PDA

View Full Version : DOF part deux



IMI
04-06-2008, 05:25 PM
A few days ago I was having a hard time understanding DOF in LW. Maybe not so much anymore, thanks to a few people here who answered with their suggestions and insights. :)

I'm still on the road towards understanding it completely, but the image I attached depicts pretty much what I was shooting for, based on what I learned. In other words, I actually managed to predict my results based on knowing my grid, object scale, focal length and distance to camera from the main in-focus part and to the background.

Now it's only the F Stop part which has me confused and is still trial and error....

Of course, this is a silly picture - just some stuff I threw together for the sake of experimenting. The objects are arranged in such a way so as to gauge the DOF effect subtleties.

Now I still wonder how to achieve perfect AA. This was the perspective camera at 24 AA passes, Mitchell, Fixed, with a threshold of .01. Took nearly an hour for my dual core AMD 64 6000, but it has reflections, area lighting and dielectric. Had to make it purty you know, not just functional.

Heh. 25 years ago this would have made me famous, not just some shlepp with shiny balls. ;)

If anyone has any suggestions to how to get that perfect AA, or any suggestions for bettering the DOF, I'd love to hear... I mean, read. :)

Exception
04-06-2008, 05:38 PM
That looks pretty nice, Imi.

I always find the Mitchell reconstruction filter to be a bit on the soft side. I prefer to use Box with a bit of oversampling, but that's personal. Oversampling also helps get artifacts go away faster.
In this scene, which has little to none really fine detail like lines or cables, you can choose a lower base AA, like 4 or so, and let the AS system determine what would be good to anti alias. An AS setting of 0.03 usually gives good print results.

Check out my guide on anti aliasing / sampling system on my site (in signature). There's lots of hints and guidelines in there.

IMI
04-06-2008, 05:40 PM
Thank you Exception, I'll do that. :)

IMI
04-06-2008, 07:46 PM
Same picture, almost. I moved the table and balls on the left up some so the front-most ball on the left is almost even on the Z axis as the rear-most ball on the right.

AA is set at 6 with Box, Fixed, and .02 Adaptive.
What's interesting is the DOF and the AA quality is nearly identical, but this one rendered at 22 minutes as opposed to 59 minutes with my previous settings.
Pretty cool. Not perfect, but definitely worth the effort in the attempt to find the way towards best AA with the least render time. :)

Exception
04-09-2008, 08:32 AM
but this one rendered at 22 minutes as opposed to 59 minutes with my previous settings.
Pretty cool. Not perfect, but definitely worth the effort in the attempt to find the way towards best AA with the least render time. :)

Cool innit?

the trick is to reduce the base AA as much as possible without it missing fine lines while letting the AS system find the problem areas where it needs refining.
I find a base AA of 4 nice since it has a passes jump at that point and is often faster than a base AA of 3.

I wish we could get user definable max-samples or at least passes control, but that might happen in the future.