The method chosen to assess the difference between modeling and simulation methods was using photos of the original space and comparing these with the same viewpoints in AutoCAD, Lightscape and Radiance. Numerical analysis of the results including computation times and file sizes will be done.
The expectation is that a 50% reduction in rendering time is significant, and certainly rates a worthwhile modelling technique.
The modelling methods that will be used for each model are as follows:
Model One:
Surface Modelling
Surface Modelling is the use of three-dimensional surfaces to define the volume or envelope of an object. (AutoCAD - 3D Design and Presentation)
Model Two: Surface Modelling
Model Three: Solid Modelling
Solid Modelling is building three-dimensional objects with solid primitives such as cubes, spheres and cones. (AutoCAD - 3D Design and Presentation)
Model Four: Solid Modelling
Model Five: Solid Modelling
Model the space using simple geometry, and then build this model using Radiance Primitives in Model 6.
Model Six: Radiance Primitive Modelling
Using Radiance Primitives to construct a simple form, representing the gallery chosen for Models 1-4. It is thought that by using a lighting program's own modelling tools, the model would be much more efficient and this would in turn optimise the rendering time and file size.
This research method has limitations as it has been developed particularly for the Menil Museum and could differ with other buildings that are constructed differently. Therefore, it is a single building case study.
Assumptions:
To be able to export the files into Lightscape and Radiance, the files need to be in .3ds format; this is done in AutoCAD using the 3DSOUT command. 3DSOUT defines the source of 3D objects, converts blocks into 3D objects, and performs smoothing and welding during the export process. (AutoCAD User's Guide)
Once the file is in .3ds format, it will be imported into 3D Studio Max to ascertain the Polygon Count in the model.
Not only is file size an indication of optimal modelling, so is rendering time. Hence, the 3DS model will be imported into Lightscape and Radiance in the following ways:
The import process may also bring to attention problems with the chosen modelling methods and may prove/disprove them as suitable. The models will have a matte grey paint finish applied to them. A low quality render will be done on each model to ascertain its computation time. The default settings will be used in all cases, so that consistency can be maintained throughout the project.
A time estimation will be given for each stage of the process. This will give some indication of which method is more useful to a user depending on their skills, time and resources.
A useful tip found by Warrick Steele in his research on Architectural Modelling in Computer Animation, was that file sizes increase in AutoCAD when you erase objects. He suggests that a solution to this is to purge the drawing, thus clearing the object instead of erasing or changing the file name. Purging also deletes any unused pen types and named objects, (such as blocks or layers), from the drawing, thus decreasing the complexity of the file and its size. This will be done to all AutoCAD files.
Once it has been decided which modelling method(s) are optimal, these models will then be fully simulated in Lightscape and Radiance. Varying settings (low, medium and high) will be used to assess ability to reduce computation times, and how this affects the quality of the final image.
Materials, textures and finishes will be applied to the models and the scenes will be simulated under a sunny sky condition.
Renders in Lightscape will be stopped at 95% of total calculation in all models. All renders will be set to render out at 800 x 600 pixels.
The Lightscape models are to be tested in two ways: using the standard GUI and using the command line interface and batch files. Literature suggests that the command line interface is faster than the GUI method of rendering. One effect is shorter computation times, as each iteration step no longer needs to be generated on the monitor.
Rayfront, which is a GUI to Radiance, will be used to produce images. However, this report is not comparing these two programs, but only using Rayfront to perform commands. Where necessary, the advantages of the GUI will be mentioned.
Simulation Method
The most important point to make is that in this process, I was not trying to achieve an accurate lighting solution in terms of a certain amount of light hitting the working plane. Instead, the aim was to produce an image that represents reality.
As the modelling methods showed, all of the geometry could not be imported into Lightscape in 3DS format. Hence, the geometry was imported in DWG format.
As a test, one of the models was imported into Lightscape and the preparation process was followed through. This included surface orientation and definition, materials, and lighting properties. Following this, a test render was performed on a low setting to establish what initial results could be found. It became evident that the rendering did not at all represent reality, as the space was not illuminated correctly and the materials looked nothing like they did in the photo.
After a long period of adjusting, it was realised that the problem was the material properties and light redirection. Due to a lack of verified information on the materiality and properties of the museum's building components, values had to be estimated from their appearance in the photos. This led to the previously discovered inaccuracies.
In the interest of optimising file size and computation time, it was decided to reduce the amount of geometry in the model to only what would be seen. Once this was done, another test render was performed. This produced very pleasing and more realistic results.