Maxwell Render | Official public forum

I've been running some tests to get to the bottom of how many polygons are needed to reach smooth caustics... smooth faces are not a big deal but caustics seem to be very effected by poly count and don't seem to be effected by smoothing like the surfaces are.

I'd be interested to hear anybody else's conclusions on this issue...



Best,
Jason.

I did another test render -- this time of the last four from the previous render: 90 sided, 120 sided, 180 sided, and 360 sided... but much larger.

As you can see 90 is still showing artifacts and to a slight degree so is 120 -- 180 looks like the minimum for truly clean results, but there does seem to be a subtle benefit to going all the way to 360 faces.



Best,
Jason.

Nice tests, Jason.
Thanks for posting those.

But you know what - as I keep looking at them and then compare those with the real glasses (some imperfections, dust/water droplets) the one with 24 sides looks sort of more realistic than the one with 360 sides (its just way to clean).
I know, the point of the test was something else, but just saying
BTW, what was the value of smoothing angle?

Certainly dust, fingerprints, etc can and should be added to the material via layers for more realism -- but for certain things (like product renders or cut crystal) that may not desirable.

I set the smoothing angle to be in sync with the face count -- eg for 360 faces a smoothing angle of just over 1 degree was used...

Best,
Jason.

Can you post comparisons with and without smoothing? I have also noticed that caustics don't seem to respect smoothed faces. High poly count and subtle imperfections make the best caustics.

High poly, 180 sides -- witness the wireframe (Sketchup was screaming for mercy when generating this )


Smoothed


Faceted


I didn't let these render very long (SL16) because I have other stuff cooking...

Best,
Jason.

Mate that's some very sexy caustics!! I can certainly imagine SU saying NOPE!