I met some problem when I do de-embedding in our software..Is there any one met this similar problem before?
According to the Apllication Note (5980-2784EN), de-embeding can be done with following equations to modify the error terms.and It's the way what we want to implement in our software.
Here's the process for the de-embedding in our software:
1. perform a full 2 port calibration, and store 12 error terms.
2. measure s2p file for the Fixture.
3. According to the metioned equations, modify the eror terms.
4. write back the modified error terms to analyzer.
Of cource, there's another way to do embeding, just use the internal application of analyzer.There's a menu to do de-embedding in internal application.This way is used to verify our software(we compare the measurement results after we do em-bedding with these different ways).
We get a wrong curve after we do de-embedding with our software, the RL is worse about 1-2 db and the shape of the curve is totally different with our reference curve.But when change to the internal application to do de-embedding, we get the correct result what we expect.
And we checked the whole test process, we're sure of that the calibration is ok, the s2p file is same whthin 2 debedding ways……..
Most likely you made a mistake in your math, or the direction you specify for the de-embedding network. On the analyzer, de-embedding defaults to port 1 of the de-embed network facing the test port (newer version of firmware allow you to select the port). The app not presumes port 1 of the de-embedding network on por1 of the VNA faces port but BUT port 2 of the de-embedding network attached to port 2 faces port 2 (confusing, yeah). so check that out.