LuxCoreRender Forums

frank_yifei wrote: Tue Feb 26, 2019 12:45 am maybe the brightness difference comes from tonemapping in camera setting, just guessing.

In my simplified scene I did indeed forget to disable autolinear tonemapping.
Here is the scene with autolinear disabled:
(this time only 50 samples instead of 100, for faster rendering)

first results of alpha 2.2

andreymd87 wrote: Tue Feb 26, 2019 10:58 am first results of alpha 2.2

Did you use hdri?

lacilaci wrote: Tue Feb 26, 2019 11:27 am

andreymd87 wrote: Tue Feb 26, 2019 10:58 am first results of alpha 2.2

Did you use hdri?

no. just white light

So it's another example where darkening is not just related to hdri...

lacilaci wrote: Tue Feb 26, 2019 11:44 am So it's another example where darkening is not just related to hdri...

This look more like small bias not the common obvious overall darkening

Sharlybg wrote: Tue Feb 26, 2019 12:17 pm

lacilaci wrote: Tue Feb 26, 2019 11:44 am So it's another example where darkening is not just related to hdri...

This look more like small bias not the common obvious overall darkening

I don't know, maybe... I'm still using 90degree normal and as low glossiness threshold as I can get away with to have most materials in cache. Helps a bit with this weak cache brightness

It is just the bias introduced by the cache. This is a debug rendering with a radius of 10cm, focus your attention on the circled areas:


and this is the same rendering with a radius of 1cm and 10x more photons traced (still 1secs pre-processing time):


As you can see, the "A" area is about the same while the "B" area is brighter with a smaller radius. This happen because a circle of radius 10cm placed on "B" area will include empty space but the values is still divided circle surface so the result is darker. Indeed, smaller is the radius and less error there is.
Note how this doesn't happen on "A" area because the 10cm circle is fully filled by the surface and fully bombarded by photons.

@lacilaci, this is also the reason why using a 90deg angle is brighter, instead of empty space, you include photons hitting the perpendicular wall.

For reference, this is a rendering with LIGHTCPU (the true reference when doing forward ray tracing, PATH is backward):


About the same of the rendering with 1cm radius.


Short version: larger is the radius more and more bias (aka error) you have.

Short version: larger is the radius more and more bias (aka error) you have.

But by reducing the radius the number of photon should be higher also to get corrected brightness ?

Also how this degree setting work. (please explain again don't really catch that one).
Thank you for the investigation.

So in a real scene, things just can't be solved using more and smaller cache entries.
Note that 1milion photons at 15cm give even brighter scene as 100 milion photons at 1 cm.

100 milion photons at 1cm render: 100 milion photons at 1cm debug: 1 milion photons at 15cm render: 1 milion photons at 15cm debug: pathtracing with 100 path depth for all path depths: