From technical ingenuity to political intrigue, a tumultuous drama has unfolded over the last eight to 12 months in the digital-TV arena. The lead roles were played by the cable, television, consumer-electronics, and broadcast industries—with the help of some "gentle" FCC directing. The action almost seemed like a pilot for a new series of exciting technological advances that will grace our living rooms in the coming years.
The drama began as early as last April, when the market caught the first of the latest second-generation, 8-level, vestigial-sideband (8-VSB) demodulators for the reception of terrestrial DTV signals. Competitive announcements quickly followed, but all fell under a cloud of suspicion as a result of a claim by a certain broadcasting company. It stated that the American Television Standards Committee (ATSC) may have erred in judgement when it selected 8-VSB over the alternative, namely coded, orthogonal frequency-division multiplexing (COFDM). If true, that mistake could have cost millions to correct. The cost of the delays to the timely rollout of DTV, however, would have been inestimable.
While hearing these arguments, the FCC was busy trying to unite the consumer electronics and cable industries. By working together, they could bring about standards for low-cost, DTV-over-cable, in-the-clear reception within the receiver itself. This would eliminate the need for an external set-top box (STB). The FCC also was working on broadcasters, encouraging them to realize 60% market penetration of DTV signals by the end of 1999. This critical mass had to be reached if consumers were to start shelling out cold hard cash for the relatively expensive DTV sets.
Reaching that mark would be moot if quality content wasn't there. Sports and primetime TV were ramping up, but movie studios were loathe to allow the digital transmission of their properties. The scene was reminiscent of the introduction of the first VCRs. Only now the argument was that digital recordings are almost perfect and don't wear out. Some kind of copy protection, then, had to be inherent to a DTV setup before the movie properties would be transmitted. Protection was not technically difficult, with the 5C (named for the five companies behind it) standard already pretty much defined. It was just a matter of getting everyone to agree on it. That's where the fun started.
All of this was playing out against the backdrop of the consumers. As a group, they're almost blissfully unaware of the actual benefits of DTV. This truth could potentially scuttle all of the above efforts.
Fortunately, sanity has prevailed for many reasons, not the least of which is the amount of money that has been invested in DTV. There's also the money that stands to be made. It will work.
The light at the end of the tunnel started to appear after William Kennard, chairman of the FCC, took the podium during a luncheon at January's Consumer Electronics Show in Las Vegas, Nev. After extolling the virtues of DTV and all of its potential, he made it clear that the FCC would no longer stand idly by and watch the cable and consumer-electronics industries squabble over standards.
In essence, he threatened to take over the process in the interests of the consumer. He would have the FCC decide how cable compatibility should be achieved. The threat of the FCC taking control was enough motivation for Robert Sachs, president and CEO of the National Cable Television Association (NCTA). On Feb. 22, he and Gary Shapiro, president and CEO of the Consumer Electronics Association (CEA), signed an agreement to achieve compatibility between consumer digital receivers and cable TV systems.
That agreement has four parts. First, the groups settled on the technical requirements for the network interface specifications that permit direct connection of consumer digital receivers to cable TV systems. They also agreed to carry Program and System Information Protocol (PSIP) data on cable systems. This would include virtual channel tables and event information data (up to 12 hours ahead) to support navigation functions in digital receivers.
As labeling was not finalized, they opted to continue discussions to devise appropriate naming when marketing various digital receivers designed to work with cable systems (cable ready versus cable compatible, etc.). Finally, the fourth part of the agreement calls for joint testing of interoperability between cable systems and consumer receivers.
Putting those particulars aside, the agreement basically states that through cable, TVs off the shelf and with the appropriate demodulators can now receive in-the-clear NTSC and digital programming. Using a point-of-deployment (POD) module, security will be supplied by the cable company to allow viewing of conditional-access programming, such as pay per view and subscription programs. This can be authorized by one-way downstream data transmission to the POD module (Fig. 1).
That module came about as part of the OpenCable specification. It calls for an out-of-band (OOB) channel to carry the PSIP. That OOB has a two-way feed, downstream and upstream, dubbed the forward data channel (FDC) and reverse data channel (RDC), respectively. The data rate on the FDC specified in the agreement can be 1.544, 3.088, or 2.048 Mbits/s to accommodate current standards and implementations. As a PC-card plug-in, the module can go into both the set-top box (STB) and receiver. But the agreement anticipates that it will now go directly into the TV.
Some minor technical issues do remain be solved. Yet manufacturers are confident that they'll have POD modules ready for retail availability by the FCC's July deadline.
HD Over Cable
The NCTA/CEA agreement provides for not only the 18 ATSC formats, but also the 17 formats that have been in use by the cable industry. These include some high-definition (HD) and standard (SD) formats that are slightly lower in resolution than the ATSC-defined formats. They save on bandwidth.
The whole issue of HD over cable has been a bone of contention for some time now. HD doesn't just require more bandwidth for the extra information. The number of receivers actually capable of displaying that information is very limited, making it hard to justify. Plus, most of the existing HD off-air receivers aren't capable of receiving the standard EIA 512 cable-frequency plan. They can only receive the UHF plan.
As a result, the cable operators who have actually put the spectrally inefficient 8-VSB on the air have had to shift their frequencies to be received by the HDTVs. That shifting has required that they use up two of their cable channels, rather than one. Aside from being inefficient, then, 8-VSB HDTV takes up twice the bandwidth.
The agreement signed by the NCTA and CEA alleviates that situation by stipulating that all signals will be sent using 64/256 quadrature amplitude modulation (QAM). So cable-ready TV receivers will be compatible with both 8-VSB and QAM.
The single-modulation scheme also cuts cable industry costs by allowing equipment reuse. According to Bill Wade at Scientific-Atlanta, "Clearly, the cable industry would like one single standard for digital transmission. QAM is the same as the standard being used for cable modems and many other cable digital transmissions. This allows us to offer an integrated set of services into a consumer's house." He's referring to Scientific-Atlanta's line of Explorer STBs for Internet protocol and digital video, which can now use the same transmission format. Also, the use of one standard increases flexibility.
A few weeks after the Las Vegas threat, the FCC made another landmark decision. It slapped down the petition to review 8-VSB as the standard of choice for DTV terrestrial broadcasts in the U.S. This affirmation of 8-VSB, along with the cable-compatibility promise, paves the way on the part of IC, STB, and TV manufacturers, as well as broadcasters. With a concerted effort, they can reach the FCC's DTV roll-out goals. The ultimate plan is to send NTSC signals "dark" by 2006. The newly released spectrum would then be auctioned off. So far, everything is on schedule.
The plan called for 24 stations to be on the air by November 1998, 40 by May of 1999, and 120 by November 1999. That translates into 53% of viewers with at least three signals.
As of January 31 of this year, 117 stations had come on the air in 44 cities, with a total coverage of over 60% of U.S. households. The next step is to have all commercial stations on the air by 2002, and all non-commercial stations by 2003 (a total of 1600 stations). Assuming 85% of consumer coverage, the transition would then be concluded.
As for the availability of equipment to date, high-definition TV (HDTV) receivers, STBs, HDTV-ready sets, and DTV PC cards are already available from more than 25 suppliers. Receivers cost from $5000 to $8000, with prices expected to drop by 50% in two years. For $649, STBs are selling briskly. They can take advantage of the ability of many high-end TVs to a-chieve resolutions higher than 480 lines. Up to 600,000 sets are expected to ship this year alone, with up to 10 million year-ly by 2006.
Surprisingly, it's the PC market, with demodulator/decoder cards available for between $200 and $300, that's likely to spur the growth of HDTV in the immediate future. The PC's relatively painless move to HDTV, over the multi-thousand-dollar requirements of dedicated TV receivers, is a result of many factors. One of the most obvious is the high cost of optics and gun technology for large-screen, direct-view HDTVs. The glass alone, in a 16:9 aspect ratio, is twice the price of its 4:3 counterpart. Factor in the higher resolutions and faster scanning requirements, and the PC becomes an attractive alternative. Also, consider that the bulk of the processing can be offloaded to the host CPU, and that computer monitors already support high resolutions.
This was the thinking behind ATI Technologies Inc.'s recent demonstration of the latest in DTV technology on the PC (see opening photo). Shown at the Intel Developers Forum, the technology was a result of cooperation between ATI, Conexant Systems Inc., NxtWave Communications, and Ravisent Technologies Inc., all leaders in their respective DTV areas.
At the core of the demo was ATI's Rage 128 Pro chip, with integrated inverse discrete-cosine transform and motion-compensation technology. That combination permits full-rate DTV decoding with minimal CPU usage.
It also used Conexant's Fusion 878A PCI video decoder, which decodes NTSC, PAL, and SECAM signals. The ATI design can take the transport streams from Nxt-Wave's NXT2000 8-VSB demodulator over the PCI bus and deliver them to the host CPU.
The company's selection of Ravisent for the software decoding of the MPEG-2 stream was far from coincidental. Ravisent and Conexant announced cooperative work back in November, which resulted in the DStreamATSC single-card reference platform for full NTSC/ATSC reception for HDTV on a PC (Fig. 2). Together, they have a proven track record of cooperation and compatibility.
Again based on the Fusion 878A, the DStreamATSC incorporated Ravisent's CineMaster HDTV all-software, MPEG-2, all-format decoder. Like ATI's design, it can decode all 18 ATSC formats. It forms an easy-to-implement platform for bringing the first HDTV broadcasts to the mainstream Pentium-III PC for under $150. Because of its flexibility and the ready availability of high-performance CPUs, the Ravisent CineMaster software may foretell the end of hardware MPEG decoding.
This is ATI's bet after working closely with both Intel and Microsoft to make its DTV solution both code-efficient and operating-system-compatible. The system requires at least a 600-MHz P-III CPU for smooth operation. While this is still a relatively high-end system, it's quickly becoming the sweet spot.
In continuing to develop its 878A chip, Conexant has made plans to add more functionality, higher resolution, and better quality to the analog portion. About the company's alignment with Ravisent and ATI, Tim Wilhelm, marketing manager for the digital infotainment division, comments, "\[DTV is\] an interesting market, as no one company can do it by themselves. Eventually, companies like ours will be able to do more, but for now it's a multiple hardware and software vendor solution."
Adding to this, Eileen Carlson, Conexant's product line manager for current technologies, says, "Traditionally, we'd do a full reference design and hand it off to the retailer to package and ship out, but not with this technology."
The selection of NxtWave by ATI was no coincidence. The NXT2000 doesn't just demodulate terrestrial 8-VSB broadcasts. It also handles 64/256 QAM for cable broadcasts. To date, it's the only DTV demodulator chip to do so. Its low power consumption of 1.25 W and its 100-pin TQFP format undoubtedly factored into the company's decision.
As it stands, the agreement between the NCTA and CEA is to remodulate 8-VSB signals over QAM and allow access to all in-the-clear channels. So it would seem prudent to include both QAM and 8-VSB. Consider how difficult it still is to get a solid DTV signal over the air. Even ATI was in a quandary as to how exactly it would ship its new board—with or without an antenna. As of press time, that board was still unannounced and remained nameless. But now the antenna issue may be moot, as a cable connection is all that's needed if the terrestrial signal is poor.
Over-The-Air Reception Still An Issue
The issue of over-the-air reception remains, however. Though the FCC remains steadfast in its support of 8-VSB over COFDM, it may have left the door open to more debate when it said that it would review the current round of second-generation 8-VSB receivers in the March-April timeframe.
There are quite a few from which to pick. Aside from NxtWave, offerings stand from Philips Semiconductors, Motorola, Oren Semiconductor, and Zenith, to name a few. STMicroelectronics has been active in the European market's Digital Video Broadcast (DVB) technology, but will be introducing 8-VSB demodulator technology later this year.
The key area of performance with respect to 8-VSB is its ability to acquire a signal under noisy conditions and/or with dynamic multipath interference. When compared to COFDM, first-generation 8-VSB receivers came up short—especially in terms of multipath. This comparison became the basis of the petition that held so many people's attention at the close of 1999. Would the FCC start the evaluation process all over again and delay the rollout by at least 18 months to two years? Not likely.
For one thing, the basic principles of the ATSC standard appealed to both manufacturers and broadcasters. For a given output power, it can cover a much wider area than COFDM—one that's more comparable to that of NTSC. As it doesn't need a double-conversion tuner, 8-VSB also is much cheaper to implement in silicon. Even if QAM is required, it's pretty similar in architecture. It could be readily added to an 8-VSB solution without the duplication in silicon required with COFDM. Finally, 8-VSB is more immune to impulse noise.
Keep in mind that the original tests were done with first-generation devices. The technology has come a long way since then in many ways. With respect to noise-handling alone, almost all available receivers have achieved the goal of a 15-dB carrier/noise (C/N) ratio.
For static multipath, first-generation devices could handle only 20 µs. Current devices can bear up to 40 µs. With other parameters, there's quite a bit of variation. Says Samir Hulyalkar, principal staff engineer at NxtWave, "The range of the equalizer has also increased. Ours is over 44.6 µs on the positive, and down to −4.5 µs on the negative—one of the larger ranges out there." Of course, this also depends on the strength of the multipath. Some first-generation devices could handle only 3 dB. Second-generation devices can handle 1 to 2 dB, depending on the echo delay.
Tracking Speed Versus Multipath
With regard to dynamic multipath, the key is how fast the device can track. There's a whole lot of variation out there, especially above 3 or 4 Hz. Between 20 and 50 Hz, the multipath is dynamic and can result from weather changes and leaves fluttering by, while people walking around a room translate to 10 Hz. So it's essential to make the system as robust as possible to this phenomenon.
As the other parts of a DTV system, such as the tuner and MPEG decoding, become better understood board and system manufacturers are increasingly turning to the demodulator makers. They want those manufacturers to provide the needed edge that will push them ahead in what will soon become a very competitive market. But the design cycles required for newer product introductions prevent a rapid turnaround of new demodulator designs.
As a result, some makers are turning to refinement, or holistic, approaches. They'll wring every ounce of possible performance out of their chip designs. The result is modules with components inside that are optimized for their particular chip design. This brings some value added to the customer.
An example of this phenomenon is Motorola. Prototypes of its MDTV module show that it's essentially a full receiver. But it can be installed in an analog TV by the OEM, making it a dual-mode TV. The module includes the company's MCT2100 8-VSB demodulator and MCT4000 transport and video processor, along with an audio DSP and a PowerPC for control.
The MCT2100 itself uses a 50-MHz clock. It accepts 25-MHz samples of 8-VSB signals at a direct tuner IF of 43.78 MHz or at a downconverted IF input of 6.25 MHz. It also supplies a fixed-rate clock to an external analog-to-digital converter (ADC).
According to Frank Eory, senior principal staff engineer at Motorola's digital TV operations, "We used a combination of blind and decision-directed equalization. But from an architecture point of view, it's a decision-feedback equalizer (DFE), with the magic in the algorithms." The chip uses 512 taps with a combination of complex and real values. The delay span is a pre echo of −3 µs to a post echo of 44.6 µs.
The use of a tap for each delay value directly opposes NxtWave's approach of sparse equalization. That type of equalization has the advantage of reduced silicon requirements and hence lower power consumption. But sparse equalization is tricky, as the designer has to locate the echoes and put the hardware there. Because the echoes come and go dynamically, this algorithm has to constantly relocate the hardware filter on-the-fly.
Says Eory, "We just built the full span for all the delays that are known to exist. Real-world data suggests that in the vast majority of cases, echo delays are in the 20-µs and below region." But in a worst-case scenario, like the U.S. city of Seattle, Wash., there are hills combined with a low antenna height on the broadcast tower. There, echoes can extend out into the upper 30-µs range.
The acquisition time on the MCT2100 is similar to the NXT2000 at 50 ms or less. This is much faster than COFDM. With the right antenna, it also allows it to support channel surfing. The chip runs off 1.8 V and comes in a 160-pin QFP.
Other offerings in the demodulator arena include the TDA8961 VSB2 from Philips Semiconductors (Fig. 3). Introduced in April last year, this device is supported by the Multi Platform One Chip, or the MPOC 8980. "In this current market of hybrid analog and digital, VSB and QAM, and VSB and QPSK, it's important to have something that brings everything to an even format. That's the purpose of the MPOC," says Philips' Giri Venkat. The MPOC receives NTSC, ATSC, and OpenCable signals. It then puts them out over a single 10-bit bus as a digital IF stream, thereby alleviating the pain of multiple standards. By hooking up the MPOC to whatever kind of demodulator, any international standard can be supported.
The setup can be used in a dual environment. An IF out goes to the QAM demodulator first, and then in parallel to the VSB demodulator. If it's QAM, the QAM demodulator acts on it, demodulates it, and sends a transport stream out to the VSB2. The VSB2 recognizes that the signal has been processed and passes it on to Video IN. The output of the demodulator is a fully compliant MPEG-2 stream.
Within the 8961 itself, the designers are particularly proud of the carrier-recovery circuit. In multipath tests, it came in at about 15.5 dB. The DFE uses blind equalization along with trained equalization from the ATSC field sync. But it features the ability to determine which has the better performance at an instantaneous sample, and then uses the better one. It does this adaptively to the tune of many times per second.
By itself, the equalizer has a range of −2.3 µs to 22.5 µs. Using external software control, that range can be enhanced to get a window of up to 80 µs. Future versions will improve on the method of equalization, as well as the ability to control and manipulate the equalizer itself. Though changes are on-going, right now no new silicon announcements are expected from Philips in this area.
While flexibility on the front end is key, it's also essential on the transport decoding side. Aware of a consumer need to be able to view multiple sources simultaneously, STMicroelectronics has introduced its ST20-TT1, a triple-stream transport decoder (Fig. 4). This device can input three different MPEG-2 transport streams simultaneously from three different front ends. it also can decode each simultaneously and then send specific information to particular decoders in the system.
Though the device is designed to handle data from QPSK, VSB, and QAM networks, it's indifferent about which one it is. It's happy as long as what gets to its input is an MPEG-2 stream.
Though many pieces are in place for an upcoming rapid acceleration of the DTV rollout schedule, two issues remain. The first of these, copyright protection, has been solved technically. As mentioned previously, the 5C digital-transmission content-protection (DTCP) scheme is named after the five companies behind it: Hitachi, Intel, Matsushita Electric, Sony, and Toshiba. It has all but vanquished any alternatives. All that remains is for it to be officially endorsed.
Essentially, the DTCP defines a cryptographic protocol for protecting audio/video entertainment content from illegal copying, interception, and tampering as it passes over the IEEE 1394 bus. The 1394 is the bus of choice for connecting digital equipment in the home. 5C will be implemented on each 1394 transceiver. It's the key to getting the movie houses to join the DTV phenomenon.
The second issue isn't technical. It concerns the inability of the consumer-electronics industry to ignite consumer interest in DTV. The fact is, most consumers have no clue what DTV is and what it can do. This has been one of the biggest barriers to penetration. But now that there's valid content, widespread product availability (both full scale and entry level), cable compatibility, and copyright protection, the industry is gearing up for a mass-marketing campaign that should put this final piece in place. So hold onto your hats. It's going to be a visually stunning and aurally exhilarating ride.
|Companies That Contributed To This Report|
ATI Technologies Inc.