Originally Posted by chienmort
I see no mention of "why this has changed form the traditional gamma of 2.2." In fact, to my knowledge EBU has never supported 2.2. That's a standard from SMPTE 170M.
As I understand it, the original work that was done to determine Gamma was from the EMI/Marconi work done in the 1930s. This was done prior to the start of British TV transmission in 1936. However there was an intermittent but ongoing process at the the BBC to revisited all their own previous research. In addition a number of new and older CRTs were re-measured using modern instruments. This was the basis of the new recommendations.
I will ask my contacts why they think the difference between 2.2 and 2.35 occurred.
With the NHK research, I was only told this in passing and I should perhaps have put that in quotation marks. I will try (I don't read Kanji so this may be a long process) to find out their reasons and methodology.
I've never seen any evidence of the 2.2 coming from anywhere except a relatively recent misreading of the 1953 NTSC gamma standard which had the encode gamma often expressed as 1/2.2. This is doubly mistaken because it's misreading the standard and ignoring the history of rendering intent and the desire for an end-to-end greater than 1.0 by assuming that the inverse of the encode is what the decode should be, in addition to the fact that if this is where it's coming from (it seems that way to me) that it's drawn from an outdated and irrelevant gamma standard that has been superseded by Rec709.
The other place that it seems to come from is from a misreading of Rec709, and that seems to be where the 2.22 figures come from. Rec709 cites an exponent of .45 for the encode gamma, but that's only for part of the whole function, and doesn't characterize the function properly which is closer to about ~.5 because of the linear tail, or more precisely about .51. But if you take only the non-linear portion of Rec709 encode gamma standard of .45 and inverse that (again wrongly assuming that the inverse of the encode should be what the decode is) then you get to 2.2222.... And this is again a two-step misreading of the standard because Rec709 isn't accurately characterized as having an encode gamma of .45 because that ignores the linear portion of the transfer function, and it continues to assume wrongly that the decode gamma should just be the inverse of the encode gamma. If the end-to-end gamma is 1.0, that returns the accurate scene tristimulus values (if unaltered elsewhere, which is generally happening anyway so accuracy to the scene isn't really an achievable goal usually anyway), but it doesn't yield good perceptual accuracy.