Tree shadows without glitch

[makanyane]

Thanks robomanus, I need to get the hang of doing this myself as I'm trying to create an entirely new set of trees. So far I seem to have managed to get rid of the odd extra bits of shadow with square edges, but have a problem with some trunk parts (which look very similar to others and aren't any smaller than them) which inexplicably don't cast a shadow....

I've checked that the faces point the right way so it's not that... any ideas?

[robomanus]

I repeat. It's difficult to understand with words. Upload your tree here. I'll check it and will say what's the problem..... in other way I can't help you. And what does it mean that you've checked that the faces point the right way as you say??? In which way???

[wilddog]

suspect I'll look at this over Christmas... hadn't really spotted the problem before but getting intrigued particularly as the problem description is so ambiguous that its hard to describe.

[robomanus]

[wilddog]

It's because some polygons of your new trees are too small or there are too many polygons at all so the game can't represent its shadow correctly. There are lots of other steps to do that can't be explained with words, it's a pretty hard work You know so if You still can't fix it upload it here and I'll try to help You

Not really sure its that much to do with too many polygons... the palm_a.cas main rectangular issue connecting the shadow to the trunk can still be shown to occur when you have very few trees and have reduced the cas model down to just 10 vertices and 8 triangles. Remove one more triangle and it disappears. They also aren't the smallest polygons on the model...

[robomanus]

one of the reasons is also the dimension of the polygons. they must not be very small

[wilddog]

as said but maybe not understood.

They also aren't the smallest polygons on the model...

.. ie they aren't very small. That's why they show the bug better.Though tripling the model size did seem to cure that.

Guess you're basically doing trial and error to resolve it. I was looking for a more simpler way. Just need to try a few other models.

Edit : Size does play a factor.. just not sure why but may be angle related. Equally interesting is why some items don't cast any shadow when a similar sized item still does...

[robomanus]

so did you solve it??

[wilddog]

Nope that why I edited 20 mins before your post saying

Edit : Size does play a factor.. just not sure why but may be angle related. Equally interesting is why some items don't cast any shadow when a similar sized item still does...

Then just thought that size doesn't account for the bug disappearing when I then removed a single triangle..

[robomanus]

and i sad that it did. try to rescale your tree from 100% to 1000% and see it ingame. it will be gigant but with corrected shadow.
P.S. I steel invite you to upload your cas here so I could test it too...

[wilddog]

and i sad that it did. try to rescale your tree from 100% to 1000% and see it ingame. it will be gigant but with corrected shadow.
P.S. I steel invite you to upload your cas here so I could test it too...

Think you are working one post behind.. as I said removing a single triangle resolved it not just scaling

[makanyane]

Hi robomanus (me and wilddog are working on aspects of this together),
would be grateful if you could look at this file regarding the fact that parts of it don't cast shadow at all.

The 'tree' is 4 sets of trunks - the two on the left were duplicated from same thing and altered slightly - the two on the right similarly... the two on the right cast a shadow but the ones on the left don't - have tried various moves / slight rotations / remapping / but none of that has changed the shadow casting glitch.

We're using ms3d converted with wilddogs IWTE tool and not 3ds. Would be grateful if you could convert and view in 3ds and see if you can spot what the issue is.... I know there are some 3ds functions (explicit normals) that aren't possible in ms3d - but the conversion through ms3d and re smoothing should have removed those!)

I can obviously just use the working chunk of trunk and duplicate it - but it would be more useful to get to the bottom of the problem (for this and the extra shadow bug)

[robomanus]

And Voilà.
The problem was because of duplicated vertixes

[makanyane]

ah thanks for that - must remember to check vertex/triangle count

in milkshape terms that was duplicated triangle sharings the same vertexes - which started me wondering if some of other extra shadow problems were to do with shared vertexes - tried giving one of the working bits of trunk a rear face - and it did indeed generate the problem where it had been fine before.

Have done further experiments with that shape and have posted the test files below.
1st experiment duplicates the same shape and rotates it 15 degrees in steps in both directions - the two sections rotated to the right don't have the shadow bug.

2nd experiment divides the trunk down the middle so the central vertexes aren't shared (A) - that still has the bug - (B) is duplicate of that but with the two sections moved apart - still has bug - (C) is duplicate of that but with bigger gap between the parts (rest is identical) - bug is fixed!
that's rather alarming as it seems to imply that problem can be generated by things being next to each other, and not just by the properties of the bits of model

Would be grateful if you have any ideas about what the deciding factors behind any of that might be....

EDIT: was thinking it might be to do with the distance/angle of planes in relationship to the light direction - discovered you can change shadow direction with descr_strategy_lighting.txt (x value) - but problem is the same on the rotation test cas file as before:

[robomanus]

Yeah, I forgot to tell You that in some cases the objects must be separated from each other, otherwise they will cause the shadow bug. Infact look at some of my new tree's models wit correct shadow:and
And rotation on my opinion has no importance...

[wilddog]

And rotation on my opinion has no importance...

So if rotating some objects makes the bug appear or disappear why do you think it has no importance?
Similarly if making the bug appear or disappear without changing the size of a polygon. Why do you think changing the size of a polygon is important?
Numbers of polygons also doesn't seem significant as we've shown

I can understand why it was difficult to put into words though... as it is an annoying bug

Think though that we've found out enough to tackle our issues (via ms3d rather than 3dsmax) so thanks for spotting the bug

Unfortunately though seems like there aren't any easy tests to avoid it... bit like the lighted shadow bug in the battle models.

[robomanus]

About rotation, according to my experience I have never seen the shadow correctly appearing when I rotated the objects. And what about resizing. When I mistaked to resize my tree from 1000 scale to 100 scale and ran the game, the tree was gigant but it's shadow was correct! I just don't know which is the correct way out, I just try them all and fix the bug.

[wilddog]

No problem robomanus - as you said its difficult to describe and seems to be a number of things to watch out for and various options to use to resolve... just nothing that straight forward.

+rep for your efforts

[robomanus]

[robomanus]

It seems I understood the reason: when 2 single objects are too much close to each other the game become unable to reppresent their shadows correctly because one shadow lays on another one and the game's engine can't understand which shadow must lay on the ground. So there are 2 solutions: the first one is to fuse the near objecs and the second one is to increase the distance between 2 separeted objects that are situated too close to each other, in that way the game will create 2 separeted shadows independent from each other.