Making a light 'soft' in a 3D scene

[Regina Henschel]

Hi Armin,

thank you for looking at the problem. [comments inline]

Armin Le Grand schrieb am 25.01.2022 um 10:13:
> Hi Regina,
> 
> On 1/24/22 15:51, Regina Henschel wrote:
>> Hi Armin, hi all,
>>
>> a light can be 'harsh' or 'not harsh' (that means 'soft') in an 
>> extruded custom shape. This is independent from the material property 
>> 'specular'. A 'soft' light is rendered in MS Office, as if the light 
>> has a wider emitting area in real world.
> 
> There are of course possible definitions for that. 'Our' light source is 
> kept pretty simple. We have a point light source with direction, no 
> position (all parallel). I remember that only a fraction of 
> possibilities - the needed ones as so often - were taken into account at 
> that time. I already had to fight to get all those 3D features we have 
> today :-) Sure technically these can be extended. We could add a 
> point/position, so make lighting position-dependent if wanted. Also 
> definitions for a size of a light source are common/possible - that 
> would lead to the light coming not only from one point, but from a 
> sphere surrounding that point. I am not straightforward sure how to do 
> that mathematically, but this just means to dig out the standards & 
> integrate them. AFAIR we tried to keep as close to OpenGL at that time, 
> so the 'old' OpenGL definitions should be most 
> doable/integratable/fitting/implementable (due to that 1st integration 
> of OpenGL 20++ years ago, had to be removed after 1y due to OpenGL not 
> being stable enough on target systems - a problem on SW even with just 
> 1% problems but scales to millions of users...).

All that does not sound, as there would be an easy, simple way to extend 
our 3D-scene.

> 
> As so often problem will be more to adapt the model data, processing, 
> file formats, UI, etc..., then the visualization I guess. The 3D 
> fallback SW renderer can/should be easily extended if you know what you 
> do, it is modular. Also always a good point is to keep an eye on 
> compatibility to standards, if we want to impl system-dependent 3D 
> renderers, too.

I see, that extending the 3D-scene would go far beyond the problem of 
rendering extruded custom shapes.

> 
> One Q stays open, though: AFAIR those defs will/may also influence 
> shadow which in case of soft light with non-hard boundaries will have to 
> soften/fade, too. That would be expensive since shadow of 3D is not part 
> of 3D scene, but generated 2D geometry that then would have to be faded 
> -> pixel operations & in theory full back/forth transform 2D/3D to do 
> that mathematically correctly. Also shadow on other 3D objects would be 
> more difficult/expensive if we would once need that (not yet needed). 
> Does our competitor do that...?

Thank you for reminding me of shadows. I have ignored them up to now. I 
have tested it just. In MS Office UI shadow and extrusion exclude each 
other. That holds not only for the old binary formats, but for the 3D 
effects in current formats too. So that in no problem for imported shapes.

> 
> HTH for the moment,

Yes. It shows me, that it is better to continue with my approach to use 
additional lights. The extruded shape has only two lights in its file 
definition, but we have up to eight lights in the scene. A test with 
shadow will then follow.

Kind regards,
Regina