[glebe digital]

this is Glebe's scene, changed roughness to 30 and bumped absorbtion to 1.0…I'll re-run this today with the same camera/angle for a better comparison. I'm thinking that surface roughness is more important than sss…
(sl13.45, 8hrs)

Thanks for having a look at that scene, yes roughness is a key factor. I'm going to try your settings with a red epoxy to start, but I'm just thinking that the 'outer face' of the epoxy could be smooth [about 15] and the 'inner face' which is -let's say- milled out of the epoxy mould will be much rougher. This make sense to me in production terms, and I want to get an accurate reality setting first and then see what it does.

BTW, that diode in my model is set on a polished metal bed as per a large percentage of modern LEDs......or so I'm led to believe.

Update: Here's the red epoxy test:

[lebbeus]

here's a question for the group: is SSS required for LEDs? Do all real-life LEDs have sss? Are some opticly correct plastics, effectively creating a lens around the emitter--for that matter do lenses use sss?

I'm just wondering, I can get a good glow using roughness without sss, but the caustics are "out of focus"…

Glebe--is the pocket milled, or is the diode implanted before the epoxy hardens? Maybe this depends on the manufacturer/application??

more tests in the works…

[ivox3]

Technically, ....the epoxy is 'water clear', but of course has a different IOR. So, ...I think SSS is not correct, but could provide a workaround for the glow effect.

The pocket is not milled, ....the diode is embedded into the liquid epoxy.

[lebbeus]

figured as much, thanks Ivox
I've got a test going now that only uses roughness and a re-modeled led based on one of those tech papers you posted…will post the final tomorrow morning

[lebbeus]

I had a flash of inspiration on the ride home today:

two layer bsdf (using Glebe's scene as a starting point), think of it as an inversion of the typical plastic material creation method

layer 1:
colors and nd per original poly material from scene
attenuation: 8mm
roughness: 0
weight: 100

layer 2:
same colors as layer 1
attenuation and nd also the same as layer 1
roughness: lambertian
weight: 60

I've got a test going that looks very promising (will post in a few hours), this seems to negate the "problem" of diffuse caustics while also creating a good glow in the LED itself

[ivox3]

...real curious to see it.

[lebbeus]

here it is at sl6…the led in the preview window is almost burning a hole in my screen have no idea how long it will take to get the actual render to that point though

[ivox3]

I think that does look pretty good, ......so, that's using the settings 2 posts above? ....and what method was used on the model?

btw: .....when the preview does that, ...I don't think the render would ever catch up to it--- ever. Maybe someday that type of discrepancy will get fixed. ---but, ....I know nothing about that, so that ends my elaboration on that.

[lebbeus]

yes, the image above uses the two-layer mat I described (0 rough 100%/lambertian 60%), as for the modeling I can't say since it's Glebe's scene that he shared with everybody. From what I can tell, there's a pocket and a tiny rectangular emitter (73lm?), set on a metal structure. I don't know if the geometry of the dome follows the info from the papers you linked.

I'm just happy i've found a way to not use sss that has crisp caustics

[ivox3]

....great. Keep it cooking.

[lebbeus]

here it is at sl9 or so



EDIT:

This cooked for 2 hrs and reached sl10.57, will turn off dispersion and let it cook overnight

[glebe digital]

That looks promising lebbeus! I'll try this approach & apply some coloured epoxy variations when I get a moment.

[lebbeus]

after 10.5 hrs or so (had a bit more to go but I couldn't wait ) reached sl14.95, no sss was used by this series

LED off:



LED on:



I'm thinking that I've set the roughness layer a bit too high, maybe it should be around a 30% weight…not sure yet

[jalokin]

Here is a test with lambertian version of lumileds

The color is ok but the intensity and sparkling effect is gone.

[lebbeus]

how is your material set up? Did you try the two layer method I described, or is this a single layer lambertian BSDF?