The latest revision (Feb. 7, 2003) of SMPTE Stan-dard 293M, 720x483 Active Line at 59.94-Hz Progressive Scan Production - Digital Representation, defines the SDTV 720x483p image sampling system. The standard covers both GBR and YCBCR formats. Its ITU counterpart, ITU-R BT.1358, covers the 525- and 625-line progressive systems.
Table 1. Picture scanning characteristics of the 525x483/59.94 format.
The formats described in both standards have a 16:9 aspect ratio. The 483 figure in the SMPTE version refers to the active number of lines per frame. The principal application of this standard is to produce enhanced standard-definition television (EDTV) signals for digital television broadcasting as per the ATSC standard. Table 1 details the picture scanning characteristics of the 720¡Á483p format.
The digital representation
Table 2 details the digital representation of the format. The digital coding is based on one luminance, E'Y, and two color-difference, E'CB and E'CR, analog signals. The specified coded signal matrix coefficients are as in ITU-R BT.601. The specified color primaries and transfer characteristics are as per SMPTE 170M. The implication here is that format conversion applications into and from ITU-R BT.709 (HDTV formats) require matrixing as well as colorimetry parameters recalculation.
Table 2. Digital representation of the 720x483/59.94 format.
The luminance sampling frequency of 27MHz is obtained from the analog input video sync signal using a phase-locked-loop-controlled oscillator operating at 858 ¡Á fH, resulting in a Nyquist frequency of 13.5MHz. The specified anti-aliasing low-pass filter has a cutoff frequency of 12MHz. The color-difference signals' sampling frequency is 13.5MHz or 429 ¡Á fH, resulting in a Nyquist frequency of 6.75MHz. The specified anti-aliasing low-pass filter has a cutoff frequency of 6MHz. The selected sampling frequencies result in an active line with 720 Y samples and 360 each CB and CR samples.
As shown in Figure 1, the digital representation assumes two separate bit-parallel datastreams consisting of:
- A digital datastream conveying a digitized luminance bit-parallel signal Y with a data rate of 27Mwords/s.
- A digital datastream conveying digitized time-division-multiplexed bit-parallel signals CB and CR with a data rate of 27Mwords/s.
Each datastream carries the active video information as well as its own TRS information, end of active video (EAV) and start of active video (SAV), and the ancillary data if present.
Figure 1. Formation of Y and CB/CR bit-parallel datastreams.
In a 10-bit system, the digital information occupies a range extending from 000h to 3FFh (0 to 1023 decimal). Table 2 shows that the luminance (Y) signal normally extends from black, 040h (64), to 3C0h (960). In order to cater to overshoot and undershoot, the allowed range is extended from 004h to 3FBh (4 to 1019). Values from 000h to 003h (0 to 3) and 3FCh to 3FFh (1020 to 1023) are reserved for TRS signals (EAV and SAV).
The EAV and SAV signals each consist of a four-word sequence:
- The three synchronizing words with hexadecimal values of, respectively, 3FF, 000 and 000.
- The XYZ word, which carries the V bit, the F bit and the H bit. These bits define the vertical and horizontal blanking. Note that the F bit is always zero as there are no fields requiring identification. In addition, bits P0, P1, P2 and P3, which assume values depending on the status of the V, F and H bits, provide a limited error correction (single error) and detection (two errors) of these bits.
The static vertical resolution, expressed in ¡°lines per picture height,¡± uses concepts dating back to the 1930s. It is equal to the number of active lines (483) multiplied by the controversial Kell factor taken as 0.7. So the 720¡Á483 format has a vertical resolution of 483 ¡Á 0.7 ¡Ö 338 LPH. This holds true for camera source signals. Digitally generated signals can individually activate each scanning line, so here the Kell factor is meaningless, and the vertical resolution equals the number of active lines.
Figure 2. Typical frequency response of the CB/CR and Y channels
Given the active line duration of this format, the horizontal resolution factor is 29 lines/MHz. The specified luminance channel anti-alias filter has a cutoff frequency of 12MHz, as per Figure 2. The horizontal resolution is 12Mhz ¡Á 29 lines/MHz+348 LPH. So the luminance horizontal resolution practically equals the vertical resolution. Any reduction of the passband will result in a reduction of the horizontal resolution. Other digital standards like the 1920¡Á1080 and 1280¡Á720 have a less critical cutoff frequency.
Table 3 compares the potential luminance resolution of the 4:3 aspect ratio SDTV SMPTE 259M format (based on Rec. 601) with that of the 16:9 aspect ratio SDTV SMPTE 293M format. As shown, the SMPTE 293M format horizontal resolution is considerably lower than that of the SMPTE 259M format horizontal resolution.
Table 3. Comparison of resolutions of two SDTV 525-line formats
This is due to the relatively low Y and C B/C R sampling frequencies. The result is stretching the Y samples (720 per active line) and C B/C R samples (360 each per active line) over a wider (16:9 aspect ration) screen. The serialization of the bit-parallel Y (27Mwords/s) and multiplexed C B/C R (27Mwords/s) results in a bit-serial signal with a 540Mb/s bit rate. SMPTE Standard 344M covers the subject.
One of the early attempts at handling a 525-line interlaced format with a 16:9 aspect ratio, while maintaining the 4:3 aspect ratio horizontal resolution resulted in a Y sampling frequency of 18MHz and a CB/CR sampling frequency of 9MHz each. The associated bit-serial signal had a 360Mb/s bit rate. The Panasonic D5 tape format could record this format, albeit with an eight-bit precision. This was the only VTR capable of recording this format. This signal format has not survived and is not an ATSC suggested format.
Because the format is progressively scanned, the reproduced pictures do not suffer from interlace artifacts such as sporadic interline flicker and movement judder. When viewed side-by-side, while displaying the same program material, the 16:9 SMPTE 293M format picture looks better than a 4:3 SMPTE 259M format picture, even though the latter has a higher horizontal resolution. This, coupled with the fact that MPEG compression is easier to carry out on a progressively scanned video signal and that a 6MHz ATSC channel can carry four SMPTE 293M programs, will undoubtedly attract a considerable segment of the broadcasting community.
Michael Robin, a fellow of the SMPTE and former engineer with the Canadian Broadcasting Corp.'s engineering headquarters, is an independent broadcast consultant located in Montreal, Canada. He is co-author of Digital Television Fundamentals, published by McGraw Hill.
Send questions and comments to:firstname.lastname@example.org