Adaptive Domain creates lines in smoke #74559

Open
opened 2020-03-09 03:10:09 +01:00 by Andrew Price · 51 comments

System Information
Operating system: Win 10
Graphics card: Dual RTX Titans

Blender Version
Broken: Blender 2.82

Short description of error
With adaptive domain enabled, noticeable horizontal (and sometimes vertical) lines appear in the smoke
weird lines.mp4

Exact steps for others to reproduce the error
Simple blend file:

  • Create a large box and set it as a gas domain.
  • Create a small object as the Flow (both 'smoke' and 'fire' reproduce it)
  • Put the flow object near the bottom of the domain
  • Enable Adaptive Domain
  • Bake
  • You should see lines like this: blender_2020-03-09_11-55-29.png

Turn off adaptive. Not only will the smoke move faster creating a totally different looking simulation (which is problem in itself) but the lines will be gone: blender_2020-03-09_12-09-10.png

It's most noticeable when the domain is large, and the adaptive box is only a small portion of it.

Increasing the adaptive margin to 24 helps somewhat, but it's still there. Any higher and it defeats the purpose of adaptive - baking takes as long as just turning it off.

Sample blend:
adaptive bug2.blend

**System Information** Operating system: Win 10 Graphics card: Dual RTX Titans **Blender Version** Broken: Blender 2.82 **Short description of error** With adaptive domain enabled, noticeable horizontal (and sometimes vertical) lines appear in the smoke [weird lines.mp4](https://archive.blender.org/developer/F8396247/weird_lines.mp4) **Exact steps for others to reproduce the error** Simple blend file: - Create a large box and set it as a gas domain. - Create a small object as the Flow (both 'smoke' and 'fire' reproduce it) - Put the flow object near the bottom of the domain - Enable Adaptive Domain - Bake - You should see lines like this: ![blender_2020-03-09_11-55-29.png](https://archive.blender.org/developer/F8396237/blender_2020-03-09_11-55-29.png) # Turn off adaptive. Not only will the smoke move faster creating a totally different looking simulation (which is problem in itself) but the lines will be gone: ![blender_2020-03-09_12-09-10.png](https://archive.blender.org/developer/F8396261/blender_2020-03-09_12-09-10.png) It's most noticeable when the domain is large, and the adaptive box is only a small portion of it. Increasing the adaptive margin to 24 helps somewhat, but it's still there. Any higher and it defeats the purpose of adaptive - baking takes as long as just turning it off. Sample blend: [adaptive bug2.blend](https://archive.blender.org/developer/F8400219/adaptive_bug2.blend)
Author

Added subscriber: @AndrewPrice

Added subscriber: @AndrewPrice

Added subscriber: @johny.zlo

Added subscriber: @johny.zlo
Member

Added subscriber: @ankitm

Added subscriber: @ankitm
Member

https://developer.blender.org/T67356#867120 are these the same reports? does the last comment by Germano help ?

https://developer.blender.org/T67356#867120 are these the same reports? does the last comment by Germano help ?
Author

No this is separate. The lines are part of the simulation, not the shader.

It's most obvious at the end of the video above, as you can see it creates waves in the plumes.

No this is separate. The lines are part of the simulation, not the shader. It's most obvious at the end of the video above, as you can see it creates waves in the plumes.
Member

https://docs.blender.org/manual/en/latest/physics/fluid/type/domain/gas/adaptive_domain.html?highlight=adaptive%20domain
I still think this is expected behaviour.. Adaptive domain saves some computation time by shrinking the gas boundary. Increasing margin gives more space to the gas, thus more computation, thus it undercuts the main benefit of AD: save computation time.
bottomline: it's a tradeoff between time saving and accuracy (accidental Heisenberg? )

https://docs.blender.org/manual/en/latest/physics/fluid/type/domain/gas/adaptive_domain.html?highlight=adaptive%20domain I still think this is expected behaviour.. Adaptive domain saves some computation time by shrinking the gas boundary. Increasing margin gives more space to the gas, thus more computation, thus it undercuts the main benefit of AD: save computation time. bottomline: it's a tradeoff between time saving and accuracy (accidental Heisenberg? )

Added subscriber: @iss

Added subscriber: @iss

Changed status from 'Needs Triage' to: 'Needs User Info'

Changed status from 'Needs Triage' to: 'Needs User Info'

This is effect of low domain resolution most likely
https://docs.blender.org/manual/en/dev/physics/fluid/type/domain/settings.html#bpy-types-fluiddomainsettings-resolution-max

Try to increase and see if issue persists.

Please provide sample file so we can actually see problem in action.

This is effect of low domain resolution most likely https://docs.blender.org/manual/en/dev/physics/fluid/type/domain/settings.html#bpy-types-fluiddomainsettings-resolution-max Try to increase and see if issue persists. Please provide sample file so we can actually see problem in action.
Author

Added subscriber: @sebbas

Added subscriber: @sebbas
Author

@iss Changing the resolution doesn't change much. It just makes the gaps between the lines smaller:
adaptive_compare.jpg

@ankitm I'd be surprised if this was the expected behaviour. If it only expands after it's needed, rather than preemptively based on the movement of the smoke then it should never be used. Coz it will always interrupt the smoke and create a totally different simulation than when disabled.

@sebbas I wonder if it's related to the Adaptive Time Steps bug , not computing multiple times between frames to create the right sized adaptive domain?

Also I've attached the .blend to the OP.

@iss Changing the resolution doesn't change much. It just makes the gaps between the lines smaller: ![adaptive_compare.jpg](https://archive.blender.org/developer/F8400221/adaptive_compare.jpg) @ankitm I'd be surprised if this was the expected behaviour. If it only expands after it's needed, rather than preemptively based on the movement of the smoke then it should never be used. Coz it will always interrupt the smoke and create a totally different simulation than when disabled. @sebbas I wonder if it's related to the [Adaptive Time Steps bug ](https://developer.blender.org/T74062), not computing multiple times between frames to create the right sized adaptive domain? Also I've attached the .blend to the OP.

Changed status from 'Needs User Info' to: 'Confirmed'

Changed status from 'Needs User Info' to: 'Confirmed'

Added subscriber: @BeckersC

Added subscriber: @BeckersC

Added subscriber: @GottfriedHofmann

Added subscriber: @GottfriedHofmann

I can confirm as well - the results are completely different with and without adaptive domain even though they should be identical.

I can confirm as well - the results are completely different with and without adaptive domain even though they should be identical.
Sebastián Barschkis was assigned by Gottfried Hofmann 2020-03-16 16:16:17 +01:00
Member

@GottfriedHofmann did you discuss it with Sebastian already?

@GottfriedHofmann did you discuss it with Sebastian already?

In #74559#892076, @ankitm wrote:
@GottfriedHofmann did you discuss it with Sebastian already?

Yes :-)

> In #74559#892076, @ankitm wrote: > @GottfriedHofmann did you discuss it with Sebastian already? Yes :-)

Added subscriber: @deadpin

Added subscriber: @deadpin

Are .blends from 2.82 supposed to be usable in 2.83? I can't get the attached .blend to work in recent builds (after bake). Wanted to see if things improved here.

Are .blends from 2.82 supposed to be usable in 2.83? I can't get the attached .blend to work in recent builds (after bake). Wanted to see if things improved here.

Added subscriber: @Adam-Burke

Added subscriber: @Adam-Burke

The problem where the old .blend didn't open has been resolved (was because it was set to openvdb/modular format). However, the issue with the lines and completely different looking simulation still seem present :-/

The problem where the old .blend didn't open has been resolved (was because it was set to openvdb/modular format). However, the issue with the lines and completely different looking simulation still seem present :-/
Added subscriber: @Iago-Diogo-de-Vasconcelos-Mota
Author

Any update on this? Sean Kennedy also ran into it.

Any update on this? [Sean Kennedy ](https://twitter.com/openvisualfx/status/1284187861105819648) also ran into it.

Been working on other areas lately. The adaptive domain is really causing lots of problems, so now is probably a good time to look at it again.

Been working on other areas lately. The adaptive domain is really causing lots of problems, so now is probably a good time to look at it again.

Just pushed 62a819202e which should improve the adaptive domain in general. In my tests I noticed that the "cutting-off" issue seems to be solved.
Maybe everyone from this thread can try out the latest master and report their findings?

There will be one more improvement commit for the adaptive domain to hopefully improve some of the blockiness - but that's just fyi.

Just pushed 62a819202e which should improve the adaptive domain in general. In my tests I noticed that the "cutting-off" issue seems to be solved. Maybe everyone from this thread can try out the latest master and report their findings? There will be one more improvement commit for the adaptive domain to hopefully improve some of the blockiness - but that's just fyi.

Added subscriber: @Creavite

Added subscriber: @Creavite
Member

Added subscriber: @SeanKennedy

Added subscriber: @SeanKennedy
Author

Nice! I'll give it a try tomorrow once the builds are updated with the new fix.

Nice! I'll give it a try tomorrow once the builds are updated with the new fix.
Member

lines_in_render_01.JPG
lines_in_render_01_v290.JPG

Today's Blender build still seems to have the issue. This is with Adaptive Domain turned on, I didn't try with it turned off yet.

![lines_in_render_01.JPG](https://archive.blender.org/developer/F8712667/lines_in_render_01.JPG) ![lines_in_render_01_v290.JPG](https://archive.blender.org/developer/F8712670/lines_in_render_01_v290.JPG) Today's Blender build still seems to have the issue. This is with Adaptive Domain turned on, I didn't try with it turned off yet.

Added subscriber: @Gadas

Added subscriber: @Gadas
Author

Any update on this one? I just tested it in 2.9 using the bug blend file attached in OP and it's still happening.

Screenshot 2020-09-25 095804.jpg

Any update on this one? I just tested it in 2.9 using the bug blend file attached in OP and it's still happening. ![Screenshot 2020-09-25 095804.jpg](https://archive.blender.org/developer/F8917671/Screenshot_2020-09-25_095804.jpg)

It's bug sprint week and a good time to look into this again.

It's bug sprint week and a good time to look into this again.
Member

My new computer arrives in a week, if there's anything I can test for you after that, please just ask. Excited to improve this!

My new computer arrives in a week, if there's anything I can test for you after that, please just ask. Excited to improve this!

An update! @AndrewPrice @SeanKennedy can you try out the latest state of the master branch (8720c66009)?
In my test the example scene from this report behaved much better. Would be interesting to see how some of your other scenes perform.

An update! @AndrewPrice @SeanKennedy can you try out the latest state of the master branch (8720c66009)? In my test the example scene from this report behaved much better. Would be interesting to see how some of your other scenes perform.

Yep, the example scene is looking a lot better! There are still some occasional lines, though:

AdaptiveLines2.92.png

But, those could be reduced by increasing the margin, after which, I ended up with this:

AdaptiveLines2.92Margin.png

So maybe part of the issue could be related to smoke getting too close to the bounds?

Yep, the example scene is looking a lot better! There are still some occasional lines, though: ![AdaptiveLines2.92.png](https://archive.blender.org/developer/F9218346/AdaptiveLines2.92.png) But, those could be reduced by increasing the margin, after which, I ended up with this: ![AdaptiveLines2.92Margin.png](https://archive.blender.org/developer/F9218350/AdaptiveLines2.92Margin.png) So maybe part of the issue could be related to smoke getting too close to the bounds?
Author

Just tried it in 2.91 using the .blend from the OP and am still seeing lines.

Screenshot 2020-11-29 063756.jpg

@sebbas Are you testing your patches on the .blend from the OP?

Just tried it in 2.91 using the .blend from the OP and am still seeing lines. ![Screenshot 2020-11-29 063756.jpg](https://archive.blender.org/developer/F9401932/Screenshot_2020-11-29_063756.jpg) @sebbas Are you testing your patches on the .blend from the OP?
Member

I also get the lines with the OP blend file

lines_in_render_03.JPG

lines_in_render_04.JPG

lines_in_render_02.JPG

I also tried adjusting the adaptive domain threshold, but didn't get results that were better. I tried @Iago-Diogo-de-Vasconcelos-Mota 's suggestion about adjusting the margin, as well. Those first 3 screenshots are with the margin set at 4.

This one is margin of 2
lines_in_render_05_margin2.JPG

This one is margin of 24
lines_in_render_06_margin24.JPG

So I got lines no matter what the margin was set to.

I also get the lines with the OP blend file ![lines_in_render_03.JPG](https://archive.blender.org/developer/F9418457/lines_in_render_03.JPG) ![lines_in_render_04.JPG](https://archive.blender.org/developer/F9418456/lines_in_render_04.JPG) ![lines_in_render_02.JPG](https://archive.blender.org/developer/F9418455/lines_in_render_02.JPG) I also tried adjusting the adaptive domain threshold, but didn't get results that were better. I tried @Iago-Diogo-de-Vasconcelos-Mota 's suggestion about adjusting the margin, as well. Those first 3 screenshots are with the margin set at 4. This one is margin of 2 ![lines_in_render_05_margin2.JPG](https://archive.blender.org/developer/F9418473/lines_in_render_05_margin2.JPG) This one is margin of 24 ![lines_in_render_06_margin24.JPG](https://archive.blender.org/developer/F9418472/lines_in_render_06_margin24.JPG) So I got lines no matter what the margin was set to.

Added subscriber: @jwu420

Added subscriber: @jwu420

Added subscriber: @god1ged

Added subscriber: @god1ged

Hello, you mean this is still not solved right ?

Hello, you mean this is still not solved right ?
Author

Correct. Still unsolved.
@sebbas have you had a chance to revisit it yet?

Correct. Still unsolved. @sebbas have you had a chance to revisit it yet?

Added subscriber: @alexisrivera

Added subscriber: @alexisrivera

Guys, I have the same lines as the video posted here by Andrew, I thought It was my computer until I found this topic

0001-0050.mp4

Guys, I have the same lines as the video posted here by Andrew, I thought It was my computer until I found this topic [0001-0050.mp4](https://archive.blender.org/developer/F9853744/0001-0050.mp4)
Author

Is there any update on this?

Is there any update on this?

Added subscriber: @ChristophWerner

Added subscriber: @ChristophWerner

In #74559#1190517, @AndrewPrice wrote:
Is there any update on this?

Nope.
Checked it today in Blender 3.0 daily build.

> In #74559#1190517, @AndrewPrice wrote: > Is there any update on this? Nope. Checked it today in Blender 3.0 daily build.

Added subscriber: @MichaelHermann

Added subscriber: @MichaelHermann
Contributor

Added subscriber: @Raimund58

Added subscriber: @Raimund58

Added subscriber: @zebus3dream

Added subscriber: @zebus3dream
Philipp Oeser removed the
Interest
Nodes & Physics
label 2023-02-10 08:47:04 +01:00

Hello guys. I'm having the same problem with both Blender 3.5 (open VDB or native smoke sim) and Cinema 4D's 2023.2 Pyro as well.
The first two images are C4D's viewport. The resolution is set really high there, but nothing matters, I get the pattern all the same. The 3rd image is from Blender's (cycles render) smoke sim, 96 res and no adaptive domain.
The video (Blender, Cycles render) is an open VDB from Embergen's default explosion template, the pattern appears in the fire once I render it in Blender.
So I'm having this pattern everywhere, regardless of resolution. Anyone know what it could possibly be?

Hello guys. I'm having the same problem with both Blender 3.5 (open VDB or native smoke sim) and Cinema 4D's 2023.2 Pyro as well. The first two images are C4D's viewport. The resolution is set really high there, but nothing matters, I get the pattern all the same. The 3rd image is from Blender's (cycles render) smoke sim, 96 res and no adaptive domain. The video (Blender, Cycles render) is an open VDB from Embergen's default explosion template, the pattern appears in the fire once I render it in Blender. So I'm having this pattern everywhere, regardless of resolution. Anyone know what it could possibly be?

Still seeing this in 3.6, anyone have any updates or fixes that make it less noticeable?

Still seeing this in 3.6, anyone have any updates or fixes that make it less noticeable?
Sign in to join this conversation.
No Label
Interest
Alembic
Interest
Animation & Rigging
Interest
Asset Browser
Interest
Asset Browser Project Overview
Interest
Audio
Interest
Automated Testing
Interest
Blender Asset Bundle
Interest
BlendFile
Interest
Collada
Interest
Compatibility
Interest
Compositing
Interest
Core
Interest
Cycles
Interest
Dependency Graph
Interest
Development Management
Interest
EEVEE
Interest
EEVEE & Viewport
Interest
Freestyle
Interest
Geometry Nodes
Interest
Grease Pencil
Interest
ID Management
Interest
Images & Movies
Interest
Import Export
Interest
Line Art
Interest
Masking
Interest
Metal
Interest
Modeling
Interest
Modifiers
Interest
Motion Tracking
Interest
Nodes & Physics
Interest
OpenGL
Interest
Overlay
Interest
Overrides
Interest
Performance
Interest
Physics
Interest
Pipeline, Assets & IO
Interest
Platforms, Builds & Tests
Interest
Python API
Interest
Render & Cycles
Interest
Render Pipeline
Interest
Sculpt, Paint & Texture
Interest
Text Editor
Interest
Translations
Interest
Triaging
Interest
Undo
Interest
USD
Interest
User Interface
Interest
UV Editing
Interest
VFX & Video
Interest
Video Sequencer
Interest
Virtual Reality
Interest
Vulkan
Interest
Wayland
Interest
Workbench
Interest: X11
Legacy
Blender 2.8 Project
Legacy
Milestone 1: Basic, Local Asset Browser
Legacy
OpenGL Error
Meta
Good First Issue
Meta
Papercut
Meta
Retrospective
Meta
Security
Module
Animation & Rigging
Module
Core
Module
Development Management
Module
EEVEE & Viewport
Module
Grease Pencil
Module
Modeling
Module
Nodes & Physics
Module
Pipeline, Assets & IO
Module
Platforms, Builds & Tests
Module
Python API
Module
Render & Cycles
Module
Sculpt, Paint & Texture
Module
Triaging
Module
User Interface
Module
VFX & Video
Platform
FreeBSD
Platform
Linux
Platform
macOS
Platform
Windows
Priority
High
Priority
Low
Priority
Normal
Priority
Unbreak Now!
Status
Archived
Status
Confirmed
Status
Duplicate
Status
Needs Info from Developers
Status
Needs Information from User
Status
Needs Triage
Status
Resolved
Type
Bug
Type
Design
Type
Known Issue
Type
Patch
Type
Report
Type
To Do
No Milestone
No project
22 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: blender/blender#74559
No description provided.