[mtripoli]

I didn't know what else to call this thread... but here goes:

I'm modeling an led (I use a lot of leds in my work)...

Here is a screen capture of the part in SW:


To be clear (no pun):
1.) The epoxy "bullet" is one body
2.) The leads are one body
3.) The led die is one body
4.) Everything is modeled to scale: this is a 3mm led.
All materials are "real world", as far as I know (I'm happy to post the whole scene if you'd like with materials).

Here is the part as shown in MXST:


Here are examples of how it renders:




You can see how the "dome" of the epoxy is being truncated and the light only coming up to the "flat" of the model and coming out the sides.
The SW object is a spline that has been revolved so that it doesn't have a vertex between the edge and the dome curve. Initially I thought that I was seeing this problem as a result of using a dome on a cylinder. It made no difference at all to the resulting wireframe.

Can anyone give me a hint to why this is happening? Is my model wrong? Am I overlooking something?

Lastly; I couldn't find the "button" for "Show tesselation"... is it gone or am I too blind from staring at this?

Thanks for your help...

[JDHill]

Hi Mike,

From the screenshot, I'm guessing that your emitter is directly enclosed in the led body. To work correctly with dielectrics, light rays need to go into, and then back out of the dielectric body- they can't start inside of it and only go out. This implies that a dielectric body with an 'inside' and an 'outside' must be one mesh, with the outer surface having normals which point outward, and the inner having normals that point inward. That might sound like a bit of a strange requirement, but in reality, there is obviously a boundary between the body and the die, and we need to model that. Due to floating-point errors, we cannot model this junction as tightly as it would be in reality (i.e. bonded) - we need to give a bit more space between the emitter and the inside of the cavity to make sure that the actual meshes aren't violating each other's space, I just use a sphere a bit larger than the emitter surface.

To show the effect of using correct geometry, here are two quick example leds:



I guess the one on the right is closer to what you're looking for. The difference between the two is that the one on the right has been constructed the way I described above. I hope that helps explain things. Oh, and you're not overlooking anything - this plugin doesn't have the Show Tesselation function yet.

JD

[mtripoli]

Thanks JD for the explanation...

Now, am I just going nuts or did the first time I read this post you said that SW couldn't flip the normals so the object had to be exported to mxst and flip the normals there? Tell me I'm not imagining this...

Could you post an image of the led you suggest as a cross section? Thanks...

[JDHill]

No, you're not losing it - I woke up and thought...I modeled that example in Rhino, where there's no notion of a solid in the inside/outside sense, so manual normal-flipping is necessary. But SW is smarter about those things, so if you revolve the body with the cavity already inside, it will have the correct normals - something like this will do it:

[mtripoli]

That's exactly what I did and it's working... sort of... I'm now working out the viewing angles... but it's getting better!