We are using the ADF4351 as a local oscillator for a converter.
We based the design on the evaluation board.
We need to reduce the noise level of the LO to meet the 12.5 KHz FCC mask at a span of 50 kHz.
The attached shows some plots.
The bottom plot shows a noise spectrum that passes the mask. The loop filter values used are based on the lowest charge pump current of 0.31 mA. We discovered that when we increased the charge pump current (to 1.25 mA), without changing the loop filter values then the noise level goes down. The sideband noise was pushed outside the 50 kHz window. So this bottom plot is with a 1.25 mA charge pump current but R-C values computed for 0.31 mA.
The top row shows plots were the loop filter values were all calculated for 1.25 mA. That is 1.25 mA charge pump current and R-C values computed for 1.25 mA. The three plots show varying loop filter BWs.
The question is; why does the spectrum look cleaner with the wrong R-C values?