Mantaflow - force fields have very low influence compare to 2.90.1 #82488

New Issue

flavien giroud · 2020-11-07T16:18:36+01:00

flavien giroud commented

2020-11-07 16:18:36 +01:00

System Information
Operating system: Linux-5.4.0-52-generic-x86_64-with-debian-bullseye-sid 64 Bits
Graphics card: GeForce GTX 970/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 450.80.02

Blender Version
Broken: version: 2.92.0 Alpha, branch: master, commit date: 2020-11-06 19:20, hash: d2c102060d

            version: 2.91.0 Beta, branch: master, commit date: 2020-11-06 16:25, hash: `rB42980abf8d0d`

Worked: 2.90.1

Force fields appear to be approximately 50X less powerful in master and 2.91.0

To reproduce the error: go to frame 1, change the resolution to 34 for updating the parameters, and play/cache the simulation in both 2.90.1 and 2.91.0 or master.

Liquid_forcefield_test.blend

**System Information** Operating system: Linux-5.4.0-52-generic-x86_64-with-debian-bullseye-sid 64 Bits Graphics card: GeForce GTX 970/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 450.80.02 **Blender Version** Broken: version: 2.92.0 Alpha, branch: master, commit date: 2020-11-06 19:20, hash: `d2c102060d` ``` version: 2.91.0 Beta, branch: master, commit date: 2020-11-06 16:25, hash: `rB42980abf8d0d` ``` Worked: 2.90.1 Force fields appear to be approximately 50X less powerful in master and 2.91.0 To reproduce the error: go to frame 1, change the resolution to 34 for updating the parameters, and play/cache the simulation in both 2.90.1 and 2.91.0 or master. [Liquid_forcefield_test.blend](https://archive.blender.org/developer/F9228621/Liquid_forcefield_test.blend)

flavien giroud commented

2020-11-07 16:18:36 +01:00

Added subscriber: @flaviengiroud

Sebastián Barschkis commented

2020-11-09 12:31:38 +01:00

Added subscriber: @sebbas

Sebastián Barschkis commented

2020-11-09 12:31:38 +01:00

Changed status from 'Needs Triage' to: 'Confirmed'

Sebastián Barschkis commented

2020-11-09 12:31:38 +01:00

Yes, this should not be like that. Fix is coming up!

blender-admin commented

2020-11-09 12:53:27 +01:00

This issue was referenced by 0e6820cc5d

This issue was referenced by 0e6820cc5d6244236dcf0087784202da4b7ff817

Sebastián Barschkis commented

2020-11-09 12:54:13 +01:00

Changed status from 'Confirmed' to: 'Resolved'

Sebastián Barschkis closed this issue

2020-11-09 12:54:13 +01:00

Sebastián Barschkis self-assigned this 2020-11-09 12:54:13 +01:00

flavien giroud commented

2020-11-09 13:05:06 +01:00

Thanks, that was fast!
Now i can test the new apic solver.

Thanks, that was fast! Now i can test the new apic solver.

flavien giroud commented

2020-11-10 15:49:43 +01:00

It appear that the problem persists, although less pronounced.
Here is a video comparing blender 2.90.1 and a build from today.

Mantaflow_force_field_tests0001-0300.mkv

It appear that the problem persists, although less pronounced. Here is a video comparing blender 2.90.1 and a build from today. [Mantaflow_force_field_tests0001-0300.mkv](https://archive.blender.org/developer/F9253009/Mantaflow_force_field_tests0001-0300.mkv)

Philipp Oeser commented

2020-11-10 16:05:32 +01:00

Added subscriber: @lichtwerk

Philipp Oeser commented

2020-11-10 16:05:32 +01:00

Changed status from 'Resolved' to: 'Confirmed'

Philipp Oeser reopened this issue

2020-11-10 16:05:32 +01:00

Philipp Oeser commented

2020-11-10 16:05:32 +01:00

@sebbas: mind checking again?

flavien giroud commented

2020-11-10 16:13:42 +01:00

Of course, with pleasure.

flavien giroud commented

2020-11-10 17:06:39 +01:00

I did a new test with a turbulence field:

add a quick liquid
disable gravity in both scene and domain
add a turbulence field
increase resolution to 34 to update parameters

This time I redid the same scene from the beginning instead of opening a scene saved in 2.90.1

Result: the turbulance field is much weaker in 2.91.0 beta

I did a new test with a turbulence field: add a quick liquid disable gravity in both scene and domain add a turbulence field increase resolution to 34 to update parameters This time I redid the same scene from the beginning instead of opening a scene saved in 2.90.1 Result: the turbulance field is much weaker in 2.91.0 beta

Sebastián Barschkis commented

2020-11-12 14:12:39 +01:00

Made some further adjustments to forces in d0c1d93b7e. Forces should be stronger than before but weaker than in 2.90.1. This change is on purpose as forces in 2.90.1 turned out to be too aggressive.

It should be noted that very, very large forces flowing out of open domain boundaries can currently cause smoke to "to explode". This will be fixed separately.

@flaviengiroud Can you try out the latest build (make sure the build is newer than this comment) and report your findings back again?

Made some further adjustments to forces in d0c1d93b7e. Forces should be stronger than before but weaker than in 2.90.1. This change is on purpose as forces in 2.90.1 turned out to be too aggressive. It should be noted that very, very large forces flowing out of open domain boundaries can currently cause smoke to "to explode". This will be fixed separately. @flaviengiroud Can you try out the latest build (make sure the build is newer than this comment) and report your findings back again?

flavien giroud commented

2020-11-12 21:04:36 +01:00

I buid 2.92.0 and I didn't see any differences
Maybe the commit is not merge to 2.92.0?
I am waiting for builbot and I will test againt

I buid 2.92.0 and I didn't see any differences Maybe the commit is not merge to 2.92.0? I am waiting for builbot and I will test againt

flavien giroud commented

2020-11-13 06:13:17 +01:00

Here is a test with the last buidbot (2020-11-12 19:14, hash: `d59fa12f2a)

Mantaflow_force_field_Nov13.mkv

The difference is really subtle, if there is any.

Here is a test with the last buidbot (2020-11-12 19:14, hash: `d59fa12f2a) [Mantaflow_force_field_Nov13.mkv](https://archive.blender.org/developer/F9271659/Mantaflow_force_field_Nov13.mkv) The difference is really subtle, if there is any.

Dalai Felinto commented

2020-11-13 10:59:06 +01:00

Added subscriber: @dfelinto

Dalai Felinto commented

2020-11-13 10:59:06 +01:00

@sebbas remember to update the release logs as well. Any intentional regression needs to be mentioned there.

Sebastián Barschkis commented

2020-11-17 21:29:15 +01:00

Short test to illustrate a problem with forces in 2.90.1. And how 2.91 improves this:

2.90.1: Scale (domain and inflow) "as in test file" vs Scale (domain and inflow) by 0.01:

forces_2901_scale_1.mov forces_2901_scale_001.mov
2.91: Scale (domain and inflow) "as in test file" vs Scale (domain and inflow) by 0.01:

forces_291_scale_1.mov forces_291_scale_001.mov

Short test to illustrate a problem with forces in 2.90.1. And how 2.91 improves this: - **2.90.1**: Scale (domain and inflow) "as in test file" vs Scale (domain and inflow) by 0.01: |[forces_2901_scale_1.mov](https://archive.blender.org/developer/F9313628/forces_2901_scale_1.mov)|[forces_2901_scale_001.mov](https://archive.blender.org/developer/F9313648/forces_2901_scale_001.mov)| | -- | -- | - **2.91**: Scale (domain and inflow) "as in test file" vs Scale (domain and inflow) by 0.01: |[forces_291_scale_1.mov](https://archive.blender.org/developer/F9313644/forces_291_scale_1.mov)|[forces_291_scale_001.mov](https://archive.blender.org/developer/F9313652/forces_291_scale_001.mov)| | -- | -- |

Sebastián Barschkis commented

2020-11-17 21:32:31 +01:00

While the new scaling of forces in 2.91 would be different, I think the benefits outweigh the cost.

Once this has been added to the release notes, I think this task can be closed.

While the new scaling of forces in 2.91 would be different, I think the benefits outweigh the cost. Once this has been added to the release notes, I think this task can be closed.

flavien giroud commented

2020-11-17 23:17:22 +01:00

what do you mean by "scaling of forces", strength of effector, domain scale or resolution?

Sebastián Barschkis commented

2020-11-18 10:11:46 +01:00

EDIT: I have edited the post with the videos to make it more clear.

ronan ducluzeau commented

2020-11-18 11:37:50 +01:00

Added subscriber: @zeauro

ronan ducluzeau commented

2020-11-18 11:37:50 +01:00

Is it possible to make mantaflow respecting Unit Scale in Scene properties ?
This is just a simple factor to take into account and that would solve the issue of having to scale down particles display size.

Is it possible to make mantaflow respecting Unit Scale in Scene properties ? This is just a simple factor to take into account and that would solve the issue of having to scale down particles display size.

Sebastián Barschkis commented

2020-11-18 15:45:59 +01:00

Ok, as we're nearing the 2.91 release I have decided to postpone the changes made in this task. I.e. 2.91 will have the same force behavior as 2.90.1.

One of the reason for this is that if the behavior of forces changes it should be in line with other physics modifiers (e.g. cloth). This is currently not the case and needs more work.

The task will remain open and hopefully there will be improvements for 2.92.

@zeauro Yes, such changes should be part of the improvements for 2.92

Ok, as we're nearing the 2.91 release I have decided to postpone the changes made in this task. I.e. 2.91 will have the same force behavior as 2.90.1. One of the reason for this is that if the behavior of forces changes it should be in line with other physics modifiers (e.g. cloth). This is currently not the case and needs more work. The task will remain open and hopefully there will be improvements for 2.92. @zeauro Yes, such changes should be part of the improvements for 2.92

Philipp Oeser removed the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Mantaflow - force fields have very low influence compare to 2.90.1 #82488