Maxwell Render | Official public forum

I created a lamp as a component with an emitter material on the bulb. When mirrored, the emitter stops working.

Yes, because mirroring points the normals of the sphere inwards. You need to explode the mirrored copy, or make it unique, and reverse the direction of the faces.

Materials don't get inverted on mirror operations... why should emitters?

Another example: I made a component for a recessed light in a hallway. I placed twenty along one wall, then mirrored it for the other side. Only the 'originals' light up. This doesn't make any sense.

Are you telling me I have to explode my component every time I flip it, if it has an emitter?

Materials don't get inverted on mirror operations... why should emitters?

Because mirroring (negative scaling) a component does not change the 'frontness' or 'backness' of its materials in SketchUp. On the other hand, since a component is placed in the scene according to its transformation, and the transformations of its parents, negative scaling physically reverses the direction of its geometry's normals. Since emitters in Maxwell emit according to mesh normals, a sphere whose normals point outward in its component definition, emits light inward, when an instance of that definition is placed in the model using a negative scale.

The only workaround I've thought of for this so far would be an explicit 'Explode during export' type of per-object flag, but I have not yet implemented this, so yes, you need to do it yourself, by not exporting negatively-scaled emitter geometries. In the scenario you mention, you would mirror the first light, rather than twenty copies, run explode/reverse-faces/make-component on the mirror, and then array the original & mirrored components twenty times.

All that aside, though, you may be better off avoiding the use of emitter geometries inside of components at all for the moment, because support for instances with emitter materials was only just added in 2.6, and currently has a bug: http://www.maxwellrender.com/forum/view ... 10&start=0, which may make it difficult to determine why a particular scenario is not working as expected. Here, this makes it a little confusing to test even a simple two-cube scene.