realistic LED advice

[andrewjohn81]

We've been showing what lighting looks like for a while now, but it's not simulation in Any way. We generally exaggerate the light so they can understand that is what we are showing them. Also, then we can show differences between LED, incandescent, fluorescent, panel lighting, etc.

My boss is encouraging me to look into a way we can more accurately simulate the light now. (more specifically using LED's, but I want to look into a solution that works for everything if possible).

He suggested Maxwell and Fry Render. I haven't had experiences with those at all, so wanted to ask what you guys thought before even attempting to demo one. My first thought was IES lights.

The big issue with LED is there are Many Many different kinds. I don't think it's really possible to just say, light it with LED lighting. Is it focused, what shape is it, how bright, what color (or how white is it)?

I have a pretty fair collection of IES definitions, but I don't think many include LED. I'd like to show him that first. Does anyone know of a good resource for those specifically? If there is cost involved, does anyone know of a better place than others? I wouldn't mind if there was a company I could just tell them a spec and they could shoot me back an ies definition of the light at a cost. That way I would know I would have the correct definition for the light they are specifying here.

Do Maxwell and Fry use IES lights? My boss suggested they used object lights, as in modeling an detailed led and lighting with that. I'd be amazing if it can do that with any efficient speed. But I'm amazing often, so it wouldn't surprise me.


Let me know what you guys think/suggest.
Thanks

[Paul Oblomov]

U mean smth like this ?

Did it for caustics test time ago.



If you interested - will up, if I find it

[rivoli]

the new maxwell version can use ies files, fryrender I don't know.
they can both simulate a led exactly as vray can, I guess, but I wouldn't try lighting a complex scene that way.
if you go over at the maxwell forums and do a search for led, it should return you a fair amount of posts, old and new, dealing with that kind of simulation. it actually went from being almost impossible in the old alpha and beta days, to a viable option for small test scenes with single light sources. more than that I think it would hardly be practical.

[andrewjohn81]

To Paul: Yeah, if you happen across that I would love to take a look at the scene and see if we could apply it to our products.

My guess is, similar to the feeling of it not really being practical for multiple lightsources in a complex scene, it's really not going to be practical.

Would you say just using IES lights right in Max with Vray would be best? Still....those may be difficult to find. Let me know if you guys find some good LED ones.

[voltron7]

I have found some LED in the past here...hth

[EricKlein]

I've looked into this a bit as we do aircraft interiors and most of the lighting is going LED these days. I'm not a scientist, and I sometimes black out during long explanations, but I can tell you that a good ole vray light set to Radiance gives us results that we see duplicated in our physical mockups. (sorry, images in this case are not from same views...)

You can use Maxwell or fry, but to model a "real" diode/lens and have it behave the same as the real world is difficult at best, and probably prohibitive for more than one led array. I tried it once with some engineering data, and it took for. ever. to render poorly.

Personally I've decided that what's most important is the light energy that comes off the lens; so .ies isn't useful as it's far field data, and doesn't accurately take into account the photometric pattern close enough to the lens. I'd love to see 'virtual barn doors' on vray lights so that I could tune the angle of disbursement, as this is a phenomena of an led light that I don't have control over the way I'm working now. (yes, I could make a geometrical one, I think someone did a while ago, but I think it would be better if samples were just 'killed' or 'concentrated' instead of bounced.)

One thing I've tried is to work with someone who uses a scientific raytrace program, like tracepro, or FRED. You can take some of the data (usually in image form) and duplicate the results, as they are quite accurate. We don't often go to this LOD except for cockpits and special critical situations.

It's definitely something I'd like to hear about as you look more into it~

e.

[Paul Oblomov]

Here we go

http://www.megaupload.com/?d=IXVHWOE8