Reblogged from MaximumPC
Posted 05/18/10 at 06:03:33 PM
by Dr. Raymond Soneira
All displays show artifacts of one sort or another when their screen images change rapidly. It’s most easily detected with moving objects, or when the entire screen moves due to camera panning. In many cases, it’s not the fault of the display. Rather, it arises somewhere in the signal path from the source, which can be caused by camera blur, interlaced scanning, MPEG compression artifacts, poor video processing, insufficient bandwidth, or insufficient CPU speed in the case of games. Further confusing matters, artifacts can occur for different reasons with CRT, LCD, LCoS, plasma, and DLP technologies. It can even occur with OLEDs, if switching speeds aren’t sufficiently high.
But when people discuss motion artifacts, they are generally talking about LCD response time. And not surprisingly, the manufacturers’ published specs for response time have become one of the major deciding issues for many consumers. As a result, in the last five years or so, manufacturers have somehow pushed response time numbers from 25ms (milliseconds) to an essentially untenable 1ms.
So what, if anything, do these specs really mean?
Behind the Basics of Blur
Motion blur arises when the liquid crystal—the active element within an LCD—is unable to change its orientation and transmission rapidly enough when the picture changes from one frame to the next. Because the standard video rate is 60 frames per second, a pixel is expected to fully update its light-transmission opacity within 16.7 milliseconds (that is, in one 60th of a second). If it takes any longer than that, the image will show some degree of lag, which appears as a trailing smear or blur whenever there is motion.
LCD motion blur is generally evaluated with an industry-standard spec called response time that (supposedly) measures the time that it takes for a pixel to go from black to peak-intensity white, and then back to black again. However, most picture transitions involve much smaller and more subtle shades of gray-to-gray transitions, which can take longer to complete.
But it gets even more complicated than that because every pixel is actually made up of independent red, green, and blue sub-pixels that have their own separate intensities, frame-to-frame transitions, and times. The upshot is that visual blur within a detailed, moving picture is a fairly complex and nebulous phenomenon.
Motion Blur: Visual Proof
Motion blur is one of the most visually tangible display problems—the evidence speaks for itself in screenshots and photography, both of which can illustrate the relationship between response time and motion blur. In this article, I’ve included high-speed screenshots of moving DisplayMate test patterns, as well as moving test photos taken of a top-of-the-line, 120Hz Sony HDTV (shot with a Nikon DSLR using a fast shutter speed of 1/160 of a second). These images were taken in 2008, but the results wouldn’t be much different today.
Sony’s published response time for this XBR model is 8ms. Since this corresponds to a double transition (from black to peak white, and then back to black again), the single transition time (from black to white, or from white to black) should therefore be about 4ms.
But is the pixel response time really that fast? To find out, I ran DisplayMate tests in which black and white squares move across the screen at measured speeds. In the examples here, one photo shows the squares racing across at 1,093 pixels in a single second. The second photo shows the squares moving about 50 percent faster, covering 1,609 pixels in a single second. The white tips seen on the edges of the ghost images are artifacts resulting from electronic overdrive processing that’s being used to try to improve the response time by exaggerating transitions.
The screen on top was grabbed from a motion speed test of 1,093 pixels per second, with the black and white squares indicating real-world pixel response times of 60ms and 40ms, respectively. The screen on the bottom shows the same lengthy response times during a faster test of 1,609 pixels per second.
As you can see from my screenshots—each a brief snapshot in time with a shutter speed of 1/160 of a second, which is less than the refresh rate—it’s possible to make out at least eight individual screen refresh cycles on this 120Hz display. Indeed, in the screenshots, each square is shifted from the other by 1/120 of a second, which is 8ms, and those ghosted squares indicate that the older images haven’t yet dissipated. The upshot is that you’re looking at a true response time of about 65ms. In fact, a response time of much less than the 8ms refresh rate would be needed for there to be no visible blur. Obviously, 65ms blur in the screen shots doesn’t jibe with the manufacturer’s single-transition response time spec of 4ms.
The DisplayMate tests clearly demonstrate that the Sony’s real LCD response time is considerably longer than its published spec would indicate. And by no means are we picking on Sony, as it actually had the best performance of all of the LCDs in our tests.
But What about Moving Photos?
It must be stated that DisplayMate test patterns are very sensitive to imaging effects—this is by design. Photographic images, on the other hand, typically consist of a very complex and varied admixture of blended picture elements. With so much going on in an image, motion blur is easily obscured and lost within the complex, variegated imagery of a typical photograph. In particular, photographs of real-world content lack uniform backgrounds, and uniform backgrounds—like those in my DisplayMate tests—make it easier to see the motion blur trails. For this reason, we would expect moving photographs to show much less visible blur than what’s demonstrated with test patterns.
To wit: If you look carefully at the magnified marching band images below, you can see a total of at least six refresh cycles in the second photograph. This corresponds to a real-world pixel response time of 50ms. But the motion blur is still much less noticeable than what we see in DisplayMate’s punishing test patterns.
Witness two photos shot by Lauren Soneira. On top, a static image—thus no motion blur. On the bottom, the image is moving at 1,018 pixels per second, and you can see six refresh cycles indicating a 50ms response time.
Photographs are static images and moving them across a screen is quite different from live video, where images are a complex and varied mixture of continually blending picture components that are themselves constantly changing in both time and position. With all this screen activity going on, we would expect to detect much less motion blur in live video than with either of the moving static photographs or test patterns.
And Now for the Tests You've All Been Waiting For
To evaluate motion blur and artifacts in live video with lots of high-motion picture content, we set up a side-by-side comparison shoot-out with 11 HDTVs, and had both consumers and experts evaluate them. The top-of-the-line LCDs from Sony and Sharp had 120Hz screen refresh rates, the top-of-the-line Samsung LCD had strobed LED backlighting, and the other units had standard 60Hz screen refresh rates. Two of the units were plasma displays, and one was a pro-grade CRT studio monitor. The goal was to determine the degree to which this varied technology affected visible motion blur.
All of the HDTVs were fed identical, simultaneous digital video using an all-digital High Definition Tivo and a Blu-ray player. They were all compared side-by-side in the configuration as shown in the photo. The content included both daytime and nighttime sporting events, TV shows, and movies, all with lots of action. If any viewer thought he or she detected motion blur on any HDTV, we would repeatedly press the eight-second Tivo backup button and watch the sequence over and over again on all of the units until we fully understood exactly what was happening on each display. We did the same thing with the Blu-ray player and its content.
Behold our 11-display shootout—with the lights on. Photo by Dieter Michel.
The conclusions from all participants were consistent across the board, and will likely surprise most consumers: There was essentially no visually detectable motion blur on any of the LCD HDTVs in any of the video content we assembled.
When people thought they saw motion blur, with only a handful of minor exceptions, the blur was either in the source video or a temporary visual illusion that disappeared when the segments in question were reviewed. Unlike what we empirically identified in moving test patterns and moving photographs, the eye is unable to detect the blur in live video because the images are much more dynamic and complex—and undoubtedly because of the way the brain processes and extracts essential information from visual images.
So, Is Blurring Even an Issue for Videos, Movies, and Games?
For all of the tests—the DisplayMate test patterns, the moving photos, and the live video—we found that there was no visually detectable difference in motion blur for the mid- to top-of-the-line LCD HDTVs. This regardless of their claimed pixel response times, 60Hz or 120Hz refresh rates, strobed LED backlighting, or motion-enhancement processing. If you find this surprising then just re-read the classic tale of The Emperor’s New Clothes.
The underlying reason why higher refresh rates don’t mitigate blurring is that the true pixel response times of displays are considerably longer than the 60Hz video frame rate, so it doesn’t matter whether the screen refresh rate is 60Hz or 120Hz, or whether the LED backlights are strobed off during the frame updating. Similarly, adjusting the electronic processing enhancements that some models offer—controls that are supposed to reduce motion blur—only served to introduce objectionable contours, edges, and other artifacts onto moving objects without reducing the overall motion blur.
So that’s the story on video. What significance do these results have for PC gamers?
First, while motion blur isn’t generally noticeable with live video, it’s more likely to be seen by gamers who intently focus on particular moving objects. For this reason, the blur illustrated above with test patterns and test photos applies.
Second, don’t pay much attention to a manufacturer’s response time specs because they are so different from the real response time and motion blur that we have demonstrated here.
Third, while 120Hz refresh rate monitors and HDTVs don’t inherently improve on motion blur over the 60Hz models, they are generally equipped with better performing panels and electronics, so they may still produce superior image and picture quality. And if you’re a movie buff, the 120Hz units should offer better motion interpolation from the 24 frames per second used in all movies shot on film. The 60Hz models need 3:2 pull-down, which produces judder, but most people seldom notice it.
Fourth, be aware that the latest 240Hz displays don’t offer any real picture-quality performance improvements, and are just a marketing gimmick taken to an absurd level.
For more information and details, see my article on LCD response time and motion blur here:www.displaymate.com/LCD_Response_Time_ShootOut.htm.
Comments
Post a Comment