Note from the Editors: The following is the first in a series of guest articles by experts from across the broadband ecosystem. Its author is Dr. John Pickens, Chief Technical Officer at Com21. John serves on several advisory boards, is a frequent industry speaker and was a major contributor to the PacketCable standard. We’re delighted to welcome John as our first contributor and hope you will find his insights as interesting as we do.
The Complex TV Universe
Twenty years ago, TVs were simple devices with cheap frequency-agile tuners and 75-ohm physical connectors capable of receiving broadcast signals. Only a few dozen channels were on the air, and content was broadcaster-selected and broadcast to regional audiences. The program channels were distributed via analog transmission - one program channel per 6 MHz analog transmission channel. The world was simple and well defined.
The rise in popularity of cable TV provided a minor perturbation to this simple world - because the frequency plan was different, cable companies provided external set top boxes for tuning cable channels. Shortly the TV manufacturers caught on and delivered "cable-ready" TVs, with integral cable tuners. TVs could receive channels over-the-air, or on-cable, and the world was once again simple and well defined (though the consumer's configuration choices were more complex and prone to error).
But the engines of technology innovation, regulation, and competition, were quietly gearing up to disrupt this simple universe.
Today (and looking ahead) the TV universe is very complex. Multiple incompatible transmission methods exist external to the home – over-the-air, satellite, cable, twisted pair, fiber-optics and wireless. A myriad of potentially usable but incompatible transmission methods now exist inside the home: coax, wireless, Ethernet, powerline, twisted pair, and Firewire. High Definition TV (higher bandwidth) is becoming more common. The simple paradigm of broadcast TV (channel surfing) is being morphed into more intelligent models of on-demand content management (VOD and PVR). The user interface model is bordering on the complexity of the PC. Broadcasters have imposed thorny proprietary encryption schemes to protect against theft-of-content (conditional access); Hollywood wants uncrackable digital rights management functionality.
External-to-the-TV set-top boxes are now needed to enable access to the new digital transport paradigms and service functionalities. Set-top box vendors see a revenue opportunity to increase the functionality of set-tops and are moving them toward the high-end paradigm of PC computing. Battles have erupted between advocates of cheap vs. expensive, thin vs. thick client, service provider vs. consumer electronics industry control, and proprietary vs. standards-based set-top box and system paradigms. And, of course, creaky legacy requirements (analog tuners in TVs) confound the need for vendors, producers, and consumers alike to move to the new all-digital world.
MPEG standards are widely used for digital TV encoding and transport. The MPEG encoding standards define digital algorithms and packet formats for digitizing and compressing audio-visual information (e.g., movies, video, music); all forms of transmission of digital TV have adopted these standards. MPEG transport standards are discussed in the next section.
Transport provides the foundation: the technology used to transport the TV stream (audio + video + control) across the backbone, through the “first-mile,” and distributed through the home.
Digital TV Transport Alternatives
There are really only three major competing networking paradigms for transport of TV – analog transport, MPEG transport, and IP transport. A fourth paradigm – Ethernet transport – is achieving gains in the first-mile; the plusses and minuses argued for IP transport are essentially shared with Ethernet, so Ethernet is not discussed further.
Analog transport is the legacy transport gorilla. It exists because it was first, and because of the massive installed base of analog TVs and VCRs. Because of its dominance even digital first mile technologies like fiber to the home (FTTH) sometimes carry analog TV in a separate spectrum band (or separate cable). Analog transport is a very inefficient use of spectrum – one audio/video channel per 6 MHz in North America (7 or 8 MHz in Europe). Today’s digital encoding technologies carry ten times the content in the same spectrum as one analog channel – with another three-times capacity multiplier on the near horizon. Nevertheless, analog continues to dominate in the installed base.
MPEG transport is a small (compared to analog) legacy gorilla in its own right. Both Satellite and Cable adopted MPEG transport in early deployment of digital video. MPEG transport defines a synchronous framing structure (including time stamps) and transmission mechanism with strict requirements for timing tolerances. Its advantage is that it is optimized for handling the real-time requirements of TV (audio/video). It can also carry IP as an encapsulated frame type (e.g. DOCSIS transmits IP in Ethernet frames encapsulated in MPEG frames). Its disadvantages are that the equipment and vendor choices for infrastructure are limited and expensive; the technology is not compatible with the evolving physical layer transport media; expensive backbone devices are required to rate shape, multiplex/demultiplex, and enforce clock synchronization of the real time stream; and physical transmission options within the home are limited.
IP transport is an emerging player in the distribution of digital TV. Some market segments, e.g. FTTH and DSL, have adopted IP for TV and are aggressively installing IP set top boxes. IP is generally accepted as the most popular convergence technology for unifying services, applications, and physical networking technologies. Its advantage for digital TV is the dramatic cost reduction and performance improvements for networking technology (see below). An IP transport switch switching frames with MPEG encoding is potentially much less expensive than an MPEG transport switch switching both MPEG and IP. The disadvantages of IP transport are lower transmission efficiency (due to IP header overhead); a primitive Quality of Service (QoS) environment (judging by the installed base of network equipment); and the large buffers the receiver (TV) needs to maintain to manage jitter and latency (increased cost and intelligence).
Why IP Transport Will Win
Ultimately IP transport will win. While it will not win overnight, and will win in some markets more slowly than in others, over time it will dominate. Since this may seem like a bold statement, some analysis of how this might happen is in order. For IP transport to win, it needs to conquer the backbone network, the first mile network, the home network, and CPE equipment.
In the backbone network the driver for long-range adoption is innovation in performance, functionality, and cost. IP backbone technology (and Ethernet) is on a rapid paced cycle of improvement. Massive strides in recent years have ratcheted IP backbone technology from 1Mbps to n*10Gbps (large “n”). A large number of systems and silicon vendors have driven the engine of innovation toward lower cost and higher performance. In the backbone, IP exhibits a “Moore’s Law” type of phenomenon. Nothing like Moore’s Law is occurring in backbone technology for MPEG transport.
One might argue that IP’s dominance in backbone networks is not absolutely assured. Technologies are appearing that make it possible to carry both analog and MPEG transport streams on separate colors of light over DWDM fiber optic networks. However, at each forwarding point from one network hop to another network hop the cost of the electronics which does analog-to-analog and MPEG-to-MPEG forwarding (and its required services of re-multiplexing and rate-shaping) is on a cost curve that is more expensive and dropping less rapidly than the technology for switching IP transport. And the dedicated bandwidth for analog and MPEG transport is unavailable for other IP services.
First-mile technologies that are not MPEG transport based (Fiber to the home, xDSL, Ethernet) have already adopted IP transport for TV. While they sometimes carry analog in an out-of-band channel as a low cost way of supporting legacy TVs, they have adopted IP as the delivery vehicle for digital TV and have deployed IP set top boxes with IP receivers.
First-mile technologies that are MPEG transport based (Cable, Satellite, terrestrial broadcast) will take the longest to evolve to IP. They are held back by their sizeable installed base of legacy MPEG equipment – thousands of headend transmitters and millions of set top boxes – both of which are generating service revenue.
MPEG-4 may play a key role in convincing MPEG transport carriers to switch their first mile technology to IP transport. Although MPEG-4 multiples the network capacity up to three times, today’s CPE hardware is not MPEG-4 capable. So a massive hardware switch will be required to take advantage of MPEG-4. Whenever CPE hardware is changed, the opportunity exists for change in transport protocols.
Another driver for change is the need to lower equipment cost in the network. Today’s narrowband MPEG networks carry 10-12 TV streams per 6Mhz with each stream capable of bursting to 25% of the total available channel bandwidth and causing overflow. Expensive DSPs are required in the network to constantly groom and re-encode the video.
The evolution to wide-band (Gigabit) delivery in the first-mile would provide a converged video/data solution and eliminate the need for expensive DSP-based switches. A fully packed Gigabit transport channel, for example, would carry 333 streams at 3Mbps each (MPEG-2 coding), or 1000 streams at 1Mbps each (MPEG-4 coding). In the home, a home gateway (the router/bridge between the first-mile network and the home network) can provide a filtering service that selectively extracts and forwards only those video streams that are being viewed in the home. Assuming that the CPE equipment is IP transport enabled, no additional hardware change is required to enable unicast delivery of TV within the home.
One impediment in home networks that needs to be overcome is Quality of Service (QoS) handling of real time TV streams. Standards (e.g. CableHome) are being defined that deliver managed Quality of Service in home networks. Switched Ethernet is, in many cases, already QoS capable (via DiffServ priority handling). Fast Ethernet (100Mbps) is certainly capable of supporting the IP transport requirement for TV even today. Other technologies are receiving QoS functionality in the standards bodies, and it is only a matter of time until the hardware is available.
The final hurdle in switching to IP transport is the CPE device itself. Today the CPE devices are the PC and set top box; tomorrow the CPE device is also the TV (or other video equipment) with an embedded IP transport receiver. If one is willing to switch physical layer technologies (e.g. from coax to Ethernet) an IP transport receiver can be much less expensive than an MPEG transport receiver. Even with a common physical layer technology, IP transport is lower cost because only one receiver is required (by comparison, today’s set-top boxes commonly have three receivers – two for video channels, and one for a DOCSIS channel).
MPEG transport advocates can stall IP transport by extending the lifetime of MPEG transport through innovation and competitive arm-twisting. MPEG standards changes may add functionality for HTML web and other IP services, and slow the migration to IP transport.
But ultimately IP transport will drive from the backbone to the first-mile to home networks and to the CPE (set top first, then embedded in TV) as a converged service. Classic MPEG transport based service providers are already replacing MPEG transport backbones with IP transport backbones, and carrying video and data over the IP backbones. The operator and the consumer both will then have wide flexibility in choosing best-in-class layer two technologies (backbone, first-mile, home network) for delivery of service. Cost will drop, performance will rise.
Whether the customer wants “plain old TV” or cool IP services, IP transport is superior to MPEG transport. It is far superior for converged services, and more cost-effective for video alone. The challenge for existing service providers is when and how to change.