ADF4351 - a cautionary tale...

[radiogareth]

I have used several of the ADF4351 PLL boards as generators of LOs for my projects, both the Greek quality item and the 'others' from eBay or Aliexpress. Some of the later boards turned up with quite a 'proper' low-noise and filtered PSU along with a 100MHz reference. I used one in my Ionica 'LNB' front-end for GB3JV reception. There was even a jumper to select the 100MHz or external via a provided SMA. As it was for DATV reception absolute precision or stability was not much of an issue so I reused a salvaged TXCO that was labelled as 18.750MHz. Controlled by an Arduino Nano with code from Robin (Programme then sleep). It worked well producing the required 2450MHz LO, which when amplified via a generic '20dB' eBay amplifier (5189 type) gave me the required LO drive for the Ionica front end.
Fast forward a few months and I started work on my combined 6 & 9 cm transceiver, This needed two LO's, one for basically another Ionica front end and a second to down-convert 6cm to 9cm. Two more boards were acquired and this time fed with 10MHz from a CTI TXCO as they have been found to be some of the best performers for very little money.
The output waveform was however absolutely foul, looking more like a comb generator than a LO. Initially I suspected an 80p Buck converter that I was using to take the strain off a LM317. However, removing that had no effect (I had added a lot of extra filtering to the buck PSU). Probing the PSU on the ASF4351 boards did show some noise but down in the low uV range and pretty much in accordance with the data on the low-dropout chip used on the ADF boards. As I had been wanting two LO's, I had stacked two boards and arranged for header pins/sockets to share the clock input. The data sheet said 100kOhm input impedance so that seemed a reasonable route. On one of the boards I'd removed the 100MHz TXCO to prevent clashes if I put the jumper in the wrong place. Isolating this board from its 100MHz equipped friend produced a clean signal. Bingo, the 100MHz oscillator seemed to be 'modulating ' the desired 10MHz possibly due to their numerical relationship as it doesn't happen with the 18.75MHz board. Anyway, long story short, I cut the power rail track to the 100MHz TXCO and normality was restored, 2362MHz on one and 2450 on the other. The same fault was apparent on my Portsdown 4 ADF4351 signal generator- a foul comb of output, useful to ident perhaps but not a lot of use at the higher frequencies as you could not see the fundamental! Must be something to do with how it generates the desired frequencies, low VHF didn't look too bad. Removing power to the 100MHz oscillator solved the problem here too. Why not use the provided 100MHz? Better stability can be achieved with the CTI unit and neither Robins code nor the Portsdown code seems to like 100MHz being entered as the reference frequency. When they were written the common on-board osc frequency provided was 10 or 25MHz.
An interesting days work/play.
Gareth