[Becco_UK]

tizxx: With your scene place glass in front of the emitter - the light then becomes a caustic - it is this that doesn't render correctly in reflections. It would be interesting to see that. I did ask the original poster if it was really necessary for the image to be projected onto the wall through glass. Your simpler method of using an emitter works ok as an alternative.

[Hybaj]

Just a typical case of path-tracer's "vampire caustics". It's been covered on the forum ages ago.

[tizxx]

Sorry I did not understand, now I download the scene and then
I can't see the image projected.

Probably this is the awaiting for RS2 render engine.

RS1 has limitided in this situation.

We must wait for Maxwell Render 2.0

[Bubbaloo]

Just a typical case of path-tracer's "vampire caustics". It's been covered on the forum ages ago.

Most of us weren't here ages ago.
...And it still seems to be a problem not yet solved. Maybe it is time for a reminder?

[Hybaj]

Bubbaloo

This is such a big and fundamental problem for all unbiased renderers using path-tracing techniques that you cannot really push NL into solving it. The guy who's coding Redqueen renderer seemed to have combined some biased and unbiased techniques into one good looking package. But the images are long gone and there's only one sample pic of it on his blog - and who knows how really biased this technique is.

For example one thing that helps caustics behind water surfaces (and let's say mirrors) is Metropolis Light Transport (MLT) - but who knows how well Maxwell implements it since I've seen a test with a free unb. renderer that can use MLT and Maxwell was less efficient with caustics behind something refractive.

[Bubbaloo]

This is such a big and fundamental problem for all unbiased renderers using path-tracing techniques that you cannot really push NL into solving it.

Not really pushing, more like a gentle nudge.
There's a solution to the problem, I just want NL to find it first. Maybe in this case a little caustic interpolation wouldn't hurt?