The example of complex fenestration worked on here is the deceptively simple case of just some common venetian blinds. Shown is a couple of 3D backward-ray-tracings in RADIANCE of an idealised infinitely thin sheet - the thick black line is just to help you 'see' it - of a complex material simulating common venetian blinds with their slats horizontal, and with the light at just two incident angles - normal, and at 75 degrees. The beam of light impinges from the right onto the sample ideal material. The impinging light is shown in red, as is the transmitted specular component. The backward and forward scattered light component is shown in blue for clarity. The 'exposures' of the images differ.

As you can see, depending on the incident angle, the light may or may not be transmitted specularly, and may be scattered from the front and the back in different amounts which are intended to correspond to the actual physical behaviour of the venetian blinds. This was a lot more difficult to achieve in practice than it looks here. The uneven nature of the scattered light distribution in the images is an artefact of the limited random sampling of light rays in the computer simulation run. In this case, the analytic algorithms of transmission and scattering were idealised approximations to the results of a slow and tedious previous 2D Monte-Carlo forward-ray-tracing study of a computer model of the venetian blinds, using hundreds of thousands of 2D rays into an aperture with just two blind slats.

The simulation modeller would get real benefits in speed if he could idealise such everyday items as venetian blinds into a simple element in the plane of the window with confidence as to the accuracy of the gross resultant light flux. At the moment, this appears to be theoretically possible. That would allow the modeller to create huge areas of venetian blinds modelled comparatively simply, so that large buildings and large spaces could be rendered without the need to model the slats individually - which would be a real killer in terms of accuracy and in computer simulation size and time.

The principle may extend to fenestrations of other types in a limited way, I hope.

To start with, in general, the desired optical properties of the fenestrations to be modelled are obtained from either: (a) test data on real sample panels, or: (b) the results of forward ray-tracing on computer-modelled sample panels, or: (c) published data if you're lucky. Suitable algorithms are then built by the lucky modeller, so as to approximate the real-world behaviour of the building fenestration system in the computer as an infinitely thin single-skin homogeneous sheet with complex properties defined by a reduced set of parameters and some sort of idealised generalised aggregated macroscopic function, all in his chosen software suite. Of the programs I've used, RADIANCE gives me the most confidence. It is a physically-based renderer which models closely the way materials and light actually behave. Usually, rendering software isn't and doesn't; its designers just try to fool the eye as well as they can, because it's easier and that's the result they're selling.

Today, using modern 3D CAD software packages, the 3D models of complex large buildings and large internal spaces are typically made and rendered - in the fool-the-eye fashion - comparatively quickly. But this type of software, although it can produce very pretty pictures in the right hands, cannot accurately predict either daylight or artificial light. For that, you need RADIANCE - or one of it's competitors.

My hope is that idealised models of complex fenestrations will be used, in RADIANCE backward ray-tracings, to generate fast economical simulations of daylighting in large areas to allow interactive informed decisions in the early stages of design evolution, when it counts the most. This matters when considering alternative architectures and fenestrations of entire large built volumes in their surroundings. If only the modelling could be easy, quick, reliable and effective ....

The issue here is more getting numerical design studies of several alternatives for 10,000 square metres of space or more, maybe over a complete year, done fast, rather than producing a desperately highly-rendered wow-inducing pretty picture of 10 square metres of space over a weekend for some client to drool over on Monday. Not that we don't do that too.

George