I'm working on a board with two AD9852 DDS's that each produce a modulated 70MHz IF output. In an effort to isolate the two channels and prevent cross-talk, we have added power supply isolating filters to the AVdd and DVdd supplies on each DDS. Each DVdd and AVdd supply has it's own local power plane under the DDS and connects to the board 3.3V supply through two LC filters. The filter topology is as follows: a shunt 10uF ceramic cap on 3.3V supply, followed by a series ferrite bead (BLM15EG121SN1D), followed by shunt 1uF ceramic cap, followed by another series ferrite bead (same part number), followed by connection to DDS AVdd or DVdd localized power plane. The ferrite beads have about 10 milliohms of DC resistance, so they can handle the current OK. Each power supply pin on the DDS is by-passed with a 0402 0.1uF ceramic cap. to a common ground plane.
The 70MHz DDS output runs through a 1MHz wide SAW BPF followed by a gain stage. The output has some spurs that are roughtly -55dB down from the 70MHz IF unmodulated carrier (~300KHz away). Other spurs can be seen. However, when we replace the ferrite beads with zero ohm jumpers, the spectral response is cleaned up considerably (spurs disappear into the noise floor). See attached Spectrum Analyzer plots.
I've seen other posts mentioning that the recommended DDS connection to power is "with minimal trace length to minimize inductance". Could this be the issue I'm seeing? I seems that adding some filtering on the supplies would be beneficial if you have a noisy board with other circuitry. Is there a recommended method for minimizing cross-talk between two DDS's on the same board? Would it help to add bulk capacitance near each DDS to help with the short-term current draw requirements?
Would appreciate any insight to this issue.