Compression Advances Alleviate Some of Pressure on SP Bandwidth

Matthew Goldman, SVP, technology, Ericsson

Matthew Goldman, SVP, technology, Ericsson

December 7, 2011 – Lost in much of the discussion about how networks will be able to keep up with the capacity demands imposed by surging volumes of video is the pace of progress toward lowering the bit rates for any given quality of experience as well as new efficiencies associated with multiscreen services.
 
At the not-so-distant cutting edge, coming up in February is a meeting of the Joint Collaborative Team on Video Coding (JCT-VC) that will mark completion of the draft of a new standard jointly developed by the ISO/IEC Moving Picture Experts Group and the ITU-T Video Coding Experts Group with the aim of cutting the bit rates currently required for delivering video via MPEG4 H.264 by half. If all goes as planned, the new standard for what is formally known as High Efficiency Video Coding (HEVC) and sometimes referred to as H.265 will be completed for final ratification by January 2013.

“HEVC is moving ahead rapidly and looks promising,” says Matthew Goldman, senior vice president of technology at Ericsson. “The standard should be technically completed by July 2012, and we should see the first devices for early adopters in 2013. It probably won’t go main stream until 2014 or 2015.”

Meanwhile, significant expansion of processing capacity for encoding on existing standards with both proprietary encoders and software systems running on off-the-shelf processors is rapidly changing the perspective on the costs of migration to multiscreen services. The advances even include new efficiencies in the legacy MPEG2 domain, where vendors are touting products that allow network operators to deliver four HD streams over a single 6 MHz TV channel at quality levels matching what could previously be done with encoders delivering three HDs per 6 MHz.

Cable Bandwidth Allocations in 2012

These developments come none too soon, given the pressure on available bandwidth coming from all directions in the changing video environment. In a recent posting on the Motorola Mobility site, Jeff Walker, director for CMTS product marketing at the company, says cable operators will soon double allocations of bandwidth to DOCSIS traffic, going from this year’s average of four to six 6 MHz channels to eight to 12 next year.

“Downstream traffic is skyrocketing at a rate of roughly 50 percent a year, thanks in large part to streaming video on the Web,” Walker says. “Upstream traffic is only growing at a rate of 20 to 30 percent annually, but that could, and probably will change drastically as consumers adopt new video chat features through social networking platforms.”

But demand for DOCSIS capacity is just part of the story, he adds. “Cable operators are still reallocating spectrum that was formerly used for analog video,” he says, “but it’s not all going over to DOCSIS delivery. Far from it.

“MSOs are using roughly eight channels for narrowcast video-on-demand streaming today, and in 2012 we’re seeing indications that operators will use anywhere from eight to 16 channels for VOD. That’s due in part to growing VOD catalogs, and in part to new experiments with network-based DVR delivery. In short, there’s growth happening in every direction, and it’s all fueled by insatiable consumer demand for entertainment and communication across multiple platforms.”

4:1 HD Channel Ratios

The ability to increase by 33 percent the number of MPEG2 HD streams carried in a TV channel offers an immediate solution for freeing up some of the required bandwidth. At the recent SCTE Cable-Tec Expo in Atlanta encoding vendors ran demos showing programs delivered at these ratios, confirming the stream count could be expanded without any loss in quality compared to what previously could be done with packing three HD streams in a single channel.

“This is very viable; it’s real,” says Yaron Raz, director of digital video solutions at Harmonic. “We have a few operators who are already operating at this channel ratio. The trick is to purposely select channels to ensure the aggregate consumed bandwidth is always within a certain threshold. But, if you do it right, it’s a really simple way to free up a significant amount of spectrum.”

As Raz notes, with the statistical multiplexing technique that allows multiple video streams to share available channel bandwidth in the most efficient way possible, operators need to make sure they don’t pack a group of high-motion video streams requiring bit rates at the upper end of the scale together in one channel. By balancing the mixes of lower and higher bit-rate consuming streams – for example, CNN with ESPN – in a single channel, operators can assure the quality performance remains high even with four HD signals packed together.

“We’re helping with choosing channels with very robust tools that look at the complexity of frame changes over time,” Raz says. “Our Iris platform tells you which combination of channels should be multiplexed to get maximum quality.”

That said, there are operator who will balk at going to four-channel density owing to concerns that quality parameters might not be met at all times on all sizes of HD screens. These typically are the same operators who resisted going to three-HD stream densities, the previous limit for MPEG2 encoders.

Ericsson, too, has leverage improvements in processing capabilities to enable the 4:1 density option but doesn’t expect everyone to go to that level, Goldman says. “With the new encoding capabilities, we can provide these people the level of performance at three-stream density they were getting at two-stream density,” he notes.

“The encoding system needs to be 30 percent better to go from 3:1 to 4:1,” Goldman adds. “We’ve had a step change in processing efficiency so we’re able to operate at unheard of bit rates. If you were satisfied with the previous generation of 3:1 you’ll be happy with 4:1. If you weren’t happy with 3:1 and did 2:1, now you’ll be comfortable with 3:1. Different MSOs will come in with different opinions of what the picture quality should be, but everyone will experience a net gain in available bandwidth.”

The Ericsson demo of the 4:1 system monitored bit rates for each stream, showing how the rates fluctuated on each stream from one second to the next. At any one moment the rate of a given stream might be anywhere from 5 megabits per second to 15 mbps. At the mean point of 10 mbps, these MPEG2 encoders are delivering HD signals at a data rate not much above the early commercial implementations of H.264 encoding but with much higher quality than those first tries on MPEG4.

Next-Gen TV Display Formats

Adding to the growing pressure points on bandwidth cited by Motorola’s Jeff Walker is the emergence of 3D, growing awareness that operator-delivered HD is going to have to match Blu-ray quality on big screen sets and the need to deliver picture quality sufficient to feed super-size screens as service providers and movie studios seek to drive a home theater market for first-run content. “We’re starting to see interest in 1080p/30 and 720p/60,” Raz says, in reference to the progressive resolution levels with frame rates that match the Blu-ray standard. “I think it’s going to start with VOD. It’s harder with linear because of the bandwidth implications.”

Nonetheless, these bandwidth-eating scenarios are coming down the pike fast, which makes the emergence of HEVC a major event. As Goldman notes, if larger screen TVs, such as the 4Ks that will again be on display at this year’s Consumer Electronics Show, take hold in the market, service providers will have to take action to get to higher quality HD. “When you start looking at screens larger than 60 inches, HD as it’s now delivered gets a lot less appealing,” Goldman says.

As reported in September (p. 16), manufacturers are now able to produce screens measuring up to 120 inches, but the current generation of display systems such as LCD and LCD LED fall short of the resolution, brightness, color density and contrast levels that are required to deliver a viewing experience on huge screen that is as good as one a 60-inch screen. And even next-generation laser-based displays are struggling to achieve the required quality levels,

4K, the successor to today’s HD standard, will address some of these limitations, although for 100-inch and larger screens brightness, color and contrast issues will still be in play pending improvements in display technologies. The 4K technology, which delivers four times the resolution of 1920 x 1080 HD with 4,000 pixels per horizontal line, is gaining broader support from manufacturers, including a new 60-inch system from Sharp and a 55-inch system from Toshiba which delivers 3D pictures without the use of glasses.

The Toshiba breakthrough, introduced at a trade show in October, relies on use of a camera under the screen to track viewers’ faces to make adjustments that keep them in the narrow 3D viewing range that is required when glasses aren’t used. The Toshiba imaging system, in contrast to how the usual stereoscopic systems with views for each eye work, directs light in nine directions to create nine parallax images to create the 3D depth effect.

Such platforms are still priced beyond the reach of the average household, although they’re getting into the $10,000 range where upscale early adopters typically start to move the market. The Toshiba Regza 55X3 3DTV set carries a price tag of just $12,000, which is on par with top-of-line early generation 3D TVs that required glasses.

“It will be interesting to see how consumers react to 4K systems,” Goldman says. “If the reaction is anything like it was when HD entered the market, it could have a big impact on services.” And on bandwidth requirements, given the impact a quadrupling of pixel counts will have on bit rates.

Operators are “showing more and more interest in the premium tier,” he adds. “It takes better encoding, and you need aggregate bandwidth allocations for this type of service. The 1080p Blu-ray and high-quality early window stuff being combined with demand for more HD than ever represents a big challenge.”

Hardware vs. Software Encoding Advances

These factors also will have an impact on how operators look at the tradeoffs between hardware-based and software-based encoding systems. Ericsson remains convinced that the combination of high-level quality requirements and the density levels that must be attained to support all the varieties of encoding parameters from super HD down to smartphone screens will assure the ascendancy of hardware systems like Ericsson’s for years to come.

Backing up this contention the company has released a new very high-density transcoding platform, the SPR 1200 Multiscreen Stream Processor, which is designed to be used in conjunction with its new NPR 1200 multiscreen formatting system for adaptive rate segmentation and DRM execution across multiple devices. “

One of the biggest challenges is performing real-time transcoding on linear programming without incurring delays,” Goldman says. “If you do things right you can have scalability with ten times the density that you get with software-based transcoders.”

For example, he adds, if an operator offers just ten linear channels for multiscreen service, each of those channels will need about ten different renditions for various models of smartphones, tablets and HDTVs, and then the system must be able to output multiple bitrates for each stream to accommodate adaptive rate requirements, with the result that those ten linear programs can require support for as many as 500 streams.

Assuming an operator wanted to support ten different end device profiles including HD per streamed channel the SPR1200, consuming just one rack unit of space, could probably handle on the order of six input sources, Goldman says. “The SPR 1200 accesses the highest quality mezzanine encoded video stream and produces whatever number of profiles you need,” he adds. “If you can do all that in a single rack unit you can make multiscreen a basic service.”

But software encoding system supplier Envivio is not about to concede the premium service linear live transcoding market to the hardware-based suppliers. The firm has introduced a live transcoding version of its Muse platform meant to support premium headend performance on HP server blades. The company says an unnamed Tier 1 cable company has already chosen the new platform to support multiscreen services.

The dense, high-quality transcoder meets the most demanding IT and headend requirements, says Arnaud Perrier, vice president of solutions at Envivio. “Operators want to leverage our transcoding capabilities for multiscreen services instead of investing in dedicated hardware,” Perrier says. “We’re the first supplier using commercial blades to do this with a Tier 1 cable operator.”

The solution, built to optimize the architecture of the HP BladeSystem c7000 and ProLiant BL460c series servers, supports delivery of live and on-demand services to iOS and Android devices, PCs with Adobe Flash or Silverlight, connected TVs and traditional MPEG-2 and H.264 set-top boxes, Perrier says. The range of resolutions for all these devices goes from mobile level to 1080p HD, he adds, noting that Muse also supports picture-in-picture, alternative audio languages, closed captions, DVB-Subtitles and DVB-Teletext.

Overcoming the Delay Factor

Another supplier asserting a software-based approach, even when it is operated from the cloud, meets premium service requirements is Imagine Communications, which over the past two years has revamped its strategy to bring its high-quality encoding technology approach to the multiscreen environment. Imagine’s ICE Cloud Software System is a low-delay, Linux-based distributed architecture solution that leverages advanced compression and pixel processing techniques with high scalability, says Imagine CTO and founder Ron Gutman.

“The ICE Cloud Software System is a ground-up development project built specifically for integration in cloud transcode farms and distributed CDNs,” Gutman says. “We took a close look at the requirements for live video transcoding in the cloud and CDN and determined that building on top of an open-source or licensed codec was not going to yield the required performance. Increasingly, live events not only require high quality and leading bandwidth efficiency, but also interactivity, and that means very low delay.”

The transcode delay issue is an important one, notes Chris Gordon, vice president of product and marketing at Imagine. “The typical transcode delay is in the five to seven second range, and then you have to add the delays imposed by fragmentation for adaptive streaming,” Gordon says. “You can end up with close to ten seconds of end-to-end delay, and now you’re out of sync with the traditional channel. If you’re trying to sync up on interactive applications across multiple devices, this is untenable.”

Imagine has cut the transcode delay to under one half second, giving operators the opportunity to introduce just enough buffer delay on the main channels of multiscreen services to keep everything in tune, Gordon explains. Even without adding interactive applications, operators don’t want their customers to be seeing a ten-second disparity between what’s showing on their TVs and on their iPads, especially when it becomes possible to make a seamless transition from viewing on the TV to viewing on a wirelessly connected device, he adds.

Adding to the benefits of the Imagine cloud-based system, the use of Linux allows customers to scale their multiscreen services across multiple types of processors, servers and memory systems,
Gutman notes. “Because available resources on a cloud transcoding farm or a distributed CDN network are always changing, it’s vital that the cloud transcoder be ready for a fully distributed network,” he says.

No doubt the dedicated-hardware versus software-based transcoding argument will persist for some time to come, but, in truth, a lot depends on what a specific operator’s requirements are in terms of how services are managed centrally and in distributed locations, notes Harmonic’s Yaron Raz. “Some operators are very interested in a software-based system,” Raz says. “At the same time we hear about concerns over the relatively lower density of software systems and the resulting higher power consumption rates, which argues in favor of the high-density hardware platform. So on multiscreen we provide both options.”

As previously reported (September, p. 19), Harmonic introduced its software-based transcoding platform earlier this year with a codec built from the ground up to meet the requirements of a software system. Now, Raz says, the company is seeing robust demand emerging for the solution, especially among operators who want to apply generic data-center server resources to handle transcoding requirements for content dedicated to specific local markets.

“Some MSOs are taking all their local channels and backhauling them to central locations, where they find they need to deal with thousands of channels,” he says. “In that situation you want maximum density. But others are doing transcoding in local regions where maybe they only have to deal with 20 channels. In those cases the ability to have transcoding and fragmentation in a single location running on multi-purpose servers is more important than density.”

Arguments suggesting software-based systems will not be able to keep up with the higher resolution requirements on the horizon are not accurate, Raz suggests. “There’s no ceiling,” he says. “It’s a tradeoff as to how many channels you can do on a blade. One of the advantages of the software approach is your ability to ride the processing benefits where Moore’s Law results in new hardware every six months. We’ve seen that with our software products, which contrasts with hardware systems where you buy a system with the expectation that it will be in place for a few years.”

The good news for service providers is that, against a backdrop of intensifying pressure on bandwidth resources, the range of solutions that promise to ease that pressure has reached a point where strategists can prepare for new service initiatives with confidence the squeeze won’t be as bad as it looks right now. By 2014, thanks to HEVC, it will be possible to introduce content formatted to 4K for 100-inch screen 3DTV sets without allocating any more bandwidth than is required today for a 1080p HD channel delivered over H.264.