how combine s-parameters
The package has two external ports and two inner ports which are directly connected with a two-port SAW device.The inner ports are defined as lumped ports with 50ohm impedance when the S-matrix of the package is calculated using HFSS.While calculating the response of the two-port SAW device,the measurement system's impedance is set to 50ohm.
I want to combine this two results using Data Items in ADS,but I am a little confused because this two results are calculated separately while the SAW device is localized at the inner ports directly as mentioned above.
Can anyone help me?
Thanks.
Hi Jiaping:
This is actually an interesting problem that more people are finding.
I think the best way to approach this problem is:
1. Build a test board with the SAW device, and measure each port with a VNA. Make sure you have a good de-embedding up to the edge of the SAW device. You need connections to the devices that are calibrated to a known impedance, for the S-parameter models of your SAW device.
2. You can simulate your interconnect, but you will find that lumped ports have error because they cross-couple if they are close, and they introduce microwave discontinuities into your model (SMD pad to ideal wire connection) that do not exist in real life. They also have extra inductance from the ideal wire of the lumped port.
Try to use only ports that are de-embedded, and characterize the SMD pads and traces in a separate EM simulation. You will need to use ports that are on the edges of your workspace. Then you can develop S-parameter models that are properly de-embedded to the correct impedance so that cascading the S-parameters makes sense.
I think I saw that Sonnet has a new release with a new type of port called Co-calibrated port that is a de-embedded internal port. I think the Co-Calibrated name means that you can several internal circuit ports that are close together, and they somehow get rid of the cross-couping between the ports without getting rid of the cross-coupling between the SMD pads. This looks very promising for simulating the board-level stuff up to the SMD.
Good luck; this is a tough problem.
--max