I've been looking through posts here, and other internet references, to try to understand the meaning of the term "measurement reference plane." My general understanding is that this refers to the place in a network (and once you attach your VNA, it, and the cables you use to attach it, effectively become part of the network), where one conducts the appropriate calibration procedures, which are intended to eliminate (or at least, significantly reduce) errors that are inevitable in the test equipment. Many of the reference materials provided by VNA manufacturers (manuals and such, which may be found on the internet) refer to the measurement being affected by the use of extensions or fixtures that may be added after calibration is completed, and how these may be compensated for without recalibration.
Okay, so far so good. (Please feel free to comment on the validity of the above, or to correct my interpretation.)
What I have been unable to find is any discussion of how much error might be introduced by use of cables, extensions, and/or fixtures that may be employed up to the measurement reference plane. To put it another way, if the purpose of conducting a calibration is to eliminate/reduce errors introduced by the measurement equipment (meaning the VNA itself, and the cables/fixtures used to attach it to the device under test), just how complicated can the "measurement equipment" be, before residual errors, after calibration, become significant?
Suppose I have a network consisting of, say, 3 connected elements, A, B, and C, where A is connected to B, and B is connected to C, like A==B==C. Suppose, further, that, for the purpose of measuring S11, I will be able to connect the s11 port of my VNA directly to element A. I begin by performing a calibration directly at the s11 port - this becomes my measurement reference plane. I then connect s11 directly to element A, and measure the impedance of the network that consists of A==B==C.
Now, I want to measure the impedance of the network without the presence of element A. Presuming I can connect the s11 port directly to element B, all well and good - I simply disconnect element A, connect the s11 port to element B, and bob's your uncle.
But, suppose, for some reason, I cannot connect directly to element B. Could I connect s11 to element A, disconnect elements B==C, perform a calibration at the far side of element A (thus establishing a reference plane there) reconnect elements B==C, and make the measurement? Element A has become, in some respect, a "test fixture" for measurement of a network consisting of B==C. Taking this a step further, could A==B become a "test fixture" for measurement of element C. And just how complicated a "test fixture" could I employ before I degrade measurement accuracy beyond any useful point? (Of course, the notion of what the "useful point" consists of will vary with the person doing the testing.)
