We are designing a synthesizer based on the LMX2820. Previous designs based on other PLLs have required an adjustable reference oscillator circuit to mitigate IBS, and we were hoping that the input multiplier in the LMX2820 would be sufficient to allow us to dispense with that circuit.
We have acquired the LMX2820 EVB and tested it in our lab, and we have also run a number of simulations in the PLLatinum simulation software. The typical scenario for the tests has been to compare the IBS generated when the VCO frequency is near the PFD frequency with the spurs generated at the same frequency offset (from the same VCO frequency) with the input doubler or multiplier engaged. In the latter case, the VCO frequency is not near an integer multiple of of the PFD frequency, and hence there should be negligible IBS. However, the PLLatinum software predicts that there should nonetheless be a spur at the same frequency offset, which it labels "Fvco%Fosc". We confirmed the presence of this spur in the lab. Both in the simulation and in lab tests, this spur was typically of the same order of magnitude as the original IBS.
From other answers on these forums, I understand that the Fvco%Fosc spurs are due to crosstalk between the oscillator frequency and the VCO frequency, and that this crosstalk is present regardless of the PFD frequency. My question is: then is there any benefit to using the input multiplier at all? Is there anything that we can do to substantially reduce or eliminate these spurs?
As an aside, the PLLatinum simulator gives rather inaccurate results when simulating these spurs. The location of the spur is accurately predicted, but the intensity is routinely off by 10 dBC, and occasionally by as much as 20 dBC.