Good HDTV: It's More Than a Numbers Game
by Randy Hoffner
The advent of digital television(DTV) broadcasting and HDTV has presented the broadcaster and the consumer with a multitude of choices. One choice that broadcasters were required to make was the selection of an HDTV scanning format. This choice might seem difficult at first glance. However, once we have cut through the myths and the hype, we find that from a technological standpoint choosing is not really so difficult after all.
Since digital HDTV broadcasting began, we have heard a lot of discourse about the two HDTV scanning formats that are used by broadcasters: 1080i and 720P. Most of us know that 1080i is an interlaced scanning format, while 720P is a progressive scanning format. Let's take a closer look at these two formats. What are their respective attributes, and what does it all mean to the HDTV viewer?
Skipping the technical details, a television picture is produced by electronically scanning a scene line-by-line, converting those scan lines to electrical signals, and transmitting the signals to a receiver, where they are converted back to scan lines that reconstruct the original picture. When we compare the pictures made by broadcast NTSC, VHS, and DVD, all of which have the same 480 scanning lines, it is apparent that the number of scanning lines alone does not determine the quality of the television picture, although the greater the number of lines, the greater the picture's resolution potential.
The earliest experimental electronic television pictures were scanned progressively. Progressive scanning is just what its name implies: the lines of the television picture are scanned sequentially, line 1 followed by line 2, etc. While this seems the logical way to scan, other considerations led to a different approach. The greater the number of scanning lines, the greater the resolution possible in the resulting pictures. However, there is no free lunch, and the greater the number of scanning lines, the greater the analog bandwidth the signals occupy when transmitted. In NTSC television broadcasting, the analog bandwidth of the visual signal is strictly limited in order to fit it into the television channel along with the sound signal.
The inventors of television as we know it found themselves on the horns of a dilemma. They wanted in excess of 400 scan lines to afford sufficient resolution, but if they scanned this number of lines progressively, they were limited to a frame repetition rate of about 30 per second in order to stay within the allotted analog bandwidth. But they also had to confront the problem of flicker, which is the sensation of the picture perceptibly fluttering or flashing. Flickering occurs when the vertical repetition rate or the number of light flashes per second is too low for the specific viewing circumstances. The critical flicker frequency, or the repetition rate above which flicker cannot be perceived, falls between 40 and 60 repetitions per second, the exact number depending on conditions that include picture brightness and ambient room light. A repetition rate of 30 flashes per second is below the critical flicker threshold under any viewing circumstances. Although motion pictures run at 24 frames per second, each frame is projected twice, bringing the light flash rate up to 48 per second. This is above the critical flicker frequency in a dark motion picture theater where the brightness of the images is relatively low, but it is below the critical flicker frequency for a bright television picture viewed in a lighted room.
In the early 1930's, someone hit upon an idea that was hailed as a great invention to solve the flicker problem in television. It was called interlaced scanning. In interlace, each video frame is scanned as two half-frames, called fields. In one field, all the odd-numbered lines of the frame: 1, 3, 5, etc., are scanned, while in the other field all the even-numbered lines: 2 , 4, 6, etc., are scanned. The fields are scanned, transmitted, and displayed sequentially. As they are displayed, the odd-numbered lines and the even-numbered lines, which are spatially offset from one another by one scanning line's height, are "interlaced" together by the human visual system into a complete frame or picture. This would, at first glance, seem to be the best of all possible worlds. Thirty frames worth of picture information is transmitted each second, while the repetition rate is doubled to 60 light flashes per second. Yet, like a David Lynch movie, if we dig below the surface, things are not quite what they seemed at first glance to be. Interlaced scanning is a compromise that trades the absence of flicker for a number of other problems. The price interlaced scanning exacts in visual quality was necessary to meet certain objectives in NTSC television, but in the digital TV broadcasting milieu, it is not necessary to pay the interlace penalty.
1080i HDTV continues the tradition of interlaced scanning, and brings with it the interlace quality penalties. In the DTV world, each scanning line is made up of samples, called pixels. In 1080i, each line is made up of 1,920 pixels, which is in some cases reduced to 1,440 pixels. There are 1,080 lines in each complete frame, and 540 lines in each field, a little more than double the number of lines in an NTSC frame and field respectively. 1080i is usually transmitted with a frame rate of about 30 frames per second, as is NTSC.
The other HDTV scanning format, 720P, is a progressively-scanned format. Each 720P line is made up of 1,280 pixels, and there are 720 lines in each frame. 720P is typically transmitted at about 60 full frames per second, as opposed to 1080i's 60 half-frames per second. This affords 720P some significant advantages in picture quality over 1080i, advantages such as improved motion rendition and freedom from interlace artifacts.
The advocates of 1080i HDTV support their cause with a flurry of numbers: 1080 lines, 1920 pixels per line, 2 million pixels per frame. The numbers, however, don't tell the whole story. If we multiply 1920 pixels per line times 1080 lines, we find that each 1080i frame is composed of about two million pixels. 1080i advocates are quick to point out that a 720P frame, at 1280 pixels by 720 lines, is composed of about one million pixels. They usually fail to mention that during the time that 1080i has constructed a single frame of two million pixels, about 1/30 second, 720P has constructed two complete frames, which is also about two million pixels. Thus, in a given one-second interval, both 1080i and 720P scan out about 60 million pixels. The truth is that, by design, the data rates of the two scanning formats are approximately equal, and 1080i has no genuine advantage in the pixel rate department. In fact, if the horizontal pixel count of 1080i is reduced to 1440, as is done in some encoders to reduce the volume of artifacts generated when compressing 1080i, the 1080i pixel count per second is less than that of 720P.
Another parameter 1080i advocates use to advance their cause is resolution. Resolution is the ability to preserve the separate components of fine detail in a picture, so that they may be discerned by the viewer. But picture quality is not dependent on resolution alone. Numerous studies of perceived picture quality reveal that it is dependent on brightness, color reproduction, contrast, and resolution. Color reproduction is identical in all HDTV scanning formats, and may thus be disregarded as a factor. A typical study assigns the following weights to brightness, contrast, and resolution:
Resolution, then, is only a factor, and not the largest factor, in the determination of the subjective quality of a television picture. This was well illustrated in an industry meeting of professional video engineers that took place a few years ago. At that meeting, two direct-view (cathode ray tube) monitors of the same size, shape, and brand were fed the same HDTV signal. One of these monitors was priced in the $40,000 range, while the other was priced in the $4000 range. The $40,000 monitor unsurprisingly had a picture tube of far higher resolution capability than the lesser priced monitor, but the lesser monitor, because of its larger pixel "dots", had the higher contrast ratio, the relationship between the lightest and darkest parts of the picture. With a single exception, the engineers preferred the pictures displayed on the lower-definition monitor. While they seem at first glance to contradict intuition, the results of this demonstration are consistent with all the published literature on the subject.
Television pictures move, so when we consider resolution, dynamic resolution is typically a more important factor than static resolution. We have seen that the goal of interlaced scanning in NTSC was to effectively provide about 480 lines of vertical resolution, while keeping the vertical repetition rate above the critical flicker threshold. The latter goal was met, but it has long been known that the former goal was not. A 480-line interlaced picture only has a vertical resolution near 480 lines when it is a still picture. In the interlaced scanning structure, the two halves of a frame are separated in time by 1/60th second, and consequently, when something in the picture moves in the vertical dimension between half-frames, vertical resolution is compromised. In the worst case, the resolution of an image that moves vertically is reduced to half, or about 240 lines. Similarly, a moving 1080i picture may have its vertical resolution reduced to around 540 lines. Thus, the real vertical resolution of a 1080i picture dynamically varies between the limits of almost1080 lines and almost 540 lines, depending on the degree and speed of motion. This resolution degeneration in interlaced scanning has been well known for many years, and its degree is quantified by application of the interlace factor, which effectively specifies dynamic vertical resolution as a percentage of the total number of lines in an interlaced frame. Progressive scanning does not have this problem, and the dynamic vertical resolution of a 720P picture is very close to 720 lines under any conditions of motion.
As long ago as 1967, a Bell Laboratories study concluded that the degree of resolution enhancement that accrued from use of interlaced scanning over the number of lines in a single field depends on the picture brightness, but at normal brightness this enhancement amounted not to 100 percent, but only to about 10 percent, corresponding to an interlace factor of about 0.60.
Results of testing done by the Japanese broadcaster NHK in the early 1980's indicate that picture quality achieved with interlacing is nearly equivalent to that achieved from progressive scanning with only 60 percent of the number of scanning lines, which is an interlace factor of 0.60. This finding agrees with the 1967 study, and also with another study that was published back in 1958. What this means to the HDTV viewer is that the vertical resolution of any HDTV pictures that have a vertical motion component is better in 720P than in 1080i. Based on the above findings, progressively-scanned images equivalent to the observed dynamic vertical resolution of 1080i may be achieved using only 648 lines. If we want to play a numbers game, 720P has better dynamic vertical resolution than 1080i by 72 lines.
Horizontal motion also causes artifacts when interlaced scanning is used. Depending on its speed, horizontal motion in interlaced scanning generates distortions that range from serrated edges, through blurriness, to double images in the extreme case.
But wait, there's more! The resolution impairments of interlace, plus the fact that progressive scanning affords far better motion rendition than interlaced scanning, make it apparent that a football game, for example, would be much more enjoyable in 720P than in 1080i. Add to this its freedom from other well-known interlace artifacts such as visibility of scanning lines, line crawl, and flickering aliases, and it quickly becomes clear that 720P is equal to, if not better than, 1080i in the representation of real-world, moving television images.
We have seen that interlaced scanning was born as a compromise to conserve analog bandwidth; a compromise that results in picture impairments and artifacts. A DTV broadcast is limited not by analog bandwidth but by digital bandwidth: the critical limitation is on the number of digital bits per second that may be transmitted. In order to broadcast DTV pictures, their bit rate must be aggressively reduced by digital compression to fit within the broadcast channel or pipeline that is available. The digital bits representing HDTV pictures must be compressed by a ratio that averages around 70 to 1 in order to fit into the 19 megabit-per-second DTV transmission channel. This creates a "funnel effect": for each 70 bits that enter the funnel's large end, only a single bit passes through the small end of the funnel into the transmission channel. Digital compression technology is improving rapidly, but it has been consistently observed that 720P HDTV pictures may be compressed much more aggressively than 1080i pictures before they become visually unacceptable. In fact, compression of 1080i pictures routinely generates visible artifacts, particularly when the pictures contain fast motion or fades to or from black. These artifacts cause the picture to degenerate into a blocky, fuzzy, mosaic, that may be observed frequently in 1080i broadcasts. The stress level to the HDTV broadcast system caused by bit rate reduction is much lower for 720P, and blockiness artifacts are seldom observed in 720P broadcast pictures. It may be expected that 720P will always lead 1080i in compressibility and freedom from compression artifacts, because progressive scanning is by its nature superior in the area of motion estimation. This gives it a "coding gain" relative to interlaced scanning, and the result will always be delivery of the same picture quality at a lower bit rate.
Finally, let's take a closer look at the display. The resolution of any type of display is dependent on its dot pitch, which effectively defines the physical size of the dots, or screen pixels: the higher the resolution, the smaller each dot must be. We see this when considering computer monitors or printers: a 600 dot-per-inch printer makes a sharper image than a 300 dot-per-inch printer, and a 0.28 dot-pitch monitor makes a higher resolution image than a 0.50 dot-pitch monitor, and of course the higher resolution printer and monitor cost more than their lower-resolution counterparts.
In order to fully resolve a 1080i picture, a display screen must have about 6 million dots, and for 720P, the figure is about 2.75 million dots. The larger the number of dots required, the smaller each dot must be, and the smaller the dot, the less light it generates. The full resolution of 720P may be displayed using dots three times larger than 1080i for a given screen size, and this gives the HDTV viewer a brighter picture with a higher contrast ratio. As an added bonus, the lower resolution display is less expensive to make.
We saw previously that the real vertical resolution of 720P pictures is better than that of 1080i pictures. It is also true that the additional horizontal resolution that 1080i boasts cannot be displayed on any currently available consumer HDTV display of any technology. Fortunately for the viewer, it is not necessary to the enjoyment of HDTV. An instructive illustration is the much-admired digital cinema, where micromirror projectors are used to project theatrical features onto screens that may be 50 feet or more wide. The horizontal resolution capability of these projectors is 1280 pixels, the same as that of 720P, and we have not heard anyone complain that digital cinema has inadequate horizontal resolution.
Micromirror projection is one of several advanced display technologies that are now available. Others include LCD and plasma flat-panel displays. All these advanced displays are inherently scanned progressively, and 720P may be displayed on all of them without the potentially image-degrading de-interlacing step.
720P, when compared with 1080i, provides better dynamic resolution, better motion rendition, the absence of interlace artifacts, and the absence of compression artifacts. It makes brighter pictures with a higher contrast ratio than 1080i. It is well matched to the resolution capability of consumer displays. It is a forward-looking technological choice that is compatible with computers, with advanced display technologies, and with the display of text and multiple windows as well as conventional television pictures. Given all this, the technological choice between 720P and 1080i is not a difficult one. The topic of subjective picture quality is complex, but the reasons ABC chose 720P HDTV may be distilled down to a simple truth: it gives the viewer better HDTV pictures.
Thoughts from Walter Graff
I see an endless masturbation on the web of discussions of 720 vs 1080. And that's what I see it as, endless masturbation. Both are simply two different ways of making a picture, with neither necessarily better than the other. 1080 is really interlaced TV. 720 is better for web production. Other than that there are advantages and disadvantages to both but that depends on the application. In projection 1080 offers slightly better resolution when discussing cameras costing $50k and above but that resolution might be lost depending on the equipment and the chain in which that signal is processed. For instance a 2/3" camera might show the slightly better resolution rendering of 1080 but that is completely lost to a sub $50k camera. In other words all this masturbation. about $9k cameras and their 'ability' shows me that those folks are neither professionals nor concentrating on the things that make one a professional. For television applications there is no real advantage or disadvantage in most cases between either 720 or 1080. It's simply personal choice. Anything else is nothing more than wannbes trying to use equipment to justify their lack of experience.
Copyright 2013 by Walter Graff. This article may be circulated and shared as long as the following reference is made: 'This article appears courtesy of Walter Graff- http://www.waltergraff.com'
Please don't hesitate to send me an e-mail if you have any questions or comments please e-mail me at Walter@waltergraff.com