CST Studio Suite: Domain Decomposition solver vs. Time Domain Solver
When designing and simulating periodical structures, like arrays, engineer wants to have reliable tool, that doesn’t require hours of setting up. Simulation time plays significant role here as well – when working with arrays we have structures much bigger that wavelength with millions of mesh cells.
Recently I have been actively using Domain Decomposition Solver (DDM) in CST Studio Suite. In this short article I would like to share my opinion on it, provide comparison with Time Domain Solver and draw some general conclusions about the new solver.
What is Domain Decomposition Solver?
DDM is essentially Frequency Domain Solver, that we have been using in CST for a long time. The difference is in meshing, as it fractions the structure into smaller blocks – domains – and adapts meshing for each domain.
Fig. 1 Cavity array with domains view.
On figure 1 you can see 2 colors of domains: orange color means base blocks, that fill evenly boundary box; pink color means element blocks, that contain an array element inside and have first priority.
Why need for an alternative solver?
Modern mm-wave applications often require big number of antenna elements in arrays. For example, some RADAR arrays consist of hundreds of elements, which makes simulation model heavy for simulation.
Figure 2. Dual-polarized cavity antenna array.
Let’s look at simple cavity mm-wave antenna array, that works at 24-32GHz range. It is 4x4 array, that gives us 16 elements in total, or 16 ports. Most of mm-wave application require dual-polarization, which multiplies number of ports by 2, resulting in 32.
Time Domain Solver calculates each port sequentially. It means that the more ports you have, the longer simulation time you get. When count goes on hundreds, it can result in days of simulation.
On the contrary, simulation time with Frequency Domain Solver is independent from number of ports in the structure. It comes with a price, of course. Direct Frequency Solver require big amount of RAM for managing big simulation, and sometimes you just do not have enough to work with.
领英推荐
By dividing structure into smaller domains DDM solver gives a good simulation speedup, as domains can be solved in parallel, depending on the number of cores. If your structure contains geometrically identical elements – like antenna arrays – DDM solver allows setting up of repetitions.
Good news – with DDM solver the peak memory relates to the number and size of domains per core. By making the domains smaller the peak memory can also be lowered.
Let’s compare
Hardware in use: 2x18 cores, 128Gb RAM
Fig. 3 Radiation efficiency comparison
Fig. 4 Radiation pattern comparison
In conclusion
Here are some pros and cons that I have noticed while working with Domain Decomposition Solver:
As always, each time engineer should decide which tool to use in order to achieve better results. Every option comes with a tradeoff. In my opinion, DDM solver is a good way of simulating periodic structure, which allows to use full potential of the solver.
R&D Eng @ EDF
2 年great job !!
Nice article !
Lead Engineer
2 年Thanks for your article ;)
RF antenna engineer
2 年Hi Katerina, Nice article. I work with large antenna arrays and this is the exact problem which I struggle with everyday: Time Domain (TD) vs Frequency Domain (FD) solvers in CST. I have a couple of questions to you to know more about CST DDM. 1) Is DDM a new feature in CST? I use CST2019 and never seen DDM. 2) How good is this new DDM solver when it comes to S-parameters simulations? In my experience, FD is very good for both radiation as well as S-parameters simulations. However, it is not always possible to use FD solver for large arrays due to memory limitations. That is why I often find myself using TD for large array. In the TD solver one need to improve the mesh setup a little bit (from defult setup) to achieve similar accuracy for radiation results. However, to achive the same S-parameters results as FD, one need to significantly increase mesh setup. This makes simulation time extremely long specially for antenna arrays where you would need to simulate the same array multiple times (depending to number of ports). A solver as accurate and fast as CST FD but which takes significantly less memory than traditionally FD solver would be very exciting for me.
Guiding robotics innovation from R&D to market success | logistics | automation | integration
2 年Katerina G. super interesting article! Thank you. Would like to cooperate with you in our project, if it is possible. :)