New Measurements and New Test Gear Part II

As you can see in the box below, we have completely revamped our video-measurements box. Gone are the plain, boring, and hard-to-understand Excel-based charts. In their place are shiny, easy-to-understand, colorful graphs. Yep, party like it’s 1994.

DataColor ColorFacts
This product lets us publish pretty color graphs instead of the archaic Excel-based charts you’ve always seen, but that’s not all. ColorFacts talks to our Photo Research PR-650 and our new Sencore VP403C (see this column, last month). It controls them both, so, with the press of a button, the TV essentially measures itself. This gives me time to, I don’t know, sleep or go play World of Warcraft or something (kidding). ColorFacts tells the VP403C to put up a certain pattern, then tells the PR-650 to measure it. It then takes that data, stores it, and repeats for the other gray-scale levels, or color points, or whatever. I’m sure you’re all for reducing the amount of time it takes me to measure products (thanks!), but there is actually more benefit to you as well.

Look at the Colors!
The biggest difference in our new measurements box is the CIE color-space chart. This may be different from those you see elsewhere. Many people in the industry still use the CIE 1931 diagram, which, as the name suggests, is closing in on 80 years old. The problem with the 1931 is that it doesn’t accurately display how your eye sees colors. Your eye is far more sensitive to changes in blue than in green, for example. So a big change in the green color point as seen on the CIE 1931 might be less noticeable to your eye than a very small change (as seen on the same chart) in blue. The CIE 1976 color space (Figure A) is more uniform to the eye’s sensitivity to color. In this way, it will be easier (and a more accurate way) to tell how badly inaccuracies in a display’s color points will affect the image. Fortunately, CIE 1976 is based on the CIE 1931 color space, so one can be converted to the other. In the measurement text, we’ll print the x,y coordinates (Figure B), so you can compare displays going forward with previous measurements. We’re also going to start printing each display’s secondary colors as well: yellow, cyan, and magenta.

The other chart (Figure C) is the same color-temperature-versus- gray-scale measurement that we’ve always had. While visually more appealing and easier to understand at a glance, it is harder to pull fine details from (like how much variance there actually is). I’ll address this below. The left side of the chart is the low end of the gray scale (a 20-IRE dark gray), and the other end is 100-percent white, measured at 5-IRE increments.

Even More Newness
There are three new bullet points that need some explanation. These address the lack of resolution and description in the gray-scale graphs. Lower numbers are better. AveDev (Figure D), or Average Deviation, is how much variation there is in the color-temperature tracking. It’s better for a display to track a certain color temperature consistently than hit D6500 at 100 IRE but be 5500K at 40 IRE. The lower this number, the more consistent the color temperature is throughout the gray scale.

It’s possible for a display to track D6500 beautifully from 30 IRE to 100 IRE but be terribly cool at 20 IRE. In this case, the Average Deviation will be fairly low (only one measurement point is off), but the MaxDev (Figure E) will be high. Maximum Deviation is how much of a spread there is between the coolest and warmest points on a display’s gray-scale tracking. You don’t want extreme jumps in your gray-scale tracking, so you want this number to be as low as possible.

Last is the average distance from 6500K, called Off 6500 (Figure F). Believe it or not, this is the least important of the three, but it’s still important. As it sounds, this is the distance, in Kelvin, of the display’s average color temperature across the 16-step gray scale.

Ideally, you want a display to track 6500K; but one that tracks a certain color temperature very well (low AveDev), without any significant jumps (low MaxDev), will be great, even if its Off 6500 is higher than another display that has a poor MaxDev and/or AveDev.

One of our most popular tests has been the one-pixel-on/one-pixel-off resolution test, which verifies that a 1080p display can actually display a full 1,920 by 1,080. We will test this over HDMI and component (which we’ve done since the beginning of this test), both horizontally (1,920, which is what we’ve done up to this point) and vertically (1,080, which we haven’t). This is all in the measurement box now (Figure G). We will still use our Leader LT-446 for analog resolution tests in case the display can’t do 1,920 by 1,080 for some reason, or if it reduces the resolution of 720p.

DC Restoration
This hasn’t really been about DC restoration on most displays in a long time. This has been more accurately renamed Black Hold (as in the blacks don’t float with content). The test itself stays the same; it just has a new name (Figure H).

More Coming
This past November, I attended a seminar at the National Institute of Standards and Technology on flat-panel measurement metrology. Mike Kahn took this course in August and wrote about it in the Hook Me Up column in our January issue. Believe it or not, this class was truly fascinating and gave me tremendous insight into how to make our measurements more accurate. As I experiment with what I learned in this seminar, I’ll roll out the changes in the coming months, and I’ll keep you apprised. As it stands, our measurements are some of the most accurate and consistent in the industry, and with the changes I have in mind, they will be even more so.

If you have any questions, feel free to e-mail us at