One thing that really hooks me as a trade journalist is when a developer or manufacturer answers technical questions on any level I want to go to. Over the years, I've found that Comsol really believes in educating the user -- and even the journalist! Yesterday, I installed a trial version of the new Comsol Multiphysics package – Version 4.1 -- and took an "optimization" minicourse taught by John Dunec, of the company's Palo Alto, Calif., office. He showed us how to set-up and solve a microwave oven circulator problem. A circulator in a microwave oven is a passive device comprising several ports (in our problem there were three ports, offset from one to the other by 120 degrees) with an iron post in the area where the ports connect. Microwaves (electromagnetic waves with frequencies between 0.3 and 300 GHz) enter a port and are transmitted to the next port in rotation. The problem is that the microwaves have to hit the iron post such that they don't bounce back and ruin the emitter. We are interested in finding out two things. One: What is the optimum material property of a transition piece (in front of the iron pole) such that the waves are directed to the next port in rotation and do not bounce back? And two: What is the optimal size of the iron post? Basically, the design objective was to minimize the waves coming back to the emitter.
We first built the model in Comsol Multiphysics 4.1. This was simple to do. You select the appropriate shape, in this case a rectangle. Place the first rectangle by inputing the x and y coordinates. Then left-click on it to select; right click to confirm, enter an angle of rotation, and press the Build button to place the next rectangle. Objects are always left-clicked first (they turn red); then confirmed by right-clicking (they turn blue). After all the rectangles are placed, we performed a Boolean operation to join the ports and placed a circle (which represents the iron post) in the center. Boundary conditions were considered "perfect electric conductors." Besides that, most of the model comprised air and was defined by parameters of an anisotropic material.
The interface of 4.0 is quite elegant, clean, and uncluttered. I found it fairly easy to follow John's instructions and successfully build and then optimize the model by running a parameter sweep. The screenshot shows results. As you can see, the waves hit the pole and nicely turn right, thus fulfilling the design objective.
Version 4.1 was just released for mass production yesterday. Upgrades will be available in a few weeks. 4.1 is a complete rebuild of 3.5, from the ground up. Comsol decided to go this route so the program could reach a broader audience (many users were and are PhDs with over 10 years engineering experience; the company also wants to target engineers with Masters degrees) and it wanted the code to be able to support future developments. These include so-called “sloppy meshing,” which basically defeatures models in an automated fashion inside the software and a .NET implementation which will give it added capabilities re Windows. 4.1 also has put the equations back so users can easily see them -- no "black box" whatsoever.