Tuning a 2.5GHz cavity filter with a home build VNA

While improving the SA I'm building I learned something about tuning a cavity filter.
The VNA I'm using is home build and consists of only 
- 2 x ADF4351 (or one SI5351, range 1MHz till 1.2GHz)
- a resistive bridge
- 2 x IAM-81008 mixers with some passive components (10 dB attenuator at the RF input and some capacitors)
- Arduino
- PC with some software



The black brick is the 10 pole cavity filter supposed to be very good, but when I measured the input impedance I got this.


Then I learned that for good input impedance measurement you have to terminate the output of a cavity filter with 50 ohm.
This improved the measurement.


When tuning the cavities the first cavity determined the reactance of the input impedance so I was able to center the circles better around Z0. 
The other cavities could be tuned to decrease the size of the loops around Z0 and finally resulted in


The first cavity still needs a bit of tuning but the S21 of the filter is now nice and flat
Passband ripple is less then 2dB and loss is also very good. 



This was the result of only one round of tuning so with some iterations it probably can be further improved.
Due to complete absence of any shielding the dynamic range of S21 measurement is limited to about 40dB-50dB. 

The IAM-81008 turns out to be the perfect component for building a quick and dirty GHz VNA