See Table 15.1, on page 143, and the associated discussion.

767 = 944 52 µs

64 µs

Widescreen (16:9) SD

Programming in SD is intended for display at 4:3 aspect ratio. Prior to (and during) the development of HD, several schemes were devised to adapt SD to widescreen (16:9) material – widescreen SD. That term is misleading, though: Because there is no increase in pixel count, a so-called widescreen SD picture cannot be viewed with a picture angle substantially wider than regular (4:3) SD. (See page 75.) So widescreen SD does not deliver HD’s major promise – that of dramatically wider viewing angle – and a more accurate term would be wide aspect ratio SD. The various schemes devised in the transition period are now obsolete. A discussion is found in Widescreen (16:9) SD on page 5 of Composite NTSC and PAL: Legacy Video Systems.

Square and nonsquare sampling

Computer graphics equipment now universally employs square sampling – that is, a sampling lattice where pixels are equally spaced horizontally and vertically. Square sampling of 480i and 576i is diagrammed in the top rows of Figures 13.1 and 13.2 on page 131.

Although ATSC’s notorious Table 3 includes

a 640× 480 square-sampled image, no studio standard or realtime interface standard addresses square sampling of SD. For desktop video applications, I recommend sampling 480i video with exactly 780 samples

per total line, for a nominal sample rate of

__

123⁄11 MHz – that is, 12.272727 MHz. To accommodate full picture width in the studio, 648 samples are required; often, 640 samples are used with 480 picture lines. For square sampling of 576i video, I recommend using exactly 944 samples per total line, for a sample rate of exactly 14.75 MHz.

MPEG-1, MPEG-2, DVD, and DV all conform to BT.601, which specifies nonsquare sampling. BT.601 sampling of 480i and 576i is diagrammed in the middle rows of Figures 13.1 and 13.2.

Composite digital video systems historically sampled at four times the colour subcarrier frequency (4fSC), resulting in nonsquare sampling whose parameters are shown in the bottom rows of Figures 13.1 and 13.2. As I stated on page 128, composite 4fSC systems are obsolete.

4fSC,PAL-I 709379

In 480i, the sampling rates for square sampling, BT.601, and 4fSC are related by the ratio 30:33:35. The pixel aspect ratio of BT.601 480i is exactly 10⁄11; the pixel aspect ratio of 4fSC 480i is exactly 6⁄7.

In 576i, the sampling rates for square sampling and 4:2:2 are related by the ratio 59:54, so the pixel aspect ratio of 576i BT.601 is precisely 59⁄54. BT.601 and 4fSC sample rates are related by the ratio in the margin, which is fairly impenetrable to digital hardware.

Most of this nonsquare sampling business has been put behind us: Most HD studio standards call for square sampling, and it is difficult to imagine any future studio standard being established with nonsquare sampling.

Resampling

Analog video can be digitized with square sampling simply by using an appropriate sample frequency. However, SD already digitized at a standard digital video sampling rate such as 13.5 MHz must be resampled – or interpolated, or in PC parlance, scaled – when entering the square-sampled desktop video domain. If video samples at 13.5 MHz are passed to a computer graphics system and then treated as if the samples are equally spaced vertically and horizontally, then picture geometry will be distorted. BT.601 480i video will appear horizontally stretched; BT.601 576i video will appear squished. In desktop video, often resampling in both axes is needed.

The ratio 10⁄11 relates 480i BT.601 to square sampling: Crude resampling could be accomplished by simply dropping every eleventh sample across each scan line! Crude resampling from 576i BT.601 to square sampling could be accomplished by replicating

5 samples in every 54 (perhaps in the pattern 11-R-11-R-11-R-11-R-10-R, where R denotes

a repeated sample). However, such sample dropping and stuffing techniques introduce aliasing. I recommend that you use a more sophisticated interpolator, of the type explained in Filtering and sampling, on

page 191. Resampling could potentially be performed along either the vertical axis or the horizontal (transverse) axis; horizontal resampling is the easier of the two, as it processes pixels in raster order and therefore does not require any linestores.

NTSC stands for National Television System Committee. PAL stands for Phase Alternate Line. (Some sources say that PAL stands for Phase Alternation at Line rate, or perhaps even

Phase Alternating Line).

SECAM is a composite technique of sorts, though it has little in common with NTSC and PAL, and it is now obsolete. See “SECAM,” in Chapter 12 of Composite NTSC and PAL: Legacy Video Systems.

In component video, the three colour components are kept separate. Video can use R’G’B’ components directly, but three signals are expensive to record, process, or transmit. Luma (Y’) and colour difference components based upon B’-Y’ and R’-Y’ can be used to enable subsampling: Luma is maintained at full data rate, and the two colour difference components are subsampled. Even after chroma subsampling, video has a fairly high information rate (data rate, or “bandwidth”). To further reduce the information rate, the composite NTSC and PAL colour coding schemes use quadrature modulation to combine the two colour difference components into a single modulated chroma signal, then use frequency interleaving to combine luma and modulated chroma into a composite signal having roughly 1⁄3 the data rate – or in an analog system, 1⁄3 the bandwidth – of R’G’B’.

Composite encoding was invented to address three main needs.First, there was a need to limit transmission bandwidth. Second, it was necessary to enable black- and-white receivers already deployed by 1953 to receive colour broadcasts with minimal degradation. Third, it was necessary for newly introduced colour receivers to receive the then-standard black-and-white broadcasts. Composite encoding was necessary in the early days of television, and it has proven highly effective for broadcast. NTSC and PAL are entrenched in billions of consumer electronic devices. However, component digital video has overtaken composite techniques, and composite NTSC and PAL are now “legacy” techniques.

By NTSC and PAL, I do not mean 480i and 576i, or 525/59.94 and 625/50!

When I use the term PAL in this chapter, I refer only to 576i PAL-B/G/H/I. Variants of PAL used for broadcasting in South America are discussed in PAL-M, PAL-N, on page 125 of Composite NTSC and PAL: Legacy Video Systems. PAL variants in consumer devices are discussed in Consumer analog NTSC and PAL in Composite NTSC and PAL: Legacy Video Systems.

Composite NTSC or PAL encoding has three major disadvantages compared to component video. First, encoding introduces some degree of mutual interference between luma and chroma. Once a signal has been encoded into composite form, the NTSC or PAL footprint is imposed: Cross-luma and cross-colour errors are irreversibly impressed on the signal. Second, it is impossible to directly perform many processing operations in the composite domain; even to reposition or resize a picture requires decoding, processing, and reencoding. Third, digital compression techniques such as JPEG and MPEG cannot be directly applied to composite signals, and the artifacts of NTSC and PAL encoding are destructive to MPEG encoding.

The bandwidth to carry separate colour components is now easily affordable, and composite encoding is now obsolete in the studio. To avoid NTSC and PAL artifacts, to facilitate image manipulation, and to enable compression, composite video has been superseded by component video, where three colour components R’G’B’, or Y’CBCR (in digital systems), or Y’PBPR (in analog systems), are kept separate. I hope you can manage to avoid composite NTSC and PAL, and skip this chapter!

The terms NTSC and PAL properly denote colour encoding. Unfortunately, they are often used incorrectly to denote scanning standards. PAL encoding has been used with both 576i scanning (with two different subcarrier frequencies) and 480i scanning (with a third subcarrier frequency); PAL alone is ambiguous.

In principle, NTSC or PAL colour coding could be used with any scanning standard. However, in practice, NTSC and PAL are used only with 480i and 576i scanning, and the parameters of NTSC and PAL encoding are optimized for those scanning systems. This chapter introduces composite encoding. Details can be found in

Composite NTSC and PAL: Legacy Video Systems.

NTSC and PAL encoding

NTSC or PAL encoding involves these steps:

• R’G’B’ component signals are matrixed and filtered, or Y’CBCR or Y’PBPR components are scaled and filtered,