Sorry, Wrong Number

A couple of questions: Might it be possible for all of the electronic processing functions of a complete television studio to be reduced to a single chip? And, if so, might that chip also be able to implement the “ultimate compression system” described here in November?

The second question is easier to answer. Whatever else the hypothetical production chip could do, it couldn’t accomplish the described compression, because there’s no such thing.

That column posited that the total number of 36-bit 4096 x 4096 frames that could ever exist would be 4096 x 4096 x 4096 x 4096 x 4096. That’s a fairly large number, close to 1.2 quintillion. Unfortunately, it’s not nearly large enough—not by a long shot.

The problem is in the third x above. It’s the wrong operator. The 1.2-quintillion number is the quantity of color values one would have to have in bins to create any 4096 x 4096 frame that could exist, like the characters kept in cases by typesetters. There are 4096 x 4096 x 4096, or about 69 billion, values in 36 bits, and there are 4096 x 4096, or about 17 million, possible locations for any of those values. There would, therefore, need to be 17 million copies of each of the 69 billion values, because every pixel in a flat field would have the same value.

That’s pixels, however, not frames. The frames depend on the positions of the pixels. If the first pixel can have 69 billion values and the second another 69 billion, then just those two pixels have 69-billion squared (close to five sextillion, five followed by 21 zeroes) possibilities. But there are 17 million pixels, not just two. So the total number of frames would be roughly 69 billion raised to the 17-millionth power, not 69 billion times 17 million.

Such a calculation is mind-boggling (and computer stumping), so try something simpler just to get the idea. Assume a trivial video system with frames just four pixels wide and three lines high. Assume each of those 12 pixels can have only 256 values (eight bits).

The number of values that would have to be kept in bins to be able to build any frame would be just 256 x 12 or 3072. But the number of possible different frames would be 256 raised to the 12th power or about 79 octillion (79 followed by 27 zeroes), and that’s for trivial, 12-pixel, 8-bit frames. There would be no compression accomplished by indexing the frames.

November’s column, therefore, was colossally, spectacularly wrong. So is the idea that electronics can take the place of operators. In other words, a mobile television production unit based around that single hypothetical chip would likely be the same size as today’s units, if not larger (the weight and heat load of today’s electronics prevent even more onsite vehicle-expansion mechanisms to accommodate more staff).

Considering just the front bench of a typical mobile or fixed control room, even a one-chip truck could still need a director, associate director, technical director and script supervisor. Room would also still be needed for the staff working on audio, communications, graphics, maintenance, playback, prompting, recording, scheduling/timing and video painting. And those staff would need to be able to monitor audio and video signals.

Although the one-chip truck is not a frequent topic of conversation, the supposed labor savings of HDTV has been. Once there was a theory that the wider, more-detailed images of HDTV could reduce the number of cameras used and the number of cuts. But close-ups are used in even highly detailed theatrical movies. In the 1968 Super Panavision 70 movie 2001: A Space Odyssey, a single human eye sometimes filled the entire screen.

Given the fine detail of theatrical motion pictures, it’s also odd that some of the best-looking Hollywood stars (e.g., Cameron Diaz and Brad Pitt) are said to look less good on HDTV. If they do, it cannot be because HDTV offers significantly more detail than theatrical film. But it can be because someone decided that HDTV didn’t need as much care or effort as theatrical film in such areas as makeup, hair styling, lighting and lens filtering.

Don’t make a wrong-number mistake like the ones in the November column. Keep the right operators, and offer them appropriate support.

Mark Schubin is an engineering consultant with a diverse range of clients, from the Metropolitan Opera to Sesame Workshop.