Академический Документы
Профессиональный Документы
Культура Документы
Report
Jean-Noël Gouyet, EBU International Training
Notice
This report is intended to serve as a reminder of the presentations for those who came to the seminar, or as an
introduction for those unable to be there. So, please feel free to forward this report to your colleagues!
It is not a transcription of the lectures, but a summary of the main elements of the sessions. The tutorial-like
presentations and tests results are more detailed. The speakers‟ presentations are available on the following EBU
FTP site via browser: ftp://uptraining:ft4train@ftp.ebu.ch
The slides number [in brackets] refer to the slides of the corresponding presentation.
1
To help "decode" the abbreviations and acronyms used in the presentations' slides or in this report, a list is
provided at the end of this report.
Many thanks to all the speakers and session chairmen who revised the report draft. Special thanks to Peter Calvert-
Smith (Siemens), Andrea Metz (IRT) and Martin Turner (BBC), who kindly allowed us to use their personal
presentation notes (§ 1.2 - § 2.2 - § 5.2). Nathalie Cordonnier, project manager, and Corinne Sancosme (EBU
International Training) made the final revision and editing.
The Networks 2005 / 2006 / 2007 seminar reports are still available on the EBU Web site:
http://www.ebu.ch/CMSimages/en/NMC2005report_FINAL_tcm6-40551.pdf
http://www.ebu.ch/CMSimages/en/EBU_2006_Networks_Report_tcm6-45920.pdf
http://www.ebu.ch/CMSimages/en/EBU-Networks2007-Report_tcm6-53489.pdf
1
More than 325!
The opening 'round table' answered the question: 'What are the strategic opportunities and issues posed for
broadcasters by developments in network technology?' It also highlighted some of the work going on in the EBU to
address these challenges.
Keeping it work!
Security & Business continuity
o We still don't have the same reliability with IP as we had in the old system, but it's constantly improving. This
is an ongoing process - we are testing new equipment and end units, doing research and measurements on
various types of IP networks.
o Some concerns about 'having too many eggs in a single technology basket'. A lot of work to do to make sure
we don't have a sudden collapse of the entire network. Network engineering can ensure business continuity and
it has to be looked throughinto.
2
cf. § 2.3
3
cf. EBU Networks2006 seminar report § 2.1and § 2.2 and EBU Networks2005 seminar report § 2.2 and § 2.3
"Technology is our word for something that doesn't quite work yet" (Danny Hillis, Douglas Adams). New network
technology is something that we have, which offers us a lot of opportunites but that don't quite work yet with many
challenges to ensure the expected performance, in a number of environments for a number of different reasons. One
of the role of the EBU Network Management Committee is to help to make it work, and this seminar should
contribute.
1 IP and MPLS
IP and MPLS has been the network technology hailed as the method of delivering high QoS media services across
both metro and wide area networks for some years now. This first session explores both the theory and the
practicalities of actually trying to achieve the business needs of the broadcast and production markets when using
this network protocol along with the lessons learned up to now.
MPLS is short for Multiprotocol Label Switching, an IETF initiative that integrates Layer 2 information about network
links (bandwidth, latency, utilization) into Layer 3 (IP) in order to improve IP-packet exchange. MPLS gives network
operators a great deal of flexibility to divert and route traffic around link failures, congestion, and bottlenecks. It has
been developed since 1999. It is complex to operate and configure for the service provider, but is simple as a "data
socket" on the wall for the user.
First, what are the video Service Level Agreement (SLA) requirements?
Throughput: requirement addressed through capacity planning and QoS (i.e. Diffserv).
Jitter: important for studio quality content. The delay variation is absorbed by the receiver buffer, which has to be
minimised in order to improve responsivity, especially in contribution environment. It is controlled with Diffserv, a
mature technology known to offer ~ <1msec network induced jitter end-to-end.
Latency: important mainly for live or conversational 2-way content.
Service Availability: the proportion of time for which the specified throughput is available within the bounds of the
defined delay and loss - a compound of the other networks and core network availability.
Packet Loss: extremely important. One lost packet may contain up to 7 MPEG Packetized Elementary Stream
packets. For example [9], with a 12-frame GOP (and a I:P:B data size ratio of 7:3:1), the I-frame loss probability
varies from 30 % to 100 % for an outage duration from 50 ms to around 400 ms. Controlling loss is the main
challenge. The 4 primary causes for packet losses are:
o Excess delay: renders media packets essentially lost beyond an acceptable bound. Can be prevented with
appropriate QoS (i.e., Diffserv).
o Congestion: considered a catastrophic case, i.e. a fundamental failure of service. Must be prevented with
appropriate QoS and admission control.
o Physical-Layer errors: apply to core (less occurence) and access. Assumed insignificant compared to losses
due to network failures.
o Network reconvergence: occur at different scales based on topology, components and traffic. Can be
eliminated with high availability (HA) techniques.
The core impairment contributors are: Trunk failures ( .0010 / 2 h) + Hardware failures ( .0003 / 2 h) + Software
failures ( .0012 / 2 h) + Software upgrades / Maintenance ( .0037 / 2 h) = a total of .0062 Impairements / 2 hours i.e.
1 impairment every two weeks. Note that the average mean time between errors on a DSL line is in the order of
4
http://www.cisco.com/
The impact of outage can be reduced but requires smart engineering. Aiming at lossless video transport, the
following deployment scenarios can be implemented, classified in Figure 1 in terms of lossless or not, of cost and
complexity of network design and deployment and application infrastructure. The corresponding techniques are
detailed in the Table 1 (Annex).
Live / Live
Network
Live / Live
No network
Fast Convergence
The aim is to deliver a glitch-free, resilient and reliable media stream with minimal latency running over networks
shared with unpredictable business data traffic.
Of the many requirements for an IP media circuit, these have been found critical to the successful implementations:
Fixed IP latency
Low IP jitter
Practically zero IP packet loss and reordering.
Minimum overall latency.
The fundamental means of meeting this requirement in IP is Quality of Service (QoS). Traffic is segregated by
setting QoS levels, either at the router port or in the coder. Our practice is to set the QoS at the router port to protect
the integrity of the network. The broadcast media traffic and VolP traffic are both given a QoS setting which is
labelled Expedited Forwarding (EF).
Table 2 (Annex) presents the experience of broadcasting engineers regarding the practical measurement [5]-[12] of
key parameters and the practical network performance [14]-[20].
55
www.siemens.co.uk/it-solutions http://www.siemens.com/index.jsp?sdc_p=c0u2m0p0
The challenge brought by video broadcasting is to maintain the QoS with the stress of a relatively small number of
flows but with large bandwidth requirements (from 2 Mbit/s up to 1.5 Gbit/s). If all IP networks are not created
equally, most IP Networks are designed for many simultaneous connections, with low bandwidth requirements per
stream,. i.e. 100 000 and more IP flows, each under 1Mbit/s.
In order to classify and prioritize traffic, different items are available at (Figure 2 -top):
Layer 2 (Data Link) : VLAN tags, MAC adresses
Layer 3 (Network): Source IP, Destination IP, Source Port, Destination Port, Type of Service (TOS) bits. TOS
flags (4 bits) in the IP Header allow prioritization and classification of traffic on a per service basis, give the most
flexibility, and are supported by most video codec manufacturers in each IP flow. The classes of network service
available are: Normal/Low delay, Normal/High throughput, Normal/High reliability, Normal/Minimize monetary
cost.
But what is 'inside the cloud', through the rest of the network (Figure 2 - bottom)?
The IP flow is connection-oriented: series of packets are transmitted based on source and destination addresses.
After a router looks up the first packet, the route is pushed to the line cards, in the hardware tables along with the
appropriate address. This is be enhanced by using basic policy based routing and/or MPLS (Multi-Protocol Label
Switching. MPLS Label Switched Paths (LSP‟s) can be used for traffic engineering, allowing complete control of
bandwidth and routing for each classification of service.
(Figure 2 - bottom/1) The Switch CPU looks up the first packet, determines the classification and priority, then all
subsequent packets matching the same criteria will be forwarded in hardware.
(Figure 2 – bottom/2) Using MPLS, the traffic can be pushed on different paths based on classification.
6
A formal method of tracking and recording proposed and actual changes to a system
7
http://www.gen-networks.com/
8
http://www.gen-networks.com/content/iris/index.aspx
Bandwidth management - An MPLS enabled network allows the creation of Label Switch Path‟s (virtual circuit) that
can tunnel traffic through the network. The Traffic Engineering implements deterministic routing and enables proper
bandwidth management. The Network Manager needs to ensure no oversubscription of high-priority flows. An
external system, such as IRIS, is required to properly manage broadcasts. This system must: know and understand
the network topology, accept and manage all high priority data on the network, understand all future bandwidth that
will be placed on the network.
In 2006 it started to design a Digital Multi-Service Distribution Network, the 'TMS' (Transport Multi-Service),
completely implemented in July 2008. It is a new MPLS Ethernet (Layer 2) network, composed of 150 routers spread
in 42 POPs all over France. No layer 3 routing is currently used in this network. The inter-POP links are based on 5
Gigabit (1 to 4) optical fiber loops [5] and the access mainly on Ethernet microwave links (155 Mbit/s), some fibers
too. The QoS management enables different levels of services on the same link, without any risk of interaction
between services and customers. 350 services are running on this network, including:
Digital distribution for regional Digital Terrestrial Television (DTT) programs [8] from the 24 regional studios of the
French regional network France 3 over 92 regional transmitters, using MPLS, RSVP and FRR.
Collecting 24 France 3 regional programs back to the national FR3 central studio in Paris [9], using the same
protocols.
Digital distribution of France 3 national programs to regional TV regional production centers for play-out [10],
using MPLS, VPLS, RSTP and PIM (at he head ends).
9
http://www.tdf.fr/groupe-tdf/filiales/
The next step is to extend this multi-service network to the European level [19].
Audio over IP equipment from one manufacturer has until now not been compatible with another manufacturer‟s unit.
This session covers applications and real-life use of audio over IP with demonstrations and hands-on.
The audio coding algorithms, which 'must', 'should' or 'may' be used, are:
The transport protocol used is RTP on top of UDP [8], as in many real-time IPTV systems.
For the session management, 3 protocols are used: SDP: Session Description Protocol (RFC4566), SIP: Session
10
See also EBU - TECH 3329 'A Tutorial on Audio Contribution over IP'. May 2008
http://www.ebu.ch/CMSimages/en/tec_doc_t3329-2008_tcm6-59851.pdf
Lars JONSSON and Mathias COINCHON 'Streaming audio contributions over IP' EBU Technical Review - 2008 Q1
http://www.ebu.ch/en/technical/trev/trev_2008-Q1_jonsson%20(aoip).pdf
11
http://www.ebu-acip.org
12
EBU - TECH 3326 (revision 3) 'Recommendation for interoperability between Audio over IP equipment'. April 2008
http://www.ebu.ch/CMSimages/en/tec_doc_t3326-2008_tcm6-54427.pdf
13
http://www.aptx.com
Audio Contribution over IP should rely on managed (private) IP networks, allowing QoS. QoS can be expressed in
Service Level Agreement (SLA) contracts with providers, but requirements must be clearly expressed concerning:
transmission performances (latency, jitter...), network availability (99.9% 99.99%), provisioning delay (1 week?
1 month?). The definition of the measurement method is also important (EBU N/IPM group is working on profiling,
measurement).
A variety of 'last mile' access methods are available with different performances:
Fiber optic Copper with xDSL Mobile (3G/UMTS, Satellite Wireless
Wimax, LTE)
High quality but SDSL: Symetrical Increasing bit rates but Long delays, often Wifi: no guaranty due
expensive uplink / downlink unreliable shared bandwidth to frequency sharing
Bit errors lead to No solutions with Inmarsat BGAN, DVB-
packet losses guaranteed QoS RCS providers
nowadays (Eurovision in the
HSDPA / HSUPA future?)
shared channel
'Audio Contribution over IP' Recommendations has already been implemented by many audio codecs manufacturers
[16]. A recent test (February 2008) between 9 manufacturers proved that earlier incompatible units can connect with
professional audio formats using the standard. Some are still premature prototypes, and not yet EBU compliant.
Marketing is very aggressive. Units are still under development. An open source reference implementation by IRT /
BBC R&D is in development (§ 2.2).
14
e.g. - sip:reportername@sr.se or telephone number
15
http://sourceforge.net/index.php
16
http://www.pjsip.org/
17
AEQ (Spain) http://www.aeq.es/eng/index.htm , AETA (France) http://www.aeta-audio.com , AVT (Germany) http://www.avt-
nbg.de/homepage_engl/start.htm , Digigram (France) http://www.digigram.com , Mayah (Germany) http://www.mayah.com/ ,
ORBAN (Germany) ) http://orban-europe.com, Prodys (Spain) http://www.prodys.net , Telos (U.S.A.) .) http://telos-systems.com,
Tieline (Australia) http://www.tieline.com/ip/index.html
Disadvantages:
Codecs and Config PC‟s have to move around testing station positions – minimal movement regime determined
Potentially, other test stations could see communication traffic to/from SIP server from codecs not involved in test
– in practice, not a problem
The testing consisted of 11 rounds of 45 permutations of devices under test for each of 4 mandatory codecs. More
18
than 180 tests were undertaken including reruns of some tests. G.711 was tested using an Asterisk SIP server;
G.722, MPEG Layer II and Linear PCM were tested without the SIP proxy. All rounds were recorded (2 GB of data
19
stored) using the open source Wireshark protocol analyser [15] [16].
The lessons learnt with the SIP Gateways. Although not strictly part of the interoperability document, are a
necessary part of infrastructure requirements for broadcasters.
Asterisk was used, although others are also available, but with a number of unexpected behaviours (for example:
INVITE queries not relayed correctly; media stream sometimes altered, promotion of G.711 to G.722; unexpected
disconnections; odd behaviour of BYE request; response by codecs handling Gateway codec presence checking
– is codec still there?).
The SIP infrastructures need to be able to handle Audio over IP, cell phones and Voice over IP units… and Video
over IP as well? Some ongoing work for broadcasters required in this area.
20
At the end January 2008 the BBC Festival of Technology took place at BBC Television Centre, London, with the
first public demonstration of the reference software with 'real' devices [23].
18
http://www.asterisk.org/
19
http://www.wireshark.org/
20
http://www.bbc.co.uk/rd/fot/handouts.shtml
Devices in action
Hardware-housed software versus all-software codecs:
Hardware codecs, are very easy to interface within the Broadcasting house, offer a lot of functionalities, but are
harder to handle for non-technicians and require upgrade attention.
Software codecs, have a friendly Graphic User Interface, are easy to use for reporters and non-technicians, offer
less functionalities, but can easily be integrated in the portable laptop.
2.4 Demos
Figure 3: Audio over IP network over satellite with DVB-RCS (future Euroradio!)
APT
APT
„Scoop4+‟
SIP AETA, (France)
„Scoopy‟ Proxy and (Unit in UK)
AETA, (France) Registrar
(Unit in Server
Geneva) (Unit in UK)
„Scoop4+‟ „PortaNet‟
Prodys (Spain)
AETA, (France)
(Unit in (Unit in
Germany) Geneva)
This session gives an overview of the work and first results of the EBU 'Video Contribution over IP Networks'
(N/VCIP) project group and two practical use cases for video contribution over IP networks.
Different manufacturers already provide solutions for real-time video transport over IP. Interoperability of equipment
is of paramount importance. A first meeting took place with manufacturers in Geneva in March 2008. A presentation
of the ongoing N/VCIP work and discussion with manufacturers is planned at IBC (September 2008). An agreement
on profiles (see below) should follow with interoperability tests. The publication of a Technical specification is
planned for mid/end 2009.
A draft of VCIP profiles (minimum set of required parameters for interoperability) is currently under specification,
including the following ones:
21
EBU – TECH 3318 'File Transfer Guidelines'. June 2008
http://www.ebu.ch/CMSimages/en/tec_doc_t3318-2008_tcm6-60595.pdf
22
SMPTE 2022-1-2007 Forward Error Correction for Real-Time Video/Audio Transport Over IP Networks
SMPTE 2022-2-2007 Unidirectional Transport of Constant Bit Rate MPEG-2 Transport Streams on IP Networks
23
http://www.t-vips.com/
The JPEG2000 bit rate characteristics are, after compression: 20-25 Mbit/s from SDI 270 Mbit/s, 60 – 120 Mbit/s
from HD-SDI 1.485 Gbit/s, 200 – 250 Mbit/s from HD 3G , Digital Cinema uses 250 Mbit/s. The lossless mode allows
a bit rate reduction of 30 to 60 %. JPEG2000 is by nature VBR - the number of bits produced by the compression
may be less than the bit rate configured. This may be used to save bit rate on transmission.
JPEG2000 combined with MXF / FEC / RTP / UDP/ IP provides a good solution for transport over IP [18]
MXF does provide a frame-based (progressive or interlaced scanned) wrapping of JPEG2000 picture, plus sound,
VANC and HANC data, and can handle synchronisation of video & audio [19]-[21]. ]. In the T-VIPS implementation
Preamble and Post-amble are skipped in order to reduce overhead (Typical overhead: 6-8 % incl. Ethernet and IP
headers).
26
The Forward Error Correction (FEC) [22] corresponds to the Pro-MPEG Forum Code of Practice #4 or to the
27
SMPTE 2022 matrix. For Variable Bit Rate (VBR), operation the FEC matrix being of fixed size, stuffing packets
are inserted [23].
The IP encapsulation supports a bit-rate range from 1 to 1000 Mbit/s,VBR, FEC, and preserves low latency.
The challenges to meet for this project were to: connect 110 sites over the French territory, make possible any-to-
any communications, keep and adapt our Network Management System (NMS), use a network operated by a data
oriented Telco (NeufCegetel), design with the provider a 'broadcast real-time' service on the network, select and
deploy more than 130 encoders and decoders.
The network [7] comprises two types of network for two types of contribution:
IP/VPN/MPLS (with QoS) for corporate data and file exchange (and managing of our NMS).
IP/VPN for real-time contribution (with MPEG2 compression): NGN28 range network, IP-based network (Layer 3)
and native multicast (no tunneling). Used for live and rushes transmission (from/to spots where files exchange
24
SMPTE 422M-2006 Material Exchange Format – Mapping JPEG2000 Codestreams into the MXF Generic Container
25
http://www.thomsongrassvalley.com/products/infinity/camcorder/
26
Peter Elmer and Henry Sariowan – 'Interoperability for Professional Video Streaming over IP Networks'
www.broadcastpapers.com/whitepapers/paper_loader.cfm?pid=221
27
SMPTE 2022-1-2007 Forward Error Correction for Real-Time Video/Audio Transport Over IP Networks
28
http://www.itu.int/ITU-T/ngn/
Encoding equipment
29
The MPEG-2 encoders / decoders 'ViBE' are products from Thomson Grass Valley , with Ethernet Interfaces
Typically, the MPEG-2 MP@ML encoding profile requires about 8 Mbit/s (IP level). The normal delay (about 900 ms,
end-to-end) is used for rushes transmission. The low delay (about 450 ms) is used for live. There is no FEC
implementation at the moment.
Example of the network management at the level of a Regional Center: the network bandwith is of 24 Mbit/s and is
managed with 8 Mbit/s one-way, making it possible to have 3 input/output links. The NMS will authorize 3 live
receptions, but will first refuse the fourth. All links being managed by a 'Scheduler', the fourth link will be proposed
after, or the operator must stop one of the 3 links.
4 HD over Networks
This session is about current status and new key technologies for High Definition Television Production in a
networked environment. The high bandwidth demand of uncompressed HD signals often necessitates some kind of
signal compression in order to enable storage on hard disks and transport over networks. This session brings some
more insight into what is lying ahead of us as well as some experience that can be gained from current real world
HD-TV production and emission.
'Contribution' is the exchange of video content from one broadcaster to another or between the regional facilities of
one broadcaster. It covers 3 main applications:
News Gathering (SNGs) implying priority to the event with best effort quality).
Off-line content exchange with high quality expected.
Live event contribution High quality expected and low delay).
It relies on 3 types of networks:
Via satellite, with limited bandwidth (depending on the transponder and the modulation technique).
Via copper links, usually for internal facility contribution, with the bit.-rate limited by cable technology.
29
http://www.thomsongrassvalley.com/products_disttrans/
30
http://tools.ietf.org/html/rfc4445
HD content suggests a higher viewing experience for end users compared to SD, and therefore HD contribution
should then meet following challenges:
Higher data rate to handle efficiently in bandwidth limited networks
Lower latencies or at least the same as in the SD world
Higher bit depth (at the moment up to 10-bit)
Multiple image formats (SD, 720p/50, 1080i/25 and potentially 1080p)
Frame rate conversion for overseas transmissions
Robustness to cascades
Etc.
The Table 3 (Annex) details the characteristics, advantages and disadvantages of the choice of codecs.
SVT produced it‟s own demanding, but not unduly so, multi-genre TV-program “Fairytale” in 3840x2160p50 (hence
31
1080p, 1080i, 720p and 576i).. 3 reference sequences "CrowdRun", "ParkJoy" and "InToTree" can be downloaded
32 33
from the EBU Web site or from the ITU-VQEG FTP site . All sequences from this set were filmed in 50 fps with
34
professional, high-end 65mm film equipment by SVT in October 2004. ITU-R BT.1122 indirectly says that, for
distribution, 75 % of such a program should stay in “excellent-range”, but 25 % (demanding scenes) may go down to
“middle of good-range”. Considering that Blu-ray has no artefacts at all, SVT believes that any scene in Fairytale
should stay for contribution in “excellent-range” to reduce cascading artefacts.
35
In our SVT's national WAN [5], we would like to use... an I-frame based codec to reduce latency in interviews,
and a codec using 4:2:2, 10-bit for further processing robustness. For the time being SVT is using the JPEG2000-
based T-VIPS products (§ 3.2) that need about 150 Mbit/s (for 720p50, 4:2:2, 10-bit), with only 100 ms in latency
(versus 2 seconds in MPEG-2 GOP at 200 Mbit/s!). SVT believes that MPEG-4 AVC / H.264 products for
31
ftp://vqeg.its.bldrdoc.gov/HDTV/SVT_MultiFormat/SVT_MultiFormat_v10.pdf
32
http://www.ebu.ch/en/technical/hdtv/test_sequences.php
ftp://vqeg.its.bldrdoc.gov/HDTV/SVT_MultiFormat/
33
ftp://vqeg.its.bldrdoc.gov/HDTV/SVT_MultiFormat/
34
ITU-R BT.1122-1 (10/95) User requirements for emission and secondary distribution systems for SDTV, HDTV and
hierarchical coding schemes http://www.itu.int/rec/R-REC-BT.1122/en
35
See also the EBU Networks2005 seminar report § 1.1.2
So, please... offer us MPEG-4 AVC/H.264 codecs Level 4.2 (SVC is not needed for contribution) codecs, with 4:2:2,
10-bit, both I-frame based high bit rate and GOP.
Eurovision has already a quite substantial experience with HD contribution [10], using MPEG2 4:2:2P bit rates from
37
34 to 60 Mbit/s:
For the coverage of big events, occasional use of satellite and fiber: 2006 Winter Olympic Games from Torino
using 20 HD encoders and 40 receivers, 2006 World cup from Germany, Euro 2008 from Switzerland and Austria,
2008 Summer Olympic Games from Beijing using 90 HD encoders and 150 receivers
For permanent connections with NHK for European customers and US networks for sports events: fiber circuits,
with a cheaper bandwidth, but still expensive on long distances, and it's easier to improve the transmission
parameters.
The bandwidth is expensive and sometimes rare. For example, for the Summer Olympic Games in Beijing fifteen 36
MHZ transponders and 7 dedicated STM4 fiber circuits have been booked!
Some issues are daily to be solved: frame rate conversion (50Hz-60Hz), Progressive or Interlaced scanning
(720p/1080i), compression cascading in contribution-distribution.
Some trends remain clear: the market is not ready to pay more for contribution links - there is an increased demand
for exclusive circuits due to the bandwidth limitations on satellite – and the cost for long distance circuits is still high.
This leads to MPEG-4 AVC/H.264 to replace MPEG-2 for live HD in the middle term (with a 50% benefit expected).
JPEG2000 will be limited to fiber or file transfer.
Tests have to be conducted (EBU, ISOG…) on: compression cascading, compression/standard conversion
36
For example, from Teranex
37
Very low for HD contribution, only on request from the client
Finally the market will decide. Since MPEG-4 AVC/H.264 equipment range for contribution is not fully available, so
MPEG-2 will survive for some time…And anyway how much broadcasters are ready to pay? File transfer is an
alternative with lower costs (and increased picture quality?).
The final two sessions will outline some recent developments in using new network technologies to support the
broadcast environment. The first session looks at the impact of networks on news-gathering kicking off with a use
case looking at the application of COFDM for ENG application, with particular attention to the monitoring of the
quality of the signals. This will be followed by some real examples of some of the innovative techniques being used
by journalists to deliver material to supporting their story-telling.
VRT has implemented 'TNG' - Terrestrial News Gathering - which is ENG (Electronic News Gathering) via COFDM
transmitters and receivers (DVB-T) for real-time audio/video. This application uses camera link equipment from Link
38
Research Ltd
Coded Orthogonal Frequency Division Multiplexing (COFDM) is a proven technique, used for DAB, DVB-T,
WiMAX… (OFDM for WiFi/WLAN 802.11a/g…). It has the following characteristics:
Double-stage FEC (Reed-Solomon and convolutional code): detects and corrects faults, but increases the end-to-
end delay by adding extra bits and data rate processing.
Orthogonal multiple carriers (2048) [6], to avoid dead spots. Choosing the correct inter-carrier distance, leading to
limiting/suppressing the interference between them.
A guard interval on the time axis, waiting to start a new symbol to avoid interference of echoes due to multi-path.
A multiple antenna diversity reception is used, with packet diversity. Each receiving antenna signal is
demodulated to ASI. The new HD/SD equipment (not used by VRT yet) uses the better performing maximal-ratio
combining, where all signals are multiplied by a weighting factor and afterwards added - the weighting factor is
function of a quality measure (amplitude, SNR…) [8].
The equipment include on the transmitter side: a MPEG-2 codec (100 or 200 mW) and modulator, a 1 W amplifier
on the wireless camera. In the TNG van, a 5 W amplifier. The transmit antennas [11]+[14] are omnidirectionnal on
the wireless camera and on the van, and directional which can be mounted at the van rear outside.
Network. At VRT, 2 frequencies in the 2.2 GHz band are used: 1 dedicated to VRT and 1 shared between VRT and
RTBF Sports. The A/V bit-rate is 6 Mbit/s (QPSK modulation) or 12 Mbit/s (16-QAM modulation), with a 1/2 FEC
(highest possible) and a 1/32 guard interval (smallest possible). There are 5 receiving points [15]: 2 in Brussels, 1 in
39
Antwerp, Ghent and on the Belgian Coast . The transport of the signal to VRT Headquarters [16] occurs over VRT
optical fiber (+ CWDM) in Brussels (ASI) and Antwerp (SDI), via a microwave radio link in Ghent and then over an
IP-MPLS service to Brussels (16].
Performances [17]. From the TNG van, transmission is possible in cities within a 6-8 km radius area (non line-of-
sight) and up to 60 km (in line-of-sight), the van becoming in this last case a "virtual feeding point". From the wireless
camera to a receiving site: 2-3 km (non LOS) or tens of km (LOS) - or hundred meters to a TNG (or SNG) van, used
as a repeater. The quality of the received signal is monitored using an application [18] [19]. The system is easy to
operate and offers a high uptime.
TNG is used for feeding items and for live “talking head” broadcasts (end-to-end delay 40 ms) for news. The TNG
teams draw maps indicating virtual feeding points and green/orange/red working areas in the cities.
38
http://www.linkres.co.uk/
39
Belgian coast network is not yet operational but will be soon
In former times it used to be simple… television journalists and crews gathered the news, processed it and delivered
it to base – either by hand, circuit or satellite. Of course, in reality it wasn‟t simple - up until the early 90s the BBC
issued its reporters with what it charmingly called “mutterboxes” connected with crocodile clips to a phone handset
or to its wall socket. Satellite equipment was bulky, it went wrong, the delivery cost a small fortune and
unaccountably, suspicious airplane passengers would sometimes refuse to accept a big bag of tapes…
Communications were difficult so for much of the time you were out of touch with head office. The return path was
rudimentary and often amounted to little more than a telex or a fax telling you how your piece had gone down.
And then everything changed. Satphones40 became more portable, speeds increased and the first store and
forward devices emerged on the scene. They were affordable, reasonably easy to use and they worked more or
less. Of course, these had an extraordinary effect on our ability to report from the scene. And then came the M4
41 42
satphone, the videophone, the undreamt of speed of 128 kbit/s. And now the BGAN , the new Thuraya system ,
43 44 45
Streambox , Quicklink , vPoint and so on. And in many places HSDPA/&UPA, WiMAX and, even in remote
areas, within months, 384 kbit/s of bandwidth. But again it‟s not that simple, because it‟s no longer a one-way street,
and what we have been building is a network. We have been extending the newsroom into the field so that many of
the activities that might once have taken place at base are now happening on the ground.
A demo video demonstrates how a reporter in the field can proceed to a federated search into the BBC's Jupiter
Media Asset Management system, import on-demand Flash Video archives files to the field and export its own files
46
after editing with a tool based on Rhozet transcoding solution.
The ability to access material in this way has revolutionised what gets on air. All this is an extraordinary opportunity
for better story-telling, but it‟s also represents a gigantic challenge both editorially and technically:
We risk overloading staff in the field with an ever-increasing amount of information which leaves them no time to
go and find things out – no time to actually be journalists. Increasingly, the point of publication will get closer to the
newsgathering process and we will need to ensure that editorial standards are upheld.
The technical challenges are just as significant. Staff is expected to have an extraordinary range of skills. Both
journalists and technicians need a comprehensive knowledge and understanding of networking principles. We run
the risk of allowing technical quality to degrade as the public become increasingly capable of creating and
delivering high-quality video themselves. We face the challenge of mixed standards, of concatenating codecs, of
equipment that simply doesn‟t perform as we expect it to. And we will often be using equipment that offers nothing
approaching the robustness of the broadcast equipment to which we have been accustomed.
The future is the network and the networking of knowledge. Audiences are beginning to experience a truly interactive
relationship with the news – and this requires the extension of the network into the field. So getting news to base is
the easy part. The challenge for the future is to ensure that we exploit the network effect to create a new form of
journalism.
How can I connect my Video Journalists best? …as flexible as possible! VJs should always be able to use the best
40
For example, with the the 'Toko', manufactured by Toko Japan (http://www.toko.co.jp/top/en/index.html ), a device that digitally
compressed video, stored it and forwarded it through a telephone or a satellite phone.
41
See EBU Networks 2007 seminar report § 5.3
42
http://www.thuraya.com/
43
http://www.streambox.com/
44
http://www.quicklink.tv/
45
http://www.vcon.com/
46
Carbon Coder http://www.rhozet.com
There are numerous ways for a VJ to get connected. The common network interfaces of the VJ equipment are:
Ethernet (≤ 1 Gbit/s), WiFi (nominal up to 56 Mbit/s), mobile (phone) connections (UMTS, HSPA…), WiMAX. Table
4 (Annex) lists the overwhelming crowd of networks worldwide and actual bit rates measured on some relevant
networks.
VPN security is important for VJs. It is useful for the protection of the reporter identity, for very critical material -
especially in 'monitored' networks (China & Co.), for full LAN-integration (with all the administrative overhead), and
for securing applications with low security levels (like FTP). But:
Transfers can be secure enough if native mechanisms are used: HTTPs, sFTP, file encryption and signing,
dedicated input server (in DMZ).
The security overhead can have massive negative impact on the transfer-throughput. Therefore individual
performance-tests are essential. For example [14], a 3.4 Mbit/s bit rate through a no VPN circuit, will become
47 48
3.3 Mbit/s with CryptoGuard VPN software (with AES encryption) and decrease to 1.9 Mbit/s with OpenVPN ,
an open source SSL VPN solution.
The VPN throughput is important for daily VJ-business, even very low bit rates are challenging the available
connections. Here is an example of calculation:
A 5-minute video material coded at 4 Mbit/s will deliver 1.2 Gbit in 5 minutes (30 MByte/minute)
25 Mbit/s will deliver 7,5 Gbit in 5 minutes (187 MByte/minute)
The Table 4 (Annex) indicates the corresponding transfer duration of this 5-minute material on relevant networks.
In earlier sessions of this seminar, you will have learned about the latest technologies which permit the construction
of networks designed to allow contribution transmissions for television production. This final session will provide
examples of the impact of new network technology on television production and broadcast.
In a production environment, with file-based acquisition – production – delivery, the network is the link between the
'clients' (desktop applications for journalists, researchers, producers, editors, librarians…) and storage (real-time
online / high-performance neartime / library archive) [2]-[7].
47
http://www.compumatica.de/cms/data/index.php?id=21&L=0
48
http://openvpn.net/
Gigabit Ethernet end-to-end uses UDP instead of TCP. Just like VoIP this type of video protocol should let the
application decide if it needs to re-request the data. A Fast Ethernet connection (100 Mbit/s) results in dissatisfied
users, especially if they move between machines with Gigabit Ethernet interfaces. Editing application can co-exist
with Video over IP corporate solutions if these are designed correctly.
Should we transfer jumbo frames or fragmented packets? Video data is in large blocks on the disks, so it should
be ideal! It is more efficient to send larger segments of data, especially with UDP, but it only sustainable in a
controlled environment (e.g. server room). Fragmented datagrams can pass over non-jumbo frames enabled
networks but need on-board memory in network interface cards (NIC).
Big chassis-based Gig-E switches have different cards with different abilities - every interface card matters!
Different chassis-based switch models from same manufacturer have different abilities - biggest is not always the
best!
NIC (Network Interface Card) - The buffers (or descriptors) required for this application would have been considered
server class just a couple of years ago, but now this class of adapter can be found on high and medium grade
platforms. On board memory is needed for fragmented packets - 32 KB on low cost NIC implementations is
49
insufficient. If a TCP-based data flow is used there will be lots of CPU requirement unless ToE -enabled NIC is
used.
A separate production network is best for the core systems and high bandwidth clients. Corporate network based
clients are possible for less demanding application and resolutions, if the corporate infrastructure can cope and has
the right products in the correct place.
Separate the Network into zones from the core highest bandwidth zone 1 (with direct storage connection) to the
lowest bandwidth (customer network) zone [21]-[23].
50
In a video production environment QoS generally relates to available bandwidth, latency & jitter. Vendors provide
different QoS tools that you should use on edge and backbone routers to support QoS. Some QoS mechanisms are
aimed at VoIP networks and low bandwidth circuits, but others apply equally to LAN. Example of QoS mechanisms
for Video over IP are: LLQ (low latency queuing), PQ (priority queuing), WFQ (weighted fair queuing), CQ (custom
queuing), PQ-WFQ, CBWFQ (class-based weighted fair queuing). A good old fibre channel is very deterministic and
at 5ms latency the effect begins to show. Design the Gigabit Ethernet system with sufficient capacity to ensure the
necessary QoS. In a corporate network ensure sufficient QoS exists - the corporate network may need to be re-
configured to correctly recognise and prioritize video traffic.
Firewalls present many challenges. They are not good at bursty, high bandwidth, fragmented real-time flows. In a
production environment we deal with REAL TIME VIDEO, not a streaming VC-1 file! Again, contain the network
design into zones and locate externally facing clients in the higher number zone, or even in the corporate network.
Buffers, buffers, buffers! Switches with dynamically shared buffers are a better choice - some manufacturers
provide 1U or 2U condensed versions of popular mid-range chassis-based switches. Switches with statically
assigned buffers limit the design scope - this limit affects some large chassis-based switches from some MAJOR
manufacturers, while smaller chassis-based models have shared buffers and are an excellent choice. Buffers in the
network cards are also critical.
In conclusion, understand every link in the chain (some design examples are presented [29]-[31]) and where that link
exists! Every switch, every speed change, every aggregation point MATTERS, and that includes the NIC in the
49
TCP offload Engine -a technology used in network interface cards to offload processing of the entire TCP/IP stack to the
network controller. It is primarily used with high-speed network interfaces, such as Gigabit Ethernet and 10- Gigabit Ethernet,
where processing overhead of the network stack becomes significant.
50
http://www.techweb.com/encyclopedia/defineterm.jhtml?term=QoS&x=&y=
The BBC network objective was to provide a shared infrastructure for all user requirements that delivers the
cheapest overall solution, for:
BBC Distribution Services ensuring the real time carriage of Vision & Audio (with multiple platforms to be
connected to the interchange points with the Transmission provider).
BBC Contribution Services: real time studio-to-studio carriage of Vision and audio; packet network for media and
business applications; Storage Area Network; telephony
English Region Cluster Sites (ERCS) - 52 sites across the UK - to enhance services and capacity in line with core
sites.
The scale of the implemented network [4] is huge:
Large national & regional sites: 18 core sites + 11 associated sites and 1000+ circuits
London BBC Sites: 10 sites and 1000+ circuits
District offices & local Radio sites: 60 small sites plus approximatively 60 further minor sites and 600+ circuits
International sites – International Bureau: 30 sites and 100+ circuits.
Raman name comes from the optical amplication technology which provides on each fiber 240 DWDM wavelengths
(=channels) x 10 Gbit/s each [5], on each „Raman Arc‟ between two RIPs (Raman Interconnect Point). Raman
technology provides both high density wavelength multiplexing and the ability to drive over much higher distances
without re-amplification. The double star topology [9] offers high geo-resilience, allowing each site to be connected
through two routes.
Implementation and testing. From 2005 to October 2007 Cable&Wireless (C&W) and Siemens SIS designed, built
and tested the network [6].
C&W managed Raman, IP WAN, Broadcast & IP Network Services, Audio
Siemens SIS managed the integration to Core Raman sites, the integration to BBC systems, LAN, and IP
telephony
C&W uses SDH transport over the Raman transport layer with following equipment [7]: Marconi SDH multiplexers
with 10G, 2.5G, STM-1, 2Mbit/s & GbE interfaces; Scientific Atlanta iLynx for SDI, PAL & ASI interfaces; utilise
existing ATM switches for ATM services - audio, both analogue and AES3 (delivered via ATM AES47); the packet
network terminal equipment is Cisco “Catalyst” in C&W core, Foundry at edges.
SIS & C&W undertook 12 weeks of circuit testing with 20 resources working with the BBC: 1300 circuits tested; bit
error tests performed; port tests performed; break tests of fibre; impacts all Layers to highlight issues ???; IP
network early gateway connection to confirm monitoring tools ??? Through this methodology the Raman network
was clean of technical issues ready for migration and it was a key work package to ensure the success of the
project. 1200 individual tests were performed over 4 months. In order to test monitoring tools and service proces
gateway connection was made in advance of migration. The tests focused on Operational Interfaces and a gap
analysis conducted on what we need to do differently. The joint SIS-C&W-BBC approach in the design and the
implementation of the tests saved time and ensured a comprehensive testing strategy. The operational model
followed the ITIL industry standard framework (Information Technology Infrastructure Library), providing a common
framework and language to all organisations.
Migration. The Broadcast Migration took 9 weeks; and the IP Network 6 weeks. Both activities were run in parallel
with the IP network starting two weeks before broadcast migrations. All migrations conducted between 20:00 &
05:30 hour. These migrations were all complete without any service outages and impact to the BBC, thanks to
working in partnership, to the SIS and C&W knowledge of the BBC‟s Transmitter, Broadcast & IT network (old and
new), and to planning… planning and more planning! The project has been delivered according to time and budget.
The network is fully operational and performing to design requirements.
English Regions Cluster Sites - This was the first transformation project after Raman. It addressed the connectivity
needs of BBC local Radio sites, to consolidate Telephony, Data and Broadcast Services, enabling initiatives such as
programme share, enabling multicast data and audio in the future.
It consisted of the migration from multiple discrete services to a fully managed, converged solution with resilient IP
over Ethernet WAN service, offering QoS, multicast, support of multiple applications, inntegration to Raman Core
6.3 DVB-H small gap fillers: home repeaters improving indoor coverage
Davide MILANESIO, Centro Ricerche e Innovazione Tecnologica, RAI, Italy
DVB-H is a system specified for bringing broadcast services to battery-powered handheld receivers. It is an
extension of the DVB-T system [3], transmitted in the UHF band, allowing multi-channel, robust with respect to
disturbances and for reception indoor, pedestrian and at high speed (train), designed to reduce the battery
consumption.
DVB-H is already operational worldwide: in Italy (since 2006), Finland (since 2007), Austria-Switzerland (June 2008),
Albania, Asia (India, Malaysia, Philippines, Vietnam) and Africa (Kenya, Namibia, Nigeria). But these networks are
mainly planned for outdoor coverage not for indoor! Therefore traditional DVB-T network planning is not sufficient
for DVB-H, because indoor DVB-H reception requires higher electromagnetic field strength, since the receiving
antenna is integrated in the terminal and not on the roof! [6]
To improve indoor coverage, the main transmitters could be completed by a number of low power urban
transmitters. But electromagnetic radiation limits have to be respected and there is a risk of interference on
traditional TV services in the existing MATV distribution systems [7].
DVB-H 'small gap fillers' are another way to improve indoor coverage using low-power on-channel home repeaters.
These consumer-grade devices can be autonomously installed by final users in their private homes, without the help
of a professional installer. It is connected to the existing in-building cable distribution system [8], Its coverage area is
this of a standard apartment, i.e. about 100 m2 enabling the interested users to be immediately reached by DVB-H
services.
Standardisation – Since these devices are radiating in the UHF band, without a specific regulation, a licence would
be needed. Therefore a new standard is necessary, also to avoid that low quality (illegal) devices appear on the
market, potentially causing dangerous interference on existing services (e.g. analogue or digital TV).
The DVB-H Small Gap Fillers Task Force, with 21 companies involved (Broadcasters, network operators, regulation
authorities, manufacturers), has prepared a Technical Specification. It has been approved by the DVB Steering
Board, in June 2008, for publication as a DVB Blue Book. It will then be submitted to ETSI/CEPT for publication as
European Norm (requiring voting by National Standards Organisations).
51
http://www.ietf.org/rfc/rfc3926.txt
Validation. The Technical Specifications have been validated in laboratory trials on real hardware prototypes [13]
52
and in the framework of the European Project 'CELTIC B21C' .
Coverage tests were conducted in the Rai-CRIT laboratories and in a real flat and proved an adequate coverage in
standard apartments (e.g. 100 m2) [14].
The disturbance on adjacent TV channels was tested using 2 reference scenarios [15], with positive results [16].
Scenario Results
1 TV / STB connected to another Video SNR degradation within 1 dB,
plug of the in-building cable not noticeable on picture
distribution network
2 TV / STB in another room, Video SNR degradation within 2 dB if OK also in pessimistic condition:
connected to an indoor using 2 SAW filters, not noticeable on Adjacent analogue TV channel received by the
amplified antenna picture. repeater with level 30 dB higher than DVB-H
1 wall separation, 3 m distance Degradation within 5 dB with more No visible impairments using repeaters
relaxed masks (out of standard) with 2 SAW filters
A CATV network could allow for a possible future extension of the DVB-H Small Gap Filler concept. The DVB-H
multiplex could be transported on the CATV network on behalf of the broadcaster [17]. The indoor DVB-H coverage
will be improved in areas where TV aerials are not very popular but where a CATV network is available. CATV would
be used only as a carrier, no DVB-C signals being involved. This would allow to reduce the DVB-H network
deployment costs.
But this involves an additional requirement, the possibility of frequency conversion. Since the CATV network might
not cover the full UHF band. Therefore, the coherence of the output frequency has to be guaranteed, i.e. the output
frequency has to be cross-checked with the Network Information Table of the incoming stream. Moreover the
network operator should guarantee the synchronisation with the traditional DVB-H transmitters as in a standard SFN
[18]. Currently, frequency conversion is not included in the Specification.
52
Celtic (‘Cooperation for a sustained European Leadership in Telecommunications’) - Broadcast for the 21st
Century
http://www.celtic-initiative.org/Projects/B21C/abstract.asp
1) Fast Convergence 2) Application Layer 3) Temporal Redundancy 4) Spatial (Path) Diversity 5) Multicast-only Fast Reroute
or Fast Reroute Forward Error Correction (FEC) (TR) ("live / live") (MoFRR)
(FRR) [15] [16] [18] [20]
[8]
Network reconverges** / reroutes* Adds redundancy to the transmitted The transmitted stream is Two streams are sent over diverse Deliver two disjoint branches of
on core network failure (link or data to allow the receiver to detect and broken into blocks, each paths between the sender and the same PIM (Protocol
node). correct errors (within some bound), block is then sent twice, receiver Independent Multicast) SSM
without the need to resend any data. separated in time. (Source Specific Multicast) tree
If block separation period is to the same Provider Edge
greater than the loss of The Provider Edge locally switch
connectivity, at least one to the backup branch upon
packet should be received detecting a failure on the primary
and video stream play-out branch
will be uninterrupted. http://www.nanog.org/mtg-
0802/farinacci.html
Lowest bandwidth requirements in Supports hitless recovery from loss due Supports hitless recovery Supports hitless recovery from loss Hitless: The Provider Edge uses
working and failure case to core network failures if loss can be from loss due to core due to core network failures if have the two branches to repair losses
Lowest solution cost and complexity constrained network failures if loss can network stream split and merge and present lossless data to its
No requirement for network path be constrained functions (e.g. Dynamic Channel path IGMP (Internet Group
The simplest and cheapest design / Management) Management Protocol) neighbors
operational approach for a Service diversity - works for all topologies No requirement for network
Provider is to have such behaviors path diversity – works for all Lower overall bandwidth consumed in A simple approach from a design
optimised by default in the software topologies failure case compared to FEC and deployment and operations
and hardware implementations and Introduces no delay if the paths have perspective
is applicable to all its services. equal propagation delays
Requires fast converging network to Requires fast converging network to Requires fast converging May require network-level techniques MPLS and Multi-Topology
minimize visible impact of loss minimize AL-FEC overhead network to minimize block to ensure spatial diversity, MoFRR Routing are options in topologies
separation period (5), Multi-Topology Routing, Traffic that do not support MoFRR
Engineering; required techniques
depend upon topology
Is not hitless - will result in a visible Higher overall bandwidth consumed in Incurs 100% overhead
artifact to the end users failure case compared to "live / live" (4) Incurs delay - longer outages
Incurs delay - longer outages require require larger block
larger overhead or larger block sizes separation period
(more delay)
© EBU Networks 2008 Seminar / 23 - 24 June 2008
Reproduction prohibited without written permission of the EBU Technical Department & EBU International Training
36
(*) According to the number of multicast groups 400 - 800 - 4000, the median / max reconvergence time for all channels following a network failure may amount to 200/290 ms – 260/380 ms
(no more than 1 frame loss) – 510/880 ms [11]. To improve it, prefix prioritisation allows important groups (e.g. premium channels) to converge first, and developments with IP optical
integration can further reduce the outage to sub 20ms in many cases (lossless in some cases) by identfying degraded link using optical data and signaling before the traffic starts failing [12].
(**) Fast rerouting: the routing protocol detects the failure and computes an alternate path around the failure [10]. But loss of connectivity is experienced before the video stream connectivity is
restored.
MPEG-2 MPEG-4 AVC / H.264 JPEG2000 Dirac Pro MPEG-4 SVC / H.264
Standard ISO/IEC 13818 ISO/IEC 14496 ISO 19446-1 SMPTE VC-2 Amendment to H.264/AVC
(Developped by BBC standard
Research)
Principle DCT based system Spatial and temporal Wavelet based (DWT) compression system coupled to an Mixing the advantages
Spatial and temporal compression with a arithmetic coder. of the wavelet
prediction with variable choice between context Lossless and lossy compression transform and motion
length statistical coding adaptive arithmetic compensation.
(VLC) coding (CABAC) or Similar wavelet filters to
variable length (CAVLC) JPEG2000
Structure 6 Profiles (prediction 7 Profiles and 5 levels Only DCI profiles exist at the moment. Follows H.264 profiles.
tools supported) and 4 For contribution (at the Broadcast application profiles under investigation in
levels (image formats and moment): JPEG.
max bit-rate supported)
High profile 4:2:2 – 10bit
For contribution:
Level 4.1 (1080i and
Main profile (4:2:2) I,P 720p)
and B frame prediction
@ High level
(1920x1080) up to 80
Mbit/s
Scalability (provided SVC Highly scalable Spatially scalable due Scalability on top of H.264
amendment) Spatially – Wavelet transform. to the wavelet. High coding efficiency
Acccess technique Bit rate (uplink) Performance of relevant networks Transfer duration
(measured) of a 5-minute material
(4 Mbit/s material)
Fixed line
Tel. modem 56 kbit/s
ISDN 128 kbit/s
ADSL 640 kbit/s
SDSL 4.6 Mbit/s Arcor Business ~ 1.95 Mbit/s ~ 10 min.
VDSL 50 Mbit/s T-Home/IRT ~ 2.94 Mbit/s ~ 6.7 min
Cable network 40 Mbit/s
Powerline (PLC) 200 Mbit/s
Wireless
WLAN (802.11b/g) 11 / 54 Mbit/s InternetCafe Munich ~ 870 kbit/s ~ 30 min.
EBU WiFi ~ 3.55 Mbit/s ~ 5.6 min.
Satellite link 1 Mbit/s
GSM 12 kbit/s
GPRS 171.2 kbit/s
HSCSD 57.6 kbit/s
EDGE 473 kbit/s
CDMA One 14.4 kbit/s
UMTS 128 kbit/s Vodafone ~ 120 kbit/s ~ 180 min.
HSUPA 5.8 Mbit/s
CDMA2000 144 kbit/s
EV-DO 5.4 Mbit/s
Flash OFDM 900 kbit/s
LTE / HSUPA 50 Mbit/s Vodafone ~ 1.35 Mbit/s ~ 14.5 min
WiMAX (mobile) 7 Mbit/s IRT-TESTnet ~ 1.15 Mbit/s ~ 17 min.
iBURST 346 kbit/s