A few linear light bugs in Raytraced Mode

cyclesbug

#1

Setup

  • Ground plane on, material white plaster.
  • Sun and Skylight turned OFF.
  • A linear light drawn a couple units above the ground plane, horizontal to ground plane.
  • “Show Lights” turned on so you can see where the light object is.

Bugs

  • SHADOW of the invisible light bulb (there’s a dark line parallel to the linear light, right where I would expect it to be brightest)
  • Banding of light levels on ground plane. (looks like I’m in an 8 color gif)
  • Lighting vanishes after toggling ‘Casts shadows’ (object properties on linear light) off then on again. Lighting does not show up again until I toggle the LIGHT ‘ON/OFF’. This glitch only happens with the linear light.
  • Light is very dim, even with color = white, and intensity all the way up. Are there any other places we can turn up the Lumans? (Other than making a light emmitting

In the Rhino Cycles capture below, notice the heavy banding and shadow beneath the light. Compare to the Rhino Render at the bottom.

Rhino Cycles:

Here is the Rhino Render:

The heavy banding and odd shadow beneath the light do remind me of magnetic field lines though! Are these bugs known?


(Pascal Golay) #2

Wow… here’s what I get if I place a linear light through a box:

Pretty cool, really.

With no box, the view is pure white… so I see a couple of bugs to log before I even get to the one you’ve pointed out.

https://mcneel.myjetbrains.com/youtrack/issue/RH-38856

-Pascal


#3

Perhaps you forgot to turn off the skylight?

Your light through a box looks way different than mine!

Here’s another comparison. The LEFT object is actually a cylinder with Self-illiumiation turned on, and Emission color turned all the way up. The object on the right is a regular linear light (note how dark it is too).

Thanks Pascal.

… and after I let it render for a while, the banding also shows up on the edges of the light from the light emitting material:

Perhaps the banding is just something that happens at transitions between very low light levels.


(Nathan 'jesterKing' Letwory) #5

@nates

The heavy banding looks weird. What WIP version are you using? Also the low linear light energy level makes me think you are using an old version. There used to be a small bug with passing the linear light intensity around, and there was a fall-of of light where there shouldn’t be any.

I actually think the banding is because of the mesh object created for the linear light.

Note the funky star-pattern on the box. I don’t recall having seen this before (and I did quite some testing). I have to revisit the mesh generation for linear light before the data is pushed to Cycles. Quick guess is Something Is Wrong With The Normals.

Linear light is a special citizen in Cycles - it isn’t a primitive light, so it (currently) behaves quite differently wrt shadow casting. Light vanishing isn’t correct, though. RH-38857

I don’t see this here, but it may be a result of improper mesh generation (if mesh edge happens to be along that line towards the ground plane).

For a shadows-only ground plane that is to be expected, as rhino lights have no fall-off. When there are no objects in the scene the entire plane gets essentially light energy, so you see the default background environment, which is solid white color.

This is probably because the linear light is actually a mesh with an emissive material on it that has no fall-off. For that intersecting bit you’ll have quite many faces contributing to light. I probably should add a backface check in the shader.

@nates, thanks for the reports!


(Pascal Golay) #6

Right, got it… thanks.

-Pascal


#7

My version info

Work In Progress
(6.0.17094.9041, 4/4/2017)
Beta

Command: RhinoCycles_ShowInfo

RhinoCycles 6.0.17094.09041 @ C:\Program Files\Rhino WIP\Plug-ins\RhinoCycles.rhp
CCSycles 6.0.17094.09041 @ C:\Program Files\Rhino WIP\Plug-ins\csycles.dll

The dark shadow under the light remains even if I rotate the light along it’s axis. Rhino WIP says it’s up-to-date though. Perhaps I have an old dll?


(Nathan 'jesterKing' Letwory) #8

Hmm, that should be fine. I haven’t tinkered with lights since. What hardware do you use to render? GPU or CPU? Brand?

I don’t use rotation information to create the mesh. Just cylinder parameters (base, length, radius), so rotating on its axis doesn’t have any effect.

That’d be bad. Can you also tell me what Windows Explorers tells you about RhinoCyclesCore.dll in C:\Program Files\Rhino WIP\Plug-ins\. You should be able to get the dll info from the properties of the dll.

/Nathan


(Nathan 'jesterKing' Letwory) #9

Added RH-38858 for the star effect.


#10

RhinoCyclesCore.dll File Version: 6.0.17094.9041 (just as RhinoCycles_ShowInfo shows)

This is on a GeForce GTX 480. It looks the same using the CPU or GPU. (Using RhinoCycles_SelectDevice).

I also noticed that the mesh of the light shows up in reflections and as a shadow when the sun is on.


(Nathan 'jesterKing' Letwory) #11

Thanks for the info. It all sounds good. I merged latest upstream Cycles code into our fork, so maybe that helps somewhat, too. It will be available in the next public WIP.


#12

Great. Yea, I almost expect ‘Linear Lights’ to actually just be a cylinder object with light emitting sides.


(Nathan 'jesterKing' Letwory) #13

They mostly are - as said there is no linear light primitive in Cycles, so it is simulated with a mesh light.


(Brian Gillespie) #14

RH-38857 is fixed in the latest WIP

RH-38858 and RH-38856 are fixed in the latest WIP