Cellular-based bonded uplink is becoming a frequent sight in sports venues. By using portable devices that bond together multiple 3G, 4G and other data connections, broadcasters can now quickly transmit live video from anywhere at lower costs than traditional uplink solutions, as well as transmit on the move or inside structures without line-of-sight or cabling restrictions. This has led to a variety of applications, such as transmitting a much greater number of live varsity games, sports news hits from the road, complete live press conferences and previously impossible live footage from inside locker rooms, golf courses, race tracks and more.
However, while allowing greater flexibility and lower costs, it is also important to understand limitations of transmitting over public cellular networks, and to use the technology in ways that produce the best results.
How it works
Bonded cellular video technology is an aggregation technique designed to group several physical links into one logical link. This provides fault-tolerance and high-speed links between the aggregation device and its paired bonding server.
The technology enables the use of several links (from two to 32) and combines them to create increased bandwidth and the data/video trunk. Should links fail, the bonding technology automatically redistributes traffic across remaining links. Automatic redistribution is done in typically less than 100ms, so no outage is noticed. In practice, a number of separate cellular or other data connections — which, individually, may not be very robust — are combined into a bigger pipe. This pipe can transmit despite any temporary individual lost connections to the cell tower or lost packets on the way to the server, for example. Data connections can include 3G and 4G modems across most major U.S. cellular carriers, as well as Wi-Fi, WiMAX, Ethernet and even portable satellite links.
The bonding device encodes video coming from a camera or a field switcher, dices it into multiple data packets and transmits each group of packets through a different modem or carrier. Information travels through cellular towers and the public Internet, and arrives in a receiver that recombines data into raw video again and then outputs it to a studio switcher (for on-air) or transcodes and sends it to a content delivery network (for online).
Typical bit rates
Transmission rates can vary dramatically based on location, time of day and number of people in a certain area. In good 4G areas, transmission rates may be as high as 10Mb/s to 12Mb/s, compared to an average of 1Mb/s to 3Mb/s in some 3G areas. Of course, it’s impossible to predict available bandwidth at a govem to,e. Therefore, another job of the bonding solution is to characterize each link every few milliseconds in terms of bandwidth, delay and loss rate. Then, it creates a Gaussian normal distribution graph for each parameter, assigns a real-time weight based on the graphs, and feeds this weight to the scheduling/queuing process. This allows for the best real-time decision regarding which link the next incoming packet should use. Typical bonding techniques use the following methods to overcome fluctuations:
- Adaptive bit rate: By changing the encoding rate on the fly, the bonding unit can maintain transmission even as bandwidth swings.
- Signal boost: The bonding unit may use special antennas that increase range compared to consumer air cards, as well as overcome multi-path fading in environments where signals bounce around different obstacles (walls, beams, etc.).
- Forward error correction (FEC): The unit will send redundant data along with the main feed, so that if some of the original packets are lost, the decoder can still recover the information. The Reed-Solomon FEC algorithm takes a predefined block of packets (FEC group) and calculates several augmented FEC packets. If, for example, the FEC group is composed of 20 packets, and we add four FEC overhead packets, any lost packet, up to four, can be recovered.
Latency control: The higher the latency is, the more likely all data packets will arrive in time. Users can select a low latency (as low as sub-second end-to-end), which carries a greater risk of occasional packet loss. Or, a higher latency (anywhere from 2 to 60 seconds, or even store-and-forward) can be chosen, which provides more time to recover lost data.
End-to-end latency has a great effect on bonded link quality. If a user selects a low end-to-end delay, say one second, then considering encoding and decoding time, and SW processing time (approximately 500ms), the remaining time for a packet left to arrive from the transmitter to the receiver may be as low as 500ms.
- It’s up to users to decide which parameters to use, based on the content transmitted. On-the-move coverage, where changes are fast and individual link properties vary, a higher end-to-end delay profile will enable better recovery from changes without data loss. But, in areas with good cellular coverage, when the camera is static, a lower delay profile can work fine.
- Changing resolutions: As resolution increases, the bit rate also needs to increase. By lowering the output resolution, one can get by with lower bandwidth in congested locations. The bandwidth control engine detects any state where the outgoing bandwidth is lower than the encoding bandwidth. It will then decide whether to let the encoder use a higher Quantization Parameter (which lowers the encoder bandwidth), or perform a video scaling (feeding the encoder with a lower-resolution image) in order to lower the encoding bandwidth.
A good bonding engine will receive from the user general preferences (low delay, best quality, etc.). Based on these, it will automatically adjust bonding parameters. This means the engine will decide on a packet-by-packet basis how much FEC overhead to use, the optimal encoding rate, what resolution to use at any given time, what frame rate to use, what packets to drop under extreme low BW cases, what packet lost recovery mechanism to use (retransmit, FEC, error concealment, bad frame skip), and how to promptly compensate the overall system knobs upon any network behavior changes that affects the uplink capabilities.
Some of the most frequent recent uses for sports coverage include:
- Live ENG: Channels have utilized cellular bonding to transmit live from road games, press conferences and events like the NFL’s Super Bowl, NBA All-Star Weekend, the Olympics and the World Cup. A reporter or photographer can set up a live on-air shot from a single camera and send footage back while covering a team or a competition. All of this is done without trucks or hardwired connections.
- Upload edited packages: Reporters can upload edited packages from the road without relying on slow hotel Internet connections or individual mifi cards.
- Live behind-the-scenes camera: Broadcasters or franchises can set up exclusive behind-the-scenes web shows while walking around locker rooms, stadium suites, etc. Similarly, rather than only clipping sound bites, many stations’ and franchise’s websites stream coach and player press conferences for the benefit of devoted fans.
- Live varsity games and specialty sports: Thanks to the lower cost of cellular transmission, broadcasters, schools, new media publishers and specialty sports groups can transmit full games and competitions to the web that traditionally would not be shown on-air. Many times, these productions vary from one camera to a fully switched production fed to the bonded unit.
- In motion, in the field: Freed from line-of-sight restrictions, cables or trucks, broadcasters are using bonded backpacks to transmit races from onboard moving cars, boats and helicopters without requiring RF solutions.
Traditional uplink solutions provide guaranteed bandwidth. Yet, these can be costly and require line-of-sight, cabling or fixed locations. Cellular bonded uplink can vary in transmission rates, but it opens a variety of sports live coverage opportunities from new locations and on new platforms by understanding needed parameters. Networks, stations, new media publishers, leagues and teams can optimize the use of cellular bonding and create a breadth of new content offerings that were once too expensive or complex to produce.
—Ken Zamkow is director of sales and marketing, LiveU. Meir Schrieber is vice president, research and development, LiveU.