Fahmi replied to he topic: Using a hammer instead of pliers

In order to create a good microstrip port in HFSS, please follow the following steps:
1- If you have the actual dimensions of the air cavity above the substrate then use it otherwise create a cavity that is at least 4-5 times the substrate thickness
2- Create a rectangle that is centered at the trace (width wise) and that extends from the ground plane up to either the top edge of the actual cavity or to 4-5 times the substrate thickness. The width of that rectangle should be either the actual width of the actual substrate , or at least 5 times the width of the metallic trace.
3-Assign a wave port to that rectangle , and make sure to define an integration line from the bottom of the trace to the ground plane, this way you are telling HFSS to calculate the port impedance along that integration line which happens to have the maximum E-filed.
I ran the example with the dimensions you provided and the port impedance was calculated as 46.45 Ohms, in 9-11 GHz range.
I will have to dig an old presentation by Ansoft (the previous owner of HFSS ) which details that, if I find it I will post it to the thread.

Microwaves101 replied to the topic: Using a hammer instead of pliers

Fahmi

Thanks for your thoughts! Can you tell me why HFSS comes up with 46.5 ohms for a line that is known to be 50 ohms?

I know very little about HFSS... but could a "lumped port" be useful in this case?

Steve

Hadrien F4INX replied to the topic: Using a hammer instead of pliers

Steve,

The lower impedance you get compared to what it should be is a telltale sign your port box is both 1/ too small and 2/ electrically shielded.

It is because the simulation method puts a box around the port, which is connected to ground, which reduces the impedance.

About the other tools: Momentum is cool for standard microstrip circuits but is awfully complex to use when you want to do fully custom parametrized layouts, both Sonnet and Momentum are slow when your metal filling factor is high (e.g. ground planes) due to the simulation method. Don't know Axiem.

Hadrien

Fahmi replied to the topic: Using a hammer instead of pliers

Steve,
I played a bit with the model, and gave it a bit more thought, the following factors may be at play here
1- the trace thickness, I used 10 um , changing that will affect the impedance
2- the air box I used was a bit small increasing its height will increase the impedance,
I worked a lot with 15 mil alumina so 4 or 5 times that is about 75 mils much more than 5 times the
100 um GaAs substrate, which makes the box small as Hadrian pointed out, given that in air the
wavelength is the same, so my previous statementabout the box height need to be revised.
3-I also did not use radiation boundaries, again due to custom, the structures I
worked with were always housed in cavities.
As for lumped ports, I did not use them before, coming from the waveguide world, I always used wave ports,.
what I can tell you is that the simulations using wave ports, gave very good results (both in waveguide and microstrp) compared to measured results for filters which are finicky to start with .
Thanks Hadrien for the hints.

Hadrien F4INX replied to the topic: Using a hammer instead of pliers

Fahmi wrote: which makes the box small as Hadrian pointed out, given that in air the
wavelength is the same, so my previous statementabout the box height need to be revised.


When I spoke about box size, it is not compared to the wavelength but to the substrate thickness. The conclusion would be the same for lower frequencies.

Fahmi wrote: 3-I also did not use radiation boundaries, again due to custom, the structures I
worked with were always housed in cavities.


For people using radiation boundaries, be careful: most often the simulator will use electric (metallic) or magnetic boundaries for the port calculation, even if radiation boundary conditions are used. And radiation boundaries don't like to be too close to the structure.

Hadrien

Fahmi replied to the topic: Using a hammer instead of pliers

What I meant was that increasing the box size will in essence allow a better capture of the field distribution outside of the substrate, the field decays exponentially and one wants to allow the fields to decay enough such that the outer box boundaries do not disturb the field too much, the more wavelengths in air the better, and yes the port boundaries in this case are PEC, so the port size will have to extend enough above the substrate for the fields to decay. I recall Sonnet recommended something like 10 times the substrate thickness.
The safest bet is to see the actual housing where the circuit will be housed in and use those dimensions, as this will mimic the actual operating conditions.
And for the original question from Steve regarding the seemingly erratic calculations, the use of integration lines should take care of that and the results should be consistent.

Hadrien F4INX replied to the topic: Using a hammer instead of pliers

The field decays exponentially, but wavelength has not so much influence in that process. The frequency dependence here is really low.

In the original question from Steve, I think that the main cause was the box influence. Integration line ambiguity has a much lower effect on microstrip lines sufficiently small.

Fahmi replied to the topic: Using a hammer instead of pliers

You are right, the box and the port surface just have to be large enough (in absolute terms not in frequency dependent fashion) for sufficient decay of the fields, but then I believe care must be taken when working with really thin substrates, as in absolute terms , the multiplier of substrate thickness for the space needed above the substrate for the fields to decay appreciably, will be different, say if we consider a 25 mil alumina vs. 100 micron GaAs substrates.

Hadrien F4INX replied to the topic: Using a hammer instead of pliers

The additional spacing changes with thickness, but the multiplying factor does not. What influences the multiplying ratio are the ratio w/h (microstrip line over substrate thickness): larger microstrip lines will need lower multiplying factors, and dielectric constant: substrate with lower dielectric constants will need more spacing.

I typically use 5 for this ratio, but I already had to use at least 10 in some simulations where I needed high accuracy on S11.

Pro tip: always look at the mode patterns. When there is field on the port boundary, it is a sign you need to increase your box. If you want to be really sure, start from 5 and increase the value until impedance does not change anymore. In all cases, be careful not to introduce higher order modes by using too high spacings. I remember a tricky simulation where I had to reduce a lot the spacing to remove this effect.