Any Old Stuff (II): RT Between Colors
There is another problem, one that is even more worrying than the interval variable. The response time measured between white and black has no connection with the time between gray and black, between green and blue, between white and yellow, etc.
For example, suppose you start with zero volts and then add 1 volt to a transistor to change the color to black and white. The voltage is then at a maximum level, the liquid crystals move as fast as possible, giving a full-on time of T.
Let's assume that you now want to display a neutral gray. The voltage needed will be half as much, which will offer less stimulation of the liquid crystals, and this will cause them to move more slowly.
Each passage from one color to another produces a different response time from the previous one.
The example used by Xtremtech is based on an LCD in which the response time between white and black (17 ms) is palpably three times faster than between white and gray (54 ms).
In such a case, what credibility could be given to a response time provided by a manufacturer or even measured with some sort of measuring instrument? The immediate reaction is: not much.
Because, let me remind you, you're no longer fooling around with black and white. As far as we are aware, 100% of the games available are in color. Whether you are playing Unreal, Space Colony or Jedi Academy, the dots on your panels will switch from pink to green, then to yellow, then to gray, etc. In short there is hardly ever a moment when the response time given for your screen will match the way in which you use the monitor in practice.
At this stage, it's amusing to watch and see what happens when a dot passes from, say, mauve (184, 128, 201 in RGB) to bright green (24, 212, 35 in RGB). Each red, green, or blue cell is merely a well of liquid crystal in front of which there is a color filter. So any of them could be considered a gray cell that eventually becomes colored. Measuring the response time from mauve to bright green is the same as measuring the time needed for three cells to change, the first from 184 to 24, the second from 128 to 212, and the third from 201 to 35. The overall response time between the two colors would actually be the longest of these three periods.
Returning to the article at Xtremtech, the authors were actually quite optimistic in their conclusions. FFD matrices had not gone into mass production yet in 2002. NEC has not abandoned the project, however, and is now talking about introducing them into LCD TV monitors in the course of 2004. Let's wait & see...
Response times measured by the ISO 13406-2 standard are merely indications. They give nothing more than a very vague idea of the quality of a screen. You can't rely on it 100%, we tested it in this comparison test, and we haven't shielded you from some big surprises.
Yes, in principle, a 16 ms screen will react quicker than a 20 ms, but the only certainty is that it changes between white and black on a 10 to 90% bandwidth of the brightness required. No promises are made for other colors, or for the speed of the system at starting up and turning full on.
All this explains that despite its less impressive characteristics, we may prefer a certain 20 ms to another that promises 16 ms or that two years ago we greatly preferred the little 15" Solarism for game-playing to all the other panels, despite its 40 ms response time, as against 25 ms claimed by its rivals.
18- And 24-bit Color Interpolations
As we emphasized last time, the AU Optronics panel does not really display 16.7 million colors, but only 262,144. It completes the missing panel by quickly alternating between the colors that are close to the missing ones, at such high speed that the eye does not detect it.
Looking more closely at the panel, it isn't the only one to adopt this tactic, far from it. Almost all of the panels sold in the mass market do not operate at 24 bits (16.7 million colors) but at 18 bits (262,144 colors). Yet all of them still claim to display 16 million colors. The minor difference, in this case, is that they don't display 16.7 million, but 16.2 million. In order to achieve this, all of them use a technique called "dithering" that, like the AU, displays alternately the two closest colors. There are now a multiplicity of algorithms that make this possible and more or less effective.
A 24-bit matrix is one in which each RGB cell works at 8 bits, because 8 x 3 = 24. The 18-bit matrices, such the one used in the AU, content itself with 6 bits per color. Where the former really do display 256 nuances, the 6-bit matrices can only distinguish 64. In concrete terms, where the 24-bit screen recognizes colors 248, 249, 250, 251, and 252 the 18-bit screen makes do with 248 and 252.