This Annex provides guidelines for MBMS senders to minimize initial buffering delay between starting of the reception and starting of rendering of media data in MBMS receivers.
When H.264 (AVC) video is in use, an MBMS sender should form FEC source blocks in which the first H.264 (AVC) access unit in decoding order is an IDR access unit. When H.265 (HEVC) video is in use, an MBMS sender should form FEC source blocks in which the first H.265 (HEVC) access unit in decoding order is an IRAP access unit.
MBMS senders should transmit all application data units for a given H.264 (AVC) or H.265 (HEVC) access unit, or audio frame within one FEC source block.
MBMS senders should set the min-buffer-time MIME/SDP parameter and the minimum buffering delay elements included in FEC source blocks to values that are sufficient to cover any required de-interleaving or de-packetization of application data units, such as H.264 (AVC) or H.265 (HEVC) NAL units and coded audio frames, from their transmission order to decoding order.
When RTP timestamps are converted to the wallclock time of the MBMS receiver, the smallest RTP timestamp among the FEC source packets of a FEC source block of a stream should be equal or close to the smallest RTP timestamp among the FEC source packets of a FEC source block of any other stream of the same MBMS streaming session.
When RTP timestamps are converted to the wallclock time of the MBMS receiver, the greatest RTP timestamp among the FEC source packets of a FEC source block of a stream should be equal or close to the greatest RTP timestamp among the FEC source packets of a FEC source block of any other stream of the same MBMS streaming session.
In hybrid streaming delivery scenarios it is assumed that a set of streaming services is available over MBMS, and a set of streaming services is available over PSS. MBMS services may not be available in some service areas, in which cases those services might be alternatively provided via PSS.
Different switching cases between PSS and MBMS and vice versa may occur in these scenarios, including user-initiated content switching with access change as well as application-initiated access change. The latter can occur when a service is available over MBMS in some service areas but not available over MBMS in other service areas, assuming that the service is provided over PSS as a fall-back.
This Annex describes some methods to improve switching times in such hybrid streaming delivery scenarios.
For switching between MBMS and PSS access while receiving the same service, comparing the SSRC values of the PSS and MBMS flows gives the UE an advantage for synchronization onto the flows as described in
clause 8.5.
In order to allow for an as interruption-free access switch as possible, the time offset between MBMS packet reception and PSS packet reception (including MBMS FEC buffering and PSS retransmission delays) should be as small as possible.
It is advantageous (but not always possible) that the same set of codecs is used for representing the same service over both accesses. If the same media bit rates are provided over both accesses, then at best the same media flows (identical media encoding) are used for both accesses. This gives the UE an advantage of being able to continuously using decoding buffers thus simplifying seamless access switching. In case the codecs or codec level/profiles change upon access switching, the UE needs to prepare a new set of decoders and possibly having them run in parallel for a short period of time.
Switching from MBMS to PSS can happen in two cases, user-initiated content switch with access change and application-initiated access change.
-
User-initiated content switch with access change: The user requests reception of a service other than the one currently received over MBMS, and the new service is only available over PSS.
-
Application-initiated switch of access: The MBMS coverage is lost, and alternative reception of the same service is possible over PSS.
The latter case may occur suddenly and without warning when the available signal strength is no longer good enough for the MBMS reception. The UE may identify the loss of MBMS access by signal strength measurements, detection of packet losses exceeding a certain threshold, or that RTP reception has stopped completely. As a result, the UE may terminate the MBMS session and initiate service reception over PSS instead.
In case MBMS coverage is lost and the corresponding PSS session is requested, the UE may use PSS time-shifting (as defined in
TS 26.234) for signalling a PSS play-out start corresponding to the time instant the MBMS access was lost, giving the advantage of synchronization of the received flows.
Switching from PSS to MBMS can happen in two cases, user-initiated content switch with access change and application-initiated access change.
-
User-initiated content switch with access change: The user requests reception of a service other than the one currently received over PSS, and the new service is available over MBMS.
-
Application-initiated switch of access: The UE recognizes that MBMS reception of that service is alternatively possible (e.g. because MBMS reception was re-gained), such that the UE may terminate the PSS session and initiate reception over MBMS.
If the UE can receive both PSS and MBMS flows at the same time, and the time offset between MBMS packet reception and PSS packet reception (including MBMS FEC buffering and PSS retransmission delays) is small, and identical media encoding is used, then application-imitated access change from PSS to MBMS without service change is possible to be made in such a way that it is seamless.