Radiance software - physically based light simulation

Light simulations and rendering at pab-opto currently use mostly the Radiance Synthetic Image System. Radiance offers light simulation based on physical laws and generating quantitative results (radiance, irradiance, luminance, illuminance).

links and talks on my Radiance and BRDF/BSDF data, models and measurements

Radiance features

• quantitative results (both images and singular values)
• validated by various studies
• considers most light paths: indirect illumination and optionally caustics (extra module)
• scales well with large scenes (millions of objects)
• code is public and open source, making it possible to verify algorithms directly
• extended rendering: depth-of-field, stereo images and animations

Alternatives to Radiance

Advantages of Radiance: Results checked over years, code considered stable, low memory footprint of core engine, supported by author and (if funds are available) by institute LBNL, support by fellow users via email list. It has been used by commercial companies on commercial scale projects over 2 decades. Since any add-on to the core has to pass through Greg Ward, the code has been very stable and therefore reliable. And, probably the most important feature: It doesn't suffer from advertising hype and wild we-can-do-all PR statements.

Drawbacks: Anyone adding to the core code suffers from its non-modular structure. That and the tenacious lack of robust automated test scenes (as of Sep 2014) limit the writers of source code for the core engine to a very small number (say maximum of 3 people over the last 25 years). Therefore, the core rendering engine is trailing current technology development. In terms of computer-graphics it is in many ways not cutting-edge. It is very difficult to get any feature into Radiance for which there is no interest by Greg Ward or funding at LBNL.
The user interface, a problem over many years, is getting better since Radiance was built into a broader framework, e.g. OpenStudio by NREL.

Alternatives: There are some commercial simulation programs out, whose advertising claims to handle everything (especially in my field of BSDF data import and models), but in fact they do not. Since there's no source code available to check, it would be pretty bad to base consulting work on that.
In optical industry, some well established, reliable ray-tracing programs, e.g. Zemax, Optis and others, have BSDF import features, which are more or less well documented. But getting the vital details requires NDAs, if they are available at all.

PBRT is probably the best alternative to look into. A modular, public domain (source available) program, with an established user base. How much this can be used for quantitative analysis is an interesting topic of investigation.

Historic perspective

Author's experience with Radiance has been accumulating since Radiance Revision 1.2 in March 1990. First at Fraunhofer Institute for solar Energy, than at pab-opto.