Question about EMDS low frequency limit

I've been looking into Agilent's new 3D solver - EMDS- aka "son of HFSS". There's a comment in their manual that the solver might become unstable at low frequencies, but nobody seems to know what that limit is. I do some antenna work down to 100MHz with planar structures under high dielectric superstrates - does anybody have any thoughts or experiences about EMDS misbehaving at 100 MHz? I ran their 30-day demo but didn't have a lot of time to explore this issue.

Hi norske,

I believe there is no fixed limit for this. The "breakdown" frequency for the matrix inversion depends on the actual matrix you are trying to solve and will therefore vary with the geometry. I have never worked with EMDS but I guess 100 MHz should O.K. especially if it is an antenna. Your structure will be still the size of the wavelength any you should not run into trouble.

Just wondering: did you also try other software? I guess the main advantage of EMDS is that is is cheap! So, if you have a very limited budget for 3DEM it might be a solution. However, If you look for more "advanced" software I would check the "usual suspects" . For what I have heard, the Interface and capabilities of EMDS are very similar to the old Agilent HFSS. I don't think they did much developments on this over the last few years.


F.

Hey RFSimulator,
That's encouraging, thank you very much. I have indeed tried a few others, but I'm as price-conscious as anyone else. It's difficult for example to justify HFSS when EMDS - a similar product - apparently does 80% of what HFSS does at a third of the price. For my application, the missing 20% isn't important.

Mefisto has a surprising amount of horsepower for its price. It's a little shy on documentation and examples (as pointed out elsewhere in this forum) and the output data usually needs a little spread-sheet-type post-processing to be useful. Defining input sources can be cumbersome for coaxial inputs (but simple for microstrip sources). There is no automatic or adaptive meshing so you need to be fully awake when you're putting the geometry together.

In my opinion, Concerto from Vector Fields is definitely a class act. It's almost twice the price of EMDS and three times Mefisto and if you can afford it, I'd say go for it. Productivity is one of the issues that sometimes gets overlooked in EM work, and in Concerto you can build and analyze pretty much any shape and kind of contraption you want - in a hurry.
Thanks again

Hi norske,

to be honest, I'm not sure if the 80% comparison is correct. Sure, for "small" models (some bond wires, small filter e.g., vias) EMDS will do the job. But for high end applications you need a more serious tool. All high end tools have improved a lot concenring the solver performance and the user friendliness over the last five years. I'm pretty sure that you can handle problems which are more then only 20% "bigger". What is you main application and what would you like to simulate. If you buy a "low cost" tool now and you figure out that you need a better solution in 6 month, paying a few 1000 S is still too expensive...

By the way: I'm missing CST in your list of evaluated software

Good evening, RFSimulator;
The 80% figure came from someone at Agilent.
My applications revolve around antenna design from 100 MHz to 2 GHz. Mostly they're planar but not always with a ground plane - esp at the lower frequencies. At 400 MHz and up it's mostly patch antennas of various kinds. There's a lot of activity with high-dielectric superstrates. The effect of thin dielectrics - conformal coating, eg - on high-Q antennas is a recurring issue.
Really sorry about CST - I'm certainly aware of the product but I've never used it. All I know about it is what I've seen in their ads and read in this forum. I'm guessing you think I should take a closer look.
thank you again for the insight.

Hi again

O.K. this explains it .

If you indeed do lots of antenna designs up to 2Ghz, I certainly would recommend that you have a look at CST. The nice think is, that it offers several solver technologies. They started with a time domain solver but added a frequency domain solver as well about 2 year ago. It s certainly more expensive but if I were you I would a least talk to them. I figured out some time ago, that the best way to test a software is to wait till you have a "real" project. Just combine the the evaluation time with the design time and run something real. This makes sure that you don't run into trouble later one. This also test the capabilities of the support they offer because they have to make sure that you do not do something "stupied" with your design when you simulate it.

.

Not even the Agilent application engineers? Strange!