Aperture diffraction, what is the right way to use it?

[ENSLAVER]

Just curious if I was doing this correctly. In the first image, no diffraction.
In the second, it's using the aperture map posted.

Is there a way to stop it blurring the grass? Only effect the very brightest areas?
Also, using an aperture map made in phtoshop, when rendered the "lens flare" around the sun was square, and cut off at the edges of the map. Was I doing something wrong?
Thanks, maxigo user.

No aperture diffraction map

indigo_grass_F400CD.jpg (215.67 KiB) Viewed 2321 times

Using aperture diffraction map

indigo_grass_F400CD_or_agfa800cd-scene1.jpg (185.6 KiB) Viewed 2321 times

(not really 2048, cut down for forum)

aperture_6blades_2048.jpg (4.13 KiB) Viewed 2321 times

[lycium]

I'm not sure if it'll solve your AD blurring issue, but if you want super sharp images I suggest to render with a high supersampling rate and to edit the inifile.xml settings to use blurring=0 and ringing=0.5 (which corresponds to the Catmull-Rom spline filter in Max).

[suvakas]

I'm not sure if it'll solve your AD blurring issue, but if you want super sharp images I suggest to render with a high supersampling rate and to edit the inifile.xml settings to use blurring=0 and ringing=0.5 (which corresponds to the Catmull-Rom spline filter in Max).

For both, splat and downsize?
Any other good settings you recommend ? Would be good to make some presets for the exporter.

[lycium]

Some presets:

* Fast preview: box splat, M-N downsample with Indigo defaults (B = 0.6, C = 0.2 I believe)
* Smooth: Gaussian splat, M-N with Indigo defaults
* Crisp: As above, but B = 0, C = 0.5 (Catmull-Rom)
* I-would-have-sharpened-it-in-Photoshop-afterwards: B = 0, C = 1 (min supersampling factor of 3 recommended)

Not really sure what to call the last one It's overly sharp in terms of signal theory*, but sometimes it can produce a beautifully crisp image for architectural scenes (especially in conjunction with glare filters which tend to blur the image). However, these settings are also in (reverse) order of convergence speed, with the sharper ones taking longer since they don't suppress noise.


* Nerdy digression: it introduces post-aliasing, i.e. degrades the reconstructed signal compared to Catmull-Rom which agrees to O(h^4) with a Taylor series expansion.

[ENSLAVER]

Shapening presets seem like a good idea, specially in terms of speed vs quality, though perhaps not entirely necessary, nice to have.

I have to admit when I first posted this, I didn't understand aperture diffraction. I had the totally wrong idea about how it worked. I thought perhaps it was just some flashy lens flare type effect that acted on the brightest points. Indigo's implementation stays true to physics and in a rendering like the one above the grass is correctly blurred.'

A little research goes a long way.