Image compression has received a great deal of attention as a key technology for making digital video practical. But compression is only one part of the answer. Putting digital video into the consumer's hands also requires a delivery method that utilizes existing wiring and cable. Digital-video transmission is one unsung technology that's evolving to fill this need.

Digital video is already demonstrating its acceptability in the marketplace. Broadcasters, cable companies, and telephone companies are actively conducting field trials in a variety of locations, including northern Virginia, the San Francisco Bay area, New York City, and Orlando, FL. These trials are using direct digital-transmission media such as fiber optics and ISDN telephone lines.

In the San Francisco Bay area, for example, television-station KRON (Chronicle Broadcasting) is using MPEG-1 compression hardware and software from FutureTel and T1 digital telephone lines to distribute KRON's BayTV cable channel to a number of cities. Digital video travels to local cable-TV headends over the phone lines. The cable stations then convert the signal to traditional analog form for local distribution.

Although these trials demonstrate the acceptability of digital video to consumers, many television-industry observers see digital telephony as only one step toward widespread digital-video distribution. Few consumers have high-bandwidth digital links to their homes, and the numbers are not increasing rapidly. Despite the publicity, ISDN telephony is still many years from widespread availability.

Entertainment providers realize that for digital video to become a commercial success, existing analog cable networks and broadcast channels must be customized to handle high-speed digital signals. As an added bonus, if those networks can handle digital data, they can also serve as pathways for computer communications, providing additional market opportunity.

Therefore, there is a relatively unheralded push to develop technology for packing digital data into existing media. The best known effort has been the drive to create a digital HDTV that is bandwidth compatible with existing broadcast TV. Other, quieter efforts are also underway to use standard telephone and cable-television services as the digital channels.

These existing channels present a number of barriers to digital-information transmission. For one, their bandwidth is limited. Twisted-pair telephone wiring has an effective bandwidth of approximately 500 kHz over a three-mile distance. Broadcast-television channels are constrained to a 6-MHz bandwidth. Satellite-transponder channels are typically 36 MHz wide. An average cable-TV distribution network has a bandwidth of 350 to 700 MHz.

Other barriers include the channels' abilities to maintain signal integrity to the precision needed. Broadcast signals suffer from reflections and multipath distortion. Telephone and cable networks generate multiple reflections from impedance mismatches that change frequently as receiver units connect and disconnect from the network. Satellite broadcasts are power-constrained, resulting in low S/N ratios.

Coding bypasses bandwidth limits

The major concern, however, is the mismatch between digital-video bit rates and channel bandwidth. Nyquist showed that the bandwidth of an ideal pulse is the reciprocal of the pulse width. Assuming that data pulses are to be as wide as possible, for example, equal to the pulse period, channel bandwidth is equal to signal-pulse rate. Thus, a straight bipolar digital signal, such as a typical logic signal, occupies a bandwidth equal to the data bit rate. A raw full-color, digital-video signal needs a minimum of 30 Mbytes/sec to achieve full resolution--a bandwidth of 240 MHz. Only the cable system would appear to have sufficient capacity to carry digital video.

The key to using the consumer signal channels, then, is compacting digital signals so that they fit within the channel's bandwidth. MPEG compression can lower digital-video's data rate to 1.2 to 1.5 Mbps while maintaining acceptable picture quality. This fits within cable- and broadcast-channel bandwidths, but it provides digital video with no particular advantage over analog. And, without further compaction, the digital bit rate exceeds the typical bandwidth of existing consumer telephone channels.

Fortunately, Nyquist's bandwidth calculations only constrain the signal's pulse width, not its amplitude. If the signal pulse employs a multilevel code, its information-carrying capacity increases without adding bandwidth. In information theory, Shannon's Limit states that the information-carrying capacity, C, of a channel with bandwidth, W, signal power, S, and random-noise power, N, is

C=W×log₂(1+{S/N}).

For a 10-dB S/N ratio channel, therefore, as many as 10 bits can be transmitted for each Hz of channel bandwidth. The trick is to encode the data within a single pulse.

Codes utilize information capacity

The choice of an encoding scheme depends on a number of factors. Primary factors are the transmission medium's bandwidth and S/N ratio. The coding scheme's susceptibility to interchannel interference, decoding costs, and ability to handle multipath and reflections are also important. To date, at least three major schemes have been developed to handle digital-video transmission.

The first and most well-developed scheme is quadrature amplitude modulation (QAM). For a number of years, this scheme has been used in microwave telecommunications. It has also been the modulation method of choice for high-speed digital modems over standard twisted-pair telephone lines. See Fig 1 for a block diagram of a QAM modulator.

QAM uses two carrier waves of the same frequency in quadrature--that is, with a 90° phase shift between them. The modulation scheme takes data in two groups of two, three, or four bits to drive two ADCs. The two ADCs amplitude-modulate the two carrier waves. The QAM output signal is the sum of the two modulated carriers.

A plot of the resulting waveform in complex vector coordinates yields a constellation of points, one for each possible data bit pattern. Thus, a QAM signal using two groups of 3 bits at each modulation step produces a constellation of 64 points in complex coordinates (Fig 2). The signal carries six data bits for each Hz of bandwidth.

A competing technology to QAM is the vestigial-sideband (VSB) modulation scheme, which Zenith proposed for spectrum-compatible HDTV. VSB technology uses multilevel amplitude modulation on a single carrier, then filters the signal to remove most of the lower modulation sideband. VSB provides only a one-dimensional vector pattern--versus QAM's two-dimensional approach. However, because VSB uses only a single modulation sideband, its spectral efficiency is equivalent to that of QAM.

AT&T Paradyne developed a third modulation scheme, carrierless amplitude-phase (CAP) modulation (Ref 2). CAP is a variation of QAM that uses direct digital-signal generation rather than modulating a carrier wave (hence, carrierless in the name). CAP modulation separates data into two groups, using one group to specify the imaginary component of the signal vector. The other group specifies the real component. The vector, which changes at the data symbol rate, passes through a pair of digital bandpass filters with the same amplitude responses--but phase responses that differ by 90ø. The final output signal is the sum of the two digital-filter-output signals passed through an A/D converter.

QAM is established technology

In the race to deliver these technologies to the consumer's set-top decoder box, QAM has a decided advantage. The technology is well-established, having been used for years in microwave links and high-speed modems. Further, there are now monolithic ICs available that implement a QAM decoder for video signals.

One such chip set comes from the joint efforts of Broadcom Corp and Scientific Atlanta. The QAMLink chip set includes a QAM demodulator IC, an adaptive equalizer, and a synchronization IC that provides digital loop filters and phase detectors for the digital phase-locked signal recovery loops. The chip set is capable of handling 64- or 256-symbol QAM, corresponding to 6 or 8 bits per Hz of channel bandwidth.

Such bit densities allow a 6-MHz video channel (5 MHz after filter rolloff) on cable systems to carry 30 to 40 Mbps of digital data--enough for 20 MPEG-1 or 10 MPEG-2 video data streams. Thus, a typical 350-MHz cable system bandwidth could handle as many as 500 MPEG-2 video channels using the QAMLink chip set. Even a 350-kHz telephone channel could carry a single MPEG-1 channel.

The impact of such information capacity on the home goes beyond the frightening ability to have all episodes of "I Love Lucy" running simultaneously. The information received at the set-top box is all digital and need not represent video. Thus, the advent of high-speed video transmission opens a doorway for digital services of all forms to reach the consumer through the TV lead-in.

1. Reitmeier, Glenn A, Richard J Knensch, and Hugh E White, "Channel coding approaches and consequences--single and dual carriers," SMPTE Journal, September 1994, pg 626.

2.Prunty, Peter F, "Delivery of TV over existing phone lines," SMPTE Journal, September 1994, pg 586.