To understand what's behind these latency figures, you have to look at what's happening on the oscilloscope. I apologize to all of you who don't like this sort of thing, but since this measurement is being used we have to show what it means. If this doesn't interest you at all, you can go directly to the next section.

In the example below, I was asking the panel to go from black (0) to a light gray (175), for four consecutive frames. My frame sequence, roughly, was 0,0,0,0, 175,175,175, 75, 0,0,0,0, 175,175,175,175, etc. Take a look at the results I got measuring pixel brightness with the sensor:

The first frame is not at 175, but 210; the second is not 175 but 194, the third is at 178, and finally the last frame stabilizes at 175, the value I'd been requesting since three frame ago!

For the published latency figures, the manufacturer has chosen the first point of the curve that crosses 175, with the rising time 4.5 ms - and that's what the oscilloscope shows. But at that point the brightness is not what was requested, and it takes one complete frame before it comes within 10% of it! So to me, the response time is not 4.5, ms but the time it takes to attain an accuracy of 10%.