Millions of voxels issue

[vanja_kapetanovic]

Hi I am trying to learb Phoenix FD and so far I have little luck. No matter how many tutorials that I watch I have hard times predicting the outcome of the simulation but biggest issue is that expendable grid. I have a scene in which I have an airplane falling and engine is on fire. I tried to manually simulate fuel explosion by assigning polygone ID of the model to the source and than trigger it from theree. I have 80 frame clip but by the time I get to frame 31 grid is like 40 million voxels already and simulation is barely moving. I mean idea is good but if I need a render farm for a 3 second simulations with this program while using this powerful computer...not sure it's a match. Not sure should I just leave it for days to siimulate these 4 seconds and get to like 10 trilion voxels or how do you guys render shots longer than 3 seconds ? am I doing something wrong? I have 64 GIG of ram, Nvidia RTX2080 and 9th generation Ryzen processor.Thank you
P.S> I got 70 milliion voxels for 50 frames. So this is pretty much suppose to be sent to render farms. Idk what kind of computers are you guys running but this is nuts. 70 mil for les than 3 sec. Ufff

[Joelaff]

You could animate the phoenix container to move along with the falling engine, thus keeping the size smaller.

You could use Confine Geometry to limit the processing to a certain volume shape (won't reduce memory usage, though):
https://docs.chaosgroup.com/display/...FireSmoke+Grid

[Svetlin.Nikolov]

Hey, Joelaff 's advice is spot on! Whenever it might help you, don't be afraid to rotate or incline the grid, or to attach it to a moving emitter or obstacle - any of these will reduce the need for huge empty grids

[vanja_kapetanovic]

I did animate the grid to follow the engine and I did make it expandable and it just keeps getting biigger every frame half a million. I start at like 20k super low res but it just adds up in like 300 frames I am already at 50 mil. I dont understand how does this work. I watched like 100 turorials and everything iis like trial and error. I set up the same units I do everything like in a tutorial and don't even get similar results. I really dont get how is this so hard. Anyway for noow I would be happy with just managing too get anything simulated but it just can't do anyything that has a smoke trial. Everyy inch of that smoke is like million voxels. So its really limited you have to follow the object to hide the fact that that it's so taxing but what If you want smoke trail or god forbid a voolcano. I can only imagine making a volcano in this thing. I would need 3000 GB ram

[Joelaff]

Be sure you have it configured to be able to shrink (turn off Expand and Don't Shrink).

Perhaps check the maximum expansion.

Check the Channel (probably want smoke) and threshold you are using for Adaptive Grid. If the threshold is too low it may have trouble shrinking (too high, and it will shrink when you don' want it to).

You could try some Smoke Dissipation as well. This might help clear it a little faster to keep less smoke active at once.

Also, perhaps you just have too many cells in your simulation. You could try reducing the resolution, at least during initial setup. This is very scene dependent, and depends on how close the camera is, etc.

If your engine is leaving a crazy long trail it may be hard to accommodate that. Are you using a particle emitter as the fluid source? That might give you some more control.

If none of this helps perhaps you will need to cascade simulators:

https://docs.chaosgroup.com/display/...een+Simulators

[vanja_kapetanovic]

ok I see what you guys are saying. Thank you very much for your inout. I will go and see what can I do with animating the grid. I am just frustrated seeing amazing results of other people but when I seem to follow and set it all the same results vary drastically.

[Joelaff]

When I am trying to figure this kind of thing out I find that having a separate simple scene in another copy of Max open can help me play with different settings and techniques to try to figure out what they do.

Simulations aren't simple. I tell my clients it's like dropping a deck of cards on the floor over and over until it falls just the way you like it, and then listening to the client say they want it a little different, and then trying to do that by dropping the cards again and again

[squintnic]

When I am trying to figure this kind of thing out I find that having a separate simple scene in another copy of Max open can help me play with different settings and techniques to try to figure out what they do.

Simulations aren't simple. I tell my clients it's like dropping a deck of cards on the floor over and over until it falls just the way you like it, and then listening to the client say they want it a little different, and then trying to do that by dropping the cards again and again

agree, droopping cards on the floor is a perfect analogy and reason why my test sims often end up as finals

the biggest thing for me was setting up cameras and basic lighting and scene geo first before doing a sim, before i started being more strict i was simming 500gb explosions that just didnt need that res or sim length. i do a lot of versioning and note taking (in a table) to try and keep track of what varibales and settings im messing with as they often have interdependencies that can cause unexpected results. smoke dissipation also useful to stop the ugly staight edges in the sim box

[Joelaff]

Yeah, definitely good to get the DOF and motion blur all set early too. This way you know if the extra detail is needed or not. Sometimes I think I need a lot of detail, but it is so blurred away in the final.

I even like to get a comp setup pretty early too because sometimes it just needs some more noise/grain to give it that extra perceived detail.

Sometimes you can even get away with a 1/2 res render and scale up in post. Sounds crazy, but you’d be surprised how well it works sometimes.

[squintnic]

yep agree - if im matching to a backplate u can def get away with low res by the time you blur and grain and artifact it to sit in nicely!
fusion comps from the first test render for me. gotta get it towards a content complete piece as soon as possible

[Svetlin.Nikolov]

ok I see what you guys are saying. Thank you very much for your inout. I will go and see what can I do with animating the grid. I am just frustrated seeing amazing results of other people but when I seem to follow and set it all the same results vary drastically.

Hey, still it would help showing your simulator settings and your viewport - there might be something specific that causes the grid to expand so much that we can spot.

[Svetlin.Nikolov]

Ah btw, which tutorial are you watching? If it's one of ours maybe it needs to be updated...

[vanja_kapetanovic]

Hi guys thank you for your responses. Joelaff Hahahah dropping cars is just perfect example. Svetlin.Nikolov The thing is I watch Redefine FX tutorials https://www.youtube.com/channel/UCH3...0Ksmi0dcEGEFRw He is like a doctor of Phoenix FD based on the ease with which he does all of his stuff. So than I follow exact same thing I literally spend 2 hours following his 10 min tutorial set the Units and everthing, every single little setting, yet my stuff looks noooothing like what he gets in his view port. So I guess he just dropped his deck of cards differently than I did hahahah. But seriously it's quite frustrating as tutorials made me believe I should get same results with same settings. I pay a lot of attention on scale as that is what matters a lot and than I improvise from there and follow tutorials. I also watched some Chaos Group official ones and all kinds on youtube.

[Svetlin.Nikolov]

Ah yes, Jesse is doing great videos but maybe he used some settings that changes over the versions, or he forgot to include something important. Which exact video did you watch? Maybe we can figure out what might be missing. Usually it's the scene scale and units setup which is crucial but often gets forgotten in videos.

And again, if you can post any settings and any shots of how your scene is laid out, it would greatly help us in giving you helpful advice - otherwise it would be too vague and general...

Joelaff , squintnic , I might be taking up the discussion from too far away, but when I was starting in Chaos 8 years ago and was trying to setup scenes with Phoenix, I had the same feeling, but over time we built up a toolbox that should help domesticate wild fluids Using Body Force or Path Follow to guide the fluid, force viewport preview to help figure out what forces will do to the sim before simulating, Motion Velocity of obstacles and Sources to help direct the emission, using the Mapper to affect the fluid with textures in different places in space, and the new Voxel and Particle tuners that allow you to cheat in many many new ways - all these help me and my colleagues get from an idea to a working setup with a lot less trial and error, and I hope they are as useful to everyone else. So it's crucial to know where does our setup fall short - what are the specific setups where these tools don't help, or whether it's more a matter of not knowing about them or how to apply them?

Cheers all!

[vanja_kapetanovic]

First of all I didn't know you are a part of the team that invented Phoenix. My bad I am new to it. I just want to tell you it's absolutely revolutionary and made for artists. Stuff you creat in Phoenix in 3 hours would take a month in Houdini so thank you for that. Secondly part of my frustration IS the fact that I see it being used in the best way possible but I am struggling to get that photorealistic look. I think the secret is also at in understanding that fire/smoke shader that I don't really get. My smoke is always set to 1 and that's the thicknes value otherwise looks think and I feel like it's overdone but I just can't get it right to save my life. Now closes I've come is this render I am working on currently. It's an airplane falling of the sky with engine on fire. Due to the moving airplane and moving camera I had to end up having voxel heavy scene because it keeps expanding as airplane keeps falling. No biggei. It should last about 300 frames. So what I think I can do is simulate with really high QUALITY setting of like 100 and plus. And then RESIMULATE to add detail. Now that I am doing that I am ggetting somewhere. Also I am sorry but I keep changing my settings every minut to try to see if it betters the outome so I can't really share any real settings as they keep changing.
One thing I noticed is that I feel like my PhoenixFD smoke doesn't rect to VRay Volumetrics and fog in the scene. Or I might be wrong and not see correctly. Phoenix is really responosive to anything in the scene soo I am sure there would be no reason to be included in VrayVolumetric environment.