- Can VNAs be calibrated somehow to show the behavior of
VNA->Match->Antenna even though it is actually physically connected as
In the general case, no. It is not even theoretically possible. A general matching device has three parameters (affecting how it cascades with other elements, so neglecting the absolute gain), S11, S22 and S21*S12. Making a single S11 measurement gives you only one equation for three unknowns. A matching device usually has at least two degrees of freedom, so it can be set to an arbitrary location on the Smith Chart.
However, there are some alternatives, and other possibilities for specific cases. Even when a route is theoretically possible, it may still be infeasible in practice, or too time consuming to develop or to execute.
Usually we match a specific antenna, to cope with variations in the antenna itself. The fact that you are contemplating setting up the matching without the antenna implies that the antenna is stable and repeatable enough to do this.
The first obvious option is then to create a dummy antenna, that you can use to load the matching circuit. Then you simply tune for a good match. You would need to create a physical model of the antenna, perhaps with another matching circuit, or lengths of line and an attenuator. 'Golden' references as they tend to be called are generally frowned upon, as they have to be created, maintained, verified, but it might be your simplest option. There would be some development work involved. Trying to get the S11 behaviour of the antenna into a small box would be a good task for an RF student, but determining whether it would meet your overall system specifications given its inevitable deviation from perfection would need a more experienced person.
A special case might be applicable (but unlikely). If your matching device is not 'general', but has only one degree of freedom, and a well defined trajectory of S parameters as you adjust it, then you may be able to ask your tuners to tune for a specific reflection. This would allow you to use a more repeatable load like a short or open, rather than a golden antenna simulator. I suggest open or short rather than termination, to give all the matching parameters some signal to 'work with'.
It is theoretically possible if you make an S11 measurement for each of three known loads, and then solve the appropriate equations. This is an analogue of how the standard one-port calibration would work. 'Tuning' may be impractical if you need to physically connect three loads to your device, but it would allow you to make one-off measurements. It's very unlikely that your VNA supports exactly this measurement out of the box, though if it can be automated, you could run it as a PC peripheral and do your own matrix inversion.
Agilent make a switchable calibration reflection device, to automate the calibration of their VNAs, or at least did a few decades ago just as I stopped working in the VNA industry. That, or something like it, would allow this measurement to operate in near real time, to allow tuning. You could perhaps make one with PIN switches, which might be feasible over the restricted frequency range of your antenna.1
With a more fully featured VNA, you could make a full 2-port 4-parameter measurement, to tune the matching circuit to a specific set of S parameters. The VNA you quote lacks the reversing switch, so can only measure S11 and S21 without reversing the device physically. You could make a reversing PIN switch over the restricted frequency range of your antenna.
As another alternative, you may be be able to devise a method whereby you terminate the matching device in an open, tune for a specific match, then terminate in a short, tune for another match, rinse and repeat. If you are lucky and this process converges quickly, it may be usable for manual operation.
1 Where the 80/20 rule bites in RF engineering is getting something to work over a wide bandwidth. Agilent spend a lot of money getting their VNAs to work down to 9k or 300k, and up to 6 GHz. If you only want to cover an octave in the middle, then all your biassing and coupling components become nice values, and something may be do-able without too much development. Peregrine Semi (other CMOS switch manufacturers are available) make nice LF to GHz switches that could make switchable terminations and paths work as hoped.