Huge object and fluid scale

[cobrax]

Hello guys,

 

I have an object that is 50 000 units wide. I am using H15 and I was thinking of dividing its size by 1000, so as to get pyro to solve properly but I thought I read that this only was an issue in previous houdini versions to 13.

 

What are your takes on the subject?

 

Thanks a lot

Xavier.

[Farmfield]

1 unit is a meter, so you got a 50 kilometer object, to big to even work in a yo-mama-so-fat joke.

[cobrax]

Hey thanks for your reply Farmfield

 

Who does yo-mama-so-fat-jokes anymore

 

The object is actually 5000 units (I mistyped), so 5km in diameter, but the biggest explosion on it will probably be 1km.

 

From tests I ran at 1km, 100m, 10m, 1m scales it seems the fluid solver is doing its job properly so far by just hitting the shelf tool.

 

Are you able to explain why might go wrong if I sim at the real scale?

 

Thank you very much

Xavier

[Farmfield]

Yo mama so fat you have to change the unit size in Houdini to kilometers simulate her correctly!!

 

Anyhow, I was in no way proposing you shouldn't use real scale as you have to for the physics to be at correct to scale, but I was more curious about your explosion, size-wise, because you need a lot of bang to get an explosion of the size you'd need a kilometre sized grid. Personally I love Nukemap as that is easy to understand the size of a blast in the bigger ranges - and surprisingly, something like a 100-150 Kton detonation will "only" create a fireball of 4-500 meters...

 

Oh, and about yo mama jokes, as a Swede, I never made/heard them in the first place, until I saw them in movies - it just came to mind when we were talking about huge scales.

Edited January 22, 2016 by Farmfield

[cobrax]

Ahah you did do the joke in the end

 

Nukemap is scary, I had nevery heard of it, but it is very informative, thanks for that

 

More or less my explosions needs to be between 1km and 250m I would say. But at that scale it might need a fair amount of frame to develop.
I need to get the timing, detail, evolution right to get the sense of scale. I don't how I am going to tackle it yet, still pondering.

[Farmfield]

Mmm, I would probably not sim at the correct scale in this case but sim at a lower size and then play with the timescale to get it looking bigger. The reason is the Pyro shelf presets are dialed in for "normal size" explosion, I'm guessing 5-25 meter fireballs, and that's why you get a lot of threads about doing nuclear bombs with Pyro - you basically have to set it up very specific to get pyro behave realistically at that scale. So even if you're doing a huge explosion, it'll be easier to dial in the look you want with a smaller sim, then use timescale to slow it down, making it look bigger...

 

In the end it's all about the look - the look you (or the client) want and/or the look the audience will buy into - and how to get there the easiest way.

Edited January 24, 2016 by Farmfield

[cobrax]

Hey Farmield,

 

So in the end I have simmed my explosion on my 5km wide object.

The size of the explosion is like 300 meters accross and it actually behaved well.

The only thing was to find the sweet spot for a few parameters, as you would normally for any sim, but it was in high numbers because of the scale.

 

Other than that all went well

[Netvudu]

Yes,one of the main reasons not to sim at proper scale are usually the units at which we have to work. Numbers tend to get quickly outside the comfort zone.

Still in the sim realm, simulations that involve particles which must show some kind of continuity, for instance water sims which are going to be meshed afterwards are much better simmed at lower scales (lower scales, though not very small or tiny scales where feedback tends to be too big and sims get unstable) so that the number of particles needed for the meshing process to be continuous is kept reasonable. Not usually a problem for pyro sims unless you are advecting particles. 

 

Another potential problem depending on the specific scene might be lighting as falloff and other similar settings depend on meters, and bigger scales might be harder to handle.

 

Finally, it is my experience that many solvers are not that much tested at extreme scales (too small or too big) which sometimes tends to end up in unwanted interactions. Not saying that happens in Houdini but it´s quite common in general.

Edited January 28, 2016 by Netvudu

[loopyllama]

I do not think you should sim at the "real" scale. The numbers work well for the things the size of the default torus. Working with huge sizes will require a balancing of possibly every parameter. You most likely run into floating point error (loss of significant digits). If you did manage to set everything correctly at a ridiculous scale, you have spent hours for what benefit? You'll still end up tuning the motion, the timescale, blending, cheating, etc just like you would at a smaller scale. If you start at a manageable res you have only to tune the motion of things to make them look big.

 

If you do decide to work at a huge resolution, please update us regularly. It would be interesting to know what happens!

[Farmfield]

Hey Farmield,

 

So in the end I have simmed my explosion on my 5km wide object.

The size of the explosion is like 300 meters accross and it actually behaved well.

The only thing was to find the sweet spot for a few parameters, as you would normally for any sim, but it was in high numbers because of the scale.

 

Other than that all went well

 

Well, if it works, it works, and that is the only thing that is important. As I said, the main issue is likely in tweaking an explosion at that scale, so if you got one look but want different behavior, I think it'll be hard to dial that in as the numbers need to change pretty substantially to have an effect at such scales. 

 

But again, if it works...