2.83 Mantaflow Smoke Simulation doesn't match 2.82 with Same Object Size when it's smaller than a particular value #79029

New Issue

REYNEP · 2020-07-17T18:10:18+02:00

REYNEP commented

2020-07-17 18:10:18 +02:00

System Information
Operating system: Windows 10
Graphics card: Radeon RX 580

Blender Version
Broken: 2.83.2, branch: master, commit date: 2020-07-09 05:50, hash: 239fbf7d93, type: Release
Worked: 2.82 (sub 7), branch: master, commit date: 2020-03-12 05:06, hash: 375c7dc4ca, type: Release

Short description of error
2.83 Mantaflow Simulation is not same as 2.82 with Same object Size when Size is Smaller than a Particular Value. 2.82 is looks less Dense in viewport Solid Mode. I Discovered This a Week ago. Then found Ram Singh (crossMind Studio) Tweeted a Side by Side Comparison about the Change from 2.82 to 2.83.

Twitter Link:- https://twitter.com/sebbas/status/1271478115198435328
I Believe Sebastián Barschkis is Looking into It
(Twitter Kills the Quality) Youtube Link:- https://www.youtube.com/watch?v=df7_yW_9-M8

UPDATE:- I Just Found this Thread: #78290 But I still Think CrossMind's Comparison would help.... And Also I did not Import from 2.82 to 2.83, I Tested Separately!

Exact steps for others to reproduce the error

On The Default cube I Added a 'Quick Smoke'
Scaled Down the Cube to a Smaller Size and Increased the Res to 96 That's All
Tests are done Separately on 2.83.2 and 2.82, No Import was Done

Like in the Picture:-
(2.83.2) Frame: 61
(2.82) Frame: 61

And Please, See The Tweet of Comparison that CrossMind did, for Side by Side Comparison on more Dissimilarities....
YOU WILL HAVE TO BAKE THE SIM
THANKS For Your Time!!!! (This is my First Time Reporting a Bug,) testMantaflow2-83.blend testMantaflow2-82.blend

**System Information** Operating system: Windows 10 Graphics card: Radeon RX 580 **Blender Version** Broken: 2.83.2, branch: master, commit date: 2020-07-09 05:50, hash: 239fbf7d936f, type: Release Worked: 2.82 (sub 7), branch: master, commit date: 2020-03-12 05:06, hash: 375c7dc4caf4, type: Release **Short description of error** 2.83 Mantaflow Simulation is not same as 2.82 with Same object Size when Size is Smaller than a Particular Value. 2.82 is looks less Dense in viewport Solid Mode. I Discovered This a Week ago. Then found Ram Singh (crossMind Studio) Tweeted a *Side by Side Comparison* about the Change from 2.82 to 2.83. Twitter Link:- https://twitter.com/sebbas/status/1271478115198435328 I Believe ***Sebastián Barschkis*** is Looking into It (Twitter Kills the Quality) Youtube Link:- https://www.youtube.com/watch?v=df7_yW_9-M8 **UPDATE:-** I Just Found this Thread: #78290 But I still Think CrossMind's Comparison would help.... And Also I did not Import from 2.82 to 2.83, I Tested Separately! **Exact steps for others to reproduce the error** 1. On The Default cube I Added a 'Quick Smoke' 2. Scaled Down the Cube to a Smaller Size and Increased the Res to 96 That's All 3. Tests are done Separately on 2.83.2 and 2.82, No Import was Done Like in the Picture:- (***2.83.2**) Frame: 61* ![blender_WWtUdjNs69.png](https://archive.blender.org/developer/F8701947/blender_WWtUdjNs69.png) (***2.82**) Frame: 61* ![blender_rDMWtk6V7X.png](https://archive.blender.org/developer/F8701996/blender_rDMWtk6V7X.png) *And Please, See The Tweet of Comparison that CrossMind did, for **Side by Side Comparison** on more Dissimilarities....* *YOU WILL HAVE TO BAKE THE SIM* ***THANKS For Your Time!!!! (This is my First Time Reporting a Bug,)*** [testMantaflow2-83.blend](https://archive.blender.org/developer/F8701992/testMantaflow2-83.blend) [testMantaflow2-82.blend](https://archive.blender.org/developer/F8701993/testMantaflow2-82.blend)

REYNEP commented

2020-07-17 18:10:18 +02:00

Added subscriber: @REYNEP

REYNEP changed title from ~~2.83 Mantaflow Simulation is not same as it was in 2.82~~ to 2.83 Mantaflow Smoke Simulation is not same as it was in 2.82

2020-07-17 18:22:34 +02:00

REYNEP commented

2020-07-17 19:16:44 +02:00

About 2.90 Comparison:- In the Tweet, Sebastian Tweeted that, 2.90 will have some Changes to Gravity. I Tried Changing The Scene Gravity, to a point where the smoke will match the height of what 2.82 and 2.83 had when they were at frame 61

When I changed to 16m/s², It Kinda looked similar to 2.83! But Still wasn't The Height of 2.83 at frame 61. But the Shape Was Similar to 2.83 as You can SEE:-

Blender Version:- 2.90.0 Alpha, branch: master, commit date: 2020-07-17 07:39, hash: 0a40c671b0, type: Release

Note:- This was a Separate test. I did not import anything, I created Cube and Domain of same size as in the prev files and same quick smoke with same changes as described at the Task Description. I am pretty Sure that Importing would have not solved it. as described on #78290

testMantaflow2-90a.blend

***About 2.90 Comparison***:- In the Tweet, Sebastian Tweeted that, 2.90 will have some Changes to Gravity. I Tried Changing The Scene Gravity, to a point where the smoke will match the *height* of what *2.82 and 2.83* had when they were at **frame 61** When I changed to ***16m/s²***, It Kinda looked similar to 2.83! But Still wasn't The Height of 2.83 at frame 61. But the ***Shape Was Similar* to 2.83** as You can SEE:- ![blender_ypnTJRDBL8.png](https://archive.blender.org/developer/F8702194/blender_ypnTJRDBL8.png) Blender Version:- 2.90.0 Alpha, branch: master, commit date: 2020-07-17 07:39, hash: 0a40c671b0b5, type: Release Note:- This was a Separate test. I did not import anything, I created Cube and Domain of same size as in the prev files and same quick smoke with same changes as described at the Task Description. I am pretty Sure that Importing would have not solved it. as described on #78290 [testMantaflow2-90a.blend](https://archive.blender.org/developer/F8702201/testMantaflow2-90a.blend)

Sebastián Barschkis commented

2020-07-20 12:14:35 +02:00

Added subscriber: @sebbas

Sebastián Barschkis commented

2020-07-20 12:14:35 +02:00

Yes, I looked into this and it turns out that there are 2 things that make the smoke look different (2.82 compared to current master 2.90):

Smoke more Dense: Is due to the changes from D6951. From 2.83 on smoke looks more dense in workbench - if you render in Cycles, for example, the result should be identical though.
Less Gravity: In general, the gravity should be more physically-correct for fluids in 2.90. With the current buoyancy settings smoke will rise less quickly (when compared to 2.82).

@REYNEP can you try out the following things:

Bake and render a simulation in 2.82, open it in 2.90 and render. The rendered results should be the same.
In 2.90 and with the file from 2.82, change both buoyancy parameters ("Buoyancy Density", "Buoyancy Heat") a bit (e.g. increase them from "1.0" and "1.0" to "4.0" and "5.0"). Then simulate. Then open those results in 2.82 and compare them to the original 2.82 simulation (with buoyancy "1.0" and "1.0"). The simulations should look very similar in workbench.

In the future and if people prefer faster rising smoke, it might make sense to increase the default buoyancy values a bit.

Yes, I looked into this and it turns out that there are 2 things that make the smoke look different (2.82 compared to current master 2.90): - Smoke more *Dense*: Is due to the changes from [D6951](https://archive.blender.org/developer/D6951). From 2.83 on smoke looks more dense in workbench - if you render in Cycles, for example, the result should be identical though. - Less *Gravity*: In general, the gravity should be more physically-correct for fluids in 2.90. With the current buoyancy settings smoke will rise less quickly (when compared to 2.82). @REYNEP can you try out the following things: - Bake and render a simulation in 2.82, open it in 2.90 and render. The rendered results should be the same. - In 2.90 and with the file from 2.82, change both buoyancy parameters ("Buoyancy Density", "Buoyancy Heat") a bit (e.g. increase them from "1.0" and "1.0" to "4.0" and "5.0"). Then simulate. Then open those results in 2.82 and compare them to the original 2.82 simulation (with buoyancy "1.0" and "1.0"). The simulations should look very similar in workbench. In the future and if people prefer faster rising smoke, it might make sense to increase the default buoyancy values a bit.

Sebastián Barschkis commented

2020-07-20 12:15:03 +02:00

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

REYNEP commented

2020-07-20 20:59:49 +02:00

My APOLOGY For Late Reply, @sebbas Sir. I checked Out, (With The Same 2.82 Version and 2.90 Version) I tried Baking, and rendering in 2.82 and then opening and rendering it in 2.90. ***They Look Similar Just As You Said, Both in Workbench and After Rendering!!***😄, (I Think I am Gonna Share Some Tests with Force Fields with 128+RES!) What I wanna Demonstrate here inside the VIDEO is that. 2.83 96RES Looks really different from other 96RES Sims. Other 96Res (Including 2.90) Sims are Really Different From 128Res, except for 2.83 96Res (2.83 96Res is Identicle to 2.83 128Res)!

Here Goes First 100 Frames (SAME Quick SmokeSETTINGS and same Quick-Smoke Default Shader only with Density 7) :-

``` ```

From 00:00	To 00:04	From 00:04	To 00:08	From 00:08	To 00:12	From 00:12	To 00:16	From 00:16	To 00:20	From 00:20	To 00:24
2.82 Res-128	2.82 Res-96	2.82 Res-128	2.82 Res-96	2.82 Res-128	2.83 Res-128	2.82 Res-128	2.83 Res-128	2.82 Res-96	2.83 Res-96	2.82 Res-128	2.83 Res-128
2.83 Res-128	2.83 Res-96	2.82Bake, 2.83Render Res-128	2.82Bake, 2.83Render Res-96	2.90 Res-128	2.90 Res-96	2.90 ByuoDens=4,Heat=5 Res-128	2.90 ByuoDens=4,Heat=5 Res-96	2.90 ByuoDens=4,Heat=5 Res-96	2.90 Res-96	2.90 ByuoDens=4,Heat=5 Res-128	2.90 Res-128

[Comparisons.mp4](https://archive.blender.org/developer/F8709918/Comparisons.mp4)

(BTW, LATE BCZ I have been doing lots of Test Sims 😃*) Last Update: Out of 10 Videos, I made 1, Where I have Put all those Side by Side in their Original 720p Video Resolution - Which Makes this Video (25601440 px)

My APOLOGY For Late Reply, @sebbas Sir. I checked Out, (*With The Same 2.82 Version and 2.90 Version*) I tried Baking, and rendering in 2.82 and then opening and rendering it in 2.90. ***They Look Similar Just As You Said, Both in Workbench and After Rendering!!***😄, *(I Think I am Gonna Share Some Tests with Force Fields with 128+RES!)* **What I wanna Demonstrate here inside the VIDEO is that. 2.83 96RES Looks really different from other 96RES Sims. Other 96Res (Including 2.90) Sims are Really Different From 128Res, except for 2.83 96Res (2.83 96Res is Identicle to 2.83 128Res)!** Here Goes First 100 Frames (*SAME **Quick Smoke**SETTINGS and same **Quick-Smoke Default Shader** only with **Density 7***) :- <table> ``` <tr> <th>From 00:00</th> <th>To 00:04</th> <th>From 00:04</th> <th>To 00:08</th> <th>From 00:08</th> <th>To 00:12</th> <th>From 00:12</th> <th>To 00:16</th> <th>From 00:16</th> <th>To 00:20</th> <th>From 00:20</th> <th>To 00:24</th> </tr> <tr> <td>2.82 Res-128</td> <td>2.82 Res-96</td> <td>2.82 Res-128</td> <td>2.82 Res-96</td> <td>2.82 Res-128</td> <td>2.83 Res-128</td> <td>2.82 Res-128</td> <td>2.83 Res-128</td> <td>2.82 Res-96</td> <td>2.83 Res-96</td> <td>2.82 Res-128</td> <td>2.83 Res-128</td> </tr> <tr> <td>2.83 Res-128</td> <td>2.83 Res-96</td> <td>2.82Bake, 2.83Render Res-128</td> <td>2.82Bake, 2.83Render Res-96</td> <td>2.90 Res-128</td> <td>2.90 Res-96</td> <td>2.90 ByuoDens=4,Heat=5 Res-128</td> <td>2.90 ByuoDens=4,Heat=5 Res-96</td> <td>2.90 ByuoDens=4,Heat=5 Res-96</td> <td>2.90 Res-96</td> <td>2.90 ByuoDens=4,Heat=5 Res-128</td> <td>2.90 Res-128</td> </tr> ``` </table> [Comparisons.mp4](https://archive.blender.org/developer/F8709918/Comparisons.mp4) *(BTW, LATE BCZ I have been doing lots of Test Sims* 😃*) Last Update: Out of 10 Videos, I made 1, Where I have Put all those Side by Side in their Original 720p Video Resolution - Which Makes this Video (2560*1440 px)*

REYNEP commented

2020-07-20 21:07:04 +02:00

Changed status from 'Needs User Info' to: 'Needs Developer To Reproduce'

REYNEP commented

2020-07-21 15:20:33 +02:00

This is a Comparison of Same 2.90 QuickSmoke Settings (with Buoyancy Density=4.0 and Heat=5.0) in Different Resolutions

2-90Res100v98v96.mp4

**This is a Comparison of Same 2.90 QuickSmoke Settings *(with Buoyancy Density=4.0 and Heat=5.0)* in Different Resolutions** [2-90Res100v98v96.mp4](https://archive.blender.org/developer/F8711507/2-90Res100v98v96.mp4)

Sebastián Barschkis commented

2020-07-21 16:17:37 +02:00

Ok, nice comparison, thx! I can see that with increasing resolution the smoke seems to rise slower. That's definitely an issue that needs to be addressed (i.e. fix scaling of buoyancy parameters).

The rightmost example in the above video (res 96) seems a bit out of place though. Are you sure it was simulated correctly?

Ok, nice comparison, thx! I can see that with increasing resolution the smoke seems to rise slower. That's definitely an issue that needs to be addressed (i.e. fix scaling of buoyancy parameters). The rightmost example in the above video (res 96) seems a bit out of place though. Are you sure it was simulated correctly?

REYNEP commented

2020-07-21 16:34:33 +02:00

Are you sure it was simulated correctly?

Yes, I did This Multiple Times, Because at First I was confused too. I was Questioning myself Too, Seeing the Changes Yesterday Night, Then i tested Again Just an Hour ago....

> Are you sure it was simulated correctly? Yes, I did This Multiple Times, Because at First I was confused too. I was Questioning myself Too, Seeing the Changes Yesterday Night, Then i tested Again Just an Hour ago....

Sebastián Barschkis commented

2020-07-21 17:35:48 +02:00

Can you quickly check if your simulation gives you a similar result with res 96? I just made this with the daily build (731d9f0bfa ) and the 2.83 test file from this report. I think from the shape it looks more similar to the results you had with res 98 and res 100.

plume_res_96_290.mov

Can you quickly check if your simulation gives you a similar result with res 96? I just made this with the daily build (731d9f0bfa40 ) and the 2.83 test file from this report. I think from the shape it looks more similar to the results you had with res 98 and res 100. [plume_res_96_290.mov](https://archive.blender.org/developer/F8711689/plume_res_96_290.mov)

REYNEP commented

2020-07-21 17:44:28 +02:00

On it, @sebbas Sir....

Adding Video, Yes Sir With The Build that you Have Specified,It Works, Just as expected....
Blender version: 2.90.0 Alpha, branch: master, commit date: 2020-07-20 20:08, hash: 731d9f0bfa, type: Release
Release20JUL_2-90.mp4
But the Build I used from 4 Days Ago (Daily Build:- 0a40c671b0), Does Not Work (Update:- I think This File's Liquid Object Size is Smaller):-
Release16JUL_2-90.mp4

(Update:- 2nd Video's (16JulyUpdate) Liquid Object Size is Smaller)
16JulyDailyUpdate.blend 20JulyDailyUpdate.blend

On it, @sebbas Sir.... Adding Video, Yes Sir With The Build that you Have Specified,**It Works, Just as expected....** **Blender version: 2.90.0 Alpha, branch: master, commit date: 2020-07-20 20:08, hash: 731d9f0bfa40, type: Release** [Release20JUL_2-90.mp4](https://archive.blender.org/developer/F8711773/Release20JUL_2-90.mp4) But the Build I used from 4 Days Ago (Daily Build:- 0a40c671b0b5), Does Not Work ***(Update:- I think This File's Liquid Object Size is Smaller)***:- [Release16JUL_2-90.mp4](https://archive.blender.org/developer/F8711758/Release16JUL_2-90.mp4) ***(Update:- 2nd Video's (16JulyUpdate) Liquid Object Size is Smaller)*** [16JulyDailyUpdate.blend](https://archive.blender.org/developer/F8712055/16JulyDailyUpdate.blend) [20JulyDailyUpdate.blend](https://archive.blender.org/developer/F8712056/20JulyDailyUpdate.blend)

Sebastián Barschkis commented

2020-07-21 18:42:31 +02:00

Changed status from 'Needs Developer To Reproduce' to: 'Confirmed'

Sebastián Barschkis commented

2020-07-21 18:42:31 +02:00

Interesting, not sure which commit made that change. Anyways, to summarize the bug for now:
The shape of the smoke plume should not differ when changing the resolution (which it does as of now, 731d9f0bfa)

Interesting, not sure which commit made that change. Anyways, to summarize the bug for now: The shape of the smoke plume should not differ when changing the resolution (which it does as of now, 731d9f0bfa40)

REYNEP commented

2020-07-21 18:51:45 +02:00

And 1 More Thing @sebbas, Sir. Here is a 2.83.2 Comparison of RES 64, 96, 98, 100, 128, 256:-

2-83Res256vall.mp4

Also I don't think 2.83 is 100% Accurate, Because when I increased to 256, You can See "That GasFluid That Comes Out of the Center", Happens Too early Compared to 128Res or 96Res I Am Gonna check This in 2.90 Now.

testMantaflow2-83.blend

And 1 More Thing @sebbas, Sir. Here is a 2.83.2 Comparison of *RES 64, 96, 98, 100, 128, 256*:- [2-83Res256vall.mp4](https://archive.blender.org/developer/F8711828/2-83Res256vall.mp4) **Also I don't think 2.83 is 100% Accurate, Because when I increased to 256, You can See *"That GasFluid That Comes Out of the Center"*, Happens Too early Compared to 128Res or 96Res** *I Am Gonna check This in 2.90 Now.* [testMantaflow2-83.blend](https://archive.blender.org/developer/F8712037/testMantaflow2-83.blend)

REYNEP commented

2020-07-21 19:22:49 +02:00

I Think I have Got Why The 20July Build was Doing Great @sebbas .... And I Think It Was Fixed because of the Size of the Liquid Object

It is a 3 Minute Long Video, But I Think It Will be Worth it....
Why20JulyGotFixed.mp4

Note:- The File That I used for 2.83, The Size of the Liquid Object was same as The Liquid Object I used for 2.90....
First I opened the File That I used for 2-90 Buoyancy Dens=4,Heat=5, Res=128
Changed the Res to 96, Baked with 2-90 20July Update, This bake was WRONG
Then I Increased Scale of the Liquid Object a lil bit, Baked with 96 Res, This bake was Good
Then I tried Manually Changing the Scale Value of Liquid object 0.151, 0.152 Did not work, but 0.158 Worked! It Looked similar to 128 Res
But that same Scale value of 0.151 Looked OKAY with 128Res
UPDATE:- I Opened the 2.90 Buoyancy File in 2.83.2 and changed the Buoyancy Values to 1.0, Then baked with 96Res, It Looked more Like (but not FULLY same) Original 2-83 96Res Bake

testMantaflow2-90Buoyancy.blend

**I Think I have Got Why The 20July Build was Doing Great @sebbas .... And I Think It Was Fixed because of the *Size of the Liquid Object*** It is a 3 Minute Long Video, But I Think It Will be Worth it.... [Why20JulyGotFixed.mp4](https://archive.blender.org/developer/F8712023/Why20JulyGotFixed.mp4) 1. **Note:- The File That I used for 2.83, The Size of the Liquid Object was same as The Liquid Object I used for 2.90....** 2. First I opened the File That I used for **2-90 Buoyancy Dens=4,Heat=5, Res=128** 3. Changed the Res to 96, Baked with 2-90 20July Update, **This bake was WRONG** 4. Then I Increased Scale of the Liquid Object a lil bit, Baked with 96 Res, **This bake was Good** 5. Then I tried Manually Changing the **Scale Value** of Liquid object **0.151, 0.152** Did not work, but **0.158 Worked!** It Looked similar to 128 Res 6. But that **same Scale value of 0.151** Looked **OKAY** with **128Res** 7. UPDATE:- I Opened the *2.90 Buoyancy File* in **2.83.2** and changed the **Buoyancy Values to 1.0**, Then baked with **96Res**, It Looked *more Like (but not FULLY same)* **Original 2-83 96Res** Bake [testMantaflow2-90Buoyancy.blend](https://archive.blender.org/developer/F8712031/testMantaflow2-90Buoyancy.blend)

REYNEP changed title from ~~2.83 Mantaflow Smoke Simulation is not same as it was in 2.82~~ to 2.83 Mantaflow Smoke Simulation doesn't match 2.82 with Same Liquid Object Size

2020-07-21 21:25:36 +02:00

REYNEP commented

2020-07-21 21:25:36 +02:00

Changed status from 'Confirmed' to: 'Needs Developer To Reproduce'

REYNEP changed title from ~~2.83 Mantaflow Smoke Simulation doesn't match 2.82 with Same Liquid Object Size~~ to 2.83 Mantaflow Smoke Simulation doesn't match 2.82 with Same Object Size when it's smaller than a particular value

2020-07-21 21:46:57 +02:00

REYNEP commented

2020-07-28 20:29:44 +02:00

Sir @sebbas, Are you There? I think latest 2.90 Still has that Problem when the Liquid Object is Smaller than a particular Size

Richard Antalik commented

2020-08-04 06:54:47 +02:00

Added subscriber: @iss

Richard Antalik commented

2020-08-04 06:54:47 +02:00

Changed status from 'Needs Developer To Reproduce' to: 'Confirmed'

Richard Antalik commented

2020-08-04 06:54:47 +02:00

@REYNEP please don't change report status once it has been confirmed.

REYNEP commented

2020-08-04 21:26:49 +02:00

@iss Sir, I think There was a Misunderstanding about the problem.... I Explained it In the Last COMMENT With VIDEO.

@iss Sir, I think There was a Misunderstanding about the problem.... I Explained it In the Last **COMMENT With VIDEO**.

Richard Antalik commented

2020-08-05 02:56:51 +02:00

In any case once developer confirms the report, don't change status yourself, otherwise it can disappear from workboard and can be forgotten.

REYNEP commented

2020-08-05 03:00:11 +02:00

Copy that, Sir.... (I misunderstood what "Confirmed" Meant, My Bad 😓)

Jacques Lucke commented

2020-08-14 15:21:23 +02:00

Added subscriber: @JacquesLucke

Jacques Lucke commented

2020-08-14 15:21:23 +02:00

In #79029#982955, @sebbas wrote:
Interesting, not sure which commit made that change. Anyways, to summarize the bug for now:
The shape of the smoke plume should not differ when changing the resolution (which it does as of now, 731d9f0bfa)

I can confirm that in the 2.83 file provided in the initial post.

> In #79029#982955, @sebbas wrote: > Interesting, not sure which commit made that change. Anyways, to summarize the bug for now: > The shape of the smoke plume should not differ when changing the resolution (which it does as of now, 731d9f0bfa40) I can confirm that in the 2.83 file provided in the initial post.

REYNEP commented

2021-03-05 02:52:04 +01:00

Just a Reminder, what was happening was Happening because of changing The size of Flow object by a really small amount, I tried to Explain that in my last comment, the Video is quite Long, but only the first 1 minute explains it too

Dalai Felinto commented

2022-05-09 11:01:54 +02:00

Mantaflow (fluid simulator) has no active developer at the moment. Moving those bugs as known issues so at least other users can find their issues already reported.

Philipp Oeser removed the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

2.83 Mantaflow Smoke Simulation doesn't match 2.82 with Same Object Size when it's smaller than a particular value #79029