need expert advice on color mixing, LED selection

I'm working on lamp project that involves (at least) 3 different LED colors; white, amber, red, + ? and have many questions. I've rethought this idea over again completely a few times and I'm starting to get a little frustrated and need some help.

This will be an uplight, no optics. I initially intended to use XMLs for the white, Rebel PC ambers and XPE reds, + maybe 1 Rebel deep red. I'm trying to create a lamp that can be adjusted from 5000K "daylight" or "full spectrum" if you could call it that, down to 1700K - 1800K candlelight. what would you use?

I'm now considering other options for the white, as the XMLs have color separation at the edges (warmer) and are less uniform. So I'm now looking for something in the Bridgelux format without the dome. 5000K - 5500K with a decent CRI, 80 or better, or in combination with other colors to fill out the sprectrum. Simplicity would be nice, but if color mixing (cyan, violet, red, etc) wouldn't be too difficult to mix and increase that CRI, then I would consider this option.

The Vf of each color (5000K, Amberish, Redish) needs to be <24V.

What combination of Amber (Rebel PC Amber or other) lumens + Red (630nm + 660nm?) lumens do I need, to make 1700K? (add white?) I was considering the red mix as 2/3 630nm + 1/3 660nm. ? advice?


Some research led me to believe that If I mix 2/3 lumens from white + 1/3 lumens from amber, I can make 3500K.

I have about 20 - 25W worth of heatsink area to work with.

I was thinking something along the lines of 900 - 1200 lumens of (5000K-ish), 400-700 lumens of amber, and 300-400 lumens of red. Does this sound about right? keep in mind, each color will be adjustable and probably won't all be at full power simultaneously.

any thoughts on LED selection or otherwise?

Thanks

From the description you give this sounds like a moderately complex project. Color mixing without optics isn't going to happen or least not to a very well executed end result. All of the LEDs you list have spectrum data available although as you mention the XM-L has a distinct color shift over the viewing axis which isn't anticipated in the data sheet. The best way to get he end result you are looking for will depend on the type of project this is. If you are designing a product for manufacture then I would think that it would justify engaging an engineering firm to run simulations. I'm certain there are plenty to be found. If this is a one of kind or hobby build and your budget will not support hiring a firm then you can get samples of the LEDs and try them out or do a search for apps or freeware that will predict color temperature based on added spectrums. Typically you get what you pay for so set your expectations low when it comes to freeware.

Just my two cents. Perhaps others will have suggestions that get you further along.