AVS Forum banner
Status
Not open for further replies.
1 - 16 of 16 Posts

· Registered
Joined
·
1,645 Posts
Discussion Starter · #1 ·
I thought people would like to read this



We love numbers. The bigger, the better. From horsepower to megabytes, from square feet to miles per hour, we use all kinds of numbers to convey the superiority of one product or system over another. Sometimes those numbers are based on facts and measured performance. Sometimes they are based on marketing hype.


It should be no surprise that the electronic-display industries are subject to the same number mongering that pervades the automobile, real-estate and computer sectors. Now that projectors are small enough, bright enough (in most cases) and have sufficient resolution for about 90 percent of their users, the latest craze is to play up their contrast performance. “2000:1!†we hear. “3000:1!†we read.


And this sort of creative number bloating isn’t limited to projectors. For better or worse, the plasma and LCD manufacturers have gotten into the act. “3000:1!†“4000:1!†Where will it all stop?


There’s no question that contrast is certainly an important display attribute, but it can be a very misleading number if used incorrectly. (Remember the “ANSI lumens†versus “white lumens†versus “peak lumens†debates among the projector crowd a few years back?)


The truth is, grayscale is the single most important attribute of any electronic display. Without shades of gray, we don’t have useful image contrast. Without shades of gray, we can’t create wide color palettes. Grayscales are where it all begins when a projector or monitor first comes to life on the drawing board.


Want color with that?


Those of us who evaluate and write about projectors and monitors are drawn to those displays that provide the most lifelike images. That means the widest possible grayscale with a virtually unlimited number of color combinations created by an equal-energy light source, such as the sun. Anything else represents a compromise, but some of those compromises look pretty darn good.


Short of using a portable nuclear fusion system to power projectors, the next best thing is to employ short-arc lamps that ionize mixtures of gases and metal halides to produce blinding shafts of light. We then force these shafts through condensers and integrators, refract the primary colors out of them, use those colors to create red, green and blue images from monochrome light modulators, and finish up by precisely overlaying the RGB images to create full-color pictures.


With flat-panel monitors, we can force light from a cold-cathode light source (such as a fluorescent lamp) through a light shutter (Active Matrix Liquid Crystal Displays (AMLCDs)) made up of pixels coated with tiny precision filters and get our color images that way. Or, we can discharge electricity through pixels filled with a rare gas mixture (plasma) and watch as color phosphors are stimulated to produce RGB color imaging.


In the old days, color imaging was accomplished by tickling phosphors with an electron gun. Surprisingly, this system produced (and continues to produce) the most lifelike images of all, which is why a small number of high-end customers still prefer Cathode Ray Tube (CRT) front projectors for home-theater applications.


That’s because CRTs are capable of a wide grayscale and can show images with very low luminance levels (shadow detail) as well as very high luminance levels (highlights) in the same scene. More importantly, when a CRT is idling, it is essentially shut off. I mean really shut off, as in black and not a deep gray, as you’ll see with LCD, Digital Light Processing™ (DLP™), and Liquid Crystal on Silicon™ (LCoS™) projectors, and AMLCD and plasma monitors.


While there have been tremendous advances in color imaging with flat-panel displays, one stumbling block still remains. And that’s the ability (or inability) to show a grayscale with the widest possible dynamic range. In some systems, brightness limits the resolution of the imaging device (CRTs). In others, it’s limited by scattered or refracted light (DLP, LCoS, LCD).


There’s a basement?


Black levels are also problematic. (The term “black level†is really an oxymoron, for there can only be one level of black, and that’s black or zero luminance. A better choice of words would be “shadow detail†or “low gray levels.â€) When viewing shadow detail with relatively high luminance levels — say, 20 percent of white or more — then we won’t notice any problems with the display.


But movies and TV programs with high key lighting are a different story. If the monitor or display can’t resolve luminance values below a certain level (say, 10 percent of white), then any detail in the program content with luminance values at or below that level simply won’t be visible.


If we raise the black levels (sorry, low gray levels) by adjusting the brightness control, then we also elevate the luminance levels at the high end and wind up compressing the grayscale at some point. Granted, we see more of the detail in the image, but not as the cinematographer or videographer intended. This type of signal processing is often referred to as “black stretch†on consumer TV sets.


To make matters worse, some funky things are now happening with the subtle shades of light that approach 100 percent gray or white. They are beginning to blend together or “crush†into the ceiling of 100 percent gray. Our display no longer has a wide dynamic range, and we’ve also clipped our grayscale, reducing the number of shades of color that can be rendered.


If the grayscale capability of a CRT-based display could be compared to the number of floors in a house, that house would have a full-sized basement and a walk-around attic. LCD, DLP and LCoS projectors will reduce that basement to a crawl space or eliminate it altogether, and the attic becomes a tight crawl space, too. We have fewer floors to work with and less space overall.


More fun with numbers


To better understand this concept, I selected a basic 16-step grayscale ramp from the DisplayMate test pattern series for illustration. All 16 steps are clearly seen in Figure 1, and this is how the grayscale would appear on a correctly calibrated CRT. Setting the step above black to about 6 percent of full white results in a contrast ratio of about 440:1 on my Princeton CRT monitor. However, with a Samsung 42" plasma, I measured only 60:1 contrast.


The difference? “Black†on the CRT monitor registered around 0.2 nits, while on the Samsung plasma “black†registered as 3.6 nits, or 18 times as bright. With a little playing around, I could expand the contrast ratio on the Samsung panel to 107:1 as “black†now measured 1.8 nits. Because my lower black level was limited by not having a “basement†to speak of, the Samsung’s 16-level grayscale resembled that of Figure 2.


You can still see the 15th and 16th steps at the high end of the grayscale, but there’s no difference between steps one and two at the low end. This is a typical grayscale rendering for plasma and LCD monitors. Keep in mind plasma and CRT displays have current limiting to prevent image burn-in and premature phosphor aging, and these circuits will limit also contrast in images having high overall luminance values.


With projectors that shutter or reflect light (this also includes LCD monitors), the value of white can be many multiples of the summed grayscale steps one and two. That’s because the resolution of the projector is not affected by brightness levels, nor is the stability of the color dichroics as sensitive to luminance values. The result is high contrast levels (great for marketing) but a loss of shadow detail (not so good for viewing).


In my November 2002 Projector Roundup, I measured some projectors with exceptionally high contrast. Several of them exhibited peak contrast ratios much higher than the 440:1 measured on my Princeton CRT. But none could come close to the value of “black†that I measured on the Princeton set, and consequently, the grayscale images they displayed didn’t have as wide a dynamic range below 8 to 10 percent of white.


Figure 3 shows an approximation of the typical LCD, DLP, and LCoS projector grayscale. Of the plasma panels I have tested, only those manufactured by Panasonic (also used in Fujitsu’s 50" product) can produce “black†levels that approach that of a CRT, and subsequently display a grayscale with CRT-like shadow detail performance. The Panasonic panels typically create a black level of 0.1 to 0.2 nits, equivalent to my Princeton CRT monitor.


As a result, these panels display wide grayscales with nice color palettes. But they also do well in the contrast numbers game, although I’ve never measured the 3000:1 contrast that Panasonic has claimed in the past. Instead, my numbers (taken after the panel was calibrated for best grayscale) typically fell in the 600:1 to 800:1 range.


Conclusions


So, just how much contrast do you need to see in an image? Empirical data suggests the human eye is limited to a dynamic range of 100:1 at any given instant. That means that if you look at a scene with objects of different luminance values, you won’t be able to discern more than a 100:1 difference between the darkest and lightest objects. Of course, the instant your eye moves, its built-in auto-iris function raises and lowers the grayscale boundaries. That’s what allows you to perceive shadow detail and also pick out a white cat scurrying along in a field of snow.


If you are watching a movie on a plasma or LCD monitor, or with a front-LCD/DLP/LCoS projector, you’ll probably be satisfied with the displayed images as long as there is not a preponderance of dark gray and black objects. But switch to a nighttime scene with high-contrast lighting, and your eyes will strain to pick out any shadow details.


Obviously there’s a long way to go to improve the rendering of “low gray levels†on projectors and monitors, but there has been progress. In addition to Panasonic’s work with plasma, Texas Instruments has made enhancements to its digital micromirror devices to reduce light scattering and refraction. This, in turn, is dropping the value of “black†and improving both grayscale rendering and contrast.


Unfortunately, polysilicon-LCD technology seems to be limited in this area. While projectors have become brighter and contrast has improved, black levels are still higher than those measured on DLP projectors by 100 percent or more. And LCoS imaging isn’t any improvement — the black levels I measured on a D-ILA projector were equivalent to several polysilicon models in the review.


Remember: Numbers are great for impressing people and can sustain a good argument for several hours. But peak-contrast claims don’t tell you everything about performance of a projector or monitor when it comes to rendering images with lifelike grayscales, only how much brighter the “whites†can be than the “blacksâ€. Caveat emptor.
 

· Registered
Joined
·
1,525 Posts
Quote:
Originally posted by tycoondog2
In my November 2002 Projector Roundup, I measured some projectors with exceptionally high contrast. Several of them exhibited peak contrast ratios much higher than the 440:1 measured on my Princeton CRT. But none could come close to the value of “black†that I measured on the Princeton set, and consequently, the grayscale images they displayed didn’t have as wide a dynamic range below 8 to 10 percent of white.
I'm not sure if you're quoting yourself or someone else - but doesn't the above imply that the projector (in particular its gamma) is miscalibrated? I assume it (you?) is saying that the CRT has a lower contrast, but it's absolute black is much lower.


But if you have a contrast ratio, and your display is properly calibrated, than the dynamic range in all areas is fully specified. A display with 100:1 contrast and proper gamma will have lower dynamic range in all areas when compared with a 200:1 contrast projector with proper gamma.


Mike
 

· Registered
Joined
·
4,525 Posts
Quote:
A display with 100:1 contrast and proper gamma will have lower dynamic range in all areas when compared with a 200:1 contrast projector with proper gamma
True, of course. However, it's the 3000:1 contrast ratio "spec" with a poor dynamic range that the author (Peter Putnum, if I recall) has an issue with.


Especially with some DLP projectors, the low light level dithering artifacts are reduced by goosing the gamma table so luminance below 10% (or some value) drops straight down to 0% with no detail in between. Although this solves the low light dithering issue, it also makes that 3000:1 contrast ratio irrelevant and misleading.
 

· Registered
Joined
·
6,092 Posts
Here we go again...


I have to make this quick but I will get into this argument latter.


Putman despite being right 90+% of the time and being one of the best projector reviewers over the last 5 years is COMPLETELY wrong in his "peak" measurements and his rationale for using them.


He wrote an article about the benefit of grayscale over contrast (which Free has just posted a link to) in which he sort of implied that the two were mutually exclusive. Which is rather silly. Anyway, his article has set in motion a general "on/off contrast is meaningless" debate recently.


Regardless, lack of shadow detail is usually a set-up issue (i.e. brightness is set low) or source material issue (i.e. a bad film transfer and the shadow detail isn't there).


I wanted to get in a few words now, but I will revisit this issue in painstaking detail later tonight.


-Mr. Wigggles
 

· Premium Member
Joined
·
25,578 Posts
I find the article annoying.


First, it's bait and switch; he gives the impression it's about CR and then turns it into a tome on the importance of gamma/grayscale tracking. They're both important, but two different things.


He implies 2000:1 to 3000:1 CR claims are bloated, when in fact they have been measured.


He says, "But none could come close to the value of “black†that I measured on the Princeton set, and consequently, the grayscale images they displayed didn’t have as wide a dynamic range below 8 to 10 percent of white." That may very well be, given the possibility of essentially infinite CR on a CRT. But again, 2000:1 CR is by definition a dynamic range of 2000:1; it sounds like the problem he saw is one of gamma.


Plus, he has a really what I think is a relatively useless, as well as nmisleading, way of measuring CR. First he measures ANSI CR, which doesn't tell you much these days, as everything is well past good enough. Then he measures the difference between the brightest and dimmest white squares, and call this peak CR. In fact it's a measure of uniformity. He doesn't even measure on/off CR, the most important one.
 

· Registered
Joined
·
4,351 Posts
What I want to know is, if you have the ability to completely remap the gamma table in your projector to get any arbitrary gray-scale you choose, and if you have sufficient resolution in the gamma table, how can you *not* have perfect, completely accurate shadow detail?


And yet I find this to be the case with my DILA, even though I can use DILARD to remap the gamma profile to my heart's content.


Andy K.
 

· Registered
Joined
·
4,718 Posts
Andy, what exactly are you seeing? Are you unable to clearly differentiate between all 11 bars in the test pattern? I find with my SX21 I can get a complete discrete greyscale by adjusting gamma.
 

· Registered
Joined
·
4,351 Posts
I can see all bars in test patterns fine, but when you watch something like Dark City you find that you are losing a lot of detail. In fact, Dark City is often held up as revealing this fact about digital projectors. I dont quite understand why but I cant deny that the problem exists.
 

· Registered
Joined
·
3,534 Posts
Grayscale and dynamic range is a common problem found in color printing. I agree completely...contrast ratio numbers spewed out by manufacturers are not very useful. What good is 3000:1 CR if you can only make out 8 squares on a 16 square grayscale? That would be awfull gamma!


In the desktop publishing industry, highend CRTs are calibrated with specific gamma levels. Macintosh monitors used a gamma of 1.8, which would give a flatter, duller image compared to the 2.3 or higher gammas used by most PC monitors.


Printers are very difficult to calibrate. The gamma is typically linear, so printer manufacturers "tweak" the gamma to try to match monitor gamma, and with the cheaper (non-photo) printers, you get this ghastly image with crushed blacks and blown highlights. And, the color of the paper makes a big difference. Plus, because printers dither to produce the proper colors, you have to worry about black point compensation, where you crush your shadow detail because the ink will spread and smudge when you try to print a dark image.


Printing is a digital process, just like digital video projection. I think the problems with color accuracy and gamma are very similar in nature. Of course, inkjet printer technology has reached such high quality that the point of diminishing returns has been reached. The most expensive 8x11 inkjet printer that rivals or even beats $50,000 photolab printers can be had for $300 US!
 

· Registered
Joined
·
4,718 Posts
Quote:
I can see all bars in test patterns fine, but when you watch something like Dark City you find that you are losing a lot of detail. In fact, Dark City is often held up as revealing this fact about digital projectors. I dont quite understand why but I cant deny that the problem exists.
That is a very good question and one I would like to know the answer to as well. I have never seen Dark City on a CRT so I don't really know if the shadow detail actually exists in the areas that are black or greyish black with a digital projector.


I am definately finding differences between the quality of transfers in DVD's with regard to black level and shadow detail. For example I see tons of shadow detail and jet blacks with Pitch Black or Below as well as the space scenes in Attack of the Clones. I wonder if Dark City really is that great in the area of absolute black and shadow detail.
 

· Registered
Joined
·
3,534 Posts
Whoops, I forgot to add that, sure, a $300 inkjet like the Canon i950 can be as good as a photolab, but like a lot of video display devices sold today, only if you calibrate it!


Out of the box, a Canon i950 or Epson 960 inkjet will have a lot of "punch", but skintones will be off and black and white pictures will have color casts. You will need to spend about $40-$100 US to calibrate for the type of paper you are using. If you plan to print on matte paper, glossy paper, etc., you will need to spend more $$$. It's akin to recalibrating your projector for a high-gain screen versus a regular screen.
 

· Registered
Joined
·
409 Posts
Quote:
Originally posted by kromkamp
What I want to know is, if you have the ability to completely remap the gamma table in your projector to get any arbitrary gray-scale you choose, and if you have sufficient resolution in the gamma table, how can you *not* have perfect, completely accurate shadow detail?


And yet I find this to be the case with my DILA, even though I can use DILARD to remap the gamma profile to my heart's content.


Andy K.
The G series of projectors have 1024 points in it's gamma table. I believe that Dilard does it's calibration using only 32 ?? points, and draws a line to connect the points. Note that the factory software only has 16 points you can adjust.

My observations are that the responce of the DILA panels is very non-linear near the extremes (Black and white points), and this is where Dilard could use more data points. Dilard also quits when it calibrates to within a certain tolerance. My 2 most wanted features in Dilard would be to specify the number of data points, and the tolerance.
 

· Registered
Joined
·
4,351 Posts
I dont know if DILARD uses 32 points either, but I would have thought that whatever was chosen was past the point of diminishing returns.


Mark, perhaps you could comment on this? Whether or not more control points would make any difference at the low end of the scale (where the light sensor starts to become erratic)
 

· Registered
Joined
·
4,525 Posts
Hi John and Andy,


Yes, Dilard uses 32 points and uses a b-Spline algorithm in each R, G and B table to connect them in the gamma table, ensuring smooth transitions between those points.


I suppose we could measure all 256 levels, and only have a small interpolation to perform in the tables, but the time expenditure would be considerable.


In my opinion, the bigger problem is the difficulty in finding a relatively affordable colorimetric instrument capable of retaining full accuracy down to 0 IRE. I'm still looking...
 

· Registered
Joined
·
4,351 Posts
The question is, how far down (in terms of sensitivity) do we need to go before we can get sufficient accuracy to nail the shadow detail?


Have you ever tried a calibration with all 256 levels and whatever top-of-the-line meter you have on hand to see if it makes a difference?
 
1 - 16 of 16 Posts
Status
Not open for further replies.
Top