ABSTRACTThis article compares for the first time three well‐known procedures for characterizing the noise temperature and conversion loss of a millimeter wave mixer. To carry out this study, a 183 GHz subharmonic mixer has been measured using three common procedures, that is, the “attenuator,” the “gain,” and the “noise injection” procedures. Furthermore, for every measurement procedure, three different detection methods have been used; a broadband power meter, a yttrium iron garnet filter working together with a broadband power meter, and a spectrum analyzer. The “gain procedure” has turned out to be the most stable one in terms of less variation between consecutive values and flatter results along all the frequency range. Results obtained with every detection method are consistent with each other, showing a similar performance of the mixer independent of the detection method used. In addition, the effect of the inclusion or omission of an isolator between mixer and intermediate frequency pre‐amplifier has also been evaluated. For the first time, clear conclusions about its influence are driven. The inclusion of the isolator has culminated in more reliable results, although the measured NT values are increased. © 2014 Wiley Periodicals, Inc. Microwave Opt Technol Lett 56:1441–1447, 2014