A Report from the WorldDMB Distribution Seminar which took place 14 May 2013, Nijmegen, Netherlands.
When deploying a DAB/DAB+ transmitter network, it is important to consider the distribution from the headend to the transmitters.
The Ensemble Transport Interface (ETI) standard was originally used for distribution on synchronous lines, but in 2009 the adoption of the Encapsulation of DAB Interfaces (EDI) standard made it feasible to distribute via IP.
Services encoded as ETI on synchronous E1 distribution lines work flawlessly if one has full control over the physical network, as is often the case with managed fibre or last-mile copper connections. However, even very short dropouts along a distribution chain can cause the affected transmitters to mute their signals, and this may initiate a recovery cycle that lasts several seconds. Such dropouts may occur e.g. on microwave hops, or when the distribution is subject to arbitrary switching by telecom providers masquerading standard data lines as synchronous E1.
The resulting transmitter errors may seem random and inexplicable, but in fact they are not and can be avoided.
The EDI standard ETSI TS 102 821 allows the encapsulation of ETI or Service Transport Interface frames (STI-D) into EDI packets.
EDI is based on the Distribution and Communication Protocol (DCP). The EDI standard is thus licence free, robust due to Forward Error Correction (FEC) as defined by DCP, and has additional resend functionality. EDI is both uni/bidirectional and uni/multicasting with transport via IP and even file storage.
The required bitrate can be much lower than with ETI, and packet spreading allows both smoothing of network traffic and extra protection against temporary congestions.
Frequent network errors will give EDI a significant increase in overhead from resends.
Resends require more bandwidth, and buffer time (end-to-end delay) is also increased depending on the delay in the network. However, interleaving with packet spreading over several 24 ms DAB frames does not increase network load even at fairly high packet loss rates, and strengthening FEC in this situation may be counter-productive by comparison as it implicitly increases the required bitrate.
The optimum choice between uni- and multicasting is largely based on the number of peer units.
Unicasting works fine for small numbers, but does not scale well as it imposes high traffic on networks close to the data source. Unicasting also places a high load on the data source itself due to the multiple data streams.
Multicasting has a lower overhead, requires less computing power in the encoder (the FEC calculation alone is demanding even on modern processors), and is easier to monitor. Multicasting also allows data receivers to be dynamically added or removed.
In addition, multicast traffic can usually be tunnelled through unicast-only networks.
Whilst routinely used for contribution (see EBU Tech document 3326) the adoption of IP schemes such as EDI for distribution has been slow, but recent embedded support for EDI in transmitters should speed that up.
One major user of IP distribution for DAB is Commercial Radio Australia and schemes such as EDI are also successfully employed in Europe, even in combination with plain old ETI.
The technical issues reported with EDI are few and far between, but running on dedicated networks is still recommended.
Note: The full report and all presentations including a comprehensive EDI introduction by Markus Prosch (Fraunhofer IIS) are available for WorldDMB members in the Technical Committee area of the WorldDMB website.