LuxCoreRender Forums

Bidir 35k samples: Path 7051k samples:

Sprocket wrote: Wed Aug 15, 2018 1:53 pm Path 7051k samples

At such high sample counts it's almost useless to continue rendering because this is exceeding the limits of floating point precision.

Nice scene, I wonder why the caustics are so different.

How can you render caustics with path engine?

Egert_Kanep wrote: Wed Aug 15, 2018 2:35 pm How can you render caustics with path engine?

If the light source can be intersected by a ray (like area lights, sun or HDR), it can be done (even if most of the times will produce just a lot of fireflies if the light is not very large and easy to hit).

Try using the path engine with metropolis sampler. In my experience, this is the best method for rendering SDS caustics such as this.

B.Y.O.B. wrote: Wed Aug 15, 2018 2:10 pm At such high sample counts it's almost useless to continue rendering because this is exceeding the limits of floating point precision.

Nice scene, I wonder why the caustics are so different.

I was starting to think it wasn't going to get any better at about 3M.

jgrover110 wrote: Wed Aug 15, 2018 7:44 pm Try using the path engine with metropolis sampler. In my experience, this is the best method for rendering SDS caustics such as this.

I'll give it a go.

how it looks in old lux for comparison:

5000k Path + Metropolis.

The absorption looks stronger in that image. Are all the values being kept the same? Also, with the old Lux image, the refraction doesn't look as strong.

jgrover110 wrote: Sat Aug 18, 2018 10:47 am The absorption looks stronger in that image. Are all the values being kept the same? Also, with the old Lux image, the refraction doesn't look as strong.

Yes they are same values, all i changed was rendering method.

The old lux scene uses a cauchy C value, the new has cauchy B. I just copied the value, but I think it should have been converted somehow.

Sprocket wrote: Sat Aug 18, 2018 12:38 pm

jgrover110 wrote: Sat Aug 18, 2018 10:47 am The absorption looks stronger in that image. Are all the values being kept the same? Also, with the old Lux image, the refraction doesn't look as strong.

Yes they are same values, all i changed was rendering method.

The old lux scene uses a cauchy C value, the new has cauchy B. I just copied the value, but I think it should have been converted somehow.

Are you sure? Any how..

I like your test pool scene. Especially the one showing caustics with path rendering.

Do you use volumes?

I got good results with the assumption: If you're medium is 2 m deep and you have an absorption of 2 medium gets the color of absorption at ground (if you look into the medium 90° from its surface).