How Big is a MPCD Unit?

[ChthonVII]

How big is a MPCD (minimum perceptible color difference) unit?

I gather (perhaps incorrectly) that MPCD is a unit of euclidean uv distance in the CIE 1960 UCS, to a given point from the nearest point on the black body curve. But I cannot for the life of me find the definition of 1 MCPD = ??? delta uv.

[KelSolaar]

Hello,

I don't think I ever seen it defined for CIE 1960 UCS, it is not typically a space where colour differences are measured and it is almost strictly relegated to colour temperature representation.

Cheers,

Thomas

[ChthonVII]

Whitepoints for CRT televisions and monitors were often specified with a color temperature plus a MCPD offset perpendicular to the black body curve. For instance, if you owned a CRT computer monitor in the 80s or 90s, there's a good possibility that the whitepoint was 9300K + 8 MCPD.

See, for instance, pp 277-78 in Poynton's Digital Video and HD: Algorithms and Interfaces, 2nd Edition:

An illuminant may be characterized by a single colour temperature number – the temperature of a blackbody radiator that exactly matches the chromaticity of the source. If the match is approximate, the term correlated colour temperature (CCT) is used.

Colour temperature is sometimes augmented by a second number giving the closest distance in the deprecated CIE 1960 [u, v] coordinates of the colour from the blackbody locus – the arcane “minimum perceptible colour difference” (MPCD) units. I consider it more sensible to specify colour temperature in kelvin for intuitive purposes, accompanied by [x, y] or [u’, v’] chromaticity coordinates.

Unfortunately, while Poynton tells what what a MPCD is conceptually, he doesn't tell us how to do the math. Hence my question: How big is a MCPD unit?

(Incidentally, by converting the commonly cited xy coordinates for 9300K + 8MPCD and 9300K + 27MPCD to 1960 UCS uv coordinates, and dividing the distance by 19, I get something around 0.0004.)

[tjdcs]

Hi @ChthonVII

I'm not familiar with that particular liturature and old CRT stuff... But based on some other historical research I've done, I can say that during the 80s and 90s there was a pretty big wreckoning to be done about what "perceptable color difference" meant and we are still going on and on about it today.

For the purpose of binning LEDs for illumination according to CIE TN 001:2014 specifies 1 JND = ∆u'v' = 0.0013, but for matching MacAdams ellipses, ∆u'v' = 0.0011. (Thanks Tim Kang for reminding me of this reference)

There are also some derived standards from the IES.

But very vitally, in truth... you can't give any single value for duv that is 1 "step" of color difference. The uv and u'v' spaces are quite bad in terms of their perceptual uniformity and the value for one step would be smaller at higher CCT values.

This kind of standard would probably be found in some old optical society of america papers. You might try searching those journals. If you find something interesting and are willing to provide an implementation to Colour, we may be able to buy the standard (no promises at this time though)

[ChthonVII]

Thank you, tjdcs.

I'm specifically looking for the archaic MPCD units historically used in the CRT industry. I am aware that the 1960 UCS is deprecated for good reason, and that these archaic units bear a poor relationship to the actual color distinguishing abilities of human beings.

Could you please provide me with a link for the 0.0011 MacAdams ellipse size? I've found this 1986 document that says a JND is about 1/3 of a MacAdams ellipse (page 22) -- which would be 0.000366... delta-uv, which is probably within a rounding error of the 0.0004 estimate I got from taking the distance between some commonly cited coordinates in my earlier post. (Assuming that "MPCD" and "JND" are used interchangeably.)