Digital TV in the U.S.

Squeeze plays ¿DTV compression and scaling

In DTV, data compression shoehorns the millions of bits of sight and sound information into the narrow broadcast pipeline and enables them to emerge at the receiving end as real-time audio and video. In the mean time, scaling-and image-shaping techniques are employed to redraw a 2-mega-pixel image on a canvas that can display only a fraction of the picture data available.

For compression, an example that illustrates the challenge is to be found in a chart of DTV formats [p. 42]. If a single frame of 1080-line interlaced DTV comprises some 2 million pixels, and there are 30 frames in each second of video, the uncompressed data transfer rate then is 995 Mb/s. Because the maximum data transfer rate for the 6-MHz DTV channel is just 19.4 Mb/s, the data must be compressed powerfully and quickly before transmission-then quickly decompressed at the DTV set.

MPEG-2 compression (used for DTV, DVD, and elsewhere) ignores redundant picture data so as to economize on bits. In a movie, for example, backgrounds that change little from frame to frame need not be re-encoded bit for bit. Unchanged blocks of pixels can be coded to repeat, and only updates within the scene need be re-encoded, such as shadows on an actor's face caused by head and lip movement.

Encoding is more easily accomplished with pre-recorded programming, such as movies, than with live action such as sports. Although used for both, predictive analysis and motion compensation are indispensable in accounting for scene changes on-the-fly in live programming. As one might imagine, live action broadcasts such as sports triage the background detail to bring the most bits to bear on the live action afield.

Meanwhile, in the pictures-by-numbers game, typical CRT-based DTV displays cannot put HDTV's entire 2-megapixel image on screen. Most so-called HD displays deliver 1to 1.5 megapixels at best, and as for the 480-line progressive scan SO displays, they field just a fifth of all incoming HD-l080i signal.

Credit mechanical tolerances and physical laws with contributing to the conundrum, at least where inherently analog CRTs of direct-view or projection-type are concerned. Digitally addressed liquid-crystal display, plasma, and other display technologies could beat the rap, but cost too much for consumers as yet.

With direct-view TVs, the beam-spot size is governed by mechanical factors. Although shadow mask/aperture grille perforations could be made finer, picture brightness would suffer. Meanwhile, compensating for brightness with greater output from the electron guns raises questions about metal-tolerance of the mask or the grille-like at what point does misshaping under hot bombardment from electrons distort the image?

This omits the issue of power consumption, or how much juice the consumer is willing to spring for unless the warmth can be recycled throughout the house and defray home-heating bills from the local utility company.

For the record, Thomson Consumer Electronics, Indianapolis, Ind., specs its 38-inch widescreen direct-view set at 1 megapixel—a compromise that, to quote tube-division vice president Tom Carson, balances mechanical and environmental factors with a resolution that is crisp at typical viewing distances (roughly 10 feet) for a set that size.

Metal masks are not an issue with projection sets, but lens-diameter size for the three (red, green, and blue) CRTs is. At present, 9-inch guns constitute the heavyweight artillery for DTV, although some manufacturers favor 7-inch batteries for reasons of reliability and practicality. Theoretically, a CRT lens can be of infinite size to yield a spot beam that is infinitesimal. But, in the real world, the set has to be light in shipping weight and able to negotiate the threshold of a typical home's front door. Again, for comparison's sake, Thomson's 6l-inch ProScan HDTV rear projector throws 1.5 megapixels onscreen—a resolution that executive vice president and chief operating officer Jim Meyer has said the company is comfortable with in terms of picture height in relation to viewing distance.

All the same, what happens when a 2-megapixel 1080-line interlaced HDTV broadcast is displayed on a 1.5-megapixel HDTV set or, for that matter, an SDTV screen with fewer than 400 000 pixels? It is not a matter of decimation—such as killing of digital-footmen on an arbitrary or pre-determined basis, according to engineers brave enough to field the query. Instead, downconversion is handled by yet another sleight-of-digits in the DTV trick-bag, generically called scaling.

Dean Malmstrom, head of Sarnoff Corp.’s DTV Professional Products Group in Princeton, N.J., told IEEE Spectrum that although typical DTVs decode every bit transmitted, an individual set selects the key pixels (actually, macro-blocks of pixels) needed to display the image according to the set's display capabilities (aperture-grille, shadow-mask, or some other type).

Subsequently, the set performs some noise shaping and other enhancements on the otherwise shortchanged picture. How well this triage-and-massage is performed differentiates one brand's DTV from another, according to Sony DTV guru Phil Abrams. Proof of the pudding? Get thee to a TV store, and compare like broadcasts or canned demos on the variety of DTVs on display. — S.A.B.