Understanding Maximal step and SPF upper/lower limits

[GrahamMac]

Using Phoenix FD 2.0

I would like to ask for confirmation if my understanding of the Maximal step and SPF upper and lower limit parameters is correct:

* For simplicity lets imagine we have a grid cell size of 1cm
* We set maximal step to 2 (meaning up to 2 cells can be traversed by fluid per simulation step)
* Based on gravity, discharge rate and other various parameters Phoenix FD calculates what the nominal fluid speed should be (let’s say around 10cm per frame)
* Phoenix FD would then calculate the SPF (steps per frame) to be 5 (10cm per f / 2cells(cm) per step)
* Therefore the OTS (optimal time step) = 1/SPF = 1/5 = 0.2 (assuming the OTS is simply the time taken for one simulation cycle)


* If the fluid becomes faster than expected, say 100cm per frame, we would need 50 simulation steps to describe this movement which may take too long to simulate.
* Suppose we set the SPF upper limit to 10. During simulation Phoenix FD would see that this fast fluid requires more than our original SPF of 5 and would use up to SPF 10 but this still only allows for up to 20cm per frame. The only option is to automatically increase the Maximal step to at least 10 cells.
* With now 10 cells per simulation step at an allowed 10 steps per frame we can achieve 100cm per frame.


* If the fluid becomes very slow, say 1cm per frame then Phoenix FD will recognise that our original SPF 5 is too much to describe this motion (a wasteful number of steps to calculate).
* Trouble is that even an SPF of 1 is still double the 1cm per frame we need.
* According to the help files if the OTS is limited to the frame duration (i.e. SPF of 1) we could get problems with undesired diffusion so clearly we don’t want to let the SPF go that low.
* Adding an SPF lower limit of say 3 would seem reasonable.
* This would mean that Phoenix FD would use a minimum of 3 steps per frame for this slow fluid and then automatically reduce the Maximal step to around 0.33 cells.
* With now 0.33 cells per simulation step at an allowed 3 steps per frame we get around 1cm motion per frame.


Thanks for any and all help to confirm this!

[Paul Oblomov]

Let's think simplier:
I think phx can handle this calculations quite good by itself. Lower 1, upper 20, quality 80 - almost everything simulates, as if it was fixed 20/20. Max step - when using adaptive grid, make sure you have enough ram, and start with relatively large box, this will give you a chance to see proper fire/smoke propagation at start, which is vital IMHO. The bigger max step - the larger adaptive grid would be.

Oh - I'm not that technical freak So, I suppose Ivaylo will answer you question

[Ivaylo Katev]

The only option is to automatically increase the Maximal step to at least 10 cells.

in general your description of the process is right, this is the only point where some inexactness can be found. actually the SPF limit does not change the *parameter* maximal step, it affects directly the time step that of course makes the fluid to travel more than the maximal step. so, it leads to the same consequences like you described, but the exact mechanism is this.
there are a big variety of reasons to use the SFP limits. in the very first phoenix version the reasons were only two - the performance for the upper one, and the scene update for the lower one (not the diffusion, it is bigger when the SPF is high). however, extending the simulation abilities to the liquids, we got one more reason to use both the limits - the simulations of settled liquids. actually there are even more reasons to use the limits in some special situations.

[GrahamMac]

Thank you Paul for your advice. I’m actually using very similar settings to your suggestion (see attached image and associated animation here: https://vimeo.com/43664511)

Ivaylo, thank you for confirming my understanding!

The reason I would like a better understanding of these parameters is because my main goal is to simulate slow motion liquid splashes, both from pouring liquid and dropping objects into liquid at normal and macro scales

normal scale examples:
http://failedmessiah.typepad.com/.a/...dc49970b-250wi
http://image.yaymicro.com/rz_512x512...ass-b9624e.jpg


macro scales examples:
http://techflaps.com/wp-content/uploads/2009/10/5.jpg
http://en.wikipedia.org/wiki/File:Wa...lashes_001.jpg

In contrast here is an example of the kind of messy splashes which I am not aiming for and tends to be fairly easy to achieve in Phoenix FD:
http://rookery9.aviary.com.s3.amazon...5d_625x625.jpg


I have had some success in creating the normal scale type splashes however in the animation (Vimeo link at top) you can see that the sheet of liquid that I want to elegantly rise up out of the glass unnaturally breaks apart near the end. The conservation was at Quality 100 for this, so perhaps it needs to be higher.

All thoughts welcome!

[Ivaylo Katev]

i think the maximal step is to blame, try with value of 2 or even 1

[GrahamMac]

Thanks Ivaylo,
Seems to be working well with maximal step of 1.