[Richard]

Brodes I reported a similar issue with HDRI made from sky export was in error - JD worked out the camera used to save out the sky must be set at an EV of 15 to get matching exposure. Workaround: Clone scene camera > set clone to EV15 and export sky > delete clone and use sceen camera thereafter with the imported maps.

[brodie_geers]

I was trying to recall what the conclusion of that was. My camera seems to be at 14.966 EV. I don't know if that's far enough off to account for all of the lighting difference. I'll probably get bored enough to run a test or two today.

In the meantime a bug confirmation from NL would be appreciated.

-Brodie

[brodie_geers]

That did indeed pretty much solve the light level issue. There's still a small difference but I think it's probably more as a result of the extra noise still in the IBL version rather than actually being darker.

-brodie

[3dtrialpractice]

but stil no joy on the caustics eh?

i remember doing a caustics render test with IBl in v2.0 and it rendered em fine(in fact very nicely as i was suprised how clean it was)

... so this must be a ver 2.5 bug

[brodie_geers]

Right, still no caustics and still more noise in the IBL version.

-Brodie

[brodie_geers]

Hopefully this will be my last test. This is a 400x300 rendering I let run overnight and it got to SL 32.20. As you can see, the caustics DO, in fact, start to come in. I'm not entirely sure when they start looking like caustics but at SL25 the white artifacts start to congregate more around the caustic areas. Sadly, even at SL32 the caustics are still quite underdeveloped.



Any comment from NL?

-Brodie

[Half Life]

I have seen exactly this behavior in certain conditions before -- this is the reason I advocate letting the renders run as long as you can... there usually is more there that just hasn't developed yet.

Glass and SSS are two areas where this is particularly true, but I see now that IBL also has a similar quirk of the priority optimization.

This was amazingly helpful Brodie, thank-you.

Best,
Jason.

[tom]

I confirm caustics generated by IBL illumination is currently not efficient. We're working on making it better.

[brodie_geers]

Thanks Tom, any comment on the noise issue as well?

-Brodie

[tom]

Noise is due to same limitation in optimization. Small sources of emission compete with ambient.

[Brendan Sharp]

hi, interesting thread.. Had a couple of posts lately trying to resolve noise- see here:

http://www.maxwellrender.com/forum/view ... 05&t=35972

http://www.maxwellrender.com/forum/view ... 97&t=35953

Tom could you please explain the limitation in optimization theory a little clearer/simpler...

at SL 16 after 3 hours I still cant resolve these fireflies in the forground of the street images I posted in the other threads.

even with individual street lights set at 178 watts, tried various emitter materials on them too. physical sky settings 5777.0 K sun power 1. planet refrlectance. 27%

everything else standard- IE: wavelenth exponent, turbidity etc.

location set to saudi arabia Dhahran & rotated the earth to be at 10.00 pm for a night scene. please see the images linked...

oh... and one more thing if someone could help me clear up what would be good camera setting for taking this kind of street view. I have seen the recomended EV values in the manual for various environments, and tried just about everything else, to no benefit

for Fstop I have 5.59, (focus should be about 5-290 meters away) ISO 100, shutter 523.68

thanks all for your time....

Brendan

[tom]

Rendering of caustics and sampling importance of emitters are two different things. Without any optimization or different strategy, the caustics could be faster or slower depending on situation. Caustics appearing faster can make the rest of the illumination multiple times slower and vice versa. There's no perfect solution which can handle every situation. Emitter importance is also similar. Roughly speaking, the optimization engages the core with sampling of powerful emissions first.

[tom]

Brodie, here's a similar test which shows no remarkable differences except the ones outlined above.

Physical Sky vs Exported HDR


Scene:
http://min.us/mv5Vsz#3

[brodie_geers]

Yeah, looks pretty much like where we left off (no caustics and more noise in IBL). Are these tests with the new release 2.5.1?

-Brodie

[tom]

Yes. An exported HDR image can never match Physical Sky in all aspects. Physical Sky is an analog model while IBL is pixel based. They have pros/cons to each other but, at least we can agree the intensity is not off (or fairly off) at all.