OPenEMS for 2.4 Ghz MIFA

[davethomaspilot]

I thought a tool to simulate a 2.5 Ghz pcb antenna would be unaffordable.  But then I stumbled on OpenEms--and open source electromagnetic field solver!

But, seems too good to be true.

Have people used it?  Can I expect reasonable results when evaluating pcb antenna (and associated ground plane) alternatives? 

Or are there too many parasitics that can not be quantified for an accurate simulation?

Thanks

[TheUnnamedNewbie]

it's an FDTD simulator, which has it's pros and cons. I've never used it but FDTD is a quite well understood method (it's a very common project for master students in EM courses to have to write their own 2D FDTD solver). Makes sense that there are open-source simulators out there.

I don't quite understand what you mean with 'parasitics that can not be quantified'?

[davethomaspilot]

So a pcb design can provide trace and ground plane geometries, but that's about it.

I can get approximate pcb thickness, trace thickness, and dielectric.  But, I'm worried there are variables that can impact the simulation results too much to be useful.

Maybe parasitics in the matching network.  I doubt I can find caps, inductors, and resistors with models for parasitics that are accurate at 2.4 Ghz. 

For example, I've read the right cap is better for a short at 2.4 Ghz than an zero ohm resistor.  Also, that 0402 are recommended over 0603, due to their smaller size (and hence lower parasitics).

So, if 0603 versus 0402 makes an impact, what else isn't modelled but important? 

[TheUnnamedNewbie]

Maybe parasitics in the matching network.  I doubt I can find caps, inductors, and resistors with models for parasitics that are accurate at 2.4 Ghz. 

For example, I've read the right cap is better for a short at 2.4 Ghz than an zero ohm resistor.  Also, that 0402 are recommended over 0603, due to their smaller size (and hence lower parasitics).

So, if 0603 versus 0402 makes an impact, what else isn't modelled but important? 

A number of capacitor manufacturers can provide you sparameter files of their devices, I have used them in the past in Keysight ADS up to 10 GHz or so. You have pad capacitance, device inductance, device resistance that all play a role (though it's mostly the fact that your caps have a lot of parasitic inductance that is a problem at 2.4GHz).

What is usually a bigger problem is that beginner antenna designers don't take into account the housing the PCB is placed into, not realizing that that housing (esp if it has a good bit of metal) might completely detune their antenna.

[davethomaspilot]

In my case, I'm worried about a metal can that encloses the pcb mounted rf transceiver.

I don't know how far the spacing from the antenna to the metal shield needs to be.  I can make the spacing larger, but at the expense of the radiation pattern being oriented non optimal.

So, I'd like to know:

1)  Is the spacing to "metal can" going to be an impact, other than what can be tuned in the pi network?
2)  Will orienting the antenna for more spacing help more than it will hurt due to radiation pattern orientation?
3)  Can I design a meander that utilizes the space available on the pcb and get better performance than MIFA designs I see on the web?
4)  Will a chip antenna perform better, given the relatively small ground plane of the pcb?

Would simulation (using OpenEMS)  be likely to help answer these questions?  Or, will it all have to done empirically anyway?  I.E, I cold just send gerber for multiple pcb designs and test with real hardware.  Not a prayer of #3 doing that...

[hagster]

I use openEMS all the time and would say it's definitely a helpful tool for your task. It takes a little while to get used to, but you can get the tutorial examples up and running in no time. Some advice...

When you design your PCB, make sure you have a nice solid ground and power plane. Use 4 layers if you need to. This will make it simpler to simulate as you can just model the PCB as a solid PEC sheets. You can add basic PEC blocks for cans and other components, but I doubt they make much difference if separated slightly from the IFA.

Enclosures can easily detune the antenna. Once you have the basic OpenEMS model running, adding some simplified version of the enclosure is a worthwhile step.

Creating the mesh is everything in OpenEMS. Follow the guides and rules of thumb.

If you plan on using a matching network, you don't need S11 to be perfect, but the closer to 50ohms you get the easier and wider band the matching network will be.

Don't try to model the matching network. This is better and faster in another package. I often use QUCS for this. You can export the S11 from OpenEMS and import that as a Touchstone file into QUCS to simulate the quality of your match .

[davethomaspilot]

Thanks for the advice!

I dug into openEMS.  It was more time-consuming to get reasonable results than I expected, but I'm now able to reproduce simulation results I see in online app notes.

I'm glad I bothered to do simulations!  I was really surprised how sensitive results were to things like where the feed element egressed from the ground plane. 

I went down a "rabbit hole" I'd advise others to avoid.  I spent time trying to get antenna layouts done in Kicad into openEMS.  I got a decent workflow from that, but I soon got tired of drawing slightly varying MIFA designs.

I found it much better to do a parameterized design.  It's really pretty simple for pcb antennas.  I used excel to define a polygon with a scattergram that showed the poly as it was being "drawn" by entering formulas in cells.  Then I created a simple script to grab the xy coordinates and create the openEMS commands required to generate the polygon in that environment.

With a parameterized antenna geometry and (parameterized) placement, one can quickly gain insight into important considerations.

FWIW, I would avoid drawing in Kicad.

Soon, l plan to export from Kicad to HyperLynx for the actual pcb design.  But, given the simulation time and attention to detail required to getting a mesh designed in openEMS, it might be quite an effort.

[davethomaspilot]

Don't try to model the matching network.

I think you mean, don't use openEMS to determine component values for a matching network?  But still include the pads and LumpedElements in the openEMS simulation once you have determined optimal component values?

Thanks!