You mentioned two scenarios:
* calibrate w/ 50 ohm, measure above RL
* calibrate w/ open, measure below RL
It would seem there is one option you left out:
* calibrate w/ short, measure below RL
In
this video of an early The Signal Path episode, Shahriar uses a directional coupler and a Rigol spectrum analyzer with a tracking generator. He does this in a few steps:
* He calibrates against a
short, explaining that this represents a perfect reflection of all incident waves at all frequencies.
* He then compares this to a crude open (crude because he leaves the calibrated open in a box, and just disconnects the short). The open appears to be roughly at the same level as the short, but has very slight bumps over and under the normalized reference level.
* And lastly, to verify the setup before connecting the device under test, he connects the 50 ohm load which represents a perfect absorber. The measurement suddenly drops about 40dB below the reference level.
I am happy to blindly trust an expert like Shahriar, so I would say this is the right way to do it. Thinking about it logically, too, it makes some sense:
* the 50 ohm load is the most expensive and difficult to perfectly match component. Its imperfections, if they exist, will limit our measurement in unforeseeable ways if we calibrate against it as a normalized reference.
* the short is the least expensive and easiest to perfectly match (and make, DIY) component. It's just a short. If it has any imperfections, they will be slight and they will probably be shared by the physical dimensions of the other devices. In any case, it's seems tremendously easier to make something 0 ohms for all frequencies than to make something 50 ohms for all frequencies.
* the open is, in a sense, the least well defined of these. I suppose it's the impedance of free space, or the environment it's in. It can pick up noise. In theory, it should be nearly the same thing as the short,
just the opposite direction in some space. I'm not sure about this, but it seems wrong to calibrate against an open by itself.
Overall, I'm not entirely sure which is exactly, right, and thinking about this I believe there are more variables, including:
* what is the range exactly (such as: if we are making the 50 ohm, the open, or the short into the zero calibration, then what is the opposite condition? above or below RL? and what does it represent?)
* what about the other measurements you can make with a TG and a RLB; with the opens/short/load in the reflected port?
* what about using 50 ohm feed-through loads instead of terminating loads?
* what about all variants on these devices, the RF bridge, directional coupler, etc...
I would love to collect enough discussion and information so that we (or I) can write up an article and/or chart about what's appropriate, what to think about, and what all of the variations are in this simple measurement setup.