> With the tee connector split and the TX side going into a > dummy load, there is no desense. If I reconnect the tee and > go to the -8920, the desense is back.
You lost me on that one. You're saying you're testing for desense by removing the tee from the antenna port of the duplexer, feeding the Tx leg of the duplexer into a dummy load, and the Rx leg gets fed by your 8920 sig gen? If that's the case, then that's not much of a test since you're no longer duplexing. You've totally isolated the Tx and Rx, so all you really know is whether or not you have "in-cabinet" desense (i.e. between the transmitter and receiver internally due to poor shielding or cable cross-coupling). Or maybe I'm totally misunderstanding how you're doing the desense test - if I have, please re-explain. The easiest way to do the desense test (while keeping the feedline and antenna out of the equation) is to connect the duplexer antenna port to a high-quality (low-noise) dummy load, with an iso-tee inline between the duplexer and load. Connect your 8920 sig gen to the decoupled port on the iso-tee, generate a weak signal while monitoring the repeater Rx local receiver, and key the transmitter on and off manually. If you have desense at that point, and it sounds "ratty" as if something is breaking down or making intermittant contact, then go do your tappin' n' wigglin' to see if you can narrow down the list of suspects. The dummy load on the 8920 RF port is OK, but I'd still be more comfortable using a good external load and isotee. Intermittant desense can sometimes be traced back to a component or solder joint in the transmitter being defective, which can manifest as arcing (however microscopic). The resulting transmitter noise may not be easily discernable on a spectrum analyzer, especially without attenuation of the carrier frequency to increase the dynamic range of the test equipment, but may still cause appreciable desense due to the broadband noise falling on the Rx frequency. Yes, half of the duplexer's job is to attenuate transmitter noise to keep it from getting to the receiver, but if a failing component causes the effective noise level to be elevated 20 or 30 dB, that's 20 or 30 dB more isolation your duplexer would need to provide to prevent desense, and often that kind of headroom doesn't exist. Have you measured the isolation of your duplexer from Tx port to Rx port with the antenna port terminated in a dummy load? What is the measured isolation at the Tx and Rx frequencies doing this test? --- Jeff WN3A