Content for TS 26.346 Word version: 19.1.0
G Guidelines for channel tune-in and switch time reduction
G.1 Interleaving for tune-in time reduction of FEC protected MBMS Services
G.2 FEC stream and channel bundling
...
...
G Guidelines for channel tune-in and switch time reduction |R8| p. 229
This Annex describes some methods to improve channel tune-in and switching times for MBMS streaming when using the FEC framework. It presents means for MBMS senders and receivers to minimize delay from a channel switch and initial tune-in time. Fast startup improvements for the layers below IP are not described here.
G.1 Interleaving for tune-in time reduction of FEC protected MBMS Services p. 229
Interleaving may be applied to source blocks before FEC encoding to re-arrange the order of transmission of the UDP packets. The target of the interleaving is to provide high priority and correctly decodable media units in a way that maximizes the resulting media duration at the receivers that tune in at that specific source block. The interleaver may operate at two different levels:
- Inter-stream interleaving: prioritize the media streams and arrange their transmission order according to the ascending order of priorities. In other words high priority data, e.g. audio data, is transmitted towards the end of the source block.
- Intra-stream interleaving: high priority media data units such as Random Access Points of a video stream are transmitted towards the end of a source block.
G.1.1 Timestamp offsets p. 229
The timestamp offset field may be used to signal a timestamp offset for the received media units in the FEC block as specified in clause 8.2.3.2.
The timestamp offsets may be used to reconstruct the presentation timeline at the UE. They may also be used to reduce the out time caused by the reception of a partially received interleaved FEC source block as shown in Figure G.1.
G.1.2 Early playout p. 230
A UE that desires to make use of the interleaving to reduce the tune-in time may start the playout earlier than dictated by the min-buffer-time value.
Instead, the UE may schedule the first media unit of the succeeding source block to be played out after the min-buffer-time. It may then estimate the appropriate time for starting early playout based on the amount of media duration that was received from the current block, the playout time of the earliest media unit of the next source block, and the highest presentation time of the media units of the current source block.
The early playout behaviour is depicted by the following Figure.
Figure G.2: Early play out of interleaved media data of a FEC protected MBMS service
(⇒ copy of original 3GPP image)
(⇒ copy of original 3GPP image)
G.2 FEC stream and channel bundling p. 230
G.2.1 Introduction p. 230
FEC Stream bundling is a method of improving the FEC efficiency and also to improve channel switching times. Several flows of one or more user services are "bundled" to form the source blocks for the FEC calculations. This means, that all flows must be received for potential FEC recovery.
If flows from more than one user service are bundled (i.e. channel bundling), then the receiver discards other services after FEC processing. When a switch is performed, media inside the same bundle is immediately available. Stream bundling can be performed on whole channels, parts of channels, or not at all.
G.2.1.1 Full channel bundling (all flows of several MBMS User Services) p. 230
When full bundling is used, more than one complete channel is bundled. A channel switch inside the bundle does not require rebuffering and can therefore be near instantaneous. However, full bundling requires the processing of all packets. In other words, at a single time instance all video streams and all audio streams are received and processed. Also, the number of channels possible is limited by the fixed bearer bandwidth.
G.2.1.2 No bundling p. 231
When no bundling is used, each channel is protected separately. In a channel switch the new channel needs to be buffered for the full min-buffer-time. At a single time instant one video stream and one audio stream is received and processed.
G.2.1.3 Partial channel bundling (some flows of several MBMS User Services) p. 231
Partial bundling may be used to combine some of the benefits of full bundling and no bundling. In this case more than one parital channel is bundled. This may be used to reduce the processing required on the terminal while retaining fast channel switching between parts of the media.
Partial bundling may be used to bundle only the audio part of channels. In other words, at a single time instance one video stream and all audio streams are received. When a switch occurs the audio is switched instantaneously as it does not need to be rebuffered. The complexity of this case is substantially lower than full bundling and the number of channels which can be bundled given a fixed bearer bandwidth is substantially increased. This is due to the fact that the audio uses a substantially lower percentage of the bitrate compared to video.
G.2.1.4 Stream bBundling (all flows of a single MBMS User Service) |R12| p. 231
When stream bundling is used, all flows of a single channel are bundled. The FEC source block is formed using audio, video and security data.
