PhoenixFD Smoke simulation issue

[CherryPicker]

Hi all. I'm hoping someone can help tell me why a phoenix smoke simulation changes so abruptly between 2 frames.

The change happens between frames 347 and 348 of the simulation (see attached screen grabs) which has run as expected up until this point (overall shot frames range is 55-428 ), but for some reason at this frame the smoke seems to stop being emitted correctly and a lot of the existing smoke seems to get culled too.

So far as I can tell the machine isn't running out of memory, as the size of the grid at this point is less that it has been earlier in the shot. The smoke is being emitted from a tycache object which seems to be behaving correctly, and it isn't getting clipped by other geo in the scene.

So far as I can tell there's no reason why this should happen. I'm stumped.

I've uploaded the max file and tycache etc to my gdrive here https://drive.google.com/drive/folde...usp=drive_link

This is a personal project so I can share any further data if needs be.

If anyone's able to take a look and enlighten me I would be eternally grateful.

[zdravko_pavlov1]

Hello,

Thank you for uploading the scene. I'll inspect it, and I'll get back to you.

Cheers.

[CherryPicker]

Thanks Zdravko for taking a look.

[zdravko_pavlov1]

Hello, I started the sim yesterday, and it quickly grew to 350 million voxels. I left it to simulate overnight and inspected the result this morning. It had calculated 180 frames, and there were no issues at that point. I lowered the resolution slightly and started a new sim. I'll wait for it to reach the 9-second mark, where the problems appeared in the preview you sent (at about frame 270), and If it reproduces, I'll dig deeper and investigate the problem. So far, all I can tell is the scene is very tidy and well-organized. Thank you for taking the time to do that.

[CherryPicker]

Morning. Thanks for the update (and the compliment on the scene). Yeah it's really frustrating that the issue only happens towards the end of a loooooooong sim. The latest restore point in the sim is frame 345, so only 3 frames before the issue happens for me. Is it worth me uploading that aur file? Would that allow you to restore from that point too? Although I guess it's helpful to have you do a fresh impartial simulation as well to see if the issue is reproduceable at all.

[zdravko_pavlov1]

Indeed, I can start it from the middle of the timeline, but the grid will be smaller, and I want to make sure that I have the right conditions for the issue to appear. A backup frame, 10-20 frames before the problem part, could be helpful.

[CherryPicker]

Understood thanks. I'm uploading some frames here https://drive.google.com/drive/folde...usp=drive_link

They're understandably large files, so will take a while to upload unfortunately.

Definitely appreciate you taking the time to look at this.

[zdravko_pavlov1]

Hello,

As an update, I can confirm that the grid shrinks unexpectedly at frame 349.
https://gyazo.com/b028175df2ff037794fcf54a02987b2a

Still investigating what's the reason for that. There are a few other things that appear strange to me.
Although there is just one source (actually, there are two, but the one doesn't have any emitters attached) that emits a smoke value of 1, there's smoke 6.63 cached in the files.
Also the cache content window lists numerous particle systems written in the cache files, which seems odd. I'll keep digging and hopefully find the reason for this.

[CherryPicker]

Evening. Thanks for the update. It's good that you're able to reproduce the error - I hate it when technology performs inconsistently.

Yes, noted on the second emitter. I was hoping to emit smoke from particle IDs 2 and 3, but so far as I could tell, it's not possible to do that with a single source - you can only choose one polygon ID to use - so I was testing to see if I could use 2 sources for the one simulation, but quickly realised that isn't possible either. I figured if it bothered me enough I would have to consolidate the IDs in tyflow and recreate the tycache.

I don't know enough about Phoenix to comment on the smoke 6.63, but I assume that the list of particles is each of the tycache elements that's being read?

I guess at this point it's trying to identify whether the problem is coming from the tycache, or a bug with Pheonix.

[zdravko_pavlov1]

Hi,

I think I have a pretty good idea of what's causing the error.

I recorded a video. Please excuse my English

https://drive.google.com/file/d/1jJ8...ew?usp=sharing

[CherryPicker]

Amazing thank you so much for taking the time to put together such a comprehensive answer. Your explanation makes total sense. I'll do some tests to work out the best approach. I hadn't even considered scattering particles along the edges.