TM-30 is based on color changes of the color evaluation samples (spectral
reflectance data) computed via CIECAM02, it is not based on spectral
decomposition, at least not directly. Which is a good thing because human
vision is also not really based on spectral decomposition of 16 spectral
bands. There are some spectrally based metrics for color rendering, such as
TLCI (can't remember what it stands for sorry!) which is based on lots of
assumptions about a rec.709 Imaging pipeline and average camera sensor
(yikes!). And the Spectral Similarity Index (SSI) which is based on the
spectral difference between the light source and the reference in the range
of human vision weighted by the visibility function. I would say both of
these are falling out of favor compared to TM-30, although SSI still has a
good amount of usage in the film (movies) industry.
Recommended reading:
This lecture by Dr. Kevin Houser: https://youtu.be/5fQkau7i51k
The standard itself (Free!) :
https://store.ies.org/product/tm-30-20-ies-method-for-evaluating-light-source-color-rendition/
Some details about how the color evaluation samples were selected and why.
A. David *et al.*, “Development of the IES method for evaluating the color
rendition of light sources,” *Opt. Express*, vol. 23, no. 12, p. 15888,
Jun. 2015, doi: 10.1364/oe.23.015888.
This paper is about the need for a new metric. I can't seem to find this
paper in my library, so I'm not sure that it's the one I'm thinking of. But
I would highly recommend anything by Dr. Houser on this topic.
Houser, K. W. (2013). If not CRI, then what? LEUKOS, 9(3), 151–153.
doi:10.1080/15502724.2013.10732183
Lastly, I've written a blog post about TM-30 with some useful
visualizations about what color rendering actually means.
https://tuckerd.info/06/what-is-tm-30/
If you are going to be in NYC in October, I am teaching a class at
Lightfair 2021 about color rendering, among other things. The details
should be in my email signature if the image makes it through, but if not
the class is titled "Optimizing Color Control with LED Lighting."
https://www.lightfair.com/browse/#/show
I hope the self promotion isn't too offensive. I'd be happy to talk more
about this topic! But don't want to distract from the original questions.
- Tucker
On Tue, Jul 13, 2021 at 8:06 PM Ben Goren <ben@xxxxxxxxxxxxxxxx> wrote:
On Jul 13, 2021, at 3:21 PM, Tucker Downs <tucker@xxxxxxxxxxxx> wrote:
A better version of CRI is called IES TM-30, which provides a specific
values for expected hue and chroma shift
in 16 hue bins. It's a great standard, the spectrum still can't be derived
but it is a much more useful measure of
color rendering.
One might hope that those 16 bins are linearly independent, which means
you should be able to get at least 16 spectral bands, which could equate to
better than 20 nm spectral resolution. The i1Pro in default mode is only 10
nm, so that’s not bad at all.
I should hastily add: I was unaware of IES TM-30 before I read your email,
and everything I know of it I know from your email. It’s not hard to
imagine the spectral sampling being too narrow or irregular or too
dependent or whatever such that you couldn’t get a spectrum (albeit a
low-resolution one). But, your description would make one hope.
I’ll have to research a bit. I won’t be surprised if they’ve solved
problems that have troubled me….
b&
