I have been using the ADF5355 Synthesizer IC in a single-channel tuner product for several years, and have had no issues until I designed a two-channel version of the product.
In the two-channel version, there are two identical ADF5355 synthesizers on the same PCB with a mechanical chassis isolating the two sections. The two ADF5355 IC's are on separate regulated power supplies, and share only the SPI interface and the 16 MHz reference source. There is no requirement for a predictable phase relationship between the two channels, although they are typically tuned to different frequencies, but could be tuned to the same frequency.
When tuning to the same frequency, the problem I observe is large spurious sidebands that pop up inside the loop bandwidth and are extremely sensitive to temperature, and proximity of the chassis. The offset frequency of the sidebands is not predicable and shifts dramatically with proximity of the chassis or your hand. The sidebands act like an oscillation caused by some sort of feedback loop or instability. They also can disappear/reappear after repeated tuning commands. Several in-between states can be seen, where the noise just increases without the discreet sidebands popping up. Installing the PCB into the chassis does not eliminate the issue. Removing power from one of the channels eliminates the problem in the remaining channel. It appears that there is some sort of interaction/feedback loop between the two channels.
Here is a list of things I've tried:
1) Added pads at the REF inputs to increase the isolation between chips.
2) Added chokes to the REF inputs.
3) Adjusted the REF input level across the recommended operating range.
I would appreciate any thoughts or suggestions you have.