Component Digital Video

We are well aware that composite NTSC video (and PAL video as well) consists of a wideband luminance signal and two chrominance signals of much lower bandwidth that are modulated onto a subcarrier, (double sideband-suppressed carrier signal, to be precise).

We know that most of the energy contained in the luminance signal is concentrated in "lumps" that recur at the scanning line rate (15,734 Hz in the case of NTSC) across the occupied bandwidth. We know, too, that the modulated chrominance subcarrier contains lumps of energy at the horizontal line rate and that its frequency is carefully chosen so that the chrominance lumps just fit in the holes between the luminance lumps.

We must transmit this composite signal over the air and - for many years - most of the equipment between camera and transmitter used it as well, although that equipment often does not have the bandwidth constraints that are necessarily imposed on over-the-air NTSC. It is no secret that the composite signal has some shortcomings and that it is not particularly difficult to degrade it in a variety of ways - one of the easiest being to repeatedly re-record it or reprocess it for a number of generations.

STANDARDIZED DIGITAL

The search for a higher quality, more robust and more "processable" video signal ultimately led us to component digital video. Back in the 1980s, component digital, standard-definition video was standardized in the CCIR (known today as the International Telecommunications Union) recommendation now called ITU-R BT.601. Rec. 601, in fact, describes a family of video signals - the most well-known, 4:2:2 component digital video sampled at 13.5 MHz - has become commonly known as "601 video."

An early application of 4:2:2, 601 video was as an input/output interface for the first digital video recording format standardized by SMPTE known as D-1. Thus, the 601 interface signal is frequently referred to as a D-1 signal, although we know that D-1 is really the name of a tape format and not of a signal interface.

Let's look at the 601 family, using as our reference the latest 601 recommendation, ITU-R BT.601-5. When the 601 standard was developed, there was a desire to make it applicable to both 525-line and 625-line television systems - and for there to be as much commonality between the 525-line and 625-line versions of the system as possible. The primary sampling frequency of 13.5 MHz was chosen, as this is a number compatible with both systems.

Rec. 601 created an extensible family of compatible digital coding standards. The most common member of this family is based on the use of one luminance signal and two color difference signals. In this case - as in NTSC and PAL - a wideband luminance signal is accompanied by two color difference signals of narrower bandwidth, although they are of much greater bandwidth than those of composite NTSC or PAL.

The (gamma-corrected) luminance signal is sampled at 13.5 MHz, while the (gamma-corrected) color difference signals, CR (R-Y) and CB (B-Y), are each sampled at half this frequency, or 6.75 MHz. Using the recommended anti-aliasing filters, the analog bandwidth of the sampled luminance signal is 5.75 MHz, while the analog bandwidth of each of the sampled color difference signals is 2.75 MHz. The signals are linearly quantized PCM at a bit depth of either 8 or 10 bits.

The family just described is called 4:2:2. Theories abound as to where these numbers came from, but the "4" represents a sampling frequency of 13.5 MHz and the "2"s represent sampling frequencies of 6.75 MHz. This sampling yields a total of 858 samples per line when sampling 525/59.94 signals and 864 total samples per line when sampling 625/50.

If we divide 864 by 858, the ratio - carried to 8 decimal places - is 1.00693300, which is exactly the ratio of 144 divided by 143. Not coincidentally, if we derive the prime factors of 525/59.94 scanning (525 x 60/2 x 1000/1001) and multiply them together, their product is 144; likewise, if we derive the prime factors of 625/50 scanning (625 x 50/2), and multiply them together, their product is 143.

The number of samples per active line is set at 720 in both the 525-line and 625-line manifestations of the family.

MULTIPLE GENERATIONS

The post-production industry discovered long ago that storing and processing video in the component digital format permitted multiple generations of recording and processing without substantial degradation of the images. It took broadcasters a bit longer to make the transition to component digital video in their plants, but that transition is well underway today.

What about the other members of the 601 family? There are two 4:4:4 members. One of these is 4:4:4 RGB, in which R, G and B analog signals are sampled at 13.5 MHz each. Another is full-bandwidth Y, CR, CB, in which the color difference signals as well as the Y or luminance signal are sampled at 13.5 MHz. These 4:4:4 signals generate very large amounts of data and there are few recording formats that can accommodate them without applying some degree of data reduction.

COMPONENT COMPROMISE

Rounding out the 601 family are the 18 MHz versions of 4:2:2 and 4:4:4. The 13.5 MHz members may be used for the 16:9 aspect ratio as well as for 4:3. This permits 16:9 signals to fit into the same bandwidth and storage devices as 4:3, but there is a compromise, of course. When 13.5 MHz sampling is used for the 16:9 aspect ratio, the 720 active horizontal samples must be "spread" over a line that is one-third longer, relative to picture height, than a line in the 4:3 aspect ratio.

The horizontal resolution is thereby reduced, although there has in the past been vigorous debate about whether this resolution reduction is visible to the average viewer. If the sample frequency is raised to 18 MHz, the total number of luminance samples per line goes up to 1144 (572 color difference samples in 4:2:2) in 525-line systems and 1152 (576 color difference samples in 4:2:2) in 625-line systems, while the number of luminance samples per active line goes up to 960 (480 color difference samples in 4:2:2).

This generates one-third more samples per active line; the net result is that the horizontal pixel density in 16:9 is equal to that achieved with 13.5 MHz sampling in 4:3, at the expense of a wider bandwidth signal. There is at least one VTR on the market capable of operating at the 18 MHz sampling rate, but this member of the 601 family has not enjoyed a great deal of use.

There are also 4:4:4 members of the family (RGB and YCBCR), but your author is not aware that either of these has ever been implemented in a product.

We still require NTSC to broadcast analog television today, but digital recording, processing and distribution systems have made it possible to operate in the component domain throughout most of the broadcast plant and it is doubtful that we will see much new NTSC (or PAL) plant equipment developed for use anywhere between the camera and the transmitter.

Randy Hoffner