What is the difference between the "TRL" and "TRL*" calibrations?

Vector network analyzers which have a four-channel receiver can do a true TRL calibration. Those with only three channels in their receivers can only do a modified form of the calibration, known as "TRL*". The primary difference is that a TRL* calibration cannot correct for the change in the impedance of the transfer switch between the forward and reverse directions. The result is that the source and load match sort of get averaged for each port.

All of the "self-calibration" techniques (TRL, LRL, TRM) require the assumption that the VNA port matches are the same whether making forward or reverse measurements. That is, Port1 Source Match = Port1 Load Match, and Port2 Source Match = Port2 Load Match. This is never true, because when the VNA source power is switched between ports, it affects their match.

However, if the VNA has four receiver channels, it is possible to make some additional measurements during calibration, and correct for this changing port match so that the TRL error correction works as intended. If the VNA has only three receiver channels, this extra correction step cannot be done, and so the VNA port mismatch is not accurately corrected.

A work-around for TRL* is to place high-quality attenuators on the test ports, thereby improving the raw port matches. This also decreases the change in port match between the forward and reverse measurements, so TRL* can work nearly as well as true TRL.