January 22, 2009 2:50 PM PST

'One million' to one: Why contrast ratio is the Dr. Evil of HDTV specs

Font size
Print
E-mail
Share
4 comments

The ANSI checkerboard seems simple enough, so why is contrast ratio so complicated?

(Credit: Ovation Multimedia)

Contrast ratio should be black and white. Taken at face value, it's the ratio of the light level (luminance) the display produces when fed a white signal to the luminance when it's fed a black signal. Unfortunately, it's probably the most misused, inflated, and ultimately misleading specification used to describe HDTVs today.

At the 2009 Consumer Electronics Show, manufacturers quoted contrast ratio specs of 1,000,000:1 or 2,000,000:1 for upcoming LED-based LCD displays (Vizio and LG, respectively), which are similar to the specs quoted by Samsung and Sony for their current LED models. Those numbers sure do sound impressive, but what do they mean in the real world?

Very little. It's true that in general, a higher contrast ratio can indicate that the display produces a deeper level of black, with all of the picture-quality benefits that brings--but then again it might not. Despite the million-to-one contrast ratios of the Samsung and Sony LED sets we reviewed, we observed better black-level performance in the Pioneer PRO-111FD. Pioneer doesn't publish a contrast ratio spec for that television, but has claimed that its black levels are so deep as to be "immeasurable."

Manufacturers are free to use whatever method they like to "measure" the contrast ratio of their displays. The big numbers you see quoted most often are for "dynamic" contrast ratio, which takes into account changes the (usually LCD) display makes to adjust for fluctuations in the brightness of the content--namely, lowering the backlight in dark scenes and bringing it up in lighter ones. Then there's the "native" contrast ratio number, always much smaller than the dynamic one, where the display doesn't perform these adjustments. Both of these numbers are usually derived from the measurement of a full-white screen and a full-black screen (so-called full-on, full-off measurements), which is obviously not representative of actual program material.

A more-representative method is the ANSI contrast ratio measurement, which uses a checkerboard of eight white and eight black squares; the average luminance of the white and black squares determines the contrast ratio. Unfortunately, you'll almost never see any manufacturer quote an ANSI number, since it's usually tiny in comparison--a few hundred or few thousand or so to one.

We're saying "most" and "usually" because there's no standard way to measure the spec. We've heard tales of manufacturers pumping all of the wattage of a display into a single pixel and measuring that to determine the "white" section, or simply turning off the display to measure "black." More plausible, these numbers can simply be pulled out of thin air for competitive reasons--you don't want to be the guy with the 500,000:1 TV when everybody else says 1,000,000:1, do you? You don't have to, because nobody's keeping track.

At CNET we observe calibrated HDTVs side-by-side and make comparative statements about black-level performance.

We have seen some signs of millions-fatigue in the CR spec game, however. At CES 2009, Samsung's press material didn't publish a number, instead using the phrase "mega-contrast" without any accompanying number. LG and Vizio use the same phrase, but did publish the numbers. We've already mentioned how Pioneer is avoiding numbers, and we hope more TV makers follow suit. When the numbers can be so inflated or simply made up, what's the point?

That's why we hope you'll pay as little attention to published contrast ratio specs as we do. We rarely mention them in reviews, and when we have to refer to them in news or blog posts we try to put them in context, comparing last year's specs from the same manufacturer with this year's, for example. We're still working on performing contrast ratio measurements ourselves as part of TV reviews, so look for that to happen this year. When it does, we doubt we'll publish anything close to "one million."