DTV over digital cable: Reaching a larger audience
Why should broadcasters concern themselves with having cable operators carry their DTV signals? Well, consider these numbers. Right now, more than 75 percent of U.S. households receive television by cable, and roughly 20 million of these households subscribe to digital cable. Over 91 cable markets now receive HDTV service and the number is growing rapidly. The installed base of ATSC (8-VSB) receivers in the United States is tiny by comparison, currently measured in hundreds of thousands.
This digital cable headend installed in a facility in China accommodates several program sources and is equipped with encoders, multiplexers and stream processors.
It takes more than a new antenna and transmitter to reach U.S. audiences with DTV. Cable carriage of broadcast DTV signals is critically important to broadcasters and station group owners who want to extend their reach to cable viewers.
Digital cable system overview
In the United States, digital cable networks are based on quadrature amplitude modulation (QAM) in 6MHz channels. There are two standard QAM implementations for cable: 64 QAM and 256 QAM. HDTV-capable cable set-top boxes are rather new and operate with 256 QAM. This modulation yields approximately 38.8Mb/s in a 6MHz channel. Many older set-top boxes work with 64 QAM at 26.94Mb/s, but these boxes rarely support HDTV decoding.
Cable headend processing
Figure 1 shows how a typical digital cable system processes incoming signals and remultiplexes them for digital cable distribution. The cable headend accepts signals from several sources including satellite feeds, analog local cable programming, off-the-air DTV broadcasts, and direct feeds from broadcast stations over fiber, microwave and WAN. Receivers and encoders process the incoming signals and convert them to asynchronous serial interface (ASI) MPEG-2 transport streams. From there, the streams undergo several processes, including grooming, rate shaping, metadata processing (e.g., PSIP data), remultiplexing and encryption. After processing, the cable system modulates the datastream (using 64 or 256 QAM) and distributes it to its customers over hybrid fiber/coax (HFC).
Figure 1. A typical digital cable system processes content from satellite and local broadcast and remultiplexes it for digital cable distribution.
From the MPEG-2 transport streams, the digital cable system selects the specific programs and services it wants to carry and groups them into different levels of service for its customers. This process is called grooming. It involves dropping programs and/or services and re-mapping packet identifiers (PIDs) and tables to prevent potential conflicts with existing PIDs and tables. In addition, the grooming equipment must create a new cable virtual channel table (VCT). And, to preserve program-guide information from the incoming PSIP, the equipment modifies the PSIP to reference the new PID numbers. Since this additional processing adds to the cost of the grooming equipment, some cable operators might opt to drop PSIP altogether.
Stations that add new services after the cable system equipment has been configured may find that their new services are not automatically passed to cable viewers unless the cable operator has provisioned them in advance.
A cable operator often must change the bit rate of incoming datastreams to fit its own bandwidth requirements. To avoid the detrimental effects of fully decoding and re-encoding these streams, not to mention the added expense of decoders and encoders, cable operators often use a process called rate shaping. it converts one bit rate (usually relatively high) into another (relatively low) by processing the signal in the compressed domain. Rate shaping may also transform video streams from constant bit rate (CBR) to variable bit rate (VBR) or it may reduce VBR to lower rates. But rate shaping also has its limitations.
The rate-shaping process first involves removing any null packets, which reduces the bit rate without degrading video quality. It then analyzes the MPEG-2 transform coefficients – the mathematical information that MPEG-2 uses to describe the video – and rounds them off in a process called requantization. This process achieves further bit-rate reductions. But requantization is a lossy process and, under stressful conditions, it may result in artifacts. Requantization is relatively harmless to MPEG-2’s B frames, but it must be applied very judiciously to I and P frames because they contain the reference information that the process uses to reconstruct the predicted P and B frames.
Removing null packets and performing requantization are the only tools that rate shaping can use to reduce bit rate. By contrast, modern MPEG encoding implements complex strategies such as motion estimation, coding mode selection, GOP structure and other compression tools. Good encoders constantly evaluate video complexity, varying their encoding strategy to maintain optimal performance.
When the amount of bit-rate reduction is small compared with the incoming bit rate – for example, when reducing the bit rate from 10- to 8Mb/s - the rate-shaping process typically yields good video quality. But when the reduction is relatively large (for example, from 3- to 2MB/s), rate shaping yields poorer results than MPEG decoding/re-encoding. Rate-shaping systems perform best when reducing rates by less than 25 percent.
Video complexity is a function of both motion and detail. Highly complex pictures (such as basketball game in HD) combine intense motion with high detail. Program material complexity varies over time. Even programs that we associate with high complexity have interludes of little or no motion. Likewise, simple programs have transitions and camera movements that create intense motion for short periods. Statistical multiplexing takes advantage of unevenly distributed peaks and dips in individual streams by dynamically allocating more bits to the programs with the most complexity, statistically averaging the complexity across several channels. A disadvantage of this process is that the coincidence of high motion and high detail in two or more program streams can adversely affect the picture quality of a third, low-motion, low-detail program. The best way to avoid this is to avoid grouping complex channels (such as HD sports channels) together and, instead, group each complex channel with a relatively simple one (such as an SD news channel).
Figure 2a. Station X transmits one HD program, one SD program and broadcast data, and encodes all three services at a variable bit rate.
How cable handles broadcast TV
Multiplexing two ATSC channels, each with a payload of 19.39Mb/s, into a single 38.8Mb/s 256 QAM cable channel is a straightforward process. Of course, the cable operator would have to remap any conflicts in PIDs. It may also have to drop broadcast PSIP tables and create program-association tables (PATs) and program-map tables (PMTs). Cable set-top boxes do not use PSIP, although this may become important when cable-ready DTV sets become available. Because PSIP processing adds to the cable companies’ cost, the companies are unlikely to implement it in the absence of cable-ready DTVs.
Figure 2b. Station Y operates a single HD service at a constant bit rate.
Figures 2a and 2b show examples of two DTV transport streams prior to processing at the cable headend. Station X transmits one HD program, one SD program and broadcast data. It encodes all three services at a variable bit rate for efficient use of the 19.4Mb/s ATSC transport stream. Station Y operates a single HD service at a constant bit rate of 18.5Mb/s for HD video. Both stations use an identical PID for their HD video streams.
The cable operator combines the streams from Station X and Station Y, modulates the combined transport stream in 256 QAM and transmits it through a digital tier in a 6MHz channel. Figure 3 shows the new combined transport stream for digital cable. The cable operator has changed the Station-Y PID for the HD service from 11 to 111 and the audio PID from 14 to 114. The operator has also decided to drop this station’s PSIP tables to conserve bandwidth. For the Station-X transport stream, the operator has dropped the broadcast PSIP tables but retains the same PIDs.
MSOs will have to do more than grooming to carry the increasing number of higher-bit-rate HD programs available. Combining three DTV/HDTV signals in a 256 QAM transport is an option. But transmitting three broadcast DTV/HDTV signals in 6MHz requires rate shaping to get the total of 54Mb/s (an average of 18Mb/s per signal) down to 38.8Mb/s for a 256 QAM channel.
Figure 3. The combined streams from Station X and Station Y
Opportunities and challenges
Clearly, broadcasters prefer that cable operators retransmit their DTV signals without rate shaping or reducing original content. But, to conserve bandwidth, cable operators must often do one or the other – or both. The question for broadcasters then becomes: What are the alternatives?
Rate shaping works well when the cable company requires only a small percentage of bit-rate reduction – typically, less than 25 percent. For the example of three HD programs in a 256 QAM cable channel, the average bit rate per HD program is approximately 12.5Mb/s. Therefore, the ideal input to a rate shaper ranges from 12.5 to 16Mb/s. Cable operators could obtain optimal results by operating the encoder at the lowest bit rate that achieves high-quality pictures prior to rate shaping.
VBR encoding of the original broadcast DTV signal may provide a more optimal input to the rate shaper than CBR. It is difficult to select the optimal CBR encoding rate. If the encoding rate is too low, the program or picture shows artifacts. If it is set too high, the rate shaper operates outside the optimal reduction range much of the time. A high-efficiency VBR DTV signal is likely to optimize the rate shaping process.
Figure 4. Three local stations could use closed-loop statistical-multiplexing encoding to form their own 38.8Mb/s cable feed.
Broadcasters could improve and thus control picture quality by cooperatively encoding a statistical multiplex pool with a direct feed to the cable company. For example, three local stations could avoid rate shaping by the cable operators if they were to work together using closed-loop statistical-multiplexing encoding to form their own 38.8Mb/s cable feed (see Figure 4). This approach provides better quality than the alternative of three stations independently encoding and distributing their programming to the cable operator. Of course, the local stations would have to cooperate closely and provide dedicated encoding for cable carriage.
Another concern is that rate shaping will worsen artifacts generated by the encoding process. If the original encoder is stressed enough to produce even slightly visible artifacts, these artifacts are likely to become more pronounced after rate shaping. Therefore, it is always a good idea to use efficient, high-quality encoders for the emission and last-mile-delivery processes. As with carrying NTSC programming, broadcasters know that carrying their DTV signals over cable is vital to reach a majority of their viewers. Now that many DTV transmitters are up and running, the time is right for broadcasters to contemplate DTV carriage over cable.
Acknowledgement: The author would like to thank Michael Guthrie at Harmonic for his contribution in writing this article.
Bill Zou is Broadcast Solutions Marketing Manager for the Convergent Systems Division at Harmonic.
Home | Back to the top | Write us