pyro banding issue

[asnowcappedromance]

Hi guys,

 

I thought maybe one of you can help me out here. I'm simulating some mid-scale fire and am getting these banding issues, both in the viewport and in the render (picture and .mp4 attached.)

It looks like the fire is being cut into thin slices, which are static and DON'T move with the flames. Unfortunately I can't post the scene file, but I'll try to describe my setup as best as I can.

 

First of all, source is static. Nothing to worry about. No crazy velocities go into the sim, in fact, this problem persists even if I set my fluid_source velocities to None.

Bounding box scale is rather big, 166, 105, 210, voxel grid is [333, 210, 421] which is already quite a bit of resolution IMO.

I'm using the pyro solver, max substeps set to 2 (I thought this would help get rid of this problem, but it didn't) and under Projection my Multigrid iterations are set to 3.

Shape settings don't seem to be important regarding my banding problem, if I switch them all off, I still get the same artefacts.

Advection type is Modiefied MacCormack, Advection Method is Trace.

Buoyancy Lift is quite high, 120, but as you can see in the video the banding appears to be static, it's not moving with the fluid, so I don't think this is the problem.

 

Any ideas? This is driving me nuts!

 

Thanks,

 

Manu

pyro_flame_banding.mp4.rar

[Guest tar]

You can set the Sub Steps on the DOP Network to 5 - see screen grabs. it's a standard sphere one with set to 1, the other to 5. 

[asnowcappedromance]

You can set the Sub Steps on the DOP Network to 5 - see screen grabs. it's a standard sphere one with set to 1, the other to 5. 

 

Hey Marty thanks for the reply! I was hoping to avoid crazy substepping due to the large scale of my simulation.

[Ankit Sinha]

well you can try adding noise(changing with time) to your source in Y-axis and if you already have noise, then try increasing rate of change of noise and also increasing amplitude of noise. Basically, play with noise of source it will solve you issue.

[asnowcappedromance]

well you can try adding noise(changing with time) to your source in Y-axis and if you already have noise, then try increasing rate of change of noise and also increasing amplitude of noise. Basically, play with noise of source it will solve you issue.

 

Hey man, a valid suggestion, I'm doing that at the moment and it's improving my results!

Cheers!

[mitul]

Check your flame height under Combustion > Flame > Flame Height (Default is 3), if you lower it too much it tends to cause banding.

 

Hope it helps.

[ChristopherC]

Sorry for reviving this thread but I'm having a similar problem and adding noise only makes it worse in my case.

 

 

I've attached a scene with a default pyro setup to illustrate the issue. Using the `Fog VDB` instead of generating the VDB from an attribute works slightly better but I'd like to use the attribute method. Having a voxel size in the VDB equal to the division size in the pyro object helps as well but the banding remains visible.

 

Increasing the substeps fixes the issue but my laptop won't like it—is this the only solution?

 

Thanks,

Christopher.

 

pyro_banding.hipnc

Edited August 17, 2015 by ChristopherC

[anim]

instead of using Dopnet substeps, try Pyro substeps as those are adaptive so you may still get acceptable speed

[ChristopherC]

The adaptive substeps seem indeed faster! Computing the 48 frames of the previous test scene takes 39 s with 2 DOP substeps versus 29 s with a maximum of 2 adaptive substeps. With no substep the time goes down to 19 s.

 

The documentation says that “Normally the pyro solver works best with no substeps”, so I was a bit unsure with using substeps on top of the speed issue. But I guess there is no way around it. Adaptive substeps, here I come!

 

Thanks Tomas!