HFSS problem: how to simulate coupled striplines

[godsound]

coupled stripline hfss

Hi,

I am a newbie of HFSS. I wish to simulate a pair of coupled stripline, and extract its even- and odd- mode character impedances and relative dielectric constants. how can i set up the simulation in HFSS?

Thanks!

 

[rautio]

hfss parallel coupled striplines

Hi godsound -- You can do EM analysis for stripline if you want to check the EM analysis. However, it is much easier to use the exact analytical solution for both single and coupled stripline. Eeff = Erel for stripline because there is only one dielectric.

Unzip the attached file and you will have an Excel spread sheet that gives you the exact Zo, both even and odd, as well as the exact S-parameters you can expect for a given length of line. The EM analysis will of course give you different results. The difference is error. The spread sheet calculates all answers to 8 significant digits, which is what I call "exact" above.

For planar circuits you are usually much better off using a planar tool rather than a volume meshing tool. I recommend Agilent Momentun for unshielded work, and Sonnet for shielded work, espeically when the highest accuracy is required. I work for Sonnet.

 

[godsound]

sonnet tutorial site:edaboard.com

Hi rautio

Thank you so much! Your advice and attached file enormously helped me.

I got one question for the Eeff. Though there is only one dielectric for stripline, there is air above lines. So should I consider there are two dielectrics-air and substrate, and calculate the Eeff through some anayltical equations?

Besides, I have looked a bunch of papers on coupled striplines, but failed to find any give the complete view of relations amony physical dimensions (Width, Spacing, Thickness, Length) and characteristic parameters (Zo, Eeff for odd- and even-mode, etc.), which like in your attached file. Would you please recommend some literatures to me on this aspect?

Furthermore, I start learning sonnet now, and got a question - Could I set the exact values of the components' sizes and locations? like set one metal box is 0.25mm away from another one.

Thanks again.

 

[rautio]

hfss coupled stripline

Hi godsound -- Stripline has the same dielectric on top and on bottom. There is no air. Both dielectrics are the same thickness and both have ground plane on the outside. It is important to make sure the ground planes are well shorted together.

If you have substrate below and air above, you have microstrip. If you have a top cover, then it is shielded microstrip. There is no exact solution for microstrip. I think there are synthesis equations (i.e., given desired Zo, calculate dimensions) and analysis equations (given dimensions, what is Zo, Eeff) in a book by Gupta, Garg, and Chadha, title is something about computer aided design. But that is kind of old, there are probably more recent equations.

In Sonnet, just draw your circuit, add ports. specify dielectric layers and analysis frequencies, then analyze. Go to Help->Tutorial. In about 45 minutes you will be almost an expert user. When you start Sonnet, there is also a Quick Start Guide, that helps you through the needed steps. Enjoy!

 

[godsound]

how set significant digits in hfss

Thanks rautio

 

[mailgvenkat]

stripline coupled line hfss

hi rautio,

normally we use microstrips for MMIC designs.But when we go for nanometer technologies the analysis of microstrips or striplines for SI becomes difficult.

Can we analyse nanometer substrates for SI using SONNETLITE?.....
Wht is the simulation technique(like MOM,FEM,FDTD,etc) used by Sonnet?.....

If it supports annalysis of nm substrates then what will be its optimum dimension?..

Pllease forward links regareding this ?...

anticipating your reply....

with regards,
venkat.g

 

[fantomf4]

si problem hfss

You did not say which version of HFSS you were using, but you should nevertheless
look at the HFSS 9.1 Manual posted on elektroda a few months ago.
I has an excelent example of coupled differential striplines.
Otherwise, to calibrate your model, you could have a look at the example file
that comes with the Sonnet, you can just download Sonnet Lite.

 

[rautio]

Hi Venkat.g -- Not sure if SI means Signal Integrity, or you mean Si as in Silicon. No problem for Si, any thickness or thinness, any conductivity.

For tools like Sonnet, what matters is that cell size be small compared to wavelength. Usually no larger than 0.05 wavelength. Even smaller if very high accuracy is required. The actual physical size does not matter. Only the size with respect to wavelength is important.

Tools like Sonnet are usually intended for high accuracy. If you are doing Signal Integrity, 5 - 10% error is often OK. If we keep the transmission lines subsectioned one cell wide, the analysis is much faster, and error is on the order of 5%, so it can be used.

If each layer is planarized prior to the next layer being added, a planar tool like Sonnet is fine. If the conductors ride up and down over dielectrics and conductors on lower layers, this is more likely to require a volume meshing tool.

Sonnet uses MoM, which divides a circuit into subsections, then fills a big matrix with the couplings between each subsection. This matrix is then inverted. We use only non-iterative matrix solve. This means you are limited to about 20,000 subsections with 2 GBytes of RAM. This takes about 20 minutes to invert on a 4 GHz PC. If you go to single precision, you can do about 30,000 subsections. Thus circuit size will be a limiting factor. We will have a 64 bit version this summer, so that limit will go away. There are some tools that use iterative matrix solvers. These tend to fail now and then, so we do not use them, but they are available.

Another problem you might encounter is that the EM tools do not include the transistors. You have to leave ports for them. If there are 1000 transistors, this is no problem (for the full version of Sonnet, deembedding turned off). If there are 1,000,000 transistors, it is not possible.

SonnetLite is limited to 16 Mbytes (about 1500 subsections), three dielectric layers, two metal layers (including vias) and four ports. There is no time limit. You have it forever. You can certainly easily try out some simple circuits, like parallel coupled lines and cross-overs. Have fun!

 

[mailgvenkat]

hi rautio,

thankyou very much for ur valuable ideas...i wish to know whether sonnet lite (evaluation version can make it).....

if so please give me the instructions for simulatiog nm interconnects?....
Also to avoid SI(signal integrity)effects Wht r the measures to be taken?..
can we analyse the fields by considering any nm substrate and dividing the substrate into subcells(so that each cell is very small compared to the wavelength)?

also can we analyse how the fields decay when any signal pulse is apllied at the certain point in the substrate with PML boudary conditions applied on it?...

with regards,
venkat.g

 

[rautio]

Many of your questions will be answered when you complete the SonnetLite tutorial (Help->Tutorial). It takes about 45 minutes. Sonnet is a frequency domain tool. In otherwords, it does not calculate impulse response. It calculates S-parameters at each frequency. You draw the circuit, then Sonnet divides it into subsections. Good luck.