Next pictures show attenuation of Narda SEM020 relays, two pieces (S/N 5782 and S/N 2037) were measured. Every picture contains a black and a red graph. The black plot is what was measured with the power head connected to the generator without the relay (all other adaptors were still in between). This plot is used as a reference. There are some ups and downs in the curves due to resonances. I did not figure out where the resonances came from, high grade adaptors were used. The red plot shows the measured attenuation thru the relay. The both straight lines are the linear fittings of the curves, the black line is the fitted reference, the red line is the fitted attenuation curve.
S/N 2037, "normal closed" (NC)-port, distance between the fitted curves is about 0.1 dB. At a certain frequency the attenuation can be much higher due to resonances. Attenuation gets smaller with rising frequency, not enough points for an accurate fit.
S/N 2037, "normal open" (NO)-port. Same as above, difference between the fit-lines 0.2 dB. One explanation for the higher attenuation is that this is the normal open port and the properties of the contact surface may be different compared to the NC port. Probably this will change if the relay is used for transmitting more often.
S/N 5782, "normal closed" (NC)-port. Difference between the fit-lines is rising with frequency as one should expect.
S/N 5782, "normal open" (NO)-port. Otherwise same as above.
SWR input to NC contact, SWR does not rise above 1:107 up to 6 GHz, there might be some resonaces above that are not within the range of the network analyser.
SWR input to NO contact. SWR rises up to 1:12 at 6 GHz. Dunno why SWR is slightly higher compared to NC port. Probably some corrosion on the NO port. Could improve after some switching cycles with power applied.