The Secret Sauces of THG's LCD Tests

Perfectionism Can Be Counterproductive

When the manufacturers got their hands on overdrive, they had a field day. Some offer values as low as 3 ms. However, they often forgot that such low values can produce undesirable effects elsewhere. And in order to achieve exceptional latency values, some manufacturers push the envelope so much that pixels sometimes stay illuminated in a transitory state before eventually stabilizing at the value required. The consequences for the user are disturbing. On all the monitors tested that offer overdone overdrive (for want of a better term), we noted dreadful video noise during DVD playback. With video the pixels change color rapidly and persistently but the overshoot measured on the pixels is significant in amplitude and in duration when compared to the refresh rate. There is, therefore, a significant risk of video noise in any areas of flat color.

Analysis: When the image is static, no problem; the pixels don't change their values much. That's the advantage of an LCD display. Now, with a moving image, the pixels go from one value to another but each individual color variation is normally very close in value. Unfortunately, overdrive causes a temporarily large variation in the value of the pixel, and as all the pixels don't react identically, some change more rapidly than the others and in the end the user sees video noise.

Here, for example, is the situation with the ViewSonic VX924. The panel is set to transition from black (0) to light grey (175), and to do that for four consecutive images. Basically, my sequence of images is 0,0,0,0, 175, 175, 175, 175, 0,0,0,0, 175, 175, 175, 175, etc... And here's what I got while measuring the pixel brightness with the probe:

The first image isn't at 175 at all, but at 210, the second is a bit closer to 175 but is still way out at 194, the third manages 178, and finally the last one settles down to ...175, the value I asked for three images beforehand.

In order to take this phenomenon into account, we're going to measure the time needed to achieve a precision of 10%. In other words, instead of measuring the latency between 10% and 90%, we'll measure the latency from 10% to anything within the band from 90 to 110%. And if the pixel is found to be in this band, we'll consider it to be stable. Indeed, it's too easy and the customer really loses out when manufacturers use this overdrive technique without control or limits.

So, a poor implementation of overdrive will have a direct impact on latency that we can now measure and this is a perfectly acceptable measurement whenever color precision is also an important requirement.