Modelling transmission lines as Capacitors

[ZincBear]

transmission line capacitor modeling

Hi there!

I am using some pretty fat transmission lines as capacitors for an input matching network. I have previously used ceramic caps, which are accurate, but are extremely sensitive to temperature, and have no room for tuning.

In a crude manner, we can simply use C=ε0εrA/D, and use an area that would give us the respective capacitance. However, this is not accurate, as the patch becomes wider(see word document). This is because the FET would see the wide patch as 2 open circuit stubs instead.... so its capacitance will deviate from the actual calculated figure from the above formula.

Has anyone have got a good idea on modelling such a fat transmission line as a capacitor?....


Thank you so very much for your help!

 

[biff44]

transmission line capacitor

If it is a low enough frequency, it looks like a lumped capacitance. If it is a high enough frequency, it looks like a multimoded blob of stuff that needs an electromagnetic simulation program (sonnet, etc) to accurately model. If you cut down the width and made the length longer, then it might look enough like a piece of microstrip. That is why most microwave engineers look funny at structures that have such a squarish aspect ratio.

 

[ZincBear]

room temperature

I can't agree further.

However, this structure is designed in such a way because of space constraints(which is always the case for odd-looking structures). Frankly, i would prefer to use a longer line rather than a fat line, since its easier to simulate.

Additionally, this matching circuit is for 8GHz....therefore, i simply cannot use a single lumped-element capacitor to model it....

I have tried with ADS momentum, but its only a 2D simulator. Moreover, im not sure if i have declared the ports correctly to provide an accurate simulation. Im really hesitant in using a 3d simulator, like HFSS which i am not exactly familiar with....hence, im seeking advice from experts in this forum...

 

[Joe_User]

lines on modelling

ZincBear,

I looked at your drawing, and Biff44 is right. You have to be really careful about assuming that this acts like a transmission line or a parallel plate cap. You did not mention the dimensions or the frequency so it is really hard to give any advice. But, not knowing any of the details I would suggest two approaches and you can decide which way to go.

The first way is the cheap-and-dirty way which involves using a planar em simulator like Sonnet. I am guessing you could do this in the free version of Sonnet Lite. You'll have to approximate each wirebond with a via that goes straight up and then bends 90 degrees into a flat transition line. If you take this approach, I would model the first cap (or pad) separately from the 2nd which gives you maximum flexibility for tweaking the design if it doesn't work that well (Sonnet Lite has a way of sweeping either the width or length of the cap).

The second way would be to use a volume-meshing EM solver like HFSS or CST's MWS. If you choose this method, I suggest modeling the entire structure (including all those squares of metal) as one simulation. This gives the most accurate result but is not as easy to tweak the design if it doesn't work very well on your first try.

If you are new to EM simulation, I suggest downloading Sonnet Lite (https://www.sonnetsoftware.com/lite) and going through the tutorial. It's a really nice way to get gently introduced into the EM world. If you are an EM expert, I would suggest both approaches mentioned above.

That's my two-cents worth.

Joe

 

[ZincBear]

modeling capacitors in ads

Oops! I should have included more details about the patch. It is 35milsX75mils, on a 5mil thick sub, and a permittivity of 40.

With regards to Sonnet's planar EM simulation, i do believe that is also a 2D simulator? I did not really consider that because ADS momentum or MWO's EM simulator should be largely similar.

ADS allows users to declare ports as "Internal" ports (See attached file). But im not sure if its correct for me to declare it so.... argh. There are so many ways to go about it, but with time constraints, i feel so messed up... LOL...

 

[Joe_User]

modeling transmission line in sonnet

With regards to Sonnet's planar EM simulation, i do believe that is also a 2D simulator?

Sonnet, ADS Momentum, and MWO's Emsight are "3D Planar" simulators (fields and currents are modeled in all 3 dimensions, but dielectrics must be planar).

I did not really consider that because ADS momentum or MWO's EM simulator should be largely similar.

I am an avid Sonnet user and don't know much about momentum or MWO. That's why I mentioned Sonnet. Also Sonnet offers a free version called Sonnet Lite which might do the trick in this case. Sonnet Lite is exactly the same as Sonnet Pro but with certain features disabled and it is limited to a certain size problem. I suggested Sonnet lite because it is easy to learn (especially for an EM novice), has an interface to both ADS and MWO, and if you get stuck, their support staff is very helpful.

ADS allows users to declare ports as "Internal" ports ...

I personally wouldn't do it this way because I would be worried that the capacitance would be a function of how you feed it. This is why I suggested (in previous post) modeling the bondwires in the EM simulation. Ths means you put the port at the end of the bond wire. This eliminates the need to use internal ports.
BTW, if you ever use internal ports in ANY EM tool, make sure you know how they are defined. If the documentation doesn't tell you, then call up the vendor and ask! You can get really bad results if you don't know the underlying definition of your internal ports.

Joe

 

[jallem]

difference between cap and transmission line

In reality, SONNET is a 2.5 dimension solver. And
I think the others two too but no sure.

 

[Joe_User]

modelling the wires transmission line

If anyone is interested, see a previous thread called "Difference between 2.5D and 3D simulation tools?" and you'll see what I mean. Pay particular attention to the posting by Dr. Rautio, most likely the one who first used the term "2.5D" in the context of EM simulation tools. He even attaches a MTT-S Newsletter PDF called "Some Comments on Electromagnetic Dimensionality".

But be careful -- most people tend to use these terms quite loosely, so you really need to ask what they mean when they say 2.5D or 3D to avoid confusion.

Joe

 

[madengr]

parallel-plate transmission line, ads

I have done this exact thing; chip and wire PA discrete FETs and thin-flim caps. I used HFSS to model the whole matching network including the bond wires. The bond wire models compared well with measurements however getting the actual capacitance of the cap material required tweaking the Er in the simulation until the simulated response matched the measured. These were custom made binary-pad caps high Er substrates. The final prototypes still required bonding out to tuning pads to achieve the 1 ohm match.

Since you are using a stitch bond maybe just go with 1 port instead of 2, then place the network in shunt. I also ended up with a stitch bond since breaking the tail bond on the cap would pull off the metalization. Maybe you won't have this problem if using Alumina. The problem with the stitch bond is you can't vary the length of your bonds but I gave up on doing that anyway since the bondwire guy had a hard enough time keeping the wire profile repeatable.

I agree you need to model the bondwires too, even as several layers of via structures in 2.5D simulator. The best thing woulds be to do it in 2.5D and 3D, comparing the results. The 2.5D won't catch the fringing capacitance on the right edge of you main plate, however if using Sonnet-Pro you could use a dielectric brick.

If you don't stitch bond and the wires go to the edge of the main pad, you could use regular edge ports and de-embed up to the main pad.

For this type of problem I would not be hell-bent on achieving accurate results with a simulator, rather use the simulator as a guide to tell you what to tweak to tune the real circuit under VNA measurement. I didn't even attach a FET until I had the matching networks tuned where I wanted them. I made up some CPW probe to microstrip transitions and calibration substrates similar to those sold by Jmicro.