skype users are typically first movers that quickly start using new technology. They
probably make a large part of their international calls using their broadband connection. If an
operator offered them access to Skype via the mobile terminal it is, in my opinion, unlikely to
have an adverse effect on revenue. By creating bundles of services it should be possible for
operators to maintain regular domestic calls in their network and have additional revenue from
data generated from VoIP applications.
As an additional option operators could launch their own softphone services. These should be
made as open as possible preferably using existing SIP-based solutions. Creating such
application is not technically difficult.
94
6 Conclusion
Voice over IP is an incredibly broad term. It carries a variety of meanings. Essentially, VoIP may be divided
into five areas. Each one of these areas poses different challenges to mobile operators.
People within the industry see the evolution towards VoIP in the core networks
as a “no-brainer”. The core network is also the most mature in which to begin
an IP-transition. Having a solid IP-infrastructure is likely to become more and
more important as services based on IMS make their debuts in mobile
networks. The massive upgrades that fixed network operators are undertaking [3.2] is thought
to have significant impact on future development towards IP based mobile networks since the
usefulness of a network is proportional to the square of the number of users.
VoIP is deployed in the fixed line world prior to the wireless world because of
the capacity and proliferation of fixed broadband access. Compared to other
services that can be delivered over broadband connections, e.g. video on
demand, VoIP is not very bandwidth intense.
On the wireless side, capacity constraints and existing efficient circuit switched solutions make
VoIP currently improbable. Due to the overhead caused by the IP protocol [2.2.1] and IP’s
95
data-centric approach, in which packet loss is handled by retransmissions, the protocol is not
efficient from a radio access point of view. This is shown as not desirable in the bandwidthlimited radio world [0]. Therefore, it is clear that there must be several improvements to both
the radio access network and VoIP implementation before operators will have full duplex
VoIP solutions, as seen in the fixed line world. For WCDMA this is not expected to happen
before 2010. At that time a more packet-friendly radio interface will be implemented which
can leverage the benefits from VoIP [4.3.3].
There are however new access technologies such as EV-DO, WiMAX, WiFi, Flash-OFDM
that do not require circuit-switched technologies. Of these, only EV-DO has any significant
deployment [4.4.1]. For a WCDMA operator it would not be crucial to stay on the alert for
VoIP alternatives before 2010. By then there are likely to be many live networks using EV-DO
and VoIP. The leading full duplex VoIP operators are thus found in the CDMA2000 world,
e.g. Verizon Wireless.
In the fixed line world, traditional telephone companies are seriously, and
rightfully so, concerned about the entrance of VoIP applications like Skype.
However, mobile operators are in a better position to meet competition
from such applications:
•
The service that traditional fixed line operators offers is very basic making it easy for
alternative VoIP operators to compete on prices. A fifty-year-old fixed line phone would
serve its purpose but a fifty-week-old mobile phone would probably not be able to use the
latest services offered by the mobile carrier. This means that the mobile community has
consistently evolved and is more used and accessible to new services. VoIP applications
will therefore meet fiercer competition from mobile operator’s own communication
solutions. It is also important to think of the large amounts of money that mobile
operators spend on advertising new services to the subscribers. It will thus not be easier
for a VoIP application to make people interested in sending photos and files or other type
of service. At the end, what will matter is that the service is attractive and easy to use the
specific platform is not.
•
Using the bandwidth starved radio interface will always be associated with a cost.
Currently, VoIP applications use resources wastefully. Compared to an operator-based
solution it may consume 3 times as many resources. It should thus be possible for
96
operators to transfer this cost of access to the subscribers, regardless of how they use their
connection. This is why Skype will not be seen as “free of charge” as it is in the fixed line
world. With intelligently designed bundles of airtime it should be possible to mitigate the
negative impact from VoIP-applications.
•
With time, more refined handsets are likely to be introduced capable of handling advanced
VoIP applications. They will however still make up only a minor part of the 1 billion
mobile phones that will be sold in 2009. Even though some people will install a VoIP
application most people will not. It is therefore likely that the uptake on VoIP application
will slow down once early adopters have made their move.
•
An enormous advantage of mobile operator is that it can offer their subscribers a very
reliable service. Considering that a VoIP application is dependent on a 3G access
technology it will be difficult to offer complete coverage.
•
A price war between operators is likely to be is far more dangerous for operator revenue.
Mobile operators, as opposed to the companies behind VoIP applications, generally have
the financial capabilities of running deficits in order to attract new subscribers, e.g.
subsidizing terminals. Therefore it will be more difficult for VoIP providers, specifically
VoIP application providers, to run deficits because their competitive advantage is in
providing a low priced service.
International phone calls are mainly initiated from either fixed line networks or by the use of
calling cards. It is thus an untapped market that mobile operators should pursue. A strategy
would be to offer a data-plan that would give unlimited calling with a third party VoIP
application like Skype. As described in [5.4.5] the market for long distance calling is
segmented. Therefore, it should be possible to attract new subscribers and increase the
demand for data services without having an adverse effect on existing voice revenue. The lack
of terminals to support a VoIP application will initially limit the use to laptops equipped with a
PC-card. This is clearly a different way of communicating, as compared to the familiar mobile
phone. This will limit the usage to specific calls and in combination with well-designed voice
bundles, operators are likely to be able to increase their total revenue from subscribers
interested in VoIP.
Existing circuit switched networks are a cost efficient way to provide voice
communication. VoIP can however introduce new services that are both more
97
efficient and exclusive to the use of IP [5.1.4.] Long before full duplex solutions are available
new services will be introduced that use VoIP.
The first service to be widely deployed is push-to-talk over cellular, PoC [5.2]. PoC is a good
example that IP-technology will be used when it provides better and cheaper solutions than
circuit switched technology. Combining different networks with UMA-like technology will also
drive the growth of VoIP [5.3]. UMA can be seen both as a threat and an opportunity to
mobile operators. UMA is thus not a clear-cut case and the success of the technology depends
on several variables. For an operator with poor in-door coverage in a country with high
broadband penetration the solution should be interesting. But, as said, it all depends on the
country-specific scenario.
As described in [3.3] there is a growing demand for Instant Messaging (IM) applications that
can provide alternative forms of communication. Overall, operators should pay attention to
which new services equipment suppliers and application developers are offering. This growing
popularity of IM applications can represent a shift in how people communicate. This is why
new systems like IMS are crucial for operators who want to be able to offer new and attractive
services. IMS will also bring efficiencies because it can blend traditional data applications that
use voice with IP solutions. IMS is also fundamental for introducing coming IP-based services.
Operators should also look at the possibility to complement their existing
infrastructure with WiFi. WiFi is an effective technology to overcome the
problems in creating high bandwidth over small areas as described in [4.1.2].
Instead of building a network with ubiquitous coverage, bandwidth is
concentrated to carefully located hotspots. The cost of building the WiFi-network can be
motivated by a reduced churn from the cellular operator’s traditional business as described in
[4.6.4]. Initially the limited supply of Voice over WiFi handsets will limit the use of these
hotspots to data sessions [4.6.1].
Because WiFi radio chip is incorporated into nearly every laptop today there is no need to
subsidize terminal equipment. This also means that after launching an access point a great
number of users will be able to instantly use it. Thus ideally giving a shorter payback period.
98
WiFi will, if not managed in the right way, pose a clear threat to operator’s desire to provide
wireless broadband connections. It is possible for a greenfield operator to launch a wireless
data service in highly profitable areas at low cost. This is illustrated in Figure 31 with a
horizontal line. For operators with 3G networks, WiFi can be seen as cannibalizing on data
revenue, example of such is Verizon’s decision to halt its WiFi rollout in favor of EV-DO
[4.6.4]. It is thus a challenge for operators to come up with a product strategy that can segment
the data market so that overall revenue is maintained.
Being able to place calls from whatever location is a great feature of mobile telephony. But
equally important is the possibility to receive calls. The limited coverage of WiFi access will not
make Voice over WiFi able to fulfill the latter, positioning the service as a non-direct
competitor to mobile telephony. However, the technology could pose a threat to some areas of
cellular operation. It is as an example possible that larger companies, as described in [5.3.3],
could move to a WiFi solution. This is specifically true for mature markets with a “wireless
aware” office environment.
New access technologies such as WiMAX will not pose a direct threat to mobile operators.
Currently, no standardized solutions are available. This study concludes that the capacity and
cost of a network becomes less and less dependent on the specific access technology used
since all networks are performing in the vicinity of physical limits. WiMAX does however have
some interesting broadband-like capabilities making it more likely to complement fixed
broadband technologies like DSL. Although a new technology can bring performance
improvements there are a number of other considerations that need to be accounted for. As
described in [4.1.8] spectrum is a limited resource. This means that political arguments have
considerable weight when deciding on future systems, e.g. the rollout of UMTS in Europe.
Even without any regulation it will still be hard to find spectrum in the sub 2.5 GHz band
where non line of sight communication is favorable.
Before a standard is widely accepted and interoperability is guaranteed between different
manufacturers it will be difficult to provide attractive handsets to the market. Considering the
time it has taken for introducing attractive and affordable handsets for WCDMA the
difficulties should not be underestimated. Such issues will with time be solved, but for now it
99
places voice delivered over alternative access technologies well below the maturity line in Figure
31
Figure 31. Threat/opportunity vs. maturity to mobile operator’s for different VoIP
alternatives.
100
7 Resources
Business Reports
Broadband Report 2004, Arthur D. Little
Emailed from Arthur D. Little
Development of voice over WiFi by integrating mobile networks, OECD
www.oecd.org/dataoecd/37/48/34741342.pdf
From GPRS to HSDPA and beyond, Rysavy Research
www.3gamericas.org/pdfs/rysavy_data_sept2004.pdf
Is WiFi wagging the 3G dog? Pyramid Research
Mobile VoIP Competitive Landscape, On World
www.onworld.com/voip/mvoipbundle.htm
Mobile VoIP – Technical Overview, Ovum
Overview and comparison of Push-to-talk solutions, Northstream
www.northstream.se/21/page.asp?page_id=4166&type=custom%2Fnorthnews&item_id=8
Signals Ahead, No.1 April 19, 2004
Telecom’s Technology Hotspots, Light Reading
WiMAX, Opportunities and Challenges in a Wireless World, Michael Thelander
www.cdg.org/resources/white_papers/files/WiMAX%20July%202005.pdf
Wireless Voice Over IP, Ovum
Literature
Delivering Voice over IP Networks, D. Minoli, E. Minoli. Wiley
UMTS Networks: Architecture, Mobility and Services, Heikki Kaaranen, Ari Ahtiainen, Lauri
Laitinen, Siamäk Naghian, Valtteri Niemi
WCDMA for UMTS, Harry Holma
Wireless Communication, Theory and Practice. T. Rappaport
Online Articles
3G Standards: The Battle Between WCDMA and cdma2000 – paper presented at Nordic
ICTR Workshop, August 2004, Helsinki
www.cict.dk/publications/workingpaper.view.php?id=100070
101
Comparison of H.323 and SIP for IP Telephony signaling, Ismail Dalgic, Hanling Fang
www.iptel.org/info/references/papers/misc/Dalg9909_Comparison.pdf
Economics of Broadband Wireless Access Systems, J. Zander
www.s3.kth.se/~jensz/Economics_of_Wireless.pdf
Low Cost Broadband Wireless Access – Key Research Problems and Business Scenarios, J.
Zander
www.s3.kth.se/radio/Publication/Pub2004/TimGiles2004_1.pdf
Research and Development of Broadband Wireless Access Technologies, Masahiro Umehira
www.ntt.co.jp/tr/0401/files/ntr200401012.pdf
Voice over Internet Protocol, Bur Goode
http://ieeexplore.ieee.org
Voice-over-IP-over-Wireless, K. Svanbro, J. Wiorek, B. Olin
http://scholar.google.com/url?sa=U&q=http://ieeexplore.ieee.org/iel5/7069/19062/00881384.pdf
News Articles
100,000 Wi-Fi hotspots by the end of 2005, Zdnet UK, September 27, 2005
http://news.zdnet.co.uk/communications/wireless/0,39020348,39222683,00.htm
AOL's Got VOIP Again, Light Reading, September 14, 2005
http://www.lightreading.com/document.asp?doc_id=80587
BT launches "watershed" fixed and mobile handset, Zdnet UK, June 15, 2005
http://news.zdnet.co.uk/communications/wireless/0,39020348,39203738,00.htm
Cable Is the Voice of VOIP, Light Reading, November 15, 2005
http://www.lightreading.com/document.asp?doc_id=84312
China may ban unregulated VoIP services; Shenzhen blacklists Skype, Forbes, September 8,
2005
http://www.forbes.com/technology/feeds/afx/2005/09/08/afx2214918.html
FCC Fines N.C. Provider For VoIP Blocking, Information Week, March 3 2005
http://informationweek.smallbizpipeline.com/60405214
Making the Most from WEP, WiFi Planet, March 6, 2003
http://www.wi-fiplanet.com/tutorials/article.php/2106281
Mobila samtrafikavgifter måste sänkas, Telekomidag.com, November 3, 2005
http://www.telekomidag.com/FMPro?-DB=artiklar.fp5&-lay=cgi_tc4&id=15329&Format=/ti/nyheter/artikel.html&-find
North American MSOs Top 1 Million Mark for VoIP Subs, Cable Digital News, September 1,
2005
http://www.cabledatacomnews.com/sep05/sep05-2.html
Packet Voice Over Broadband, Light Reading, March 3, 2005
102
www.lighreading.com/document.asp?doc:_id=53864
Skype Rules North American VOIP, Light Reading, June 16, 2005
http://www.lightreading.com/document.asp?doc_id=75833&site=lightreading&WT.svl=news1_3
Sony Ericsson Mobile Phones Push to Talk Wireless, 3G, February 20, 2005
http://www.3g.co.uk/PR/Feb2004/6603.htm
Sprint Nextel Preps Wireless BB, Unstrung, August 5, 2005
http://www.unstrung.com/document.asp?doc_id=78580&WT.svl=news2_2
Study: Teenagers favor IM over e-mail, Zdnet, November 10, 2005
http://news.zdnet.com/Study%3A+teenagers+favor+IM+to+e-mail/2100-9588_225944265.html?part=rss&tag=feed&subj=zdnn
UMA's Stepping Stone to IMS, Wireless Week, July 15, 2005
http://www.wirelessweek.com/article/CA626287.html?spacedesc=Departments
Verizon Talks Cellular VOIP, Unstrung, March 31, 2005
http://www.unstrung.com/document.asp?doc_id=71150
Verizon, Cingular get hooked to IM, Zdnet News, August 5, 2004
http://news.zdnet.com/2100-3513_22-5298633.html
Verizon Wireless Cuts NYC Wi-Fi, DevX.com, April 29, 2005
http://www.devxnews.com/article.php/3501611
Vonage to sell Wi-Fi phone,, Zdnet News, December 12, 2005
http://news.zdnet.com/2100-1035_22-5992450.html
WiMax/802.16 Revealed, WiFi Planet, September 21, 2005
www.wi-fiplanet.com/tutorials/article.php/3550476
White Papers
Combinational Services, Ericsson
http://www.ericsson.com/ericsson/corpinfo/publications/review/200 ...
Evolution of WCDMA, Ericsson
http://www.ericsson.com/products/white_papers_pdf/wcdma_evolved.pdf
Gigabyte Performance, Flarion White Paper
http://www.flarion.com/products/whitepapers/Gigabyte%20Performance.pdf
Operator Options Beyond 3G, Northstream
http://www.northstream.se/page/custom/northnews/get_news_file.asp?id=86
Softswitch in mobile networks, Ericsson
www.ericsson.com/products/white_papers_pdf/3025_softswitch_mobile_A.pdf
Understanding WiMAX and 3G for Portable/Mobile Broadband Wireless, Intel
whitepapers.silicon.com/0,39024759,60118993p-39000800q,00.htm
103
Interviews
Anna Kristoffersson, Ericsson, September 15, 2005
Jens Zander, November 24, 2005
Joakim Enerstam, Effnet July 14, 2005
Johan Sköld, Ericsson Research, September 14, 2005
Marcus Nylén, Alpha Telecom, November 9, 2005
Martin Rönnlund Nokia, September 19, 2005
Mats Nordström, Ericsson Research September 27, 2005
Patrik Wikström, Netcom Consultants, July 15
Stefan Hagbard TeliaSonera, September 14, 2005
Åke Andersson, Tove NilssonVodafone, August 24, 2005
Websites
ABI Research
www.abiresearch.com
ArrayComm
www.arraycomm.com/serve.php?page=cellCooper
Boingo
www.boingo.com
Bredbandsbolaget
www.bredbandsbolaget.se/portal/FORETAG_INTERNETACCESS
British Telecom
www.bt.com
www.bt.com/btcommunicator/index.jsp
CDMA Development Group
www.cdg.org/worldwide/index.asp
Cisco
http://www.cisco.com/univercd/cc/td/doc/product/access/acs_mod/1700/1750/1750voip/intro.htm
DSL Forum
http://www.dslforum.org
Gartner
http://www3.gartner.com/5_about/press_release/pr17feb2004a.jsp
Global IP Sound
www.globalipsound.com
Government of District of Columbia.
www.phila.gov/wireless/briefing.html
104
GSM World
www.gsmworld.com/news/media_2004/
www.3g.co.uk/PR/April2004/6982.htm
OnWorld
www.onworld.com
Qualcomm
www.qualcomm.com/technology/1xev-do/solution.html
www.qualcomm.com/technology/1xev-do/revA.html
Swedish National Post and Telecom Agency
www.pts.se
Webinar Light Reading
www.lightreading.com/webinar_archive_home.asp?webinar_id=27493
Nokia
www.nokia.com/nokia/0,8764,46740,00.html
Northern Light
www.centerformarketintelligence.com/analystviews/20050927-WeeklyReport.htm
Skype
www.skype.com/company/news/2005/boingo.html
Tele2
www.tele2.se
TeliaSonera
www.teliasonera.se
TNS Infratest
www.tns-infratest.com
UMA Technology
www.umatechnology.org
UMTS Forum
www.umts-forum.org/servlet/dycon/ztumts/umts/Live/en/umts/Resources_fastfacts
Vodafone
www.vodafone.se
WiMAX Forum
www.wimaxforum.org/technology
105
8 Acronyms
3GPP 3rd Generation Partnership Project, WCDMA
3GPP2 3rd Generation Partnership Project, CDMA2000
BER Bit Error Rate
BSC Base Station Controller
CDMA Code Division Multiple Access
FLASH-OFDM Fast Low-latency Access with Seamless Handoff OFDM
GPRS General Packet Radio System
GSM Global System for Mobile Communication
HSDPA High Speed Downlink Packet Access
HSUPA High Speed Uplink Packet Access
IMS IP Multimedia Subsystem
Node B The base station in a UMTS network
OFDM Orthogonal frequency-division multiplexing
PoC Push-to-Talk over Cellular
PSTN Public Switched Telephone Network
QAM Quadature Amplitude Modulation
RNC Radio Network Controller
UMA Unlicensed Mobile Access
UMTS Universal Mobile Telecommunications System)
UNC UMA Network Controller
VoIP Voice over Internet Protocol
WCDMA Wideband CDMA
WiFi Wireless-Fidelity
WiMAX Worldwide Interoperability for Microwave Access,
106
9 Appendix
9.1 Link budget
A link budget is calculated as the difference between signal power sent by the transmitter and
the received signal at the end point. This amount must at least be equal to the receiver
sensitivity.
A typical link budget could be as found in WCDMA for UMTS
12.2 kbps voice service (120 km/h)
Transmitter (mobile)
Max. mobile transmission power (W)
As above in dBm
Mobile antenna gain (dBi)
Body loss (dB)
Equivalent Isotropic Radiated Power (EIRP) (dBm)
0.125
21.0
0.0
3.0
18
a
b
c
d = a+b-c
Receiver (base station)
Thermal noise density (dBm/Hz)
Base station receiver noise figure (dB)
Receiver density (dBm/Hz)
Receiver noise power (dBm)
Interference margin (dB)
Total effective noise + interference (dBm)
Processing gain (dB)
Required Eb/No (dB)
Receiver sensitivity (dBm)
-174.0
5
-169.0
-103.2
3
-100.2
25.0
5
-120.2
e
f
g=e+f
h=g+10*log(3840000
i
j=h+i
k=10*log(3840/12.2)
l
m=l-k+j
Base station antenna gain (dBi)
Cable loss in the base station (dB)
Fast fading margin (dB)
Max. Path loss (dB)
Log normal fading margin (dB)
Soft Handover gain (dB) multi cell
In-car loss (dB)
Allowed propagation loss for cell range (dB)
18.0
2.0
0.0
154.2
7.3
3
8.0
141.9
n
o
p
q=d-m+n-o-p
r
s
t
u=q-r+s-t
107
Varying parameters
Different radio propagation environments will alter the parameters such as Soft Handover and
In-car loss while different equipment will have effect on parameters such as base station
antenna gain.
When the link budget is established a propagation model for calculating the range is applied to
the link budget. As an example there is the Okumura-Hata propagation model. These models
need some additional information such as sending and receiving antenna height. The model
below is for a suburban system where the base station antenna height is 30 m and the mobile
antenna height reaches 1.5 m with a carrier frequency of 1950 MHz
L = 137.4 + 35.2 LOG(R)
Where R = Range and L is the link budget.
It should be noted that this is a simplified model and to get accurate results one should include
more parameters and consider where in the cell you want to calculate the link budget. An
example is the loss provided by the in-car parameter. Such loss is only relevant when the
received signal is close to the thermal noise floor. Close to a base station the attenuation will
affect both signal and noise equally thus not affecting the signal to noise ratio in significant
way. But nevertheless the model will give a number by the hand to use for further
investigations.143
9.2 OFDM
OFDM is a modulation technique that up until recently has not been available for consumer
products. The inherent complex calculation, e.g. FFT, requires advanced microprocessor, has
up until recently not been possible to manufacture at competitive prices.
The OFDM technique takes the data stream and divides it into smaller stream. A 1Mbps
stream could be divided into 100 streams of 10kbps. Each of these streams are then mapped
143
Interview Netcom Consultants
108
onto a unique frequency and combined together using the Inverse Fast Fourier Transform to
obtain the time domain waveform to be transmitted.
OFDM is especially efficient in environments where multipath delay is common, e.g. urban
and indoor environment. By carefully selecting the tones Intra Symbol Interference can greatly
be reduced since the delay spread represents a much smaller fraction of the lengthened symbol
time144
9.3 CDMA (Code Division Multiple Access)145
In a CDMA system all users transmits simultaneously. This results in that a user receives
interference from all other users in the own and neighboring cells. To be able to distinguish
the right signal each user is given a unique code, hence code-division multiple access.
Chip rate
When making a voice call the codec produces a data stream of roughly 16 kbps. Each bit of
this stream is then multiplied with the unique code that the network has assigned to the
subscriber. When sending a ”1” the code word, typically 64 bits long would be sent, a ”0”
would be the inverse of the 64 bit code word. This process will effectively broaden the signal
since instead of sending one bit, 64 bits are sent. The output transmitted through the air equals
16 kbps x 64 = 1024 kbps. This rate is called the chip rate and is held constant; a lower bit rate
is thus multiplied with a longer code word and vice versa. The process of multiplying is called
spreading.
Spreading and de-spreading
At the receiver the reverse will occur. The 64-bit word representing one bit has to be ”despreaded” by multiplying the received word with the same code word used when ”spreading”
the signal. The advantage of such process is the signal will receive a signal strength
enhancement of 64 making it more tolerate to interference.
144
Flarion White Paper, OFDM for Mobile Communications
145
WCDMA for UMTA, Harry Holma
109
These codes should ideally be orthogonal. The result of multiplication of two orthogonal
codes is always zero and this is the base of the interference robustness of CDMA systems. A
user with a different spreading code will not affect the de-spreading of another user. The
problem is that the number of spreading codes is limited; there are as many orthogonal codes
as there are spreading codes. This would not be a problem when studying an isolated cell but
when receiving interference from neighboring cells the number of orthogonal codes would
quickly run out. Using different frequency in neighboring cells could solve this problem but
this would limit the available bandwidth.
110
To overcome this problem, CDMA uses semi-orthogonal codes; these are called Pseudo-noise
or PN codes. Multiplying two near orthogonal codes will not give 0 but close enough. The
mayor advantage of PN codes is that there are more or less an unlimited amount of them.
Processing gain:
One of the advantages of using a high chip rate is that there will be a processing gain.
chip _ rate 10 log
user _ bit _ rate As an example, a 12,2 kbps speech codec is commonly used in WCDMA systems. Here the
processing gain will be
3.84 Mcps 10 log
= 25dB
12.2kbps This means that the required signal to noise ratio in the radio interface Eb/N0 can effectively
be much lower. It is obvious that this processing gain becomes less significant with higher
bandwidth and at 2 Mbps the processing gain will only be around 2.8dB.146
This processing gain alone does not come for free, however, since it requires more bandwidth.
Instead there are other parameters that give CDMA advantages.
The wideband nature in conjunction with the processing gain gives the possibility to use a
frequency reuse factor of 1, which enables high spectral efficiency.
With a wideband signal, different propagation paths can be resolved with higher accuracy in
comparison with signal that uses lower bandwidth. The result is a higher robustness towards
interference.
Cell breathing
When more users are added to a cell the noise increases. Since the signal to noise ratio has to
reach a certain ratio each user has to transmit at a higher output power. Terminals are very
146
Harry Holma, WCDMA for UMTS
111
limited in their output capabilities the only solution is to be closer to the base station. This
means that the cell size shrinks an effect that is called cell breathing.
112
9.4 List of Codecs
Codec
Algorithm
Usual rate
Comments
64 kbps
58,56 or 64
kbps
32 kbps
Universal use
Wideband coder
ADPCM
Frame Size
/Lookahead
0.125 ms/0
0.125 ms/1.5
ms
0.125 ms/0
G.711
G.722
PCM
G.726
G.728
LD-CELP
0.625 ms/0
16 kbps
G.729(A)
G.729e
CS-ACELP
Hybrid CELP
10 ms/5 ms
10 ms/5 ms
8 kbps
11.8 kbps
G.723.1(6.3)
MPC-MLQ
30 ms/7.5 ms
6.3 kbps
G.723.1(5.3)
ACELP
30 ms/7.5 ms
5.3 kbps
IS-127
RCELP
20 ms/5ms
AMR
ACELP
20 ms
Var. 4.2 kbps
average
Var. 4.75-12.2
kbps
High quality, low
complexity
High quality in
tandem;
Recommended for
cable
Widespread use
High
quality/complexity;
Recommended for
cable
Video
conferencing
origin
Video
conferencing
origin
WCDMA,
CDMA2000
113
9.5 Frequency bands
Frequency Europe
USA
700 MHz
Used for other
purposes
608-746 MHz
(lower)
746-794 MHz
(upper)
850 MHz
Used for other
purposes
824-894 MHz
900 MHz
890-960 MHz
Used for other
(GSM)
purposes
880-960 MHz
(Extended
GSM)
1710-1880 MHz 1710-1755
MHz paired
with 2110-2155
MHz
1800
MHz
Americas excl.
US
Used for other
purposes
Asia
Comment
Used for other
purposes
824-894 MHz
824-894 MHz
890-960 MHz
used in some
Latin American
countries e.g. in
Brazil and Chile.
1710-1770 MHz
paired with 21102170 MHz
1710-1880 MHz
e.g. Brazil
1850-1990 MHz
890-960 MHz
(GSM)
880-960 MHZ
(GSM)
The
broadcasting
industry is in
transition from
analogue to
digital systems
“Cellular
band” used for
TDMA,
CDMA, GSM
and WCDMA
“GSM 900
band”
1900
MHz
Other spectrum
arrangements
1850-1990
MHz
2 GHz
1920-1980 MHz
paired with
2110-2170
MHz;
1900-1920
MHz, 20102025 MHz
unpaired
Used for
aeronautical and
military services
Other
spectrum
arrangements
Other spectrum
arrangements
1920-1980 MHz
paired with 21102170 in Brazil
2300-2400
MHz
DRS and WCS
2300-2400 MHZ
Used for other
purposes
2400-2483.5
MHz
2500-2690 MHz
IMT-2000 /
UMTS
extension
2400-2483.5
MHz
2500-2690
MHz AWS
including IMT2000
2400-2493.5
MHz
2500-2690 MHz
IMT-2000
Used mainly for
MMDS
2.3 GHz
2.45 GHz
2.5 GHz
1710-1880
MHz in most
Asian
countries.
“GSM 1800
band”
Other
spectrum
arrangements
“PCS band”
used for
TDMA,
CDMA, GSM
and WCDMA
1920-1980
“Core band”
MHz paired
/IMT-2000
with 2110-2170 Partly possible
MHz
“WiMAX
2010-2025
band” (TDDMHz unpaired version)
2300-2400
MHz
WiBro in
Korea, China
TD-SCDMA
2400-2483.5
MHz
2500-2690
MHz
IMT-2000
extension
WiBro, similar
to WiMAX
Possible future
WiMAX band
Bluetooth and
WLAN etc.
“IMT-2000
expansion
band” 25002690
114
3.5 GHz
3410-3600 MHz
Partly allocated
for FWA.
Nomadic use
5 GHz
5150-5350 MHz 5150-5350
5470-5725 MHz MHz
5470-5725
MHz
5.8 GHz
Used for
military
purposes
Source: Northstream
2400-2650
MHz Partly
military, partly
FWA
5725-5825
MHz
3400-3650 MHz
allocated for
FWA
Unclear, no
homogenous
usage in region
5150-5350 MHz
5470-5725 MHz
5150-5350
MHz
5470-5725
MHz
5725-5825 MHz
Used for other
purposes
WiMAX band
(USA)
FWA and
BWA band
“WiMAX
band” (FDD
version)
WLAN etc.
coexisting with
radar.
Possible
WiMAX
WLAN etc.
Possible
WiMAX
115

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project