We in the television technology field have been increasingly hearing about these two topics: "1080p" and "3 Gbps." What is all this about, anyway?
The scanning format 1920x1080 progressively scanned has in fact been with us for awhile, in the form of 1080p/24 frames per second (fps). The appearance on the market of so-called "1080p TV sets" which have 1080p displays, but many of which cannot accept 1080p input signals, has raised the awareness of 1080p in consumers. Add to this 1080p Blu-ray DVDs (which are also 1080p/24 fps) and 1080p video game material, and you have the two sources of 1080p material available to the consumer.
FILM FRAMES AND VIDEO FRAMES
Several years ago, video mastering and archiving of film-based material for the various, and proliferating, video applications, migrated to 1080p/24 (well, really 1080p/23.98). This was a good thing, for several reasons. 24p video establishes a one-to-one relationship between film frames and video frames, and its 1920 pixels per line horizontally by 1080 lines vertically afford the maximum spatial resolution available in the video world, as opposed to the file world. It is, in fact, nearly 2K resolution. Further, by adding the appropriate pulldown (2/3 in the 60 Hz world or 2/2 in the 50 Hz world); subsampling as required; and, for the 50 Hz television world, speeding up to 25 fps, it may be readily converted to any required output format, with very high quality results. Along with 24p video acquisition, 24p editing and post production of 24 fps film-sourced material has been the industry standard for some time.
It is at least theoretically possible to broadcast 1080p/24 material using the ATSC system for terrestrial digital broadcast that we employ in the United States. Along with 1080i at 30 fps and 720p at 24, 30, or 60 fps, 1080p, in the 24 and 30 fps frame rates plus their 1/1.001 variants, are listed ATSC formats. It is, therefore, technically feasible to broadcast both 1080p/24 (or 23.98) fps, and 720p/24 (23.98) fps, and ATSC receivers operating in film mode should at least theoretically be able to decode it, add 2/3 pulldown, and display it at 60 (59.94) fps. Parenthetically, the same may be said for 1080p/30 fps and 720p/30 fps.
We now know, however, that for all the once much-discussed 18 (or was it 36, or 360?) possible ATSC scanning formats, HDTV has only ever been transmitted in the United States as either 1080i/29.97 fps or 720p/59.94 fps. And for all the talk about megapixels and such, the uncompressed data rates of the three HD scanning formats that are actually used: 1920x1080i/29.97; 1280x720p/59.94; and 1920x1080p/23.98; are, by design, comparable.
The data rates for these scanning formats in 4:2:2 color difference component form (Y, R-Y, B-Y), at a bit depth of 10 bits, are as follows: For 1080i/29.97: 1920 horizontal pixels × 1080 lines × 2 components (R-Y and B-Y are each 1/2 Y) x 29.97 fps × 10 bits = approximately 1.243 Gbps. For 720p/59.94: 1280 horizontal pixels × 720 lines × 2 components × 59.94 fps × 10 bits = approximately 1.105 Gbps. For 1080p/23.98: 1920 horizontal pixels × 1080 lines × 2 components × 23.98 fps × 10 bits = approximately 0.995 Gbps. Each of these signals fits nicely into the SMPTE 292M HD-SDI, which, with overhead, has a nominal data rate of 1.485 Gbps. For terrestrial transmission, these data rates are usually reduced by being subsampled as 4:2:0, typically at 8 bits, and significantly compressed to fit into the ATSC transport stream, but we see that they all start out at relatively comparable data rates.
WHY NOT SWITCH?
Given that 1080p/23.98 fps has a one-to-one relationship between video frames and film frames, and that much primetime television programming originates on 24 fps film, and that further, some small degree of bandwidth savings might be effected by transmitting 1080p/23.98 as opposed to either 1080i/29.97 or 720p/59.94, it logically raises the question, "Why not broadcast 1080p/23.98 when possible?"
As we have just seen, the data rate for 1080p/23.98 is only slightly lower than for the other two HDTV formats, and, the degree of redundancy in video that contains 2/3 pulldown is so great that the compressed data rate savings is considerably less than might be expected. The most important reason, however, is that television networks and stations don't transmit only material that was originally captured on 24 fps film. Film or 24p-video sourced material accounts for a considerable amount of primetime programming.
However, the balance of primetime and other daypart programming, plus commercials, promos, sports, live events, upconversions from SD, 30 fps film-sourced commercials and a plethora of other things originate as 29.97 or 59.94 fps video. Intermixing 24 fps material with 60 Hz video material would thus require the network or station to constantly hot switch between scanning formats and this is never done. Switching scanning formats on the fly in this way would at best generate a huge number of switching glitches.
Given all this, it is not difficult to see why although 1080p/23.98 is a terrific mastering and archiving format for film-sourced and 24p-video sourced material, it really has no future in television broadcasting per se. It raises the further question concerning the recently speculated use of 1080p/59.94 video. It is currently impossible to broadcast this kind of video terrestrially, and broadcasting it terrestrially would require some kind of fundamental change to the ATSC standard that would have the potential to turn current ATSC receiver screens dark.
In Europe, a different landscape exists. Because Europe is really just launching HDTV, they were never tethered to MPEG-2 video coding for HD, and have opted for MPEG-4 AVC coding, which we understand can yield video quality comparable to MPEG-2 at compressed data rates as little as half those required for MPEG-2. This would seem to put 1080p/50 fps within the realm of feasibility for European HDTV broadcasting.
In the United States, there are too many MPEG-2 HDTV sets in the field to contemplate shifting broadcasting to MPEG-4, in the absence of some sort of backward-compatibility scheme. Development of such a scheme is of course possible, but it is not a reality today. Possible solutions that have been proposed include duplicative parallel services, i.e., simulcasting both 1080i/30 and 1080p/60. This approach would seem to be a nonstarter for a number of reasons.
Another proposal discussed is called "scalable video coding," or SVC. SVC involves transmission of a base layer, say 1080i/30 or 720p/60, and an enhancement layer. When the two layers are combined, the result is a full 1080p/60 signal. [Does this sound a little familiar to those of you who were around in the early analog days of HDTV development?]
Two serious impediments to this idea immediately arise. First, the two layers must use the same encoding scheme: an MPEG-2 base layer cannot be combined with an MPEG-4 enhancement layer. Secondly, the transition from an interlaced base layer to a progressive enhancement layer would be significantly difficult. This approach, while not theoretically impossible, would be difficult, to say the least.
1080p/59.94 also has some serious implications for production and post production, one of which is the fact that it generates a data rate double that of the current HDTV scanning formats, or around 2.5 Gbps, requiring an interface with twice the bandwidth of today's HD-SDI. It can be done using a dual HD-SDI link, or by the implementation of SMPTE 424M, which specifies a 3 Gbps interface that is effectively a rewrite of SMPTE 292M at twice the data rate. This is easier said than done, and has some serious implications for television technologists, system designers and engineers. In the future, we will take a look at some of the challenges of working in this high-speed realm.
Randy Hoffner is a veteran TV engineer. He can be reached through TV Technology.