[konjin]

The main issue is the blue scattering. If you can get away with pre-scattering the light before it hits the dish adding an IR lamp is no issue since you're using a parabolic mirror which has no diffraction. I have no idea if it 'feels' right doing that though.

For full immersion you would also want a number of different light sources since leds are have very narrow spectra.

[lelandbatey]

The method proposed in the video already accounts for the narrow LED spectra by using lights that have excellent color reproduction (the part where he recommends a CR rating of 95%+)

[konjin]

It's not really good enough: https://www.oscars.org/science-technology/sci-tech-projects/...

Mixing violets and greens makes the spectral curve closer to that of sunlight.

Again, without seeing it in person there is no way to know how much impact this has.

Using a parabolic reflector however is the only way to go since you will not get diffraction or blocking of wavelengths.

[andreareina]

Without going into the pdf report, the multiemitter doesn't look better than the blended phosphor.

[konjin]

No, you would need to pick better diodes, and wider variety of them. 3 greens and two violets + phosphorus at least.

[konjin]

To answer anyone looking at this now: the Rayleigh scattering needs to produce diffuse light [0]. Which actually makes the project easier to pull off since you just need to place secondary lights all around the roof and walls of the room that emit AM1.5 diffuse light globally and not worry about any sort of filtering on the main light, which you need to match to AM1.5 direct.

[0] Applied Photovoltaics, Earthscan Publications, 2007: http://www.eng.uc.edu/~beaucag/Classes/SolarPowerForAfrica/B...