Video distribution: Home networking's ultimate end
Today, the very tip of the home networking iceberg is all that is showing. Slowly moving beyond the "early adopter" stage that typically precedes the proliferation of any new personal technology, home connectivity gear is becoming increasingly essential for even casual technophiles.
But hidden below the surface is what has cable system operators paying close attention–namely the massive potential that home networking presents, which takes several applications way beyond simple data-centric tasks like file and printer sharing and connecting multiple PCs to one broadband connection. Ultimately, moving beyond simple data applications means one thing–pumping video over home networks–and it's the endgame operators seek, while today they're busy testing the first real wave of home networking gear.
But how far away is a video-centric home-networked world? The answer to that question lies in the capabilities of the protocols powering today's home networking gear (both wireline and wireless) and closer inspection reveals various states of progress and potential.WHERE THE WIRED GROUPS ARE
Much of the first wave of home networking equipment that consumers have come in contact with uses home wiring, like power or phone lines, following the logic that consumers are more likely to install a network if they don't have to mess around and run new wires throughout the house. So far, in this early rollout stage, it has been relatively easy for consumers to set up shop using ubiquitous connections in their homes, plugging devices and modules into phone jacks and power outlets to create basic home networks.
But of the wired networking technologies, specifically the HomePlug powerline protocol and the HomePNA phoneline transmission scheme, none can really serve the needs required for video distribution . . . at least not today. Video's sensitivity to delay means that the transmission scheme must be reliable enough to guarantee delivery the first time, as retransmission isn't an option. A viewer just can't "wait" for video to be delivered.
"I think that this concept of applying packet switched principles to the home is probably not the best," explains Steven Gardner, an engineer with chipmaker Conexant, who doubles as vice president of the board of directors with the HomePlug group. "With home networks, the vast bulk of the data that is going to be transmitted is constant bit rate kinds of stuff, which is very different than computer data."
That inherent difference becomes even more marked when it comes to video bandwidth engineering. With the wired protocols, bandwidth limitations are perhaps the biggest reason they're not ready for prime-time video transmission. On the lower end of the spectrum, the current HomePlug 1.0 specification ships data around at an estimated 11 Mbps–a rate more than adequate for basic file sharing and simple data transfers, but nowhere near the levels most operators would require for video distribution. One stream of standard-rate video requires a clip of between 4 Mbps and 5 Mbps, and most systems will require multiple streams, along with corresponding Web and data traffic, which pushes the number well beyond HomePlug's capabilities.
The other wired technology making hay–the HomePNA phoneline transmission scheme–touts a throughput closer to the threshold required to do video. The HomePNA 2.0 specification claims a line rate of 32 Mbps; however, when accounting for things like overhead and retransmission, the rate is closer to 20 Mbps. With 20 Mbps, an operator could potentially get two or three video channels transmitted without too much trouble, but not much else.
But while the phoneline network might be a convenient way to connect data-centric devices, entertainment boxes like TVs and stereos aren't exactly phone jack-friendly. In response, HomePNA chipmakers like Broadcom are adapting the HomePNA technology for the coaxial environment, which should make it much more attractive to cable operators who will undoubtedly gravitate to technologies designed for their network.
"Because HomePNA was originally designed with reasonable power levels and to be spectrally-friendly, all we had to do was take the exact same chips, working in the exact same ways, with just a little bit of impedance matching with a few little passive components, and immediately you could light up the entire coax network as well," says Steven Palm, principal engineer with Broadcom's home networking division. Broadcom's iLine 32 chip is one of the most adopted HomePNA chipsets available today, with both phoneline and coax versions of the technology.
HomePNA has also announced a roadmap to get to a 3.0 specification, which is designed to zip data around at a speedy 128 Mbps. Operators will need a scheme that can transmit video at or near that speed if they have eyes for HDTV carriage, or for distribution of the five or six standard streams required for advanced set-top applications like shared PVR.
But while the technical transition of HomePNA to the coax environment may have been relatively simple, how the technology will perform on the new platform is a lingering question. First, phoneline topologies are quite different from coax, and the question of how to overcome signal attenuation caused by splitters is still floating in the ether. Even more importantly, industry watchers are wondering whether the scheme is truly optimized for coax in terms of performance. HPNA was designed for twisted pair, using low frequencies, which can be problematic in coaxial environments. HPNA uses frequencies below 50 MHz, potentially even below 20 MHz, which means inevitable issues with DOCSIS traffic, and conflict even with household appliances operating in the lower frequencies.
These issues with coax transmission are yet to be solved, but if the ultimate goal is to have an entertainment-centric distribution network available to consumers, working through issues with coaxial plant is a necessity. The issues associated with splitters and attenuation are easily dealt with if you're in broadcast mode, but there are some significant barriers when you start talking about two-way and high-speed communication over those wires.
And there is a lot of work going on–much of it by stealthy developers–to overcome the issues with coax. One company just emerging from stealth mode, Tiaris Networks, has been working at a coax solution that can yield raw data rates of 100 Mbps, with a realistic net payload of about 80 Mbps. That's more than enough headroom to carry the multimedia traffic operators aim to deliver. In contrast to the lower frequency HPNA-over-coax platform, Tiaris technology uses higher frequencies, running above existing services so as not to interfere with them. In basic terms, it's looking to develop a coax-to-Ethernet bridge, and ideally it would take the form of a chipset integrated into one of the rich media center boxes the cable industry has been developing.WHAT ABOUT WIRELESS?
But when it comes to potential, there's nothing sexier than wireless. All of the issues that plague wireline schemes are pretty much non-issues for wireless–there is no infrastructure to deal with, attenuation is a non-factor, and its mobility means it can be set up virtually anywhere.
That said, wireless has its own set of distinct obstacles to overcome, especially when it comes to transmitting video over those wireless links. Today, there are more than 6 million networks in homes and offices that utilize the 802.11 wireless networking scheme; unfortunately, the basic fundamentals behind 802.11 inherently undermine attempts to transmit video over those links.
To begin with, 802.11 (be it the slower 802.11b or the developing 802.11a wireless standard) wasn't designed to route video traffic. At the core of 802.11 is the "fairness" principle, where in order to support multiple users, the network is consistently splitting the available bandwidth to allow additional users to join the network. As one design engineer explains, this fairness principle cuts against the grain when it comes to delivering video in an environment that must guarantee a certain level of bandwidth for successful transmission.
"Fairness is somewhat at odds with the transmission of video," says Mahesh Balakrishnan, vice president of business development for the emerging business unit of Philips Semiconductor, makers of more than 50 percent of the world's 802.11b radios. "If you want to have high-quality video transmission, you have to be able to reserve some bandwidth, and then stick to it."
Therein lies the problem with 802.11 for video. Bandwidth levels are notoriously in flux with 802.11 wireless networks. Homes are one of the most challenging environments for wireless devices. Walls and furniture create signal loss through a phenomenon called multipath, and moving traffic makes for a fading of signal strength as a function of distance. All of this contributes to a culture of ambiguity in terms of guaranteed bandwidth . . . essentially a dealbreaker if your aim is video transmission.
To address some of these issues, the standards body defining 802.11 technologies is attempting to add quality of service (QoS) parameters to the spec by establishing sets of priorities for 802.11 traffic. A separate group within the body, the 802.11e group, has been working for the past couple of years to specify a QoS priority scheme for wireless LANs, but so far, it has made little progress.
But others in the market are trying to solve the issue through proprietary product development. These issues related to QoS in 802.11 networks are specifically in the sights of Magis Networks, a company with a chipset based on the 802.11 RF template, but with a synchronous MAC layer more amenable to QoS functionality. By diverging from the 802.11 standards group at the MAC layer, and adopting a synchronous MAC more in tune with the HyperLAN2 set of specifications for wireless transmission, Magis thinks it has the best of both worlds–the use of the increasingly popular 802.11 RF spectrum, plus the QoS capabilities found in the HyperLAN2 standardization effort.
"In a sense, we've used chunks from (802.11a, HyperLAN2 and IEEE 1394) to try and come up with the most cost-effective and appropriate way to do video wirelessly," explains Pete Fowler, executive vice president at Magis.
Another developer working higher up in the software stack is ViXS Systems, a Toronto-based chipset maker aiming to bring QoS to video delivery, but with a twist. Rather than treating video traffic like traditional data bits, and trying to guarantee a certain level of bandwidth, ViXS guarantees video delivery based on a frames-per-second scheme. Whereas Magis is attempting to widen the available wireless pipe to assign priority to video traffic, ViXS essentially accepts the fact that the wireless pipe will be constrained, and guarantees at least 30 frames per second of video–even if bandwidth is at a trickle.
At the center of the ViXS strategy is the ability to read and adapt to fluctuating network bandwidth. A software piece monitors each stream as well as the bandwidth demands of devices at either end of the network. The ViXS IC (called XCode) then modulates the associated bit rates, resolution and format of multiple MPEG video streams in real-time to fit that environment.
ViXS executives point out that their XCode solution isn't limited to wireless transmission schemes, and is more like an application that would sit above any wired or wireless IP network. As such, it could be applied in conjunction with a solution like the one from Magis, or the wireline schemes previously discussed.MEDIA CENTER STRATEGIES
All of these advanced transmission development efforts are aiming to achieve real estate, specifically board space in the reference designs of the coming fleet of set-top "media centers" that are currently being developed. Operators are anxious to deploy systems that use a client/server model to distribute signals to multiple TVs in consumers' homes. This would allow them to deploy one central smart box, and multiple "dumb" (read: cheap) thin clients at other TVs around the house in an effort to lower their cost-per-TV.
One company developing a distributed media center strategy is Ucentric Systems. The company has demonstrated a compelling shared PVR system, based on the distributed client/server model, and is now testing available signal transmission options. New Ucentric CEO Michael Collette says any transmission scheme must excel in three areas: bit rate, distance and cost.
But in assessing wireless technologies, specifically all the current variants of 802.11, Collette says, "they tend to fail on one or more of those dimensions." He adds, "You need that magical trio in the wireless domain, and none of (the 802.11 technologies) are there just yet."
As a result, Ucentric has focused on a wireline approach to meet its transmission needs in the near term. The company, along with partners Broadcom and Comcast, is leading a field test of the HPNA-over-coax technology, which could lead to a full-scale field trial of a client/server system if the technology proves ultimately capable in the real world.
But perhaps the biggest fish in this pond is the Digeo group, developers of a media center system based on the Moxi platform. The Moxi-Digeo plan has been slightly overhauled from its first iteration, which was plagued with high costs. Digeo has since embraced analog as a way to cost-effectively distribute video throughout the home.
"We pretty much concluded that the cheapest way to deliver (S-video and stereo audio) to additional TVs is to stay in the analog domain," says Digeo CTO Toby Farrand. "While we have demonstrated and can support digital networking solutions, the cheapest way to do it is through analog."
Farrand also points out that no matter what type of medium operators choose for digital communication–be it 802.11a, or Ethernet via Cat5 cable, or digital networks over coax–all will require some kind of radio as the network connection, an MPEG decoder, graphics capability, a processor and memory in the client. That approach to client design would be expensive, and would defeat the general purpose of the client/server model overall.
"Over time, the digital solution will decrease in cost through integration. But we can't wait five years for this to become a sub-$50 solution for digital," says Farrand.
Operators will be keen to see if Digeo can deliver a client/server set-top model that can rival the pricing of a comparative two-box solution. That might just mean analog is the only possible plan of attack in the earliest stages of media center development. And if digital transmission technologies want on those reference designs, they'll need to progress a lot farther than they are now.