why s12 isnt equal to s21 ??

[moreho]

when i measuring a pd constructed using wilkinsons, the measured s12 and s21 must be equal, as in simulation in ADS, and because it is a reciprocal network.
but in reality, in the VNA Agilent, the shape of s12 isnt equal to the shape of s21.

why?!

 

[volker@muehlhaus]

You are correct, this can't happen with a reciprocal DUT. Something might be wrong with your VNA calibration.

 

[moreho]

but my calibration is true. i use full2port coax cal. after it, s12&s21~0.01db and s11&s22<-60/-70dB.
but i wonder why this happens.
sometimes the difference between s12 and s21 of my reciprocal device is large and sometimes is little.

of course sometimes after my tests finished, the difference between s12 and s21 of my device is equal to the same of when the ports of vna is trued and this means that the calibration also has found a problem.
what problem and how to resolve it? ...

 

[BigBoss]

You will probably see the same effect S31==S13.I guess your Wilkinson Divider is assymmetric on XY-axis.( Mostly manufacturing faults) since your frequency of interest is high.(Small structures are much more sensitive to manufacturing defects/tolerances)

 

[volker@muehlhaus]

I guess your Wilkinson Divider is assymmetric

You explain why transmission to different output ports can be different, e.g. S21 different from S31. However, that isn't the issue here. The question was about results that see to be non reciprocal.

- - - Updated - - -

but my calibration is true. i use full2port coax cal. after it, s12&s21~0.01db and s11&s22<-60/-70dB.

This doesn't mean that your cal is correct. Testing only with testcases that are identical to the cal standards doesn't tell us much about the quality of calibration. It only checks repeatability.

of course sometimes after my tests finished, the difference between s12 and s21 of my device is equal to the same of when the ports of vna is trued and this means that the calibration also has found a problem.
what problem and how to resolve it? ...

I don't understand what you mean.

 

[conn_find]

did you terminate unused ports with a 50 ohm load?
I have seen when they are not terminated sometimes these things happens, if it's not a manufacturing fault.

 

[moreho]

did you terminate unused ports with a 50 ohm load?
I have seen when they are not terminated sometimes these things happens, if it's not a manufacturing fault.

thanks...

yes i terminated all other output ports with same loads (50 ohms).
somebody may think that tolerances between loads can show significant error in results, but i tested it and found that only the load that is nearest to the tested Port may have that big effect.
but at finally, it isnt seems to be for termination loads differences.

- - - Updated - - -

You will probably see the same effect S31==S13.I guess your Wilkinson Divider is assymmetric on XY-axis.( Mostly manufacturing faults) since your frequency of interest is high.(Small structures are much more sensitive to manufacturing defects/tolerances)

thanks...
same as volker@muehlhaus, i declared that i want to Obtain test Results that Confirm my network is Reciprocal, as in theory it is.

- - - Updated - - -

I don't understand what you mean.

thanks..
i will draw a diagram to explain it.

This doesn't mean that your cal is correct. Testing only with testcases that are identical to the cal standards doesn't tell us much about the quality of calibration. It only checks repeatability.

As you mean, i think my network (VNA) has a problem that i dont know what it is.