Dear All,

I tried to calculate the complex propagation constant of a microstrip line with CPW-to-microstrip transition using uncalibrated S-parameter measurements on two identical lines with different lengths. I used T matrices and calculated the eigenvalues in order to get the propagation constant for the microstrip line.

I found that the phase constant calculation is in very good agreement with my HFSS simulation. However, the attenuation constant from the measurement is much larger than the HFSS simulation. My HFSS simulation took the surface roughness, the dielectric complex permittivity frequency dependency into consideration. I suspected that the difference between the simulated and measured ones on the attenuation constant is caused by the uncalibrated S-parameter measurements that were not corrected with switch term correction.

I also tried to calculate the complex propagation constant of CPW lines using the same approach as I used for microstrip lines. I ended up having the same problems. For CPW, the calculated phase constant from uncalibrated S-parameter is larger than the ones measured with Multiline TRL and the attenuation measured is also larger than Multiline TRL measurement.

If my guess on the error source is right, how can I correct the uncalibrated data with the switch-term correction without adding any measurement complexity?

Thank you

I tried to calculate the complex propagation constant of a microstrip line with CPW-to-microstrip transition using uncalibrated S-parameter measurements on two identical lines with different lengths. I used T matrices and calculated the eigenvalues in order to get the propagation constant for the microstrip line.

I found that the phase constant calculation is in very good agreement with my HFSS simulation. However, the attenuation constant from the measurement is much larger than the HFSS simulation. My HFSS simulation took the surface roughness, the dielectric complex permittivity frequency dependency into consideration. I suspected that the difference between the simulated and measured ones on the attenuation constant is caused by the uncalibrated S-parameter measurements that were not corrected with switch term correction.

I also tried to calculate the complex propagation constant of CPW lines using the same approach as I used for microstrip lines. I ended up having the same problems. For CPW, the calculated phase constant from uncalibrated S-parameter is larger than the ones measured with Multiline TRL and the attenuation measured is also larger than Multiline TRL measurement.

If my guess on the error source is right, how can I correct the uncalibrated data with the switch-term correction without adding any measurement complexity?

Thank you

But, I'll try to guess at what you're doing and answer.

First, no, I doubt very much that what we call "swtich term" is the cause of any trouble. It depends a very great deal on your particular analyzer, but in general, source match, load match and transmission tracking will be much greater values.

Further, the loss correction of microstrip is notoriously difficult to get right. Did you include the edge roughness of the line (the fact that the edge is not a straight line, but rather a jagged line?). In my experience, this is as significant as surface roughness. Also, how did you measure your surface roughness. I have seen in general that it is greatly under-estimated. What about the dlelectric loss tangent of the substrate? And, did you include the additional loss of the glue for the laminate (Here I'm assuming it is a PC board, but that's all I can guess at). How did you make connections to the the microstrip line? Did you use coaxial adapters or wafer probes. In any case, why would you not do full calibration at least in the coaxial case before attaching your microstrip to minimize errors? Its really a very poor idea to rely on uncorrected performance of any quantitative measurement, and surely not recommended.