Having got the signal generator from ebay seller ayanhu81 running it was time to measure Phase noise. I decided to start at the low frequency end, so 42MHz was chosen. The controller was set to this frequency and the lock light came on. However, looking on the spectrum analyser no signal was found. Winding out the span a signal was found at 2668MHz (42x64), very strange! Next 116MHz was tried. The controller was set to 116MHz and the lock light came on. No signal was found at 116MHz but a signal was found at 3712MHz (116X32). I next tried 144,432, 1296 and 2320MHz and the output was produced on the expected frequency. Checking further the IC only worked correctly above 137MHz
As a check I used the AD435x software tool to tell me what the register values should be for 42 and 116MHz and modified the arduino software to load the hex values directly, but still got 2668 and 372MHz. Its not the code, its a problem with the chip
Looking ath the Analog Devices datasheet for the AD4351 it has a divide by 64 and divide by 32 circuit in the chip, which my IC apparently didnt have. I checked the markings on the chip and it said AD4351. I did notice from it's datasheet that the AD4350 does not have these dividers, behaving exactly as my AD4351 was doing currently. Obviously my board had an out of specification AD4351 or a mislabelled AD4350
At this point G4JNT made a post on the UK microwave reflector that he was experiencing exactly the same issue with exactly the same module from the same Chinese supplier
I have raised a "not as described" case with ebay against the seller. He was forced to refund the cost of the module!
Moral: You get what you pay for!