Originally Posted by kjgarrison
Thanks for thoughtful answer. I really have no idea what you do to make these discs. I thought the dE bump at 70 was probably real, maybe software, maybe probe. Never thought it might be the DVD until your post. I posted the same question on the ColorHCFR thread. Thanks again.
It is probably real. I would question the TV first, the probe second, and the software and test disc last.
The software and disc would be the last part of the chain I would question, mainly because you have people right here that can verify things for you. alluring and I can verify if there is a problem with the disc (like the near white issue), where as the maker of your probe/TV set is much harder to reach. In other words, errors in AVS HD (and Color HCFR) are much more likely to be found and fixed quickly because we are closely related to the community and we listen to your feedback carefully. AFAIK
, the 70% pattern is accurate on RC1. The only known issue is the near white pattern.
Anyway, there is roundoff error as alluring has described, but you have to keep in mind that this error is very small. Take 70% gray for instance. We figure the digital level for 70% like this:
(0.7*(235-16)) + 16 = 169.3
We can't do 169.3, so we round down to 169. We figure the percentage error from rounding (this is just one way to get an idea):
(0.3)/(169) = .001775 = .1775%
So you see the deviation from the real 70% is extremely small. It is basically small enough to ignore given that we are working with consumer displays and, at the very best, ISF level measuring devices.
Even if you take the "incorrect" near white level pattern on RC1, you quickly find out that even that error is realatively small.
As for verifying the patterns, we don't do a whole lot of that anymore. When we first started working on AVS HD, everything was checked by hand in the digital realm at each stage... the pattern was checked for correct RGB, the encoded material was checked in YCbCr, and then the encoded pattern was decoded by a reference decoder and the RGB was checked again. After we did this a whole bunch of times (it can be time consuming), we had faith in our process and we quit checking every single pattern. The only time we have an error is when alluring or I make a human error, and those get sorted out pretty quick. I think the near white error is the first error we have ever had on an official release, and it was found within a few days of the release and reported.
So, as much as I hate to say it, it is probably
your set and not your testing equipment. At one time I had questioned the accuracy of my own DisplayLT because the Red primary on my A3000 measured undersaturated where everyone else's was closer to spec. Just to make sure I sent my meter to alluring for comparision... even our relatively "cheap" DisplayLT meters were remarkably close to each other. I just had to accept the fact that my red primary was undersaturated, at which point I began developing my sinister plans to steal DaveHancock's set