# Unexpected results with a thin dielectric while generating BSDFs

Hi,

I am trying to model a thin pane of glass with `dielectric` and create a BSDF for it. I think I am setting the transmittance parameters for the dielectric incorrectly as they appear to have no impact on the BSDF data generated through genBSDF.

I compared two cases: with transmittance of 0.1 (file) and 0.9 (file).

My geometry sample size is given below:

I also did some checks to ensure that the surface normals are pointing outwards.

The BSDFs generated in both cases are nearly identical. The direct incidence is shown below, but I compared other bins too.
For glass with 90% transmittance (`void dielectric thinGlass 0 0 5 0.9 0.9 0.9 1.52 0`)

For glass with 10% transmittance. `(void dielectric thinGlass 0 0 5 0.1 0.1 0.1 1.52 0)`

Thanks,
Sarith

(PS: I am aware that the `glass` primitive exists, but this exercise is meant to be expanded further into modeling more complex shapes which a single surface geometry will not be able to handle)

Hi Sarith,

The transmission coefficient for dielectric and interface types is per unit distance, so if your glass is 1 unit thick, then normal-angle transmission will equal the value you specify. For arbitrary thickness, the transmission will be:

``````trans = mat_color ^ thickness
``````

There will then be the normal front and back surface reflectances decreasing the transmittance further, similar to glass. As you go towards grazing angles, the transmission drops both because these interface reflections become more pronounced according to Fresnel’s equations, and because the interior distance becomes slightly longer.

The point is that you may only be getting a tiny amount of absorption on top of the usual reflections with your specified transmission per unit coefficients. If your glass is only 0.001 unit thick for example, you will find the difference between a coefficient of 0.1 and 0.9 to be quite minor (99.99% vs. 99.77%).

Cheers,
-Greg

Hi Greg,

Thank you for clarifying this.

Regards,
Sarith