Digital component video background
Key Concept
Digital component video is digital video that uses three separate color components, such as R9G9B9 or YCbCr. In digital component video, the video signals are in digital form (YCbCr or R9G9B9), being encoded to composite NTSC, PAL, or SECAM only when it is necessary for broadcasting or recording purposes.
The European Broadcasting Union (EBU) became interested in a standard for digital component video due to the difficulties of exchanging video material between the 576i PAL and SECAM systems. The format held the promise that the digital video signals would be identical whether sourced in a PAL or SECAM country, allowing subsequent encoding to the appropriate composite form for broadcasting. Consultations with the Society of Motion Picture and Television Engineers (SMPTE) resulted in the development of an approach to support international program exchange, including 480i systems.
A series of demonstrations was carried out to determine the quality and suitability for signal processing of various methods. From these investigations, the main parameters of the digital component coding, filtering, and timing were chosen and incorporated into the ITU-R BT.601 standard. BT.601 has since served as the starting point for other digital component video standards.
Coding Ranges
The selection of the coding ranges balanced the requirements of adequate capacity for signals beyond the normal range and minimizing quantizing distortion. Although the black level of a video signal is reasonably well defined, the white level can be subject to variations due to video signal and equipment tolerances. Noise, gain variations, and transients produced by filtering can produce signal levels outside the nominal ranges.
8 or 10 bits per sample are used for each of the YCbCr or R′G′B′ components. Although 8-bit coding introduces some quantizing distortion, it was originally felt that most video sources contained sufficient noise to mask most of the quantizing distortion. However, if the video source is virtually noise-free, the quantizing distortion is noticeable as contouring in areas where the signal brightness gradually changes. In addition, at least two additional bits of fractional YCbCr or R′G′B′ data were desirable to reduce rounding effects when transmitting between equipment in the studio editing environment. For these reasons, most pro video equipment uses 10-bit YCbCr or R′G′B′, allowing 2 bits of fractional YCbCr or R′G′B′ data to be maintained.
Initial proposals had equal coding ranges for all three YCbCr components. However, this was changed so that Y had a greater margin for overloads at the white levels, as white level limiting is more visible than black. Thus, the nominal 8-bit Y levels are 16–235, while the nominal 8-bit CbCr levels are 16–240 (with 128 corresponding to no color). Occasional excursions into the other levels are permissible, but never at the 0 and 255 levels.
For 8-bit systems, the values of 0x00 and 0xFF are reserved for timing information. For 10-bit systems, the values of 0x000–0x003 and 0x3FC–0x3FF are reserved for timing information, to maintain compatibility with 8-bit systems.
The YCbCr or R′G′B′ levels to generate 75% color bars were discussed in Chapter 2. Digital R′G′B′ signals are defined to have the same nominal levels as Y to provide processing margin and simplify the digital matrix conversions between R′G′B′ and YCbCr.
SDTV Sample Rate Selection
Line-locked sampling of analog R′G′B′ or YUV video signals is specified for SDTV. This technique produces a static orthogonal sampling grid in which samples on the current scan line fall directly beneath those on previous scan lines and fields. It ensures that there is always a constant number of samples per scan line, even if the timing of the line changes.
Another important feature is that the sampling is locked in phase so that one sample is coincident with the 50% amplitude point of the falling edge of analog horizontal sync (0×0). This ensures that different sources produce samples at nominally the same positions in the picture. Making this feature common simplifies conversion from one standard to another.
For 480i and 576i video systems, several Y sampling frequencies were initially examined, including four times Fsc. However, the four-times Fsc sampling rates did not support the requirement of simplifying international exchange of programs, so they were dropped in favor of a single common sampling rate. Because the lowest sample rate possible (while still supporting quality video) was a goal, a 12 MHz sample rate was preferred for a long time, but eventually was considered to be too close to the Nyquist limit, complicating the filtering requirements. When the frequencies between 12 MHz and 14.3 MHz were examined, it became evident that a 13.5 MHz sample rate for Y provided some commonality between 480i and 576i systems. Cb and Cr, being color difference signals, do not require the same bandwidth as the Y, so may be sampled at one-half the Y sample rate, or 6.75 MHz.
Insider Info
The “4:2:2” notation now commonly used, and discussed in Chapter 2, originally applied to NTSC and PAL video, implying that Y, U and V were sampled at 4× , 2× , and 2× the color subcarrier frequency, respectively. The “4:2:2” notation was then adapted to BT.601 digital component video, implying that the sampling frequencies of Y, Cb and Cr were 4×, 2×, and 2× 3.375 MHz, respectively. “4:2:2” now commonly means that the sample rate of Cb and Cr is one-half that of Y, regardless of the actual sample rates used.
With 13.5 MHz sampling, each scan line contains 858 samples (480i systems) or 864 samples (576i systems) and consists of a digital blanking interval followed by an active line period. Both the 480i and 576i systems use 720 samples during the active line period. Having a common number of samples for the active line period simplifies the design of multistandard equipment and standards conversion. With a sample rate of 6.75 MHz for Cb and Cr (4:2:2 sampling), each active line period contains 360 Cr samples and 360 Cb samples.