At the recent ‘Digital Show’, I was not surprised to see those silly little purple ‘light checker strips’ being trotted out, once again. These things just will not die, and they drive me a little nuts.
If you haven’t seen one, they look like this (corporate logos removed to protect the guilty!):
In human terms, there are two purple-ish colour strips next to each other, and the idea is you put this checker under your light source and if the two purples appear to be the same colour, then you have a good print viewing light, and if they appear as two different shades of purple you have a bad print viewing light. In more technical terms, the strips use a metameric pair of colours that look the same under 5000K light and not other light temperatures. The idea being that this then means the light source is a good version of a D50 light source and thus suitable for print proof work (with pretty much all of colour management being built around the D50 standard light source in the printing world.
The only problem with this test is…it doesn’t really work – it’s both practically and conceptually flawed. It is conceptually flawed because it only pays attention to light temperature and not the overall ability of the light source to accurately replicate colour. That is, it only tells you (loosely) if the light source has a whitepoint of 5000K, but tells you absolutely nothing about the ability to reproduce colour across the full spectrum of light, or how close the light is to the spectral distribution defined by D50.
It is demonstrably flawed in practise as well. I know of three excellent light sources to evaluate prints in practise – daylight itself, Solux 4700K black backed lamps (near universally regarded as the best colour rendering artificial light source available), and for a cheaper but very effective alternative, the Grafilites. All three of these benchmark quality light sources fail this test (although not badly) – and yet all three of them actually have excellent overall colour rendition (technically, very high CRI scores – that’s Color Rendering Index, and it’s a measurement of the light’s ability to reproduce colour accurately *across the whole light spectrum* – and not just for the whitepoint).
In fact, you can have a truly awful light source with huge spectral spikes that is quite awful for colour rendering, but happens to have a white temperature of 5000K, and despite being useless for colour proofing, it would sail past this test.
(You will note on the reverse of these strips they have a ‘get out of jail caveat’ (‘metamerism is only one aspect yadda yadda etc’) – but they still claim it’s a ‘very good indication’ – well, for my money, it’s not at all!). In fact, given the fundamental principle of colour management is chromatic adaptation (that is, the eye will balance happily to different whitepoints) – the actual whitepoint colour of the light is (within limits) pretty much the least relevant aspect of the light source!
We’re not at all alone in our views on this – here are a few colour experts talking about this on the public ColorSync mailing list:
these indicators are good tools to warn you if you are in non D50 light.
They are not suitable to determine if you have a good D50 simulator.
If you do not see or only barely visible differences it _could_ be that you have a good D50 simulation.
If you see big differences between the metameric pairs you definitely do not have a suitable D50 simulation.
The second application is what these strips are meant for, the first not.
sorry, this is the most basic of basic color science. Tri-stimulus matching depends on metamarism. Two colors are assumed to match if the spectrum reaching our eyes triggers the same response in our retina.
Our eye’s spectral response is modelled by the standard observer matching curves. A painted surface will reflect a spectrum that is the product of the illuminant spectrum and the reflectance spectrum of the paint.
So the two paint colors on the test strip will only perfectly match if the illuminant spectrum times the paint spectrum returns a metameric pair – two spectra that trigger the same response in our retina.
None of the test patches seemed to specify what spectrum they are expected to match under. A color temperature certainly isn’t a spectrum. Now it’s possible the test patches were setup to match under “typical” viewing booth illuminant, but who knows, they aren’t saying what that is.
There is also a limitation imposed by what ink formulations are available/practical. Even if it was intended to match exactly under your particular lamp, it may be too technically difficult to formulate an ink pair that will do so perfectly.
> What exactly is ‘sufficient’ to achieve a match with a metameric pair?
The usual – 1 delta E !
[In practice there is a whole family of illuminant spectra that will give a metameric match. The interesting question is how similar that family is to standard D50 viewing illuminants! ]
> If these things are so inadequate why does UGRA/FOGRA/GATF market them?
I’m sure they have a purpose, but they are being rather sloppy in the specifications, and misleading in when they can be expected to “work”.
- Jeff R -
Thanks for the profiles. I knew there would be a difference because everything I've read said there would be, but I guess seeing is believing. Even though I expected a difference I'm astonished at the difference in output quality - well worth the expense.