I am trying to determine the uncertainty specification for differential S parameters on the ENA 5071C. My machine has options 485 and 790. In particular I am concerned about a measurement of Scd21 which of necessity employs the Fixture Simulator and internal matrix math for four

of the two port parameters: S31,S32,S41, and S42. I would expect Agilent

to derive and publish the uncertainty values for all such four port differential parameter measurements to assist customers in making

meaningful measurements and being able to publish accurate results.

of the two port parameters: S31,S32,S41, and S42. I would expect Agilent

to derive and publish the uncertainty values for all such four port differential parameter measurements to assist customers in making

meaningful measurements and being able to publish accurate results.

To get meaningful results, you must realize that the measurement uncertainty depends upon the particular values of the underlying S-parameters. For example, there is less uncertainty on an S21 trace that has a value of -20 dB than there is on one that has a value of 0 dB. Further, since the computation of Scd21 depends upon 4 S-parameters, you must know the values of all 4 to compute the uncertainty of each, then combine the uncertainty of these underlying parameters to determine the Scd21 uncertainty.

Since Scd21 is typically a small number, due to being the subtraction of two big numbers, a good guess would be to take the uncertainty of S21 (something like0.05/_1 deg), and S31, and subtract the worst case values; but the problem is S21 is probably a large number and the angle error is most significant, so you would need to take S21 with a plus angular error, and S31 with a minus angular error, to discover the Sdc21 error as something like 1/2(tan(2*angular uncertainty). Take this times 2, (for the other two S-parameters) and you'd be close. For a 1 degree angular uncertainty on S21, you'd end up with about 24 dB as the "floor" for an Scd21. But of course, this computations completely fails for small values of the underlying S-parameters.