PDA

View Full Version : EIAS scale issue in Caustics…


Hiddenman.fr
08-16-2006, 10:01 AM
Ok, I've just lost some time with Caustics tests to really understand what was going on.
The problem that I encountered was that Caustics tutorials were always working whereas my projects don't. And the problem was a scale issue. EIAS base projects are always made with big models (don't know why. If you have an answer do not hesitate to explain it to me) - say 100 meters (in real life that type of object is not very popular). I'm very sorry to live in a "real" meters world but my models are often 1 or 0.1 or even 0.01 meters.

So let's start.
If you test a 100 meters object with a Photon Density of 1 all is ok and Caustics are rendering fine.
If you test a 1 meter object, the Photon Density has to be set to 10000 to obtain the exact same result. Not very easy to find even if I now understand why.

So the aim of my post is : Couldn't developers create their tutorial projects with NORMAL scale objects to prevent us to waste too much time to figure out what is going on?
Scale issues are really annoying but common things in EIAS plug-ins life. I've already had big issues with objects scales in Blaster, Rodeo, and EIAS.

halfworld
08-16-2006, 02:07 PM
Caustics are probably the most scale sensetive effect in EI so I agree with you that greater care should be (or have been) taken with the preparation of those particular tutorial projects.

I suspect the reason for the inordinate scales is the fact that these tutorials are set up quickly to achieve an effect rather then to achieve an effect at an working environment scale. Other tutorial projects (some of the GI tests for example) are set up at scales you would expect to see in a workplace which is great for everyone.

As for plug-ins, I find rODEo bends over backwards to accommodate the different scales people work at, but as for Blaster, you are correct, it is rather unyielding (no option to change the scale it works at unlike in rodeo). It is still pretty cool though ;)
Ian

Igors
08-16-2006, 02:48 PM
Hi, StephaneEIAS base projects are always made with big models (don't know why. If you have an answer do not hesitate to explain it to me) - say 100 meters (in real life that type of object is not very popular). I'm very sorry to live in a "real" meters world but my models are often 1 or 0.1 or even 0.01 meters.A model with size 0.01 and less is potentially dangerous: a distance between vertices easily can be out of 4-byte float precision. For example, a subdividing/triangulating of such model produces a lot of non-planar geometry

If you test a 100 meters object with a Photon Density of 1 all is ok and Caustics are rendering fine. If you test a 1 meter object, the Photon Density has to be set to 10000 to obtain the exact same result. Not very easy to find even if I now understand why.The square of quad with side = 1 is: 1*1 = 1
The square of quad with side = 100 is: 100*100 = 10000
That's less intuitive than linear , but square is square.

So the aim of my post is : Couldn't developers create their tutorial projects with NORMAL scale objects to prevent us to waste too much time to figure out what is going on? Scale issues are really annoying but common things in EIAS plug-ins life. I've already had big issues with objects scales in Blaster, Rodeo, and EIAS. "NORMAL" = what's Ok for Stephane in his current prj? Or what? :)

FelixCat
08-16-2006, 04:47 PM
I suposse Stephane thinks that Normal size is the real size of objects in the real world. Personally i try to set the objects with real sizes in mind: 1,8 mts for a man and 0,15 mts for a pen.

FelixCat

Hiddenman.fr
08-16-2006, 05:09 PM
You have right Felix, for me "Normal Sizes" is "Real Sizes". Why working in 10 or 100 times bigger than life scenes? (and no Igors I'm not working on a project, it is just R&D on Caustics but scene units bother me since 9 years). If a 3D application is unable to work at real resolutions cause floating things, why not having a pref. to set the working scale inside this application and it converts them itself to what it suits it best to prevent errors. That way we wouldn't have to make perpetual conversions to know exactly what we are doing. It is easier to work in real units cause we know them, we know what they are. Or it is just me?

halfworld
08-16-2006, 05:58 PM
I think what the Igors are meaning is:

People work at different scales...

For some people 1 scene unit is 1km or 1m or (like me) 1 cm.
So what is ten metres wide for you could be 10 cm wide for me....

No one scale is correct. Going back to the tutorials, they seem to have no coherent scale, that being the issue.
Ian

Igors
08-16-2006, 06:11 PM
Hi, Stephane, Felix

I suposse Stephane thinks that Normal size is the real size of objects in the real world. Personally i try to set the objects with real sizes in mind: 1,8 mts for a man and 0,15 mts for a pen.But, Felix, 1.8 mts for you and us is "6 foots" for Brian (or something other in yards ?)

You have right Felix, for me "Normal Sizes" is "Real Sizes". Why working in 10 or 100 times bigger than life scenes? (and no Igors I'm not working on a project, it is just R&D on Caustics but scene units bother me since 9 years). If a 3D application is unable to work at real resolutions cause floating things, why not having a pref. to set the working scale inside this application and it converts them itself to what it suits it best to prevent errors. That way we wouldn't have to make perpetual conversions to know exactly what we are doing. It is easier to work in real units cause we know them, we know what they are. Or it is just me?

The 3D geometry is fully "similar" or "homothetic". The render should produce exactly same results for objects 1 and 100 meters, but, of course, IF ALL in scene is changed correspondly (distance to camera and lights are first in this list).

Let's back to you caustics example.If you test a 100 meters object with a Photon Density of 1 all is ok and Caustics are rendering fine. If you test a 1 meter object, the Photon Density has to be set to 10000 to obtain the exact same result. Not very easy to find even if I now understand why.That's fully true, but did you change a distance from object to light correspondly? As we understood, no, you use same "0" (AutoDetect). And if so, please agree: same object produces very different caustics depending from how it closer/far to radial light (and requires very different count of photons). Direct your caustics light into your object center and type this distance in Light Window/Caustics Tab. Now scale object and scale distance to light correspondly - you should see exactly same result with same photons density. That's how geometry works from times of Euclid yet :)

Igors
08-16-2006, 06:54 PM
Hi, IanNo one scale is correctIanExactly. Maybe it's better to say "no one scale is absolute/ultimate". And if so, why need to burden app with a lot of "units" instead of giving a single but universal one that each one can treat as he wants?

plehrack
08-24-2006, 07:51 PM
EIAS base projects are always made with big models (don't know why. If you have an answer do not hesitate to explain it to me) - say 100 meters (in real life that type of object is not very popular). I'm very sorry to live in a "real" meters world but my models are often 1 or 0.1 or even 0.01 meters.

Animator uses Granger units, not meters. Who is to say that the default models are not 100 cm?

Peter

halfworld
08-25-2006, 08:33 AM
Most of them look to be at a reasonable scale (just looked again), but that would still make the glass of sherry over two meters across...

GU = MM? 22cm across....

It would be fair to say that either it's an enormous glass, or the scene has no scale...
Ian

CGTalk Moderation
08-25-2006, 08:33 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.