EBU Technical Review : No. 279 (Spring 1999)
Networks in the production environment
Many broadcasters are converting from analogue to digital in TV production centres by installing SDI (Serial Digital Interface) networks, typically operating at 270 Mbit/s. These networks can be used to carry uncompressed digital video signals, as defined in ITU-R Recommendation BT.601. Last year's Report of the EBU-SMPTE Task Force considered the more complex problem of interchange of compressed video signals. The Report noted that the SDI infrastructure could also be used for SDTI (Serial Data Transport Interface) applications, including real-time and faster-than-real-time transfer of compressed signals.
SDI and SDTI are probably today's best choices for in-studio applications, but other types of network will become important because of technical and commercial developments in other areas, particularly in the telecommunications and computer industries. Are broadcasters sufficiently aware of these developments?
In the past 15 years, many organizations have learnt that "office automation" did not yield the promised gains in efficiency nor, indeed, achieve the nirvana of "paperless offices". Perhaps their expectations were unrealistic – but, in many cases, the real problem was that office procedures based on paper were simply transferred to computers. Are broadcasters making the same mistake by opting for a direct digital replacement of existing analogue networks? Is an important opportunity being missed?
Telecoms operators are rapidly adopting packet-switched networks in preference to traditional circuit-switched networks, but broadcasters are reluctant to embrace packet-switched networks. Is this reluctance justified? Or is it another example of the "not invented here" syndrome?
Broadcasters have long believed that their needs are very different to other users of telecommunications networks. This was certainly true 20 years ago, when the bandwidth of video signals was huge in comparison with other types of traffic, mainly telephone calls. Nowadays, broadcasters do not have the monopoly on wide bandwidth signals: data traffic now exceeds telephone traffic on many networks.
In the production environment, today's networks are based on uni-directional transmission of programme material from a source (e.g. a camera) to a destination (e.g. a video recorder). In practice, a typical operation, such as replaying a video-tape from a machine within the same building or at a remote site, generally requires supplementary bi-directional links – either telephones used to co-ordinate the actions of human operators or digital links allowing full remote control.
In the future, broadcasters will adopt computer technology for programme production: editing of audio and video material is now routinely performed on desk-top computers, whilst networked servers are increasingly being used for production, especially for news programmes and for archiving. These "distributed" production processes require network topologies which have little in common with existing networks. Efficient use of server technologies requires faster-than-real-time transfers – as well as needing bi-directional links for control purposes. Existing networks cannot offer such facilities. Does this imply that broadcasters will need two or more independent networks in their studio centres? Would it not be better to have a single multi-purpose network?
Ethernet is the most popular type of computer Local Area Network (LAN). It was originally designed to operate at 10 Mbit/s, but later variants operate at 100 Mbit/s and, now, at 1,000 Mbit/s (so-called Gigabit Ethernet). Ethernet allows point-to-point transmission within the LAN by operating in a "broadcast" mode: all terminals receive all transmissions, but individual terminals ignore transmissions not intended for them. As only one transmission can use the LAN at any instant, special measures need to be taken to avoid "collisions" which occur when two terminals transmit simultaneously. This mode of operation may seem crude, but it is surprisingly effective if the average utilization of the network is low. However, because the LAN can accommodate only a single simultaneous transmission, it is difficult to envisage widespread application of Ethernet in the production networks.
IP (Internet Protocol) is the best known packet-switching protocol. It is much more sophisticated than Ethernet. In conjunction with higher-level protocols such as TCP (Transmission Control Protocol), packets of data can be delivered through complex networks using "intelligent" routers. As successive packets of data may be delivered from one terminal to another through different routes on the network, packets may be significantly delayed and can even arrive "out of order". The TCP/IP protocol suite can re-assemble the packets in their correct order, as well as requesting re-transmission of packets with errors. Despite having many advantages, IP is regarded with cynicism by many engineers – probably because they have all suffered poor performance on the Internet. In fact, this is unfair because many of the problems are caused by congestion on the Internet, rather than being an inherent feature of IP.
IP can be characterized as a "best effort" protocol but, in practice, IP cannot offer any guarantee of adequate performance. The issue of "quality of service" (QoS) will be addressed in future versions of IP and, perhaps, more importantly, by Asynchronous Transfer Mode (ATM) networks.
During the last five years, ATM has been proclaimed as the next big development in broad-band telecommunications. ATM's ability to provide "bandwidth-on-demand" is attractive to broadcasters for "dial-up" contribution circuits. Unfortunately, the introduction of ATM has been less rapid than envisaged, mainly because of the increasing dominance of IP.
Networks designed for computer applications are unlikely to be suitable for audio or video applications. Data transmissions generally consist of short bursts of data. Even if the peak data-rate is high, it is rarely sustained for a long period. Most data transmissions can tolerate intermittent delays of a few seconds since they are not used for time-critical operations. On the other hand, audio and video signals in the production environment need to be available immediately – without any subsequent interruptions to the signal flow which may need to be sustained for several hours.
Despite these disadvantages, packet-switching enables easy expansion of a network, in contrast to the inflexibility of a network based on a matrix switch for interconnection between m sources and n destinations. If more sources or more destinations are needed, it is often impossible to expand the matrix switch and, hence, secondary matrix switches or manual connections are needed.
Packet-switched networks can be designed to be resilient to failure of parts of the network (e.g. by using dynamic routing in IP-based networks). In circuit-switched networks, the whole network can be paralysed by the failure of key elements, such as the central matrix switch.
Circuit-switched networks satisfy many of today's requirements, but packet-switched networks will become more and more attractive to broadcasters.
When will packet-switching become the preferred choice for in-house production networks? I believe that the answer to this question is "much sooner than expected by most readers of EBU Technical Review"!
Philip Laven
Director
EBU Technical Department
European Broadcasting Union
