Singular Matrix ERROR in MW0

Hi,

I simulated a filter in EM sight of MW0,and sweep freq is 7-25GHz.
When freq steped to 24.8GHz, singular matrix error occured.

I set X-grid division in 40um,Y-grid division in 20um.
and the minimum size of microstrip section is over 100um.

Can you tell me what action should I take to avoid this problem?

Thanks in advance.

option1: try changing the de-embedding length. if it is zero (no de-embedding), add some.

option2: check box resonance at that frequency.

loucy,Thank you for your response!

What is the de-embedding?Pls explain it!How to add it?

I didn't add the box in simulation.
I specified the EM structure in two layers:air(up),substrate(down),and add PEC above air layer and under the substrate.

Was I wrong?

check your setting for the port, look for the word de-embedding.

or, upload your project file here (don't save the results to reduce the size).

Hi,

Here is my filter project in MW0 EM structure.

Pls check it and correct my settings.

Thanks.

Sorry!
.emp file was not allowed to be uploaded here.

If someone so kind want to help me,pls post your mail.
I will send the file to you.(only 5k)

Here is the .rar file of the project.
Pls check it.

MWO reports singular matrix for the first frequency (7GHz) with the current project setting.

However, after I reduce the de-embedding length of port1 (on the left) by a small distance (don't remember the value), it runs ok to 9GHz, and I stop the simulation.

I think this confirms that the singular matrix happens when solving the de-embedding standard for the left port. MWO should have handled that autmatically and correctly. to by pass the problem, just change the de-embedding length.

To change the de-embedding length, zoom in on the left egde of the feeding microstrip, select the port, you would see a black line running along the center of the microstrip. The line ends at a perpendicular dashed line, place your mouse cursor over the dashed line, then you would be able to drag it to the left or right. this change the de-embedding length.

I am not sure how MWO sets up the de-embedding standard. Remember that it has to find the propagation constant and characteristic impedance of the feeding line in oder to shift the reference plane (which is part of the de-embedding procedure). So this de-embedding standard refers to a separate EM problem where the structure is just a simple microstrip line (with the same width and substrate as the line to which you set the left port). This EM structure is again bounded by a conducting box, thus box resonance is possible. In this case the resonance means that you don't need an excitation entering one end of the transmission line to get some output from the other end. In other words, the de-embedding length happens to be mutiple of half elecric wavelength, or the size of the box happens to be the correct value to support a cavity mode at that particular frequency for which singular matrix is reported. As a result, it has trouble finding the propagation constant.

Thank you loucy.

Tomorrow I will follow your instruction to have a try.