What is the best EM software to purchase?

I know that this topic has been beaten to death, but I would like to know "what is the best EM software". Since this is very specific to the application, desired accuracy, cost, etc ... let me explain my situation. My company has two needs which, as far as I can tell, are not completely complementary. The first requirement is that we want to do IC package modeling of bump and wirebond leads for SER/DES I/Os from 1.25Gbps to 10Gbps+. Accuracy here is importance, but a couple percent error is acceptable (<5%). My impression is that a simulator like HFSS, Fidelity, Feko or CST MS would be best for this. The other requirement is modeling of planar IC inductors at frequencies from 2GHz-10GHz+. Accuracy here is critical. It seems like simulators like Momentum, IE3D or Sonnet would be best for this. Assume for now that cost is not the limiting requirement, is there a single simulator or package that can handle both needs or do we need to look at two solutions from two different vendors?

I appreciate any help on this.

Hi, Green:

IE3D should be excellent for both IC and packaging. IE3D has the Automatic Geometry to IE3D Flow (AGIF). AGIF is integrated into Cadence Virtuoso for IC designs, AWR for microwave and planar circuit designs. It can also convert Cadence Allegro and APD files (.BRD and .MCM) automatically into IE3D models including wire bonds, bumps, vias, thick traces and ports. The converted IE3D model is ready for simulation. I am from Zeland (producer of IE3D). Thanks!

If you want an unbiased (non-vendor) view, then for your application (which is similar to work I have done before) would be a combination of sonnet and CST/HFSS.

certainly for spiral inductor type structures sonnet proved to be the quickest and closest to our measurement.

as for a 3D tool, we're split between CST and HFSS, both do the job well, although I personally prefer the CST user interface. Also you need to look at time domain versus frequency domain. If you have many ports then TD requires a run per port, however FD needs careful choice of mesh/sampling frequency.

good luck

Instead of Sonnet, you might consider getting Ansoft Designer, another program using a MoM solver.

In my opinion, combination of Ansoft HFSS and Designer would suit you well.

As they are widely used in the EDA Comunity, you should be able to calibrate the models using a varitey of tutorial samples.

:D

Thanks for the input from everyone. We are considering packages from Ansoft, Sonnet and Zeland right now. I understand the offerings from Sonnet and Zeland but I am a bit confused about the Ansoft packages. Sonnet and Zeland have separate tools for 3D structures (packages, etc...) and planar structures (inductors) and the analysis and viewing tools are built in. Ansoft seems to only have a 3D tool and all the viewers and analysis tools are add ons (reminds me of Cadence). HFSS is a 3D FEM solver that can be used for package and SI analysis, but I do not see which tool would be used for planar structures (i.e. IC inductors). Fantomf4 recommended the combination of HFSS and Ansoft Designer, but as far as I can tell Ansoft Designer is just a front end tool. Is this right? Or do you use HFSS as the simulation engine for all structures and then do the analysis with Ansoft Designer? I may have my facts wrong, but I did not think that tools like HFSS that use the FEM were appropriate for planar structures.

Thanks again for the help. I am also going to pose these questions to the Ansoft sales rep today, but I figured I would ask here for an unbiased opinion from people who use the tools.

Green Goblin

Well, you probably got an answer from the sales rep already, but I'll try anway.

I've used both HFSS and Designer, but I'm not an expert on either. After reading reading this forum, I've discovered we should have been using Designer for some of the situations we used HFSS for.

In my non-expert understanding, Designer and HFSS are independent products. Designer is a planar EM solver based on MoM, whereas HFSS uses FEM. Designer and HFSS can be linked for co-simulation, but this is optional. Designer is best for circuit layout and planar antennas, and HFSS is good for 3D structures.

Good luck. I've got to call some sales reps myself soon.

You are correct. You'll get much better accuracy and simulation time using a 2D simulator for a 2D problem.

However, even for 2D MOM tools there are generally two types:

Closed boundry MOM - Sonnet, AWR EMSight - Solved with FFTs, gridded, strength is high dynamic range. Good for filters, inductors, couplers.

Open Boundry MOM- AWR Axiem (So new I have not even demoed it), Agilent Momentum, Ansoft 2D (I believe) - gridless, intergal solver, good for antennas and highly radiating structures (especially off the edge)

Here is a stripline benchmark I did a while back. BTW CST does very well but I have not released that yet.

http://www.rfpoweramp.com/papers/str..._benchmark.xls

You should read this book before commiting to a package:

Microwave Circuit Modeling Using Electromagnetic Field Simulation

http://www.artechhouse.com/default.a...nent=NO&State=

IMHO for your application you need two simulators. Closed form MOM for inductors and CST time domain 3D for packages (sicne packages are non-resonant). However maybe you should demo CST and IE3D for your packages.