Multiplying occlusion in last in light pipes.
I can understand the motivation if in a old real time engine wven so its not the best strategy. But it just don't make sense to mult the final cg image with ambient occlusion last. Theres a lot of tutorials out there that do this. but theres very little motivation to do this.
Imagine a pipe, now what does ambient occlusion of the pipe tell us? it tells us how much the internals of the pipe is obscured by geometry. And that is the shadow of the ambient surrounding!
So it would be motivated to multiply it by the diffuse pass and then add lights to it.
However if you do something like Compositing Ambient Occlusion and channel based lighting
where ambient lightning is actually read in 2 times, granted the render of the diffuse environment by the ibl is overexposed because most likely the render was not accounting to linear light but rather has a 2 times occouring 2.2 gamma flaw. So there would be motivation to multiply the ibl pass with the occlusion maybe. but all in all they represent the same thing! Occlusion is just a computational shortcut to the same thing.
However multing the occlusion in at thos stage defies all other possible logic except it sometimes works to make the image look good.
Now lets go back to the pipe which is demonstrates this easily imagine you have one key light and then the rest of the sit and the key shines inside the pipe. If you multiply the image with the occlusion you end up killing most of this key lights impact inside the pipe. So it oviously is wrong. And its common, its so common in fact that you can measure the effect in almost all midrange 3d shots. so much that it serves as a forensic fingerprint off many 3d artists.
Possible motivation 2: Well granted there is a nother possible motivation and tahts dirtying things up even so it doe snot hold up either, dirt also acts on light naturally so again its more motivated to use the occlusion to color grade the image up front in the diffuse color area this gives more smart optiosn for the comp person too, because the grime culd be textured.
possible motivation 3: As a fix for over exposure, caused by wrongly done global lightning. in this case there would be a slight motivation but it would be better if it was a keymix using occlusion as mixer and a revers gamma plate. Even so the depth and breadth of most peoples occlusion isnt deep enough. The entire idea of fixing this in post is insane tough granted needeed quite often becase of sloppy mistakes.
Its interesting that first dozen or so sientific an sigraph papers on occlusion never mention this kind of work flow, tough i know you can easily misunderstand it. But it would be nice to know who actually coined this wierdly motivated work flow first.
Anyway comp and 3d seems to be so mystified to most persons taht wrong placed comments spread into well accepted truths very fast. A hallmark of something where many workers are new i suppose.