I have some CDM discharge lights (along the same lines as Cool Lights' CDM150 fresnel) which are sort of roughly tungsten, and some others that are sort of roughly daylight. The cooler ones are about 4200K (datasheet) and the warm ones are about 3000K (datasheet). They have reasonably good CRI in the high 80s for the warm and a very good mid-nineties for the cool. For comparison, Osram's HMI bulbs are usually specified as 90 CRI. The warm CDM produce an attractive, golden light that's very nice on people's skin. It's an efficient way to warm light and I like them a lot.
It'd be nice, however, to be able to match them with reasonable accuracy to conventional 3200K tungsten or 5500K daylight. A bit of mathematics for the warm ones:
(1000000/3200) - (1000000/3000) = 312.5 - 333.3333 = -20.833
The mired shift value of eighth CT blue is -18, which should put us within 25K of correct. This is almost three mireds off, which is likely to be visible as a warm cast, but I'm not aware of a better way to do it.
Regarding the 4200K types:
(1000000/5500) - (1000000/4200) = 181.82 - 238.094 = -56.274
Again, the nearest option is not quite perfect - a layer of quarter blue, and a layer of eighth blue has a total shift of -53 which puts us within 3.2 mireds of correct.
If anyone has experience of using this sort of mathematics, I'd be grateful for any input. The accuracy with which any of this works is also likely to be modulated by the imperfect CRI of the lights, especially the warm ones which have a fairly pronounced green spike:
The 4200K types are better:
...but it seems worth a try.