[seghier]

Hello
i made comparison between the old method to create liquid inside glass and the new feature Nested Dielectrics
now which one is physically correct and give real result ?
First test : The "infinitesimal gap" solution





Second test : Nested Dielectrics ( priority glass and liquid = 0 )



Third test : Nested Dielectrics ( priority glass = 0 and liquid = 1 )





Fourth test : Nested Dielectrics ( priority glass = 1 and liquid = 0 )



Comparison:

[Aniki]

so the ND approach is still errorneous?

[ababak]

Seghier, you should set liquid to 1, not glass.

[photomg1]

Seems to be working well, have seen something a bit odd with thousands of little replicators but need to make a better test scene before confirming as a bug. (I wasn't convinced a boolean operation was taking place over a certain density of droplets)

my quick test

[seghier]

Seghier, you should set liquid to 1, not glass.

Are you read my post

[seghier]

Seems to be working well, have seen something a bit odd with thousands of little replicators but need to make a better test scene before confirming as a bug. (I wasn't convinced a boolean operation was taking place over a certain density of droplets)

my quick test

It's not question of bubbles ; but : liquid inside glass
Can you make comparison between old method and the new ?

[ababak]

Yes, I've read your post. Just didn't understand why you test this:

Fourth test : Nested Dielectrics ( priority glass = 1 and liquid = 0 )

[seghier]

Yes, I've read your post. Just didn't understand why you test this:

Fourth test : Nested Dielectrics ( priority glass = 1 and liquid = 0 )

because it is comparison ; and study of the new feature

[photomg1]

there is water in my glass

[dariolanza]

Hi Seghier,

Definitely the white border you get in the [glass:0 liquid:0] and in the [glass:1 liquid:0] tests is incorrect and is caused by the small horizontal plane that you would get if you "booleanize" the objects in the incorrect way (if the liquid mesh removes glass).

Between the other more correct ways, the new approach is more correct as you can get the caustics of both the glass and the liquid medium, where the "infinitesimal gap" workaround provides a poorer glass caustics solution. This is because the inf. gap approximation can not consider the double material wall as the new method does.

So, the winner is the new method with glass:0 and liquid:1 priorities.

Greetings

Dario Lanza

[seghier]

there is water in my glass

can you made comparison without bubbles ? before you confirm that it work fine
-------
my comparison :
A: old method
B: setup for nested dielectrics / glass and liquid : 0
C: setup for nested dielectrics / glass has priority
D: setup for nested dielectrics / water has priority
----------
B and C : give the same result
A and D : give the same result at level of Gap
B , C , D : give the same result for bubbles inside liquid ( the same position but D have better result )
----------
at level of ground :
A : rendered fine
B , C , D : intersected with ground and we can see triangulation and we also see the same refraction

we can not say it's fine ; the method it's fine and what about the results ? what about all renders of liquid inside glass before Nested Dielectrics ; are they all wrong ?
i hope see other comparison and explanation from developpers to confirm which method is better and physically correct and please test than judge

[seghier]

Hi Seghier,

Definitely the white border you get in the [glass:0 liquid:0] and in the [glass:1 liquid:0] tests is incorrect and is caused by the small horizontal plane that you would get if you "booleanize" the objects in the incorrect way (if the liquid mesh removes glass).

Between the other more correct ways, the new approach is more correct as you can get the caustics of both the glass and the liquid medium, where the "infinitesimal gap" workaround provides a poorer glass caustics solution. This is because the inf. gap approximation can not consider the double material wall as the new method does.

So, the winner is the new method with glass:0 and liquid:1 priorities.

Greetings

Dario Lanza

thanks Dario
so all renders done before are wrong !
what about distance between glass and ground ? with the old method we don't see triangulation but we see it with the new method

[seghier]

other question about animation
how we can create animation with this setup ? realflow for example create real results

[seghier]

the new approach is more correct as you can get the caustics of both the glass and the liquid medium,
Dario Lanza

the second example with glass and liquid priority = 0 give the same caustics as the new approach
and this setup give the same caustics with previous maxwell versions ; except the gap

[seghier]

"For our algorithm nested objects should be modeled in such a way that ensures they overlap"

in this case how we can create animation if the overlapping don't happen in reality ?