wave length

[Svetlin.Nikolov]

Oh, I'm seeing the blocky grid, some smaller holes in the mesh that pop up occasionally, and also some black longer artifacts that might be happening near the simulator border or the image border of the simulator does not stitch well with the infinite ocean. Looks more like a rendering issue with the Ocean Mesh mode. Does it show up if you render in Cap Mesh, or Isosurface which should be veeery smooth? If you enable the particle preview of the liquid and disable the mesh preview, does the Liquid particle mass also have such a blocky shape?

[olika]

Hi Svetlin.Nikolov
I know what you mean, the smaller holes comes from mesh smoothing, i think they would not even been visible when rendered.
The big black ones in the back is form ocean mesh mode and i would also say it's disappears while rendering. All good, but i'm really focusing on that few frames barely before the wave breaks and stasrts to develop the splashes.
Because i think before that point all seems smooth already, i would not raise spf at all. But right when the splashes start to show up and the wave breaks then the most ahead mesh starts to jump somehow. Sorry, don't know how to find better words for it

I'm running a sim with 4 spf and droplet surfing completly turned off. let' see whats happening .. but 4 spf's could take some time now

[Svetlin.Nikolov]

Yess, I'm still gravitating towards a render-time solution since it would be quicker... I think the particle preview could really tell if it's a simulation thing though...

[olika]

Cool! I'll bring one up! I just thought maybe it's a easy one, like yeah for sure it's the spf or something hehe

[Joelaff]

Could it possibly be the storage quality of your grid? I have done some larger sims (large grid) and with lower storage qualities values get quantized and kind of jump.

is the storage quality actually the number of bits used to represent the values? It seems to act like that.

[olika]

morning Joelaff , oh thanks for pointing out to this! I actually never thought about that one. It is quite a large grid i'd say hmmm...
That would be a classic, would'nt it! Simming for days and then crunching it while saving

[Joelaff]

Worth a try. I have done just that. I had lowered the quality a little bit, plus the camera got real close to it. Specifically, the camera was inside a bunch of bubbles.

If that storage quality really is the number of bits (anyone confirm or deny?) then the default of 14 would give you 16,384 positions in each dimension (assuming it’s stored as an integer). This would only be a big deal with a bunch of cells, especially in one or two dimensions only (because three would likely be so huge you couldn’t/ wouldn’t simulate it at that resolution). Also, having the camera close could exacerbate things.

Apparently the storage quality also applies to particles, which is where I was really noticing it.

I could be totally wrong here, and this certainly may not be your issue. Assuming I am right each increase by one of the storage quality will double the number of positions, and increase space on disk and load time (Roughly) linearly.

[Svetlin.Nikolov]

Hehe, indeed, it IS the number of bits for most of the particle and grid channels - for some of them the meaning is different. And also yes - this means that particles could form lines when the quality is low: https://docs.chaosgroup.com/display/...ssionartifacts

[Joelaff]

Thanks for confirming this. In some tests I believe I saw results that indicated that the data was quantized at each frame and the next frame simulated would use the quantized days as the starting point.

I think I know why it is done this way (because you could never resume accurately). Part of me would like to simulate at the higher accuracy and then just store the quantized data art the lower bit depth.

This would be like doing all your color corrections in 16bit and then saving your output image as 8bit. Better results than converting to 8bit after each color adjustment.

Willie you certainly would get different results if you stopped and resumed would this not reduce the particles forming lines in the sim? Sorry, just thinking out loud here.

[Svetlin.Nikolov]

Oh, the simulation does not read back the compressed caches. The state of the solver remains in RAM with no quality changes. Only if you choose to Restore a simulation from cache after you've stopped it or even closed the host app, will the solver be initialized with the lower quality data. But also - the default compression quality is chosen as a good balance between size and quality, and we very rarely see scenes where 14 is bad (to put it more specifically - less than once a year).

[Svetlin.Nikolov]

Ah, the solver Does also read back caches for Resimulation as well. Could it be that the setup you saw the interesting results with used that?

[Joelaff]

Oh, the simulation does not read back the compressed caches. The state of the solver remains in RAM with no quality changes. Only if you choose to Restore a simulation from cache after you've stopped it or even closed the host app, will the solver be initialized with the lower quality data. But also - the default compression quality is chosen as a good balance between size and quality, and we very rarely see scenes where 14 is bad (to put it more specifically - less than once a year).

I guess I was simply mistaken in my assessment.

I wouldn't say I have ever seen 14 look bad. I would say I have seen higher numbers like 16 or 18 look a tiny bit better. It is all very scene dependent of course.

[olika]

Oh wow! ahuumm i just turned storage quality to 20... now to see it's tenfold in size! hui! With 14 it was around 300mb, with 20 is about 3-4gb, so be aware of that! i mean per frame

[Joelaff]

Yeah, they can get pretty big.

Did it help with the issue? Or haven’t gotten it simulated yet?

[olika]

Joelaff i wasn't able to check if there's a difference because of storage quality to be honest. I found a few other minor quirks so i'm running a new sim now with quality set to 16.
But the more i test, the more prominet it is, that in that special setup here, droplet surfing may cause the strange mesh movement. As i say it's a special scene where i push things to extremes. So maybe it's because normally the surfing is a vertikal thing in terms of splash is coming from above. And here when the wave staerts to crash it becomes more of a horizontal move because of the speec etc.. Don't know if that makes sense, just loud thinking