S matrix in HFSS microstrip simulation

[arthury]

simulation of microstrip transmission line hfss

Hi everyone:
I am working on a minor study of microstrip simulation in HFSS.
I created a model of a microstrip line and add 2 waveports as the excitations .
In the results it shows the port Z0=59ohm@40GHz. the length of the microstrip is set to 2.5mm
The strange thing is about the S matrix data:
S11=-0.58+j0.34 and S21=0.092+j0.26
If i did not consider in a wrong way, S11 should be 0 and S21 should be 1, since the waveport is supposed to simulate the same microstrip line when it is connected tomy microstrip line. And the structure should have no reflections at the input and the output.

I saw some similar discussions on this site but still can not figure out where is the problem. Am I thinking wrongly? Why the S11 is a non-zero value



Btw: from the result I could also find the complex propagation constant Gamma. And according to my previous reading, the real part is the attenuation constant Alapha and the imaginary part is the propagation constant from which I could extract the effective dielectric constant of the transmission line. These two data match the results I get from another simple microstrip transmission line calculator: TXline.exe.

Thanks in advance

 

[arthury]

hfss microstrip line

anyone can help me out?
I checked pozar's book Microwave Engineering.
When the impedance of the feed line and the load is equal to Z0 of the microstrip.
The S matrix is like
S11=0 S12=exp(-betaZ)
S21=exp(-betaZ) S22=0

However in the HFSS simulation the S11 and S22 two non zero values.

 

[YoungEng]

microstrip transmission line calculator

hi there arthury, S11 will never be zero because no matter how well your ports are matched there will always be some small reflections and loss. Only mathematical equations will give you zero. Also make sure that your waveports are not renormalized, this will make it seem as if the ports are "perfectly" matched.

 

[arthury]

hfss microstrip

hi there arthury, S11 will never be zero because no matter how well your ports are matched there will always be some small reflections and loss. Only mathematical equations will give you zero. Also make sure that your waveports are not renormalized, this will make it seem as if the ports are "perfectly" matched.

Dear YoungEng
Thanks very much for your reply
I think you are right in the sense that S11 will never be zero due to minor reflections, but the S11 data I get is fluctuating in a broad bandwidth (0-80GHz)and the magnitude can be as large as 0.5.
Actually this morning I just did a very simple coaxial line simulation to debug the problem and seems that I got some clues.
This copper coaxial line has a larger size than the microstrip one I did before.
The result shows that the S11 is nearly zero (~0.01) in a very broad frequency range. And the magnitude of S21 is nearly 1 (~0.997) in the same frequency range.
This results makes some sense since the S matrix of a transmissionline which shares the same impedance as the feed and load will be
[0, Exp(gammaL)
Exp(gammaL), 0]

and the magnitude of Exp(gammaL) should be unity.

I guess my orignial strange results of the microstrip is possiblly due to the size of the structure or the mesh accuracy.

I will make further investigation on this problem and post some pics of the results.

 

[micky74]

non-zero result s21

There is definitely something wrong with your simulation. Even if you don't get perfect 0 for s11, it should still be nearly so. Why don't you post your file and then we can take a look at it?

 

[arthury]

hfss simulation

There is definitely something wrong with your simulation. Even if you don't get perfect 0 for s11, it should still be nearly so. Why don't you post your file and then we can take a look at it?

Thanks so much, Micky74!

I gathered some data of the simulation of Microstrip I did.

The model is contained in an air box with radiation boundary.

The two ports are waveports without re normalization. Then the physical picture is a source and a load which perfectly match the microstrip in the model.

Therefore S11 should be 0

The result of Z0 and effective dielectric constant can be verified by another simple 2D line calculator TXline2003.exe

The abnormal thing is the S-matrix data which is shown in the following pictures.

At 10GHz the magnitude of S11 is 0.69

and I also upload a frequency sweep up to 80GHz. S11 is fluctuating heavily.

This is definitely not the physical picture of a source/load matched transmission line.

 

[arthury]

hfs s;s

There is definitely something wrong with your simulation. Even if you don't get perfect 0 for s11, it should still be nearly so. Why don't you post your file and then we can take a look at it?

This is my HFSS file

 

[micky74]

hfss simulation of microstrip transmission line

I had a brief look at your simulation and found a few mistakes. Since you're simulating over such a broadband and your structure is so tiny, why don't you break your simulation into different bands? Your absorbing boundary is definitely not big enough at the lower frequencies, so effectively whatever that is radiated by the microstrip will be reflected back. Also, the length of the line will be so short at the lower frequencies that I advise putting more modes in case any evanescent modes are there. Your substrate layers below the ground and above the microstrip are totally unnecessary.

 

[arthury]

microstrip transmission line in hfss

I had a brief look at your simulation and found a few mistakes. Since you're simulating over such a broadband and your structure is so tiny, why don't you break your simulation into different bands? Your absorbing boundary is definitely not big enough at the lower frequencies, so effectively whatever that is radiated by the microstrip will be reflected back. Also, the length of the line will be so short at the lower frequencies that I advise putting more modes in case any evanescent modes are there. Your substrate layers below the ground and above the microstrip are totally unnecessary.

Yes, I also guess it's because of the size and the reflections.
Actually I did a coaxial cable model with larger size and the S11 problem does not appear.

Thanks so much!

However, this is only a minor experiment. I believe the problem can be solved

Here my main question is as following:

Since the port mode solution gives information complex propagation constant which has a real part and an imaginary part.

The real part (alpha) of the complex propagation constant should be the loss coeffcient of the mode.

Then the problem is can I use this alpha and the length of the transmission line to calculate the RF loss along the transmission line?

Another way is to extract the RF loss from the S-matrix (i.e.S21) by doing a 3d simulation of a transmission line .

My guess is that if there is no reflections and higher order mode problem, these two methods should give the same results of the RF loss along a transmission line.

Am I right?

 

[micky74]

hfss microstrip transmission line

That is right. Using alpha to calculate loss will give you a value very similar to what S21 gives. Provided no higher order modes...