We use an ADF4360-8 at 125MHz with a successive AD9513 to source 2 AD9268 125MHz/16bit ADCs and a FPGA.
The Lext is 180nH, the reference frequency 24MHz, PDF=1MHz, loop filter 3rd order passive, 105kHz and 44.4deg phase margin.
Almost the recommended values of ADIsimPLL.
The ADC input is driven by an ADA4938-2 differential amplifier. Measuring a grounded input signal we sometimes get the expected good noise data with some 9.5 LSBs FWHM distribution. But then, we as often get additional 282kHz (and harmonics) that distroy the good ADC data.
Fig 1: GND noise and its FFT when good
Fig 2: GND noise and its FFT when bad
We found that this seems to come from the PLL circuitry as a finger on the 2 inductors can fairly influence and improve this behaviour.
Any ideas where this really comes from and how to avoid this?
The inductors are at a 90° angle as recommended.