RF Level Characterization

While trying to measure the frequency response levels of a mixer, I was having problems accounting for the observed mixer gain.  I do not have a high quality wide ranging frequency source.  I use synthesizers built from previous efforts.  These work fine with respect to frequency accuracy and phase noise.  I knew they were not designed as signal generators and had some level variance.  For my most purposes this was fine. With this activity I needed a better estimate.  Below is the output power level at the fundamental of a ADF4351 based synthesizer.

This was taken using a SA0314 using the max hold functionality while scanning 1GHz at a time.  Beyond the expected roll off with frequency (largely due to PCB layout and board) there is a ~90MHz swing of +/- 2dB.  There is a slightly larger variance at select frequencies.  My first thought was to calibrate it out and use the 3dB power output steps of the part to get something a little better.   Unfortunately the variance changes with load and between parts.  I knew this from reading the support forums but also observed it by comparing the maximum at different frequencies through different mixers, amplifiers and load combinations. So if you are using these items in their intended application (i.e. a LO with only a minimum output requirement) its fine.  If, on the other hand, you are using it as the RF source and sweeping them across a fixed LO you end up measuring the source variance (+/- 2dB), not the mixer gain.