London facility On Sight uses the Alchemist Platinum Ph.C from Snell & Wilcox for upconversion from SD to HD of general material for U.S. and UK consumption.
To paraphrase a popular statement, the nice thing about formats is that there are so many of them. In our industry, we have analog and digital formats with different frame rates, coding schemes, scan formats, bandwidth, aspect ratios, compression systems, bit rates, bit depths, file formats and interconnection media. Increasingly, production personnel expect that any format can be converted to any other. That may well be precisely true, but as in all things technical, it is not always practical or affordable given the economics of any particular application.
The most important example today might be conversion between the various HDTV and SDTV formats. Manufacturers of hardware select from support silicon to find high-quality and low-cost solutions, or they use considerably more hardware processing power to craft a dedicated solution at the highest quality possible, perhaps at the expense of low cost. The result is that there are an array of single-board solutions at modest prices ($3-$10,000), as well as a family of much more expensive products that can cost five times that much for the highest possible performance.
Inside each of these products, one must do several functions, none of which is intrinsically simple. Converting picture formats, for instance 486×720 to 1920×1080, used to be the province of high-cost standards converters, especially when HDTV signals were involved. That is just the starting point. Color space conversion is often needed, with HDTV standards using a wider gamut than older 525 and 625 scanning systems. Progressive to interlace conversion is well understood and important. Lastly, and most difficult, is frame rate conversion.
The right way to do frame rate conversion is to interpolate missing time samples to recreate smooth motion in the target format. It is cheaper to duplicate or drop frames when data underflows, or overflows, with predictably jarring results. When the frame rates are closest together, as in 60Hz to 59.94Hz conversion, the repetition rate is quite long, and the difficulty of doing a smooth temporal conversion is highest. When there is an integer relationship, as in 24 frame to 30 frame, the conversion is much more straightforward, and the number of frames needed to do a thoughtful interpolation is more restricted (actually 23.976 to 29.970 in real-world implementations, both being related to the integer rate by the vertical offset of 1/1.001).
In a production environment, it is often the case that aspect ratio is the most obvious conversion. Deciding how to handle the active picture content is not always straightforward. For instance, take the case of commercials in a downconverted HDTV program. As long as the content is consistent, you could choose to either letter box or crop the sides. But if a commercial that came from a 4:3 source run in the HD program is presented in “pillar box” mode, the result of the down-conversion and letter box would be a commercial that originally occupied 80 percent of the screen area as a pillar box, but only 45 percent when converted back to SDTV as a pillar box inside a letter box.
A second complication arises when a production is intended for “dual release,” i.e. in both 16:9 and 4:3. For example, Hollywood for years has been shooting television content in 16:9 with 4:3 framing protected (technically 16:9 is “protected” when shooting 4:3). Now that a serious amount of sports content is being produced for dual release, it has become quite clear that graphics are a major challenge. When a lower third fits the 4:3 release, it often looks odd in 16:9 release. On the Super Bowl this year, FOX did an excellent job of finding acceptable compromises that left both releases of the program looking normal, though 4:3 was obviously the target that was favored. Think about a replay transition; when the picture first hits the edge of the 16:9 frame and the audio effect is heard, it might be several frames before the 4:3 picture shows the transition, which might not be acceptable.
Cross-conversion between the two predominant HDTV formats entails several simultaneous conversions. 720p to 1080i, the most common conversion in at least some facilities, involves both progressive to interlace conversion as well as line and vertical formats, and frame rate as well. At least the color space and aspect ratios are the same! Making an interlace frame from a progressive is not hard, for the time samples exist on each line of each frame in the 60p source. Making interlace amounts to throwing away half of the lines in a simplistic view, and then rescaling the picture up to 1080 lines by 1920 samples. Going the other way requires a de-interlacing step, which is not as friendly to the new format as information on successive fields were not sampled at the same time. The practical results achieved by modern hardware are quite acceptable going both directions.
At one time, when the P vs. I “battle” raged early in the HDTV transition, I was involved in a major effort in 720p which the producer had sold to NHK, decidedly in the 1080i opposite camp. Even then everyone was quite surprised that the end result was remarkable and quite acceptable to all. That same result is available in much cheaper hardware today.
Also remarkable is the variety of form factor that format conversion devices come in today. Some are plug-in options in videotape recorders. Others are several rack units and offer noise and grain reduction, and advanced temporal interpolation. Several manufacturers have created single-board converters using integrated image scaling silicon, which has been on the market for several years.
Our industry has benefited from research and manufacturing done for the consumer electronics industry. Essentially, every HDTV consumer display, or at least set-top receiver, must be able to handle a variety of formats. To get the cost to acceptable levels, a great deal of research has been done, and the quality of products is remarkable. For the highest quality conversion, I am sure designers can still eek more performance out of more complex products, but as professionals we benefit from the economies of scale that consumer electronics bring to bear (again).
John Luff is senior vice president of business development at AZCAR.
Send questions and comments to:firstname.lastname@example.org