Incomplete Bokeh shapes

[^Lele^]

Yeah, make the ice shader a bit rough (you currently have it at 1.0 gloss) and the "artefacts" will be gone.

[fixeighted]

Aha! Problem solved

[^Lele^]

Apologies for the short reply.
We're busy busy, these days.

It's the "usual" case of low-probability, high-energy event.
In this case, it's one which is quite perfectly resolved (it wasn't the bokeh shape that was incomplete. it was geometric shadowing of the highlight from the mesh, before it turned into a bokeh...), but wasn't very pretty to look at.
You wouldn't be able to see the highlight without DoF because the Confusion Circle was needed, and big enough to be able to actually see the highlight from some whacky angle.

Here, lowering gloss from 1.0 has two direct effects:
a) it spreads energy a little bit, softening the highlight and reducing its peak brightness (contrasting the high-energy)
b) The highlight footprint grows and it's more easily "found" (contrasting the low-probability).

In this specific scenario, a) was why lower gloss was needed, as the results were clearly well sampled already.

[fixeighted]

That's great knowledge and really adds to at least my understanding. The sort of stuff that needs to be added to docs (when not super busy of course )
Doesn't it make a good case for the default glossiness to be .98? So we don't forget and it's like a 'recommended' setting.
I habitually make everything .98 to start, but I bet many don't, then wonder why odd things occur.

[^Lele^]

That's great knowledge and really adds to at least my understanding. The sort of stuff that needs to be added to docs (when not super busy of course )
Doesn't it make a good case for the default glossiness to be .98? So we don't forget and it's like a 'recommended' setting.
I habitually make everything .98 to start, but I bet many don't, then wonder why odd things occur.

There is a considerable performance difference between 1.0 and 0.98, if nothing else because of the many more rays spawned.
In normal situations (i.e. no DoF/Moblur, and so on), this may not be ideal.
With the presets on the vrayMtl (which we can expand and/or refine) we have a way to show users how to make a valid shader, and let them choose for themselves how they'd want their defaults (via, f.e., maxtsart.max file).

I'll make a note to add this to the docs, although to be honest i'd prefer we had an option to take care of this trick automatically (I am *not* making friends with the devs, right at this moment... ^^).

[fixeighted]

Sorry, not sure I get you...the performance hit is greater for 1 than .98 or vice versa?
I would have imagined it to be for .98 and lower - but then we don't have any option for most surfaces other than shinies.
I just assimilated the use of.98 for the shiniest stuff, purely to get away from perfectly reflective.

Quote: (I am *not* making friends with the devs, right at this moment... ^^).

Naughty step occupied then?

[^Lele^]

A perfect mirror (that is, 1.0) traces one ray, any value lower than 1.0 will trace more (potentially many more: think of inter-glossy-reflections. Some very clever maths is at work to avoid the number of rays per pixel to exceed the quintillions.).
The lower the gloss, as a general rule, the broader the cone to cover, and the more rays to be traced.
As various optimisations come into play, however, this isn't cast in stone once one is working at a shot, but rather a useful guideline.

[kosso_olli]

I can confirm that lowering the reflection glossiness for the ice fixes the issue. I would not have thought of that as the culprit. Thanks again!

[fixeighted]

Do I get cake?

[kosso_olli]

https://en.wikipedia.org/wiki/The_cake_is_a_lie

[fixeighted]

Hehe...I played Portal to death...love it
Portal, The Movie is out next year...