Cycles volume low density cutoff #104504

New Issue

Thomas Kole · 2023-02-09T11:43:52+01:00

Thomas Kole commented

2023-02-09 11:43:52 +01:00

System Information
Operating system: Tested on Windows 10 and Ubuntu
Graphics card: Tested on various NVidia cards, as well as CPU rendering

Blender Version
Broken: 3.3.0
Worked: unknown

Short description of error

When rendering large volumes, such as this 2x2 Km fog around a terrain, super low densities (<0.000 01) are desired for atmospheric effects.

However, there seems to be a seemingly arbitrary cutoff at around 0.000012, where nothing is rendered.

Exact steps for others to reproduce the error
In this example, I am using a gradient from left to right to drive density from 0.00003 to 0.0.
As you can see, there's a sudden cutoff.

I can imagine that this is some sort of performance trick, so that it doesn't get hung up on low densities, but being able to adjust this cutoff value would be very handy.

Here's two BlenderArtists threads that also describe this issue.

https://blenderartists.org/t/low-density-volumes-dont-render-in-cycles-blender-3-0/1356018
https://blenderartists.org/t/volume-atmospherics-rendering-is-there-a-cutoff/1448963

**System Information** Operating system: Tested on Windows 10 and Ubuntu Graphics card: Tested on various NVidia cards, as well as CPU rendering **Blender Version** Broken: 3.3.0 Worked: unknown **Short description of error** When rendering large volumes, such as this 2x2 Km fog around a terrain, super low densities (<0.000 01) are desired for atmospheric effects. However, there seems to be a seemingly arbitrary cutoff at around 0.000012, where nothing is rendered. ![image](/attachments/32c4f82f-9f6b-4d75-af0f-24dd6affec83) **Exact steps for others to reproduce the error** In this example, I am using a gradient from left to right to drive density from 0.00003 to 0.0. As you can see, there's a sudden cutoff. ![image](/attachments/f96eb33b-13fe-4f64-b541-b56007d97fce) I can imagine that this is some sort of performance trick, so that it doesn't get hung up on low densities, but being able to adjust this cutoff value would be very handy. Here's two BlenderArtists threads that also describe this issue. https://blenderartists.org/t/low-density-volumes-dont-render-in-cycles-blender-3-0/1356018 https://blenderartists.org/t/volume-atmospherics-rendering-is-there-a-cutoff/1448963

Atmosphere_LowDensityVolume_Bug.blend

2.1 MiB

image.png

232 KiB

image.png

48 KiB

Thomas Kole added the

labels 2023-02-09 11:43:53 +01:00

Thomas Kole changed title from ~~Cycles volume density cutoff~~ to Cycles volume low density cutoff

2023-02-09 11:45:05 +01:00

Christoph Lendenfeld added a new dependency 2023-02-09 12:21:32 +01:00

#104505 Fix #104026: Click-Drag to select graph editor channels no longer working

Christoph Lendenfeld removed a dependency 2023-02-09 12:21:36 +01:00

#104505 Fix #104026: Click-Drag to select graph editor channels no longer working

Iliya Katushenock commented

2023-02-09 15:25:22 +01:00

Hello. this not a bug, but just math. Final effect of volumetric on light trace can be approximate to limit:
color = Lim{step->0} E{n = 0; n += step; n->max dist} volume.color * volume.density * step
in this case, when step is non-zero, result can be very large and you have to control this by using gradients with some curvature like smoothstep. In your example, result very different between 0.001 and zero.

Hello. this not a bug, but just math. Final effect of volumetric on light trace can be approximate to limit: `color = Lim{step->0} E{n = 0; n += step; n->max dist} volume.color * volume.density * step` in this case, when step is non-zero, result can be very large and you have to control this by using gradients with some curvature like smoothstep. In your example, result very different between 0.001 and zero.

Iliya Katushenock added

Module

Render & Cycles

Status

Needs Info from Developers

and removed

Status

Needs Triage

labels 2023-02-09 15:25:44 +01:00

Thomas Kole commented

2023-02-09 16:19:39 +01:00

I think there might be a misunderstanding.
The box spans the entire image above, the hard clip happens at 0.000012, any value below is the same as 0.0

Here's the same scene, but instead of 20Km, the terrain is 20m, and the density is interpolated from 0.003 to 0.0 to compensate. Now there is no such cutoff.

It's worth emphasising that those extremely low values (0.00001 and lower) are not a weird edge case, but necessary for rendering large terrain realistically.

I think there might be a misunderstanding. The box spans the entire image above, the hard clip happens at 0.000012, any value below is the same as 0.0 Here's the same scene, but instead of 20Km, the terrain is 20m, and the density is interpolated from 0.003 to 0.0 to compensate. Now there is no such cutoff. ![image](/attachments/f76987ed-e528-4d8c-bc1e-8f2fd79a1f47) It's worth emphasising that those extremely low values (0.00001 and lower) are not a weird edge case, but necessary for rendering large terrain realistically.

image.png

363 KiB

Iliya Katushenock commented

2023-02-09 16:26:44 +01:00

Ah, yeah, sorry. Yes, i see that. But that look like limitation.
Need information from devs/

Ah, yeah, sorry. Yes, i see that. But that look like limitation. Need information from devs/

example video.mp4

2.9 MiB

SamuelWarren commented

2023-02-19 19:14:35 +01:00

I've been having the same issue with large planetary atmospheres where the volume scatter density exponentially decays with altitude. Below a certain density threshold, the volume is seemingly ignored.

This seems like some sort of threshold that's been hardcoded in somewhere rather than a floating point precision issue (floats should be sufficient?).

I've been having the same issue with large planetary atmospheres where the volume scatter density exponentially decays with altitude. Below a certain density threshold, the volume is seemingly ignored. This seems like some sort of threshold that's been hardcoded in somewhere rather than a floating point precision issue (floats should be sufficient?).

Aras Pranckevicius referenced this issue

2023-05-30 21:28:16 +02:00

IO: remove OBJ and PLY Python add-ons (#104503, #104504) #108440

Aras Pranckevicius referenced this issue from a commit

2023-05-30 21:41:32 +02:00

IO: remove OBJ and PLY Python add-ons (#104503, #104504)

Iliya Katushenock added

Status

Confirmed

and removed

Status

Needs Info from Developers