CST MWS - Simulating CBCPW 50ohm line

[Geoffrey_85]

Hello everyone:

I have an issue with understanding which set of S-parameters should I use from CST MWS simulation of a CBCPW 50 ohm line. I come from an ADS background (both Momentum as well as Schematic), and CST is fairly new to me.

I want to test a CBCPW 50 ohm line up to 300GHz. I saw that by default in the transient solver, in Specials, there is an Option AR-Filter which is set to on. When finishing the simulation, in 1D results pallete I don't have only S-Parameters block but as well S-Parameters(AR). Which is the difference between the two, as the results are not exactly identical (very similar but not identical)? Which of the 2 should I consider? S-parameters or S-parameters (AR).

I want to mention that I use GaAs, 100um thickness. with lateral top GND width of 100um to avoid as much as possible resonant patch-like behaviour. The waveports are defined with lateral dimensions to be on the lower limit, such that higher order modes don't get excited.

Any feedback on which set of S-parameters should I consider (S-Parameters vs. S-Parameters(AR)) is greatly appreciated!
Thanks in advance to all,

George.

 

[Geoffrey_85]

Can anyone provide any feedback on the topic?
Thank you!

 

[johnjoe]

The AR-Filter (Auto regression) is mainly used for high resonant structures to save simulation time. The problem of AR-Filter is, you have to set it correctly, otherwise the results would be wrong. If you don't need, deactivate the filter and use normal S-parameters. Or if you don't want to rerun the sim, use the s-paramters without the AR and brackets!

 

[Geoffrey_85]

Thank you JohnJoe:

Meanwhile looking for a feedback on the topic I reasearched more and found that, indeed the Ar filter is used when having resonant structures. And probably I have to use it since I am simulating CBCPW at very high frequencies (up to 300GHz), where finite ground widths can transform into patch like antennas and resonate. I know that best way is to use VIAs to shortcircuit parallel plate mode, but I am doing some trials in order to evaluate the ground width. I am trying to optimize the top GND width such that the first higher order mode has the cut-off frequency very close or beyond my highest frequency of interest.
The result curves I have in S-params are typical for a 50 Ohms T-line, with the difference that these results vs. S-params (AR) have some ripples (like small oscillations), and from what I was able to document myself this is cause of some resonances taking place. And it's exactly what's happening with ground widths comparable with Lambda/2 at the operation frequency. Looking at S-params(AR), they are ripple free, and I believe this is the more accurate frequency response!?! Please correct me if I am wrong.

Another aspect is the "Steady state criterion". In some cases (when the finite GND widths are comparable with lambda/2) the criterion is not met, but in the end I managed to avoid this warning and reach the "Steady state criterion" threshold by increasing the number of pulses in the Steady State window from menu Specials in Transient solver (default is 20 pulses, I raised it to 60 such that energy balance occurs and steady state criterion is reached).

If anything of what I said is wrong please correct me, as I am new to CST! Wainting for a feedback.
Many thanks,

George.

 

[johnjoe]

Sounds ok, but be careful the AR-Filter it is a signal processing procedure which is performed AFTER the actual simulation. You can also increase in "Time Domain Setup" the Accuracy to e.g. -40dB or even more to avoid the ripples. But depending on your simulation it can increase the simulation time tremendously. You can also watch your signals in "1D-Results" to see the amount of ripples.

 

[Geoffrey_85]

Sounds ok, but be careful the AR-Filter it is a signal processing procedure which is performed AFTER the actual simulation. You can also increase in "Time Domain Setup" the Accuracy to e.g. -40dB or even more to avoid the ripples. But depending on your simulation it can increase the simulation time tremendously. You can also watch your signals in "1D-Results" to see the amount of ripples.

Thanks for replying. I forgot to mention in the other post that besides increasing the number of pulses I also increased the AR accuracy to -40dB when performing the simulations. I was also keeping an eye on port signales in 1D (o1,1 and o2,1) to see for ripples, and at the begining of time lapse you can see some small oscillations but as time increases they dissapear and the signals flatten.

Many thanks JohnJoe!!!

 

[johnjoe]

Just one thing: The accuracy in Time Domain Setup, first page, has nothing to do with AR-Filter.

 

[Geoffrey_85]

Just one thing: The accuracy in Time Domain Setup, first page, has nothing to do with AR-Filter.

My mistake, thank you for the correction !