I am working on a design with the ADF4360-3 driving a 10W power amplifier through a lattice balun at 900MHz. I have configured the registers to output ~900MHz and the loop locks when the PA is not powered. Spectrum analyzer confirms the correct frequency output. The SimPll software confirms the loop filter is stable at 20kHz bandwidth and 45 deg phase margin. The PA datasheet can be found here: http://www.mitsubishielectric-mesh.com/products/pdf/ra13h8891mb.pdf
When the power amplifier supply is switched on, the ADF4360-3 immediately loses lock and the N counter goes unstable. At first, I assumed this was noise coupling into the loop, however I shielded the synthesizer with grounded copper foil as well as inserting the PCB into an aluminum case grounded to the PCB ground plane. The problem persists when hooking the PA output to a 50 ohm dummy load (no antenna).
This design contains a full ground plane and the suggested decoupling. The PA is isolated from its supply by decoupling and ferrite bead. I'm at wits end, hopefully someone has fresh ideas...