All source formats are transmitted using the same nominal 1.485Gb/s serial bit rate.
The capacity for carrying ancillary data varies from format to format.
February's column described the manner in which various versions of the 1920×1080 HDTV format are organized in order to achieve a common image format (CIF). This month's article describes the manner in which various HDTV formats, including the exclusively North American 1280×720, are organized to achieve a common serial bit rate, as described on the most recent version of the ever changing SMPTE Standard 292M, Bit-Serial Digital Interface for High Definition Television Systems. I would tentatively call this approach the common serial bit rate (CSBR).
Table 1 details the essential source format parameters. They are identified as A through M and are referenced to four SMPTE standards as follows:
- SMPTE 260M: The legacy HDTV format with 1125 lines per total frame, 1035 active lines per frame 2:1 interlaced with a 30Hz or 30/1.001Hz (NTSC friendly) frame rate
- SMPTE 259M: A European HDTV format with 1250 lines per total frame, 1080 active lines per frame 2:1 interlaced with a 25Hz frame rate
- SMPTE 274M: A (ever expanding) family of formats with 1125 lines per total frame, 1080 active lines
- SMPTE 296M: A uniquely North American HDTV format with 750 lines per total frame, 720 active lines per frame progressively scanned with a 60Hz or 60/1.001Hz (NTSC friendly) frame rate
As shown in Table 1, all formats are transmitted using the same nominal 1.485Gb/s serial bit rate. This is obtained by adjusting the number of total lines per frame and words per total line while maintaining the appropriate number of total active lines per frame and words per active line. All nominal frame rate formats (A, C, D, F, G, I, J, L) are operating at a bit-serial data rate of 1.485Gb/s. Formats with an NTSC friendly frame rate (B, E, H, K, M) operate at 1.4835Gb/s (1.485/N). The divisor N has a value of 1.001. The difference between the two data rates is unimportant. This article focuses on the D format, which is typical of other formats.
The D format source data consist of two bit-parallel datastreams operating in tandem:
- A bit-parallel Y datastream with a resolution of 10 bits per sample and a data rate of 74.25Mwords/sec; and
- A bit-parallel time-division-multiplexed Cb/Cr datastream with a resolution of 10 bits per sample and a data rate of 74.25Mwords/sec.Each of the two datastreams carries the following data in the horizontal blanking interval:
- Its own set of four-word end of active video (EAV) and start of active video (SAV) timing reference signals (TRS) for a total of eight words. In a manner similar to that used with the SDTV format, as described in SMPTE 259M, each TRS comprises three synchronizing words (3FF, 000, 000) and a fourth XYZ word. The XYZ word of each of the two datastreams carries the F, V, H information as well as the P0, P1, P2 and P3 protection bits.
- Two words of line data (LN0 and LN1) with line number information in a binary code
- Two words of cyclic redundancy codes (CRC). Two separate CRCs are calculated for the luminance data (YCR0 and YCR1) and color difference (CCR0 and CCR1)
Figure 1 shows details of the horizontal blanking interval for the luminance (Y) channel. Note that the capacity for carrying ancillary data varies from format to format.
Figure 2 shows a conceptual block diagram of an HDTV serializer. The serializer performs several functions:
- Each of the two bit-parallel datastreams is fed to a co-processor that inserts line number and CRC data. An additional co-processor may be used to insert ancillary data.
- The formatted bit-parallel datastreams feed a multiplexer. The two 74.25Mwords/sec bit-parallel datastreams Y and Cb/Cr are multiplexed word by word into a single 148.5Mwords/sec 10-bit parallel datastream in the order Cb, Y, Cr, Y, Cb, Y, Cr, Y and so on as shown in Figure 3. In this drawing the first row shows details of the horizontal blanking interval of the Y datastream. The second row shows details of the multiplexed Cb/Cr datastream. The structure of the multiplexed Y, Cb, Cr datastream is shown in the bottom row.
- The output of the multiplexer feeds a parallel-to-serial converter whose output is an nonreturn to zero (NRZ) coded bit-serial datastream with a 1.485Gb/s bit rate.
- The NRZ bit-serial signal feeds a scrambler that randomizes long sequences of zeros and ones.
- The scrambled NRZ datastream feeds an NRZ-to-NRZI converter, which converts long runs of ones to transitions, thus further helping the clock recovery process in the receiver.
The digital representation of analog video signals results in a specific number of samples per total line and a smaller number of samples per active line. The difference between the total line duration and the active line duration is an analog television legacy related to the necessity of formatting and transmitting an analog horizontal as well as vertical synchronization signal to ensure a correctly timed reproduction of the picture. Component digital television replaces the analog synchronizing signals with TRS. The EAV and SAV TRS have a duration of four words each or a total of eight words ensuring proper synchronization of the digital video signals. This leaves a considerable quantity of unused words in the horizontal and vertical blanking intervals for other uses such as the transport of ancillary data.
Source format D has a total line duration of 2200 samples, of which 1920 are active. This leaves 2200 - 1920 = 280 unused samples in each of the two datastreams (Y and Cb/Cr). As shown in Figure 3, the combined EAV, line number data (LN0 and LN1), error detection codes and SAV data amount to 12 words, leaving 268 Y words (numbered YA0 to YA267) and 268 Cb/Cr words (numbered CA0 to CA267) available for carrying horizontal ancillary data (HANC).
The vertical blanking duration is 45 lines for each of the two basic datastreams. Each active data line has a duration of 1920 Y data words (numbered YD0 to YD1919) and 1920 multiplexed Cb/Cr words (numbered CBD0 to CBD959 and CBR0 to CBR959), which can be used to transmit vertical ancillary data (VANC).
In a future article we will describe the manner in which ancillary data are multiplexed into an HDTV datastream.
Michael Robin, former engineer with the Canadian Broadcasting Corporation's engineering headquarters, is an independent broadcast consultant in Montreal, Canada. He is the co-author of Digital Television Fundamentals, published by McGraw-Hill.