More modeling techniques

The basic concept of subdivision surfaces is not new as it is originally published in a paper written by current Pixar's CEO Ed Catmull in 1974. However the technology did not get popular until it was used by Pixar in modeling Geri's head in a movie. They had realistically designed the creases both hard and soft in Geri's head, noses and knuckles and clothes.

New technologies arise from university labs. One such example is the photogrammatery software that creates 3D geometry from analyzing multiple 2D Photographs by interpolating corresponding points between these photos.

It is all about the design data and its applications. If it is only to visualize (say in video games or movies) and not for manufacturing, there is no need for a lot of geometry. With little geometry and detailed textures and sophisticated bump and displacement maps shaders can represent a wide variety of models.

Unlike geometry based on some primitive shapes or NURBS or just plain old polygonal meshes, mathematically described implicit surfaces also known as isosurfaces are the basis for functional modeling. Surface is defined using the superposition of Gaussian potentials representing individual atoms. Many design software like 3DS Max and Maya have a feature called metaballs thas implements this technique.

Constructive Solid Geometry is the technique widely used by MCAD software and B-Rep (boundary representation) is used by CATIA, Autodesk Inventor, etc. In CSG, complex objects are represented as boolean operations on simple primitive solids like cone, sphere, tori and blocks. Some CAD/CAM applications use this technique.

In B-REP, objects are defined using topology and geometry. Topology represents the connectivity between various entities like SHELLS, FACES and EDGES. Geometry actually defines the data underneath the FACE or EDGE.

I will talk about animation in computer graphics tomorrow.

Computer Graphics Fundamentals

Computer graphics has three main areas of concentration.
1. Modeling
2. Animation
3. Rendering

Subdivision surfaces are the latest in the geometric modeling area. There are some unknowns in this area and is an active research area in the graphics community. There are many open problems for PhD ideas available in this area. We need breakthrough results in this area to have it being used in commercial products for manufacturing. Not having C1 is a major letdown. Also, booleans operations and other solid modeling operations on a subdivision surface are still a major challenge. Currently these are used in products aimed at video games and movies. These models will visually look pleasing and smooth but are not mathematically smooth. Manufacturing these surfaces will be an issue.

Inverse kinematics is an active research area in animation. Maya is the world leader in providing animation modules. Cloth animation and fur animation, particle simulation, crowd effects are recent additions.

Photorealistic rendering and fast and accurate global illumination for lighting effects is very active in the area of rendering. Many university labs are doing active research in improving rendering techniques. Cornell university comes to mind immediately.

More on further research studies in a future post.

Geometric modeling - mathematical representations

There are many ways to define the geometry of a digital model. Some popular ones are
1. Polygonal meshes
2. NURBS (Non-Uniform Rational B-Spline)
3. Subdivision surfaces
4. Implicit surfaces

There are software packages entirely based on one of the above to internally represent their models. Rhino, AliasStudio are pure NURBS based modeler. Some software like 3DSTudio and Maya have both meshes and NURBS representations.

Meshes vs NURBS have their own advanatages and disadvantages. For example, to represent a perfect sphere as a mesh object, we need thousands of tiny triangles to get the smooth curved shape of a sphere. NURBS can represent a sphere with just 9 control points and a few knot values. However, NURBS need to have rectangular surfaces (as they are parameterised in 2D space). Trimmed NURBS are common to represent freeform surfaces but they have leak and gap problems between surfaces.

Subdivision surface came as the rescue to solve both these issues. We can represent arbitrary free form surfaces with less control points. However, mathematically some regions of the subdivision surface is not smooth (not C1 continuous).

There are pure subdivision surface based modelers out there. Modo, Silo are some examples. They have huge limitations in the kinda models they can represent. They work great for certain applications.

IT is a great opportunity

Data Translation

CAD data can survive many changes and iterations and still retain its integrity only if the chain of data remains unbroken and all departments have access to it in appropriate forms. Miss this point and you open the door to the probability of lost time and resources. Companies have to find reliable distribution and translation methods to reduce loss of data quality.
Before exploring the various distribution methods, companies have to determine exactly who needs design data in various departments. The answer may be obvious but careful examination might bring some surprises. The purchasing department may send out hard copy drawings for quotes or even for tooling. If so, the data will most likely be redrawn or recreated on a CNC machine somewhere with all the possiblities for simple human error. The sales department may be sending drawings to customers - make sure they are sending the correct versions and are tracking what they send. The minute a CAD model leaves drafting it may no longer agree with the design data so there is some definite potential for noncompliance with the released design unless they are properly controlled.
IGES (Initial Graphics Exchange Standard) is a graphics file format for 3-dimensional models. IGES files contain surfaces, so they are a natural choice for distributing the CAD data to anyone who does machining. However, IGES translations can be difficult and frustrating due to a variety of differences in the CAD products and the vagueness of definitions of the file format. Properly done, though, IGES translations do contain accurate model data and are compatible with most machining packages. Newer CMM and other digital inspection hardware are able to compare model data with acquired data points to generate reports and 3D color plots that visually display feature compliance.
STEP (Standard for the Exchange of Product Model Data) files, which contain non-parametric solid model data are smaller than IGES files and are usually easier to translate. Companies doing original designs for their customers can transmit design data back and forth in STEP format while retaining the engineering properties of the objects.
STL (Stereolithography) format is commonly used in rapid prototyping business.

Geometric Modeling

The core strength of all the CAD products is their design capability. All these products started out as B-REP (boundary representation) modelers. As they evolved, they developed more modules to better serve the need of customers. They developed modules for simulation and analysis, finite element analysis, computational fluid dynamics, tubing and piping, cable and harness, electric circuit design, and data management solutions.

Companies that deviated too much into other categories and lost focus on their core modeling are in big trouble. PRO-E can be cited as an example.

Usually customers use a surface modeling program like AliasStudio for conceptual design. Then they use a solid modeling program like CATIA for mechanical and manufacturing design. There are some programs out there that can do hybrid modeling ie. surface and solid modeling. CATIA has a surface modeling module. Their recent release of expensive Imagine and Shape module sounds very promising.

They have combined mesh modeling into the manufacturing design process that is pretty innovative in my opinion. I can talk hours and hours of NURBS based modelers and subdivision surface modelers. Imagine and Shape is the first product out there that combines the best of both worlds.

I have to admit I did not have a chance to try the program myself to write any review. But the idea sounds very interesting and as it evolves more, I see other companies do the same after this technology gets proven.

Interoperability between CAD products

A long standing problem in the CAD industry is the interoperability issue between CAD products. Usually users cannot do everything in just one product. They start the conceptual design in one product (say, in AliasStudio), take the data to a solid modeler like Autodesk Inventor, finish up the engineering design, take it to CATIA for more design and some simulation analysis, etc.

When data gets translated between systems, they lose their integrity with every data translation. Many products support industry standard neutral formats such as IGES or STEP, CAD vendors also support direct translators.

For example, PRO-Engineer can read CATIA files directly. However, the native file formats are not open...meaning, the file format is not published. In order to read CATIA files, pro-e has to reverse engineer CATIA's file format.

Autodesk's most popular format is DWG and it is not open. It is one of the most reverse engineered formats. http://www.opendwg.com is a forum just to reverse engineer dwg file format.

CAD Products

Manufacturing CAD (MCAD for short) is a growing market due to explosive economic growth in the BRIC nations (Brazil, Russia, India, China).

Dessault, Parametric tecch (PTC), Unigraphics (UG) and Autodesk are the market makers in this industry.

Dessault develops CATIA (high end MCAD product)
PTC develops Pro-Engineer
UG develops UG-NX
Autodesk develops Inventor (mid-range MCAD product).

Other mid-range products include Solidworks, SolidEDGE, etc. (smaller companies like Think3, Alibre exist too).

Autodesk Inventor shows tremendous potential to win over Solidworks and SolidEDGE and become the leader in mid-range CAD products. During the course, it looks like to win over high end products like PRO-E and CATIA. Autodesk's billion dollar cash reserve and its powerful marketing muscle may acquire good talent and propel Inventor to new heights.

Given that ADSK at $39 per share looks like a good buy. I personally think it may reach $50 in a year.