desktop video conferencing - cellmigration.org · users to participate in a conference, regardless...

Desktop Video Conferencing

How do I choose a Desktop Video Conferencing system?Well, you could just buy the same as the person you want to have a conference with. However, this might notbe the best solution. There are essentially two types of systems, proprietary and standards based. If theperson you want to have a conference with uses a proprietary system, then you must buy the same as them,or persuade them to buy something different.

There are many questions that must be answered as you steer your way towards identifying which is the bestDesktop Video Conferencing system (DVC) that meets your needs. This document is about standards basedsystems and is intended as a guide to their selection.

The following technical papers are also available to provide more information:Video Conferencing Standards and Terminology.■

H.323 Terminals, Gatekeepers, Gateways & MCUs.■

H.221 Framing used in ISDN Conferences.■

H.323 Dial Plan and Service Codes used by Gatekeepers etc.■

International Telecommunications Union & The Internet Engineering Task Force.Telecommunications standards are set by the United Nations agency, International TelecommunicationsUnion (ITU) and the Internet Engineering Task Force (IETF). Products that adhere to these standards allowusers to participate in a conference, regardless of their platform. These standards for desktop videoconferencing ensure compatibility on a worldwide basis. The ITU has developed the H.320, G700 & T.120family of standards whilst the IETF has developed Real-Time Protocol (RTP) & Resource ReservationProtocol (RSVP). These standards apply to different transport media.

Before you start, it is useful to understand what types of media are available. ISDN, LANs and POTS (PlainOld Telephone Service) and now the Internet are the popular transport media used in desktop videoconferencing. They all have strengths and weaknesses that should be considered carefully before decidingupon which one to use.

Integrated Digital Services Network (ISDN).

What you really need to know about Desktop Video Conferencing Systems

file:///F|/WebSupport/WebDev/phase12/video.html (1 of 9) [18/09/2001 11:36:59]

There are two available ISDN connections, Basic Rate Interface (BRI) and Primary Rate Interface (PRI).Essentially, a BRI provides two 64kbps B channels and one 16kbps D channel. A PRI in Europe provides 30B channels and one D channel.

New ISDN connections usually aggregate the BRI and share the same number for both B channels. Knownas ISDN-2, this provides a line speed of 128kbps is typically used in a desktop conference over ISDN. Forincreased bandwidth, ISDN-6 provides a line speed of 384kbps and is typically used in room-basedconferences over ISDN. With ISDN-6, the sequence in which the lines are aggregated must be known andadhered too! Furthermore, if the connection is going to use some form of 'switch', this must be configured topass both voice and data!

In the past, most of these would have been between just two participants as ISDN is essentially apoint-to-point connection. However, multipoint technology now makes it possible for groups of people toparticipate in a conference and share information. To hold a multipoint conference over ISDN, participantsmust use a Multipoint Control Unit MCU.

H.320 is the ITU standard for ISDN conferencing and includes H.261 video, G.711, G.722, G.728 audio andT.120 data.

Local Area Network (LAN) & Intranet.10 Mbps Ethernet LANs are used in most companies today, and these have enough bandwidth to supportdesktop conferences. With a LAN offering significantly more bandwidth than ISDN, the video quality within aconference is much higher and can approach that of television. Communications advancements such as FastEthernet (100 Mbps) and ATM (155 Mbps) have increased bandwidth, whilst MultiCast technology hasreduced network loading in conferences involving more than two participants.

Unlike ISDN networks, LANs use TCP/IP protocol and the H.323 standard defines how to assemble theaudio, video, data and control (AVDC) information into an IP packet. To hold a multipoint conference over aLAN/WAN, participants must also use a type of Multipoint Conference Unit MCU

H.323 is the ITU standard for LAN conferencing and includes H.261 & H.263 video, G.711, G.722, G.723,G.728 audio and T.120 data.

Plain Old Telephone Service (POTS).The standard telephone system is the most readily available form of transport media for use in desktop



conferencing systems. With modems giving transmission speeds of up to 56 kbps, there is sufficientbandwidth available to support audio, video and data sharing with this media, especially when used inconjunction with the latest CPU's, compression techniques and technologies like DirectDraw.

H.324 is the ITU standard for POTS conferencing and includes H.263 video, G.723 audio and T.120 data.

However, the use of standard telephone lines for POTS based H.324 conferencing has given way to the everincreasing popularity of the Internet.

Internet.Like LANs, the Internet is another form of TCP/IP network and hence can be used as a transport media indesktop conferencing systems. Not to be confused with POTS, the Internet uses a modem as a TCP/IPdial-up adapter in order to gain access to the network. What the users must do is to get their Internet ServiceProvider (ISP) to provide them with a fixed IP address. Alternatively, NetMeeting users can register theirpresence with a Microsoft Internet Location Server (ILS). This is how you determine the address of themachine that you want to conference with. Obviously, speed is limited to that of the slowest link. The ITU hasnow ratified a 56 kbps standard; the V.90. If buying a new modem, you should select a V.90 compliant one;existing K56Flex and X2 modems maybe upgradeable to V.90.

H.323 is the ITU standard used for Internet conferencing and includes H.263 video, G.723 audio and T.120data.

Issues to Consider.Each of the above technologies has there own strengths and weaknesses that should be consideredcarefully before deciding upon which one to use.

The trade-off factors involved in determining the best system for your situation will be:

Expectation levels Available bandwidth

Acceptable quality Number of participants

Location of participants Installation costs

Expectation Levels.A crucial area in selecting a system is to discuss and then set the expectations levels of the users to beattainable. What is a realistic frame rate and window size for the available bandwidth will most probably bemuch lower than that expected by the users. However, for the users to get the most out of a DVC system,they must be content when using it. The expectation levels must be aligned to what is realistic from theavailable systems.

Don't expect full screen motion video via the Internet, the numbers just don't add-up!

Required versus Available Bandwidth.DVC is a form of communications involving the transfer information between two or more locations. Theconnection between these locations is the communications channel and is called the network. It is thenetwork loading in terms of required bandwidth that needs to be considered. Bandwidth is the resource of a



network. It is the term given to the rate of transfer of information, usually in bits/second; it is like the speedlimit of the network and cannot be exceeded.

Analogues can be made between speed and bandwidth. If you wanted to know how long it would take totravel 1000 km when you are limited to 50 km/h; then a simply calculation of 1000 divided by 50 shows that itwould take 20 hours. Likewise, if you wanted to transfer one megabyte of data across a network with anavailable bandwidth of 9600 kps, it would take just over 14 minutes.

Available bandwidth is the limiting factor with desktop conferencing and sending video creates lots of data.Consider a typical CIF video image size of 352x288 pixels; then this represents 101376 pixels or 304128bytes of information (if the image depth is 24 bits) per single frame. Now consider how many frames persecond that you want to see; this is called frame rate. The human eye perceives 25 frames per second ascontinuous motion, therefore using this rate in our calculation means that if we wanted to see continuousmotion video at this image size, we would need to transfer 304128 x 25 or approximately 7.2 Mbytes persecond.

It is clear from the above example that video can place enormous demands on the network and hence whyavailable bandwidth is the bottleneck with desktop conferencing systems.

There are essentially two ways of reducing the impact of this bottleneck. One is to use a faster method ofcommunications that increases the available bandwidth to the conference; the other is to utilise methods thatreduce the amount of data to be transmitted.

Acceptable Quality.There are several steps that can be taken to reduce the amount of data that has to be transmitted whendesktop conferencing. The obvious combination is to use the smallest acceptable window size with theminimum acceptable frame rate. Applying compression can then further reduce the resulting amount of data;but which type has a crucial effect on quality.

There are two types of compression techniques, lossless and lossy. GIF files are an example of losslesscompression. With this technique, the exact data that went into the algorithm comes out when the data isuncompressed. But GIF compression is applied to still images and is not applicable to conferencing.

Take the previous example of a CIF sized video at 25 frames per second; to transfer this over an ISDN-2 lineat 128 kbps, we need to apply a compression ratio of approximately 500:1. To achieve this, DesktopConferencing Systems use lossy compression. The amount of lossy compression applied, or lack ofcompression, is sometimes expressed as a 'quality' percentage, with 100% quality meaning leastcompression. The use of lossy algorithms creates a compromises between acceptable image quality anddata volume. An important consideration in the selection of a compression technique is the interoperabilitybetween the desktop systems being used in the conference.

JPEG Compression Effects.The JPEG compression algorithm was created by the Joint Photographic Experts Group, for still images. Ittakes 24 bit colour data and applies a combination of both lossless and lossy compression techniques in asliding quality scale, with 100% quality equating to lossless compression. JPEG provides good qualityimages with good compression ratios ranging from 3:1 for lossless images to 100:1 for lossy images. Theacceptance level of the resultant image determines the amount of compression applied. JPEG is asymmetrical algorithm, needing the same compute power and time to compress or decompress the image.Since JPEG was developed for still images, it does not use motion prediction.



Figure 1

Figure 3

Figure 2

Figure 4

The above JPEG images of Elle MacPherson are 24 bit colour, 176 x 144 pixels in size, (QCIF) and show theeffects of applying varying amounts of compression. This image size as raw data is 76032 bytes. Figure 1 isa 100% quality, lossless JPEG image, it is reduced to 27285 bytes. Applying lossy compression createsFigure 2, with an acceptable image at 25% quality of 2507 bytes, a 30:1 compression. Figure 3 has a 10%quality value applied, but does not provide an acceptable image at 1598 bytes, a 47.5:1 compression. Figure4 has a 5% quality value applied, which reduces the image beyond recognition. These samples demonstratehow much data is stored in QCIF size images along with a realistic value of how much JPEG compressioncan be achieved before the images becomes illegible.

But if we take the above QCIF sized video at 25 frames per second; to transfer this over an ISDN-2 line at128 kbps, we need to apply a compression ratio in the region of 120:1. Clearly we need to apply someadditional technique than just lossy compression in order to achieve the reduction in data we need toeffectively conference at this bandwidth.

DVC uses H.261/H.263 Compression.The H.261 and H.263 compression algorithms are designed for use with low bandwidth communications andincorporate motion prediction as well as lossy compression to further reduce the amount of data to betransmitted. The H.261 and H.263 images are also limited to CIF and QCIF sizes. The basic technique of



motion prediction works by sending a full frame followed by a sequence of frames that only contain the partsof the image that have changed. Full frames are also known as 'key frames' or 'I-frames' and the predictedframes are known as 'P-frames'. Since a lost or dropped frame can cause a sequence of frames sent after itto be illegible, new 'I-frames' are sent after a predetermined number of 'P-frames'. It is the combination ofboth lossy compression and motion prediction that allows H.261 and H.263 systems to achieve the requiredreduction in data whilst still providing an acceptable image quality.

The difference between systems mainly lies with the extent and how well the algorithm is implemented andwhether it is performed in hardware or software. Be cautious of systems that claim to achieve high framerates over low bandwidth lines using software implemented compression; they may, but is the image qualityacceptable?

Number of Participants.H.320 conferences are essentially a point-to-point connection and need to use a Multipoint Control Unit(MCU) to link and manage all the ISDN lines in order to hold a conference with three or more participants.

The MCU's basic function is to maintain the communications between all the participants in the conference.MCU's are hardware based as they need to connect to all of the ISDN lines from each participant. Forexample, to manage a conference between four H.320 systems, each at 384K (3xBRI), the MCU needstwelve BRI connections.

Most H.323 systems support IP multicast and use this to send just one audio and one video stream to theother participants. To allow three or more participants into a conference, most H.323 systems usually requirea Multipoint Conference Server (MCS). This is also referred to as an H.323 Multipoint Control Unit (H.323MCU). This is not the same as an H.320 MCU; hence it is important to be clear about what you mean whenusing the term MCU.

The H.323 MCU's basic function is to maintain all the audio, video, data and control streams between all theparticipants in the conference. Whilst most H.323 MCU's such as the MCU-323 from RADVision arehardware based, software versions for both Windows NT 4.0 and Solaris platforms are available. MostH.323 MCU's work in conjunction with and include a basic Gatekeeper functionality.

Gateways.



H.320 and H.323 systems can interoperate with the use of a Gateway. Essentially, the Gateway providestranslation between circuit-switched networks (ISDN) and packet-based networks (LAN), enabling theendpoints to communicate. Most Gateways have multiple BRI connections and can support severalconferences simultaneously. For example, a Quad BRI Gateway such as the L2W-323 from RADVision cansupport four conferences at 128Kbps, two at 256Kbps or one at 384Kbps.

Most Gateways work in conjunction with and include a basic Gatekeeper functionality. A real world H.323implementation of a Gateway working in conjunction with a Gatekeeper is in a Multimedia Call Centre wereneeds-based call routing and a variety of other automatic call distribution features are used.

Gatekeepers.Gatekeepers, such as the NGK-200 from RADVision, are an essential tool for defining and controlling howvoice and video communications are managed over the IP network; they provide call control services toH.323 endpoints. Gatekeepers are the intelligence for delivering new IP services and applications. Theyallow network administrators to configure, monitor and manage the activities of registered endpoints, setpolicies and control network resources such as bandwidth usage within their H.323 zone. Registeredendpoints can be terminals, Gateways or MCU's.

Only one Gatekeeper can manage a H.323 zone, but this zone could include several Gateways and MCU's.Since a zone is defined and managed by only one Gatekeeper, endpoints within a zone that also contain abuilt-in Gatekeeper must provide a means for disabling this functionality. This ensures that multiple H.323endpoints that contain a Gatekeeper can all be configured into the same zone. Both the L2W-323 Gatewayand MCU-323 MCU from RADVision have a built-in Gatekeeper that can be disabled, this allows the zoneto be controlled by a more powerful Gatekeeper such as the NGK-200.

Location of Participants.The first question that starts the process of identifying your DVC system is concerned with who and whereare the people that you want to conference with. It is a networking issue that determines how the participatingplatforms are going to be connected together and hence which is the applicable standard that you shouldconsider following. As indicated above, there are effectively three standards used in desktop conferencing,H.320, H.323 or H.324. After you have decided which standard you want to adhere too, you can start lookingat the platform, price and performance equation.

Do you want to conference just within your organisation, or with suppliers or with the world?

If it's just within your organisation, do you have an existing LAN? If so, then you already have a network inplace that can be used for DVC. Go to the section titled H.323 based DVC systems. Otherwise you need tocreate a corporate LAN or look at alternative network types. Go to the section titled ISDN, POTS or theInternet?

If it's to suppliers, are they on your corporate LAN? If so, then you already have a network that can be used.Go to the section titled H.323 based DVC systems. If not, then you need to extend the corporate LAN toinclude your suppliers or look at alternative network types. Go to the section titled ISDN, POTS or the



Internet?

If it's the world you want to conference with, then your options are ISDN, POTS or the Internet.

ISDN, POTS or the Internet.The decision on which of the above networks to use is essentially a trade between quality, cost and topology.At the higher performance and cost end there is ISDN, whilst at the low performance, low cost end there isthe POTS or Internet option. They are all effectively standalone systems, but all adhere to a differentstandard. H.320 and H.324 both have a 'star' topology.

There are now systems that include 384 kbps ISDN cards and offer the highest performance within the H.320standard. Obviously, there is the initial investment of installing ISDN lines to all the sites involved and this isprobably the most common reason for not going H.320. The big question here is 'What is an acceptableframe rate and window size'? If it is 2-6 fps at QCIF, then the Internet option and H.323 systems willprovide a solution. However, most professional people will demand much higher frame rates at CIF size andas such, ISDN linked H.320 systems are the only solution. These systems can achieve 10-25 fps at CIF size,depending upon the hardware.

H.324 is the newest member of this group of standards and whilst virtually everybody has access to POTS,there are very few H.324 systems available. Time will tell, but this low cost end of the market will probably bedominated by using your modem for Internet access. Hence, this document deals with just H.320 and H.323systems. Internet systems can achieve 2-6 fps at QCIF size.

H.323 based DVC systems.H.323 is the most flexible member of this group of standards. By using TCP/IP as its transport protocol,H.323 based DVC systems range from low cost Internet connected PC's at 28.8 kbps through to powerfulmachines using 155 Mbps ATM links. H.323 systems also cross the divide between the PC and UNIX worlds.

SUN Micosystems Inc have recently introduced SunForum 3.0, an H.323 standards compliant audio & videoconferencing system with T.120 based collaborative tools that are compatible with NetMeeting™. Toolsinclude whiteboard, application sharing, file transfer and chat. SunForum 3.0 is available for Sun Ultra 1 orbetter workstations with Solaris 2.5.1 or higher operating systems and is free to download. TheOsprey®-1500 capture card can be used to provide the audio & video inputs to SunForum 3.0, hence userscan now interoperate with other H.323 standards compliant systems.

The lowest cost H.323 software package is NetMeeting 3.0™ from Microsoft™, it is free and is includedwithin Windows 98™, Windows 2000™ and Internet Explorer 5.0. NetMeeting™ 3.0 provides audio, video,whiteboard, chat, file transfer and application sharing tools. NetMeeting™ was written to take advantage ofthe MMX technology in Pentium processors. It is also available on Windows 95™ and NT 4.0™. All that isneeded to complete a system is a TCP/IP networked PC with a video source and video capture card. ThisTCP/IP network could be a LAN, or an Internet connection. (See Internet section above for how to get an IPaddress). Performance will be very much dependent upon the power of the PC and the available bandwidth.

CU-SeeMe™ from White Pine Software offers similar functionality to NetMeeting™ and is also available onApple Mac.

Both NetMeeting™ and CU-SeeMe™ implement software codecs within the H.323 standard and are aimedat Internet users, although they will work over a LAN. Either an Osprey®-100 video capture card or anOsprey®-210 audio/video capture card should be used for maximum performance.

TCP/IP based LAN networks range from 10 Mbps Ethernet, 16 Mbps Token-Ring to 100 Mbps Fast Ethernetand 155 Mbps ATM links. They can be joined using ISDN, V.35, T1 or E1 lines between routers that maintainthe TCP/IP packet integrity. In all these cases, the available bandwidth is limited to that of the slowest link.

The VCON product range is targeted at business users with a PCI based Pentium PC and is very flexible.Most of the products are available as both H.320 and H.323 systems. They use NetMeeting™ to provideapplication sharing, whiteboard and file transfer.



http://www.sun.com/desktop/products/software/sunforum/index.html

http://www.microsoft.com/windows/netmeeting

http://www.wpine.com/

For an overview of available products, please see Video Conferencing Products.

H.320 based Conferencing systems.The Cruiser range provides H.320 solutions for PC desktop systems; whilst the MediaConnect 8000 Seriesprovides a Super VGA-based H.320 solution for group meetings and the MediaConnect 6000 Series orFalcon provide TV-based H.320 solutions for groups.

TeamSolutions can advise, supply and support whichever system is appropriate to your needs. For moreinformation and help in selecting the right system, please email: [email protected]

Subject to change without notice; no liability accepted for use. Company or product names mentioned aretrademarks of their respective owners. ©2001 TeamSolutions (UK) Limited. Updated: 14 September 2001.



mailto:[email protected]?Subject=Video Conferencing

desktop video conferencing - cellmigration.org · users to participate in a conference, regardless...

Documents