[feynman]

I have to do lots of catalogue images of tables with glass tabletops. A basic on-white setup: Only a Constant Dome and a bit of Sun. A real glass material and powdercoated metal with a slight bump are the only materials used. No additive layers. A 100% diffuse black plane behind the table hidden from GI and camera acts as reflection flag for the tabletop, just as one would do in the studio.

A (GI off for the tabletop) renders very fast to noiseless, but lacks shadows from the glass. B (GI on for the tabletop) renders dreadfully slow with much noise, but has some shadow from the glass, albeit not where the shadows are seen through the glass.

Do I understand the "AGS trick" correctly that one has to "encase" the real glass tabletop in an AGS glass copy scaled fractionally larger all around?



If one instead uses the alpha/shadow channel technique to obtain the pure white background, with A = GI off and B = GI on, the result is the same, with the added downside that also A shows murky glass now and the subtle colour cast on the floor is no longer present.



So, the version A on top (with GI off on the glass and no alpha/shadow channel compositing technique) plus that "AGS trick" will probably yield the best result in the least amount of render time?

Tack!

[Mihai]

Try using B, but in render options, set Refracted Caustics to "None", under the Illumination & Caustics section. Those dots you are seeing is the famous "caustics through glass" limitation. The light falling through the glass is caustics and eventually it will show properly, but will take a loong loooong time.

This will then give you the shadow from the glass, but avoid the caustics dots.

[feynman]

Thanks! A loooong time I do not have, and no budget for a render farm service, it's just too many images, so I try that next. Speed is in need!

I read much about the famous "caustics through glass" limitation over night. Maybe I print myself a T-shirt or a bumper sticker with "caustics through glass" on it and thus meet fellow insiders : )

[Mihai]

Actually try instead Direct, instead of None and see if that helps.

[feynman]

So, above is A again (GI off for the tabletop). Below is B again (with GI on for the tabletop) with Refractive Caustics set to Direct (Illumination and Reflective Caustics unaltered). No alpha/shadow channel compositing used, to capture some coloured light bounced off the structure onto the floor.

A renders clean to SL 20.1 in 20 minutes.
B renders noisy to SL 18.8 in 20 minutes.

It is the best B-version so far. Still, A is preferable in terms of cleanliness and speed, and the table's shadowing etc. has to be faked in Photoshop, I suppose.

[eric nixon]

...

[feynman]

C with the infamous AGS glass fudge on top of GLASS glass. Better than no shadow at all and it clears dead quick.



For laminated glass (transparent coloured film) or lacquered with transparent colour or dichroic coating; how could one obtain a coloured shadow with the AGS glass & GLASS glass method? Not needed now... just curious.

[eric nixon]

Coloured shadow is not possible.

For your invisible shadow-making geometry, you could drop the reflections entirely, just use a black object with around 5% layer opacity, and if the camera is fixed, then delete the top faces.. small optimizations though.

[rusteberg]

Just to let you know that there is a very simple solution to this, so keep thinking about this. I do think its a shame that NL couldn't give you that solution already, maybe they will remember the answer or any other moderately skilled user could chirp up ofcourse.

Sunshine, what is the solution....?

[feynman]

No coloured shadows with the AGS glass fudge... oh, well, too bad. Maybe one day we'll see a speed bump via using the graphics cards or some other trick and can use only the GLASS glass proper.

[Fernando Tella]

Do I understand the "AGS trick" correctly that one has to "encase" the real glass tabletop in an AGS glass copy scaled fractionally larger all around?

The AGS could also be moved instead. The point is that the AGS faces and glass faces should not be coincident.

You could also explore with bigger sun radius which speeds up those caustics but blurs shadows. Maybe it works for your setup.

http://www.maxwellrender.com/forum/view ... sun+radius

[feynman]

That's what I did in the end; the AGS glass copy wraps the GLASS glass original with a 0,1mm gap.

What do you mean by exploring a bigger sun radius? Do you mean in conjunction with the B versions, meaning GLASS glass setups with GI on?

[choo-chee]

I'd try to render twice: 1 time the fastest "A" and the other, a very low opacity lambert model of the glass solid and merge the 2 MXI files.
So you have a fast render that is the product + another fast render that is "the glass shadow".

[feynman]

Yes, I could try that next time... had to be finished today though; too many variations/colours at high resolution were needed for a printed catalogue, so I just had to pump all the work out over night like in version C (AGS glass and GLASS glass) with SL 19 @ 2400 x 1600 px @ 1,5 h per image/node. The glass shadow reads well in CMYK proofing, so it all looks good.

[Fernando Tella]

What do you mean by exploring a bigger sun radius? Do you mean in conjunction with the B versions, meaning GLASS glass setups with GI on?

When you make sun radius bigger caustics converge faster. Just use normal glass with GI and everything. In the thread I posted all pool caustics are produced by a normal water material.