Detail loss in smoke

[ttdww]

Any tips on how to resolve the detail loss I'm getting? It seems wherever the smoke slows and starts to fade off, it's not responding to the turbulence helper very much until I really crank up the strength and then of course the whole sim is too noisy overall. I've tried every variation of massive-vort on/off, low/high vorticity, large scale from 0-1.0, and I just can seem to get this sim to where it needs to be. This is PFD v3.x.

What I'm going for is how the sparse smoke around the edges looks in this image: https://finalight.com/maindir/webpages/order/SM003LMS_HD.php (Hover with mouse to magnify, or can download HD still)

Trying to give this a go instead of using.. the other package ;D

Thanks!

[Svetlin.Nikolov]

Hey,

Can you help me with the Phoenix version and the specific settings that you used and with more info about anything specific happening in this exact setup?

During Phoenix 3 I did a few changes to how massive vorticity works, and I also changed PCG and the behavior of moving simulators a few times - this is why the exact Phoenix version from the About dialogue really matters. Also, is the simulator moving in the scene, because this might be causing some smudging of the smoke too?

Otherwise, the general advice is - use as few steps per frame as possible (ideally 1), use multi-pass advection, and use higher resolution. How many voxels do you have in this simulation?

Ah, and on Turbulence - gotta check if this is on the docs site and update it if it isn't - if your Turbulence helper is in Pressure mode, it will not take into account the density and will uniformly affect all voxels. However, if it's Not in Pressure mode, it will act more strongly on the dense voxels and more mildly on the ones where smoke is thinner.

Cheers!

[ttdww]

Hey Svetlin,

Thx for the speedy reply. Much appreciated. Your note on 'pressure mode' caught my attention. Will give that a whirl.

Here is the data on the sim:
Workstation: 192GB ram dual hex core, 3ds2019 perpetual holdout
PFD v3.12.00
Sim box settings - see attachment - however to answer a few questions...sim box not moving / PCG method / steps per frame: 1 / multi-pass adv. / fairly high res per the detailing in the rolling parts of the smoke ;]
If not included in settings..
PHXSource: emitting from simple plane 55x55m / mode: surf force / channels: smoke / emitting smoke and velocity from texmap: round gradient with noise texture
PHXTurb: strength: .25 / size: 50 / depth 2 / last stage/scale: default / rate of change: .15 / create pressure: unchecked
PlainForce: strength: 2 (pointed to left)

This is frame 535 after the Resim pass.

Thanks!

Best,
Brian

[Svetlin.Nikolov]

Oh, is this a resimulation?

So 3.12 is missing just a change to massive vorticity that went in 4.0, so this should be okay.

The preset file does not contain the grid resolution though, can you share the number of voxels?

500 frames are deep in the timeline, so indeed some dissipation is expected at that point, but if this is a resim, there might be ways around it...

[ttdww]

Yep, it's a resim, however in the default sim, though less detail of course, it also shows the same issue - the smoothing and lack of detail once the smoke slows out of it's rolling. As with other simulators, I've noticed the 1st pass sim is usually a sort of foreshadowing of what can expect in the resim but with additional detail - unless there are some tricks you can suggest - like changing the params for the resim.

I've tried pressure mode in the turbulence and it didn't resolve the issue. The 1st pass sim was ~240million voxels about 710m x 560m x 610m tall, .6m spacing. The resim pass was 1.0.

[Svetlin.Nikolov]

A note just in case - the Turbulence, nor any other force, will work during resim - you have to include it in the base sim. While Sources and obstacles work during resimulation, none of the factors that change the fluid's velocity are enabled, so you don't get a completely different simulation from the base sim.

The grid resolution looks great. The vorticity is quite low at 0.12...

[ttdww]

Roger that regarding simming 1st and 2nd pass with the Turb helper, and howthe sources and obstacles function - makes sense. Glad the grid res is good. I have massive vorticity checked, so the classic vort of .12 is greyed out. This sim is using the .5 smoke surface vorticity.

Well, I haven't been able to get that detailed noisy burn off. I'm wondering if the only way to get it in PFD is with a tuner, i.e. if smoke density < x value, increase turbulence. Thoughts?

[Svetlin.Nikolov]

Ahh, sorry, indeed your smoke surface vorticity is at 0.5 which should be good enough.

So as far as guessing goes, all I can say is that the massive vorticity and turbulence Should work, but if they don't, gotta do some experiments. Yes, a tuner could do it, but I think you should be able to do without it as well.

Does changing the turbulence to pressure mode have absolutely no effect? Could it be that its stength is too low?

[ttdww]

Copy that -pressure mode doesn't resolve it unfortunately. Have RnD'd quite a bit more today and am stumped on how to get detail in those areas. Here's another screenshot - just a sliced sim - and not the best version, this is Buffered at 50% quality I believe, however the results is always the same shortcoming. This sim has 2 turbs active covering pretty much from super large noise all the way down to tiny as a just in case. (i set this up with 2 instead of 1 while I was testing so just left them both on). Once I crank up the strength on the turbs enough to get detail in that sparse area, the whole sim is destroyed.

I've also tried all Conservation and Advection methods. Is there something in the code that is deliberately trying to keep sparse slower moving smoke from having barely any Turb applied? If so, I could see the logic behind that, however it's keeping the realism from being achieved. Note that this sim is smoke-only with no temperature nor fuel. Let me know if that would have something to do with it. Thanks!



Here's the example of what I'm trying to achieve. Around the edges where the wispy smoke is breaking up and thinning out, it's affected by the turbulence enough to where if feels natural. Actually the breakup looks better than the sim itself. I realize this is somewhat comparing apples/oranges since the sims are quite a bit different ;]

[Svetlin.Nikolov]

Negative, there isn't such logic in the sim.

Would it be possible to send the setup over so we can see what's up?

[ttdww]

I see - thx for that info. Sure, what email can I send it to?

[Joelaff]

Not sure about your loss of smoke you have been discussing, but the effect you show in the image of a wispy layer outside the main sim can be achieved by first simming your main sim. Then create a New simulator (duplicate existing perhaps) with a new Source. Set that source to the existing base Simulator as the source. Adjust the outgoing velocity to taste (I found low looks nice and drags behind). Configure this new sim to include only the objects you want (include/exclude). Add a turbulence or two at different scales. Adjust Smoke Buoyancy and add a little Smoke Dissipation. Now greatly reduce the opacity of this new simulator for rendering (I used Simple Smoke at about 0.05 I think).

Tweak to taste of course

You end up with something like this. This is super rough, just a teapot default "explosion" with the extra simulator added to it.


Above is BEFORE



Above is "After" with the extra simulator.

These are crappy quality, but you get the idea...

EDIT: Added the test scene here if anyone cares... Simulate the PhoenixFDFire001 first. Then PhoenixFire002 second.
TestSmoke.zip

[ttdww]

Hey Joe, Thanks very much for that workaround solution. It's not off the table - to go with your 2 pass method. I've been thinking about tuners however would be great if I could get this working with v3.x. Thx again!

[Joelaff]

The nice thing about the two simulator method is that you can get the base sim all set and approved and then tweak the second wispy layer to taste afterwards.

This way you avoid making tweaks for the wispy parts that manage to change the macro simulation which usually leads to the client saying they preferred the version from a week ago over the new one, but the wispy stuff looks good.

[ttdww]

Great point. I'll admit I've been fighting what was already a great looking pyroclastic roll to try to bring in that nice breakup of fading smoke, plus having separate control, like you mention, is a good advantage. I guess where the heavy smoke is fading, I can clamp off the sparse areas before they show up. Will see how it turns out.