Ad Insertion within a statistical multiplexing pool: Monetizing your content with no compromise on picture quality Pascal Jezequel, May 2013 Operators or broadcasters can increase their ad revenue by specifically targeting national and regional ads, while offering local news and programs. However, bandwidth and cost constraints should also be taken into account. Thomson Video Networks describes an ad insertion solution preserving the best video picture quality thanks to its statistical multiplexing system, while also achieving bandwidth optimization. www.thomson-networks.com
1. AD INSERTION... 3 2. CHALLENGE OF AD INSERTION WITHIN A STATISTICAL MULTIPLEXING POOL... 4 3. FLEXSPLICE FOR AD INSERTION... 6 4. CONCLUSION... 11 2
1. AD INSERTION Television delivery has undergone rapid change, with new networks, new ways of accessing television content, and more local channels. Advertising is still important for financing television and advertisers are seeking to get closer to end viewers. For digital pay-tv operators, one solution for earning additional revenue and better monetizing content consists of replacing original advertisements with more targeted ones in live channels. This is known as ad insertion. For many years, some operators, particularly in cable TV, have used ad insertion mainly on CBR streams, but today, all digital pay-tv operators for cable, terrestrial, satellite and IPTV want to personalize ads in their offers to earn extra revenue. In some cases, ad insertion is performed directly in the end user s set-top-box. But in most cases, this is achieved either in the pay-tv central headend or downstream in a local or regional headend. The first case will be referred to as centralized ad insertion and the second one, when an ad is replaced in a local headend, cable hub or common SFN area, as regionalized ad insertion. - Centralized ad insertion can be used in any application (satellite, cable, digital terrestrial TV (DTT), IPTV etc.) where it makes sense to replace ads from incoming services (international turnaround channels received through satellite for instance), as most of the time incoming ads do not fit the pay-tv operator s audience. - Regionalized ad insertion is mainly used, sometimes in combination with regionalization (local breaks), for any system with local headends or several transmission sites, such as DTT, Cable, and IPTV. In this case, ad replacement is carried out as close as possible to the end user to target a specific area. Today, for most digital TV distribution systems, bandwidth optimization is key and more and more efficient statistical multiplexing solutions are used to meet the end user s demand for picture quality while optimizing bandwidth. This white paper focuses on solutions to perform ad insertion for broadcast digital TV in central or local headends where statistical multiplexing is used. 3
2. CHALLENGE OF AD INSERTION WITHIN A STATISTICAL MULTIPLEXING POOL A typical ad insertion system is composed of a splicer and an ad server. The splicer detects cue tones in the incoming stream and replaces the content signaled by these cue tones with a stream from an ad server. Splicing one stream by another in a splicer seamlessly is challenging especially when the stream is part of a statistical multiplexing pool. Seamless Service Splicing: A must have but known to introduce delays Advertisers want their ads to be played with optimum quality. This is the reason why advertisement swapping in a TV channel must be done seamlessly. An advertiser will refuse to pay for an advertisement clip if the transition between the service and ad clip shows artifacts or black screens. A seamless switch between the service and advertisement is achieved with service splicing. Splicing is based on a switchover performed at the video and audio level (elementary streams), offering the benefit of seamless switching for the receiver. Consequently, splicing consists of switching from one program to another with the constraint of preserving stream continuity during program insertion: linear PCR and PTS, same PID and SI/PSI tables, management of the video/audio transition, and T-STD buffer occupancy. Splicing is therefore a relatively complex process. Without modifying the GOP structure of the streams it is not possible to switch from any one image to another (GOP-accurate splicing), or, if frame-accurate splicing is needed, the GOP structure of the streams must be modified (for example by converting a B picture to an I picture). SCTE 35 with I frame insertion ˮnot transmittedˮ period in primary service Ads I Splice out Splice in at the end of the ad playlist Splice to nearest I image in primary service Ad 1 2Ads 3 4 5 6 7 After Splice Ad 1 2Ads 3 4 5 6 7 Consequence: Delay in primary service In this scheme, without specific processing splicing introduces delays for a few frames. 4
This is not an issue on CBR streams but it becomes truly problematic as soon as this occurs on a VBR stream, i.e., on a stream that is part of a statistical multiplexing pool. It simply desynchronizes the entire pool and creates bitrate overheads as explained below. The following example shows a statistical multiplexing pool of seven programs/services. The bitrate of each service varies but the total bitrate of the pool is constant (in this example at 25Mbit/s). Allocation of bitrate for each service is performed based on a per picture policy. If we now assume that in this pool, service 1 needs to be spliced for ad insertion, splicing occurs downstream of bitrate allocation for statistical multiplexing and the splice introduces a delay of a few frames as explained in the previous section. In this case the resulting statistical multiplexing pool will look as follows: Serv #1 before delay The total bitrate of the pool is no longer constant and in certain places exceeds the maximum bitrate 5
allocated to the pool. This creates bitrate overflows on output, resulting in blocks or even black screens on all services (not just service 1). Traditional methods to solve this issue: Traditionally, there are three methods to solve such an issue, but none of them are particularly satisfactory: - Having the services spliced in CBR all the time - Provisioning enough null packets to absorb the overshoots - Splice with transrating / rate shaping to rebuild the statistical multiplexing pool. Services in CBR all the time: Having the services spliced in CBR all the time has two main disadvantages: it reduces the number of services in the statistical multiplexing pool and thus reduces statistical multiplexing efficiency. Moreover, the services in CBR have a variable quality and this can greatly affect the end user s experience. Provisioning enough null packets: Provisioning enough null packets amounts to nothing more than wasting multiplex capacity, and hence money. Splice with transrating / rate shaping to rebuild the statistical multiplexing pool: With transrating, in order to achieve exemplary VBR pool reconstruction, it is necessary to transrate the spliced service along with any other services present in the VBR pool. This has an impact on the overall quality of all the services present in the multiplex as the bitrate of all services needs to be reprocessed. In addition, if too much strain is put on the transrater in order to reach the required bitrate, picture quality could be seriously negatively affected. Furthermore, transrating works only in MPEG-2. H264 requires complete transcoding and this obviously affects picture quality and increases latency. As none of these traditional methods are particularly satisfactory, Thomson Video Networks developed an innovative method to splice within a statistical multiplexing pool: FlexSplice. With FlexSplice, performing ad insertion within a statistical multiplexing pool is no longer an issue. 3. FLEXSPLICE FOR AD INSERTION FlexSplice concept FlexSplice resolves the challenge of performing ad insertion within a statistical multiplexing pool with seamless splicing, but without the drawbacks explained above. As the ads are stored in a server in CBR, FlexSplice requires the service to be in CBR during the splice period only but it remains in VBR continually outside the splice period. This solution does not require any additional trigger but reuses the same existing triggers used for splicing (mainly SCTE based signals) as a way of also controlling statistical multiplexing. This makes it a very efficient solution for performing ad insertion within a VBR pool with the following key advantages: - No compromise on picture quality - No need to transrate or transcode with the inevitable degradation of picture quality on all services in the pool - No need to add extra stuffing packets: null packet rates remain at a low bitrate just as with a standard multiplexer (and in any case below 100 kbps) 6
- Splice and statistical multiplexing control are performed using the same trigger, based on SCTE 35 tables FlexSplice is based on two pillars: Ad insertion architecture with FlexSplice - Transparent & seamless splicing that does not cause any delay between services or any overshoot in the statistical multiplexing pool bitrate - Single trigger to control statistical multiplexing as well as the splice itself FlexSplice: transparent and seamless splicing With FlexSplice, services become completely transparent outside the advertisement slots. This means that there are no bitrate fluctuations or bursts from splicer input to output, and that all services have the same latency. This is based on advanced processing within the splicing engine of Thomson Video Networks advanced multiplexer and splicer, the NetProcessor 9040, and this processing is eased by the use of SCTE 35 messages. The SCTE 35 messages are associated with an I (Intra) image on splicing for efficient and frameaccurate switching to the secondary service (advertisement) and back to the primary one. Consequently, a statistical multiplexing pool will not be affected by crossing the splicer outside of the splicing time. And during the splice period, all non-spliced services will remain in VBR. Since there are no timing changes between the VBR pool services, the pool is not affected in any way. The SCTE 35 messages associated with I images ease the transparent splicing process, but are also used to trigger the service to CBR during the splicing period. FlexSplice: splice triggering and Flextream control The splice and Flextream control are achieved by the same SCTE 35 message. Having a single control to perform both splicing and Flextream control is paramount. With different controls for splice and Flextream, overall workflow management would prove problematic in terms of timing. The switch to CBR must be performed before splicing to the advertisement, and the return to VBR must be performed after splicing back to the primary service. Using the same SCTE 35 tables to control both splicing and Flextream guarantees perfect timing consistency and is totally operator-transparent. 7
SCTE 35 tables are either already present in the incoming streams or generated in the central headend. Generation can be easily performed through SCTE 104 commands in the SDI feed or over IP. At the end of the advertisement slot, to revert to the primary service and VBR, triggering can be either via a SCTE message, or simply the end of the advertisement slot. This provides valuable flexibility for managing overall workflow. This solution naturally works for both centralized and regionalized ad insertion architectures. The next two sections describe how our FlexSplice solution can be implemented in the two architectures using existing Thomson equipment, namely the Sapphire ad server, ViBE EM encoders and NetProcessor 9040 serving as multiplexer, statistical multiplexing (Flextream) bitrate allocator and splicer. Splice triggering and Flextream control for centralized ad insertion For centralized ad insertion, the same NetProcessor uses the SCTE 35 table to trigger the VBR/CBR transition as well as splicing. 8
Splice triggering and Flextream control in regionalized ad insertion For regionalized ad insertion, the NetProcessor in the central headend uses the SCTE 35 table to trigger the VBR/CBR transition and transmit the same table on output. Flextream is then configured to offer a CBR slot to allow different regions to insert advertisements. The SCTE 35 table is used by several downstream local headends to perform the splice, and each one is equipped with a NetProcessor connected to a Sapphire server. 9
FlexSplice: splice triggering and Flextream control compatibility with legacy analog cue tone signaling SCTE 35 or 104 messages are still not yet commonly deployed and legacy DTMF analog cue tones (or any alternative method, such as cue tones carried by VBI signals) are still widely used. It is therefore important to be able to use FlexSplice with legacy signaling systems. A specific feature has been developed within the Thomson Video Networks XMS management system to convert a GPI from a DTMF decoder (or any legacy cue tone decoder generating a GPI) to an SCTE 104 over IP message. The SCTE 104 messages are processed by central headend encoders and converted into SCTE 35 tables to be processed by FlexSplice in order to: - Perform splicing at the right moment. - Ensure an appropriate CBR of the primary service for matching with the advertisement slot. This GPI SCTE 104 SCTE 35 process managed by XMS is fully compatible with n+p encoder redundancy. If the nominal encoder is replaced by a redundant one, the corresponding SCTE 104 message is routed to the right encoder. 10
4. CONCLUSION Based on open standards, FlexSplice offers the best compromise for picture quality thanks to statistical multiplexing preservation, allowing the operator to take advantage of ad insertion while achieving bandwidth optimization. FlexSplice is a unique technology that allows for splicing within a VBR pool without transrating or additional null packets. The whole process is triggered by a single control, based on SCTE 35 tables, and compatibility is ensured with legacy DTMF cue tone signaling. FlexSplice technology does not affect picture quality and allows for seamless and transparent splicing within the VBR pool. In addition to FlexSplice, Thomson Video Networks ad insertion solutions benefit from the advanced Sapphire ad server. The Sapphire manages ad playout according to the playlists and is interfaced with the NetProcessor splicer through SCTE 30. The Sapphire also manages assets with automatic clip provisioning and can be interfaced with traffic systems or ad insertion management systems using various formats and standards. These ad insertion solutions in Flextream VBR pools using FlexSplice, Sapphire and NetProcessor have been deployed by Thomson Video Networks and are on air. CONTACT INFORMATION Please contact your sales representative or visit our website for all contact details. http://www.thomson-networks.com This document provided by DigitalGlue on www.digitalglue.com. Contact us: sales@digitalglue.com Copyright 2013 Thomson Video Networks. All rights reserved. All other trade names referenced are service marks, trademarks, o r registered trademarks of their respective companies. Specifications subject to change without notice. CDT-5188D 11