UDcast recently released a white paper, titled "Open Statistical Multiplexing Architecture for Mobile TV," discussing the principles of the company's Open Statistical Multiplexing Architecture (OSMA) approach to optimizing video content in a DVB-H network.
Broadcasters and operators have long used statistical multiplexing to gain 30 to 40 percent bandwidth efficiency on a given transmission channel. Transmitting video and audio content via DVB-H presents a new challenge, however, when it comes to being able to efficiently use the bandwidth gain achieved by statistical multiplexing. The DVB-H IP encapsulators (IPE) have to adjust their burst parameters to the bandwidth of the incoming stream to optimize the use of the transmission link and take advantage of DVB-H's time-slicing mechanism.
According to the white paper, this issue has been addressed in the past by placing a connection between the encoders and the IPE to transmit information back and forth. It is necessary, however, to use a proprietary communication protocol between the encoders and the IPE, forcing the operator to use a single vendor.
In addition, the IPE and encoder may need to be co-located, which may not be practical for a particular network's architecture. According to the white paper, in a typical commercial DVB-H deployment, "there are potentially several tens of IPEs located in the distribution network just before the transmitters, whereas the encoders are located in one or more places in the IP network." This setup makes it necessary for the statistical multiplexing on the IPE to be "independent of its physical location and of the encoder's manufacturer."
UDcast's proposed solution allows groups or multiple groups of video services encoders on the IP encapsulator side to be allocated a specific amount of bandwidth. Individual services potentially containing multiple IP sessions within this group are defined with a minimum guaranteed and maximum allowed bandwidth. Each encoder group is defined at the IP encapsulation level. In this scheme — called hierarchical multiplexing — it is not necessary for the encoders to be located close to the IPE. In addition, the minimum bandwidth defined for any service cannot be taken away by any other services. At the same time the "maximum bandwidth that is defined for a service will be used as the bandwidth limit of the service, as well as the basis to calculate the boundaries of the burst geometry."
Congestion can be avoided by giving each content session a priority, allowing each session to be sent according to its priority level. Content sessions with lower priority are buffered at the IP encapsulator input and transmitted when more capacity becomes available.
The hierarchical multiplexing scheme is also beneficial when it comes to scenarios requiring filtering of local and regional content at the transmitter level. By combining such content into a single, statistically multiplexed group, the hierarchical multiplexing scheme leads to less "non-deterministic behavior" when using a DVB-H local content adaptor such as the UDcast iSplicer. The regional or local content groups can then be encoded in constant bit rate (CBR) or different groups of regional content can be kept in separate statistical multiplexing groups, making it possible to keep the fully deterministic behavior in the case of DVB-H content filtering.
To ensure bandwidth efficiency, it's necessary to arrange the encoders to communicate with one another to be able to commonly commit to a maximum usable bandwidth. This can be implemented in whatever scheme an operator finds most useful for his network architecture. Typically, a master encoder will direct the communication between the encoders.
For more information, visit www.udcast.com/products/downloads/WP_OSMA.pdf.