Page 1 of 2
Another lens test
Posted: Sat Mar 31, 2007 8:16 am
by ryjo
This is a test I did, inspired by another lens test in this forum. The light in the upper part is a flare.
Indigo 0.7t5
Posted: Sat Mar 31, 2007 9:04 am
by manitwo
unbelievable clean

rendertime?
superb flare!
Posted: Sat Mar 31, 2007 9:26 am
by Kram1032
yeah... really clean...
looks like an inversed reflectioning ball, instead of a transparent ball...
Posted: Sat Mar 31, 2007 9:55 am
by ryjo
This is the latest entry from the log:
Time elapsed: 47h, 32m, 3s
Done 9932600000.00000 samples (18393.70370 samples per pixel)
58043.58855 samples / second (17.22843 micro-seconds / sample)
Shutting down...
But it looked pretty ok after about 10 h.
In order to set the focus I moved the image plane to 48.5 mm (it ought to be about 50 mm).
Posted: Sat Mar 31, 2007 1:13 pm
by IanC
Wow, that's amazing! I never managed to get a decent image out of it.
How hard was it to alight the lenses? I could never get the focal point right (I think) with simple lenses, so decided not to go for the cooke triplet.
Posted: Sat Mar 31, 2007 2:10 pm
by zsouthboy
Is it me, or could we(by which i mean exporter writers) automate this once someone gets the hang of it?
Posted: Sat Mar 31, 2007 9:10 pm
by ryjo
IanC wrote:Wow, that's amazing! I never managed to get a decent image out of it.
How hard was it to alight the lenses? I could never get the focal point right (I think) with simple lenses, so decided not to go for the cooke triplet.
Hi,
The way I did it was to take an image of the design and then reconstruct based on this. I added circles that "overlapped" each lens surface, and then generated spheres from these circles. Using CSG i could make the three lens elements. Then I added an aperture behind the middle lens (where the aperture stop is). I exported the lenses (and a box) as obj and did the xml by hand.
Each lens is about 35000 vertices.
There is this program, zemax, and the demo has some example designs.
Edit: The theoretical focal length is 50 mm for this lens, but that came out blurry in the tests, so I adjusted the image plane to 48.5 mm, which is what is rendered.
Posted: Sat Mar 31, 2007 11:26 pm
by dougal2
wow. your modelling must be somewhat more accurate than mine.
good work
Posted: Mon Apr 02, 2007 12:11 am
by ryjo
dougal2 wrote:wow. your modelling must be somewhat more accurate than mine.
good work
Thanks! I think your result were as good, and my scene isn't very interesting. It would be nice to see a real scene 'shot' with a modeled camera and see if it looks different, or more realistic (as a photo).
I have attached the sources.
Cheers
Posted: Mon Apr 02, 2007 8:34 pm
by 8bstudio
Great....but I don't understand how you use your lens project instead the camera in Blender for example!! Can you explain me please!!
Posted: Mon Apr 02, 2007 9:54 pm
by ryjo
8bstudio wrote:
Great....but I don't understand how you use your lens project instead the camera in Blender for example!! Can you explain me please!!
Hi,
I don't use Blender, but I'll try to explain how to set up set camera: The indigo camera is placed inside the modeled camera and is pointed towards the rear end (opposite to the lens, at the front). The rear end must be diffuse white (1,1,1), all other walls in the box must be black (0,0,0). The lens will project the world outside the box onto the white screen. Set the FOV of the camera so it captures the screen (back of the box), but no the walls (they will just be black). See layout.
Hope this helps!
One additional comment: In order to focus the image plane (back) must be moved slightly. For focus at infinity move i 1.5 mm closer to the lens. For focus on something closer than that (say 5 metres), move even closer. This is trial and error process (even though there are some equations...)
Posted: Mon Apr 02, 2007 10:04 pm
by 8bstudio
Thanks a lot "ryjo". I understand now and I'll try soon and surely!
Posted: Wed Apr 04, 2007 6:49 am
by ryjo
This is another result of using the simulated camera. The cleaner image is the reference pic (rendered with indigo camera only).
The "simulated camera" image is left/right-flipped.
Time: 20h / image.
The model is the sponza atrium.
Posted: Wed Apr 04, 2007 6:55 am
by Kram1032
looks good

Posted: Wed Apr 04, 2007 7:47 am
by Ricky
That is really a great test...!