There was a time when video formats were NTSC, PAL or SECAM. Today, it's not that simple. Not only do broadcasters have to contend with the 50Hz and 60Hz standard-definition systems, but there is also a plethora of HDTV formats in 16:9 aspect ratios, and a few SDTV 16:9 formats as well. To these, add progressive-scan and interlace variants, 24Hz adaptations, and many formats related to the NTSC color subcarrier (60/1.001). It presents an interesting bouquet of possible production and air formats.
As with any conversion between standards, format conversion can be either simple or very complex, depending on which two formats you choose to convert. Converting properly scaled analog components into digital components sampled to meet ITU-R BT601 (SMPTE 259M) is relatively easy. Filter properly, sample, quantize and code the samples for transmission.
Figure 1. Gennum’s GF9320 scaling processor makes converters less costly and complex. Click here to see an enlarged diagram.
But pick a different conversion and the degree of difficulty increases dramatically. Take, for example, a true 60Hz 720p60 HDTV signal and try to jam it into a 1080i59.94 transmission system. Spatial and temporal samples don't line up nicely for the conversion process. Several things have to happen. First those pesky progressive frames must be sliced and diced into interlace fields. Seven hundred and twenty lines must become 480. Secondly, in the next frame, make sure the vertical samples are interpolated from the correct lines in the 720 frame so they end up a smidge lower to line up with the even 1080i field. After that, begin again. The frame rate doesn't match; so temporally interpolate a few 720 frames to come up with a candidate frame to use for the vertical interpolation to the 1080i (540-line) field. With each successive pair of conversions, the temporally interpolated frame moves a smidge, so its control has to be sophisticated. The line rate is also a tad off frequency as well (1/1.001).
The math involved in the conversion process is complicated. Gennum, a Canadian chip manufacturer, sells a single chip solution to mathematically convert pictures. The GF9320 scaling processor replaces what was once a Ω rack of heat buildup. Arguably, a design using individual components that are optimized for the task might do a better job. But, for a first approximation, this is a pretty sophisticated and high-quality solution. Figure 1 shows the chip's capabilities.
There are several opportunities left to create great conversions or poor ones. The quality of any A/D and D/A involved in the process is a major factor. Getting the filter mask to match the standard precisely still requires good old-fashioned design talent. Other considerations include integrating a de-interlacing solution and sundry other required components. Clearly, this is no amateur design project for the basement. Many recent, modestly priced conversion boxes offer such integrated solutions, some of which are now confined to a single module in a D/A tray. One manufacturer, Cobalt Digital, sells a box a little larger than my Palm Pilot that does a pretty good job; it is designed using precisely this type of integrated solution.
Sophisticated, high-quality solutions are set apart by the quality of the optimized code that controls them, or, in some cases, by a custom integrated solution. For several years, Teranex has offered some of the best conversion equipment available. Its products come as an outgrowth of a processing engine designed initially for government applications. I don't suppose that was used for converting the images from the Hubble Telescope. But, in any event, they designed a box that used multiple processors on multiple boards with more computing power than any of us would know what to do with. This year, the company introduced a new solution, the XM series modular product, which fits in a D/A tray and produces about the same results.
It's pretty impressive, but it's clear that the science and engineering involved are becoming mature. When that happens, competition pushes prices down rapidly. This is happening now, just in time for the serious implementation of HDTV production and distribution systems.
It is valuable to remember why conversion products exist. They are the bridge between eras of technology or between competing economic interests. The 525 and 625 standards coexisted for many years with little interchange before the first converters were put in service in Europe in the 1960s, when Telstar first beamed live television between continents. Telstar 1 launched in July 1962. It permitted only minutes of live connection, but it required a bridge more immediate than kinescope copies sent by commercial aircraft. Today, we see the overlap of NTSC and PAL with HDTV. At some point, other media will supplant HDTV. In all cases, the bridge period requires a new device to enable both technologies to succeed simultaneously. So long as we don't have one world, standard conversion will remain a fixture in the engineer's toolkit.
It is logical to look to a product range that can be used for coversion in both directions. For instance, today we have converters with single inputs that accept either analog or digital inputs and convert to the opposite without reprogramming. The same is true of standards converters that auto sense 625 or 625, or 1080i and 720p, or combinations thereof, allowing outstanding flexibility. This kind of auto-adapting bridge is being moved into master control and routing switchers and even production switchers. At some point in the not too distant future, you will be able to plumb up a system without being too worried about the boxes talking to each other, leaving you to concentrate on programming them for the functions you want them to perform. System integration will become “function integration”…and then it'll be time for me to retire!
John Luff is senior vice president of business development for AZCAR.
Send questions and comments to: firstname.lastname@example.org