Having agreed the first specifications for 3DTV broadcasts the DVB has begun thinking about further phases of standardisation more applicable to the needs of public service broadcasters and that of multiview autostereoscopic displays, reports Adrian Pennington.
As more broadcasters gear up to launch 3DTV channels the DVB Project has responded by agreeing the commercial requirements for the format. The group is also preparing the technical specification itself and according to DVB 3DTV Chair David Wood, “this is on the home straight.”
While the group has opted for the Frame Compatible Plano-Stereoscopic System, which is already operational by Sky, Orange, ESPN, DirecTV (and will be by Canal+ when it launches) there could be room for another set of standards within the DVB Project which meet other requirements, notably those of PSBs.
“Other DVB members have expressed the potential need for a set of standards that are appropriate to a different set of commercial requirements,” states its document. “These commercial requirements are in the process of being discussed and agreed.”
The Frame Compatible system arrays left and right images in a ‘spatial multiplex’ compressed variously side-by-side, chequerboard or top/bottom (MPEG-2 in the US, MPEG-4 AVC in Europe) into one HD picture for broadcast and requires that the customer only invest in a 3D-ready TV to decode the images on reception.
In the longer term the industry would prefer to move to a Service Compatible format based on MPEG MVC in which the left eye is broadcast as a standard MPEG AVC picture and the data for the right eye is derived using the left as reference (or 2D+Delta).
“Frame compatible can address existing HD STBs and TVs and is a very quick means for broadcasters to deliver 3D to the consumer,” says Simon Gauntlett, DTG technology director (pictured). “The downside is that because each image is compressed by half of the horizontal resolution you are getting a lower resolution of 3D.”
The side-by-side version favoured by Sky cannot be viewed on a 2D set and requires broadcasters to operate two separate services (for 2D and 3D), which for Sky necessitates two satellite transponders. A Service Compatible system on the other hand would require new decoders within the STB but mean that that the right eye data would be ignored on transmission to a 2D set.
“Colour television would not have got off the ground if the picture wasn’t able to be viewed on legacy black and white sets, and it’s a similar situation here,” observes Bill Foster, senior technology consultant at Futuresource Consulting.
The MPEG AVC delta signal still incurs an overhead, currently anything from 30% to 70%, with the variation largely dependent on the type of content.
“With Frame Compatible methods, if you have a 3D and an HD service you would need 8Mbps for each of the services whereas with the 2D+Delta model you’d require 8Mbps plus 3-4Mbps in extra capacity so it is more spectrally efficient,” says Gauntlett.
“Broadcasters need to service new 3D and legacy 2D viewers, therefore, saving costs by reducing the bandwidth needed is one of the key factors,” agrees Manuel Gutierrez Novelo, CEO, TD Vision which claims to have invented and patented (in 2003) the 2D+Delta method and has its IP incorporated into the MVC standard. “Instead of using 200% (100% to service 2D legacy and 100% to send a frame-compatible 3D viewed by a small part of the population) we optimise the 2D+Delta to service both with around 130%-140% bandwidth.”
Broadcasters are betting, not without reason, that the amount of overhead will reduce significantly over time as compression technologies improve. MPEG-2 SD for instance was introduced at 15Mbps, but can be below 2Mbps these days.
“MPEG MVC is the standard used by Blu-ray and is a strong contender for broadcasters because it’s apolitical,” says Foster. “There would still be a license fee but it won’t be a proprietary system with which broadcasters are uncomfortable.”
However there are still a number of questions as to how widely adopted 2D+Delta will become. According to Matthew Goldman, Head of Compression Technology, Solution Area TV, Ericsson, “it is favoured by bandwidth-restricted networks such as DTT to overcome the inability of these networks to support separate 2D and 3D versions of the same service but there is currently no industry-wide agreement on how to provide the ‘Delta’. This is a further obstacle in DTT where public broadcasters cannot fully control the STB population.”
To optimise 2D+Delta for transmission it would be useful to have MPEG MVC adopted. However, what is less clear is how large the demand for 2D plus difference is going to be.
“In both contribution and transmission, the jury is still out,” says Mark Cronin, Technology Director, Arqiva Satellite & Media. “It may be that the developments in L+R mean that 2D plus difference has less advantage than was originally hoped for.”
While the DVB recognises the main MPEG candidate for a 2D Service Compatible signal to be MVC, there is an MPEG call for proposals for a more efficient system. “One of the issues for DVB members could be whether to run with MVC, like Blu-ray, or wait, however many years, for a more efficient system,” reports Wood.
However the move to full 3D channels rather than isolated events based on MPEG MVC would require a massive upgrade to infrastructure beginning with contribution circuits but eventually encompassing the broadcast centre. “It will be a challenge to manage production of a whole channel including continuity and commercials remotely from an OB truck where most 3D events are currently produced,” notes Foster. “You’d want the incoming left and right feeds to be ingested within a stereo-capable broadcast HQ rather than created side by side at the point of capture.”
The extension of this is that MVC – or multiview video coding – can be ramped up to accommodate more than today’s stereoscopic 3DTV’s two views. This will come in particularly handy when auto-stereoscopic TVs go mainstream and MVC could potentially be required to handle fifteen or more simultaneous views – something it is in theory capable of (the first iteration will use the stereo profile).
Realtime MVC encoders
“Quite how content will be created to cover multiple views is something no-one’s really started to plan for,” Foster observes. A more immediate concern for broadcasters considering a 3DTV launch based on MPEG MVC is that there are, to the best of TVBEurope’s knowledge, no commercially available realtime MVC encoders.
TD Vision is perhaps closest, debuting an nVidia PC-based realtime encoding platform at NAB2010 and launching the TDV 3D Quantum Licensing Early Adoption Programme along with Magnum Semiconductors last January to provide “a reference design for a broadcast quality realtime stereoscopic encoder,” says Novelo. “It delivers an amazing 1920×[email protected] per eye using 2D+Delta in around 12-16Mbps, a format that is not even currently supported by 3D Blu-ray discs.”
Harmonic says it already has multichannel HD capability within the Electra 8000 which could potentially cater for dual channel or MVC encoded 3D content and that it is listening to customer requirements to understand how it should develop this capability.
“VoD delivery can extend the prospects for frame compatible mode through the use of 1080p 50/60, provided an HDMI connection can support the increased clock rate,” says Ian Trow, the company’s director of broadcast solutions. “This overcomes the realtime bandwidth overhead associated with broadcast services. However, whilst it’s relatively simple to incorporate 3D into VOD, which works well for 3D movies, it doesn’t address live sports.
“When we look beyond the frame compatible mode this raises the issue of broadcasters’ desire to improve quality and address 2D/3D compatibility and MVC can address these needs,” says Trow. “But this requires significant standardisation work and broadcaster investment at both the head-end and STB to make a reality.
Another crucial issue is the acceptance of 3D shot material in a 2D environment since the two production techniques are very different. The prospect of achieving greater compression efficiency through combining services has previously cropped up in SVC encoding within MPEG-2, allowing a base layer to be built on through additional enhancement layers.
“A typical scenario would be a SD base layer that is enhanced to produce an HD channel,” explains Trow. “While such systems can achieve modest compression gains, the increased system complexity and the restrictions in terms of frame rate and compression standards between the base and enhancement layer have deterred broadcasters from adopting such a strategy to date.”
Standards in the works
Dovetailing with the work of the DVB is SMPTE’s development of a file format needed for producing and transporting 3DTV in the studio. This will probably need to include all Frame Compatible and other formats.
“Meanwhile the HDMI consortium has agreed formats for inputs to 3D displays with the introduction of HDMI v1.4a, and the [email protected] consortium is seeking to get display manufacturers to agree to an interoperable shutter glass signalling system,” says Wood.
An alternative and probably faster route to market are ‘universal’ glasses, like universal remotes, that work with any brand of TV. XpanD is about to launch these. Telecommunications standardisation body the ITU-T has also been busy. Its classification system contains several ‘Generations’ of 3D, and four levels in the first generation on which it will initially focus.
“The four ‘Generations’ are four new technologies, which may emerge as a result of research at intervals of possibly ten years,” explains Wood. “Needless to say, this is a guess, but you have to start somewhere. The first generation will be the two channel world of left and right signals, used with a display and glasses. The second generation will be the multiview environment with auto-stereoscopic displays. The third may involve multiview in both the horizontal and vertical directions. The fourth may involve recording a continuous object wave passing through a given area.”
With that we’re into the realm of volumetric or holographic displays—and that’s a whole different ball game.