Video Asylum

Posted by Homer (A) on March 30, 2002 at 02:31:54

In Reply to: How is it 35Mhz? posted by lwang on March 29, 2002 at 08:38:08:

Video systems don't generally use pixels; they use video samples. This means that the bandwidth of a reconstructed signal will really only be half the sampling frequency.

Pixels are good for computers, because they're mathematically easy to work with and conceptualize, and the harmonics let you get cleaner edges, at a given resolution. The price you pay is that aliases get into your signal. So, for example, when a full-bandwidth image is being slowly panned, you'll notice a certain amount of popping and noise. To compensate for this, people have developed "anti-aliasing" techniques, which selectively limit the bandwidth of the signal, in order to reduce these display artifacts. Pixels are also a fairly accurate model for the limitations of image reproduction, on LCD or DLP-based displays. (Note that this doesn't mean you need higher bandwidth cables, between your DVD player and LCD projector - only between your computer and LCD projector.)

Conversely, pixels aren't good for video. Within a scanline, a video signal is reconstructed from samples using a sinc reconstruction filter (as opposed to a box). This means you can freely use the full theoretical bandwidth, without having to worry about artifacts introduced by aliasing. For signals along the vertical axis of the screen, more aggressive pre-filtering should be done, since the result is reconstructed with a gaussian (the beam shape) FIR, which does admit a certain amount of aliases.

Therefore, you should only need nominally more bandwidth than 31 MHz, in order to account for horizontal and vertical blanking. When you're looking at bandwidth specs, be sure that the passband is qualified by the amount of allowable attenuation (i.e. -3dB). Otherwise, the number is meaningless.

This whole issue seems to be the greatest single point of confusion, between the general public (and even some professionals and reviewers) and the engineers who write the specs and build the equipment. I guess basic signal processing concepts really need to find their way into high-school curriculums.