An integral aspect of radio broadcasting is the need to send program material from studios to transmitter sites. While this is a relatively straightforward task for small, standalone radio stations where production and transmission facilities are co-located, the bigger the broadcaster, the more complex the task becomes.
For a major national and international broadcaster, such as the BBC, ensuring the right material is delivered to the right transmitters reliably and at the highest possible quality is of vital importance.
For many years, the BBC relied primarily on a range of analog twisted-pair "music circuits" provided first by the Post Office and later British Telecom. These circuits were of variable quality and unpredictable reliability. Especially problematic was stereo information, which was often degraded by phase incoherence between the left and right signal paths.
During the 1980s these analog circuits began to be replaced by permanent digital audio circuits provided by BT along with NTL Broadcast (now Arqiva) and National Grid Wireless (Crown Castle). The BT system used was known as MusicLine 2000 and was based on E1 2 Mbps circuits, terminating at RE Barco codecs using the ITU-T J.41 companding standard.
Using this standard required 384 kbps for a mono 15 kHz circuit (six 64 kbps channels) doubling to 768 kbps (12 64 kbps channels) for stereo circuits. These J.41 circuits used a 14-bit word depth, providing 84 dB of dynamic range.
A particular benefit of these circuits, of vital importance in live broadcasting, was their impressively low latency, at just 4 milliseconds end-to-end. This latency was a fundamental requirement for real-time applications and a legacy from the era of twisted pair analog circuits which MusicLine 2000 ultimately replaced.
Over an E1 circuit, MusicLine 2000 is capable of delivering two 15 kHz stereo signals plus an additional 15 kHz Mono signal or, alternatively, a combination of voice communications (engineering talkback, etc.) and data (RS-232, contact closures, alarm relays, etc.).
As a technology and solution, the BBC installed several hundred of these circuits to service its audio and data requirements throughout the United Kingdom. Using E1 circuits offers great confidence to a broadcaster in that it is a clocked synchronous network, guaranteeing delay down to microseconds with the connection remaining absolutely stable.
Despite the success of E1-based circuits, since the early part of this decade, telecommunications providers have begun to move away from the use of synchronous network use in favor of Internet Protocol (IP) network operation for broadcast audio applications.
Although such networks provide greater flexibility than their synchronous counterparts, the main reason for the change is, of course, cost-savings.
However, despite the undoubted benefits for telecommunications operators, such a change is not without its potential problems for broadcasters. IP based networks are fundamentally different from their synchronous counterparts, both in terms of path latency (which is no longer fixed and not necessarily of low duration) and because they do not employ a central clock system (such that each individual data path is independent and therefore variable).
Not surprisingly therefore, despite exploring the use of audio over IP for some time, broadcasters have been somewhat reluctant to give up the predictability and synchronized latency of their program circuits and to embrace IP for delivery of key program links.
Despite such potential problems, the fundamental benefits of IP networks offer tremendous potential for cost savings and other benefits to larger broadcasters. Therefore, it is no surprise that the BBC commissioned in 2007 a research project to examine the integration of IP networks into its operations, as well as the practicalities of migrating from existing synchronous networks.
Siemens, having recently acquired the BBC Technology department, was appointed to project manage the research, integration and hand-over of an IP-based network to the BBC. In parallel, Energis (now part of Cable & Wireless) was chosen as network provider for the project.
Both the BBC Technology department and Energis had previous experience working with Broadcom Ltd., the U.K. distributor of Belfast, Northern Ireland-based APT, a manufacturer of audio codec and multiplexing solutions for both IP and synchronous networks.
Because Broadcom had previously conducted various IP audio tests with Energis, it was a natural outcome for Broadcom to be one of those consulted about suitable equipment for use with this large-scale project. Its recommendation for the terminal equipment was the recently launched APT WorldNet Oslo codec. According to APT, this 3 rackunit modular codec was developed specifically to meet the needs of both broadcasters and telecommunications providers, as well as to ensure both redundancy and long-term reliability.
Designed to be transport stream agnostic, the WorldNet Oslo has both E1 and IP interfaces. The latest version of the Oslo’s IP transport card offers increased processing power, plus dual Ethernet ports and the ability to address fully the BBC project’s fundamental requirements in terms of latency and multiple routing support.
BBC Scotland’s Pacific Quay Home in Glasgow, Scotland In terms of setup and control, the codec interfaces with PC-based software, either Codec Management Software (CMS) or, for larger and mixed codec environments, Network Management Software (NMS). CMS/NMS is used to setup and monitor the performance of individual units, but neither is required in relation to maintaining ongoing operational performance.
Remote operators can access the WorldNet Oslo through the CMS/NMS, with different levels of control and reporting including SNMP and Alarm status monitoring. The final piece for this project was the selection and integration of a proven IP router. Siemens selected a Foundry Networks switch to interface among the various IP circuits and codecs.
Although a geographically small country, the United Kingdom is, in fact, a union of four constituent nations: Scotland, Northern Ireland, Wales and England. The BBC, as the national public broadcaster, operates within each of these, providing some services for the whole of the United Kingdom and others for the individual constituent home nations.
After a series of laboratory and controlled test evaluations, it was decided to use Scotland as the location for the first large-scale pilot project.
This project was based around a 100 Mbps Ethernet ring network linking three of the main cities in Scotland — Glasgow, Aberdeen and Inverness. Coming off the ring network were a number of 10 Mbps spurs to connect with the Highlands and Islands and other lesser populated areas in Scotland.
At each of the main sites, the primary requirements were space efficiency, power resilience and the ability to feed multiple smaller sites.
According to Greg Massey, APT chief technical officer, a key objective was that: "The audio performance and network performance needed to look and feel just like the old synchronous infrastructure, but with all the cost and efficiency benefits of moving to a new IP based network."
Assessing the requirements of the pilot project prior to roll-out, it was concluded that the WorldNet Oslo was perhaps somewhat over-engineered for some of the smaller sites involved. Typically, these smaller sites only need to interface with a single stereo signal and the WorldNet Oslo, with its ability to deliver up to 24 discrete audio channels, was not cost-effective for such installations.
After due consideration, Siemens opted for alternative APT product, the WorldCast Horizon, a stereo input/output IP audio codec designed to service less onerous audio density requirements. For added resilience, at least two stereo units were placed at each location — three if the site could not be accessed within 24 hours.
A fundamental driver behind the selection of the WorldCast Horizon was that this unit can also be controlled remotely and that an operator is able to view both WorldNet Oslo and WorldCast Horizon codecs through the same NMS software suite.
By early 2008, all the various circuits were commissioned across Scotland and the WorldNet Oslo and WorldCast Horizon units were installed along with the required Foundry Networks switches. After several weeks of testing and commissioning, during which the networks were aligned and performance was carefully monitored and with the overall IP deployment running in parallel with the existing synchronous networks, the decision was made to go live.
As previously mentioned, broadcasters have generally been reluctant to migrate key program links away from the safety of a synchronous circuit to the vagaries of "packetized" IP networks. The key concerns behind such reluctance tend to be:
- Data over IP does not have a fixed latency;
- Data over IP does not necessarily have low latency;
- Without a central clocking system, every link is independent and therefore variable.
However, several technical solutions can be implemented to overcome such fundamental difficulties. Specifically, the following criteria needed to be satisfied:
- • Installation of an MPLS (MultiProtocol Label Switching) network;
- • Allocation of priority of service to packets on the network;
- • Reduction of the jitter buffer to an absolute minimum;
- • Shaping of the packet size for optimum efficiency.
- • Generation of a clock from the information contained in the packet to ensure a consistent latency with minimal drift and optimal settling time on boot up or reconnection.
For the BBC project, the MPLS requirement was dealt with by Cable & Wireless and the packet prioritization requirements were resolved by the use of the Foundry Networks switches.
With both these requirements in place it was then possible to ensure the jitter buffer in the WorldNet codecs could be set to a nominally low figure, 20 milliseconds.
The WorldNet Oslo codecs, as the master units employed at the major nodes of the network, were then able to successfully address the issue of packet efficiency, as well as ensure consistent and repeatable latency across the various links involved.
Packet size and efficiency were aided by the use of the Enhanced apt-X compression algorithm used by broadcasters for many years. This algorithm offers well-proven audio response figures and word depths that support 16-, 20- and 24-bit audio.
In addition, it has several key features that make it particularly suitable for use in conjunction with audio over IP delivery:
- • Latency as low as 1.9 milliseconds for encode/decode processing time (45+45 samples at 48 kHz);
- • Seamless and glitch-free start-up or reconnection in the event of network drop out, as a result of being ADPCM based with a built-in word identifier (auto sync feature);
- • High-quality audio algorithm, arguably in practice similar to linear PCM while requiring only a fourth of the bit rate, i.e. 576 kbps for a stereo signal sampled at 48 kHz with 24-bit word depth (offering a maximum theoretical dynamic range of 144 dB).
Defining the size of the audio packet was a key design decision because it has an impact on all the other characteristics of the system as each packet traverses the network.
Having been given an estimate of the packet size that best suited the characteristics of the Cable & Wireless network, the next phase of the project was to optimize the codec’s ability to manage the audio over IP characteristics.
The key requirements of the system in operation were to ensure that it:
- • Recovers quickly and effectively;
- • Provides consistent latency;
- • Minimizes the effect of network jitter; and
- • Minimizes end to end latency.
Because some of the locations involved were somewhat isolated, with less than 100 percent reliable power availability, the ability to return to a stable condition after a power outage was an important requirement for the equipment employed.
Information relating to the characteristics of any network can be gathered easily over a long period of time and adjustments made accordingly, but, logically, over a shorter period of time, much less information can be gathered and assessed.
For this particular project, the aim was to ensure the audio link should always return to a stable state in less than 30 seconds. This objective was achieved through the ability to derive synchronization information from the packets arriving at WorldNet Oslo and WorldCast Horizon units and thus predict quickly and effectively the behavior of network.
Consistent latency throughout the network was always a primary benefit of synchronous audio networking, and it was a major concern that this feature could be lost when migrating to an asynchronous IP-based architecture.
To ensure maximum consistency requires the careful management and monitoring of network characteristics and the use of the data collated to configure the audio distribution system for best performance on a real-time basis.
Similarly, ensuring minimal network jitter is related to the effective assessment of network characteristics over time and the use of that information to control the operation of each audio codec in the network.
Correctly implemented, the damping effect the use of this information has on jitter in the network results in stable and predictable audio reproduction system across an IP network.
The use of Enhanced apt-X as the audio compression algorithm provides a predictable and ultra-low encode/decode delay. The clock management provides a very stable transport system together with equally stable transport latency across the MPLS network.
Also, packet size needed to be set as small as possible to minimize packetization delay. The processing of 24-plus audio routes with a high rate of small packets would undoubtedly put pressure on the whole system, so APT invested heavily in the WorldNet Oslo to ensure a very powerful processing platform with careful optimization and prioritization of the packetization system.
This has lead to a very low end-to-end latency with minimal variations, even with more than 24 routes active simultaneously, meeting customer expectations.
Careful management of the network characteristics and the inclusion of this information in the control and processing systems of the audio codecs can deliver a very stable and predictable audio over IP system. However, in all-IP-based systems a reference point is required leading to a master-slave type relationship where clock drift information is derived from a common point.
In creating the BBC network, a solution was sought to an "any point to any point" audio distribution problem where audio placed into the network could exit at any point without carrying additional information with it. The presence of a common AES house clock at each point in the network provided a common reference point that could be used to frequency lock all units together, removing the need for clock management.
By locking the WorldNet Oslo to an AES house clock, it was then possible to effectively synchronize the IP network. Stability in the system was derived from the use of a hardware solution to accurately track any deviation/alteration in the house clock and feed that information directly to audio codecs, helping ensure reliable and stable audio transport from any point to any point in the network.
A further advantage of this approach was that it had the effect of being able to connect regions across the United Kingdom on an ad-hoc basis without the need for user intervention or excessive reconfiguration of the underlying network or codecs. In addition, it meant that it was also possible to further reduce remaining receive buffer levels.
Bringing together all aspects of the MPLS network, traffic shaping, jitter buffer, Enhanced apt-X algorithm and the processing power and functionality of the WorldNet Oslo master codecs, the total end to end latency was less than 50 milliseconds.
By using this system, built around APT WorldNet codecs and the Enhanced apt-X algorithm, the BBC was able to migrate from its traditional E1-based solution to an IP-based alternative.
This migration means the corporation can now take advantage of the various benefits, cost savings and flexibility associated with packetized networks, while at the same time ensuring that its core concerns relating to the reliability of program delivery and the minimization of path latency were properly addressed by this fundamental architecture redesign.
The BBC is one of the first high-profile public-service broadcasters to take this course of action, helped by the project management of Siemens and facilitated by Cable & Wireless. Following the successful deployment in Scotland, these companies are currently progressing with further installations in Wales, Northern Ireland and the English Regions.
Greg Massey said that, in his view, "The end product is a fantastic success as it brings together a lot of new technology into one of the most demanding markets."