We made an observation with regards to the two tone SFDR measurement (IMD3), where as the tone separation narrows from 3MHz down to 10KHz, the SFDR performance degrades. This is issue is more pronounce in the 2nd Nyquist zone.
I attached narrowband 2-tone data around IF=185 MHz with CLK=250 MSPS. Two RF generators were combined via high isolation power combiner and then filtered with 200 MHz LPF which was then connected to the EVB.
Data does show IMD does gets better as 2-tone separation increases. My best guess is that it could be related to fact that a two-tone signal in the time domain exhibits AM modulation with modulation or beat frequency being equal to the delta difference between the tones. Perhaps low frequency impedance on supply domain or internal ADC mode (i.e. reference buffers) could be issue.