I am currently using the Phase INC/DEC function with the Si5338.
1) The datasheet says that the accuracy of phase steps is "20ps" MAX.
However I have measured the repeatability of these phase steps to be <4ps RMS. Also, the granularity is more of the order of 3ps.
2) And for example on this post, we have an example of how to set up a 20ps phase step size
(which is questionable if the phase step accuracy is 20ps. You would not want to have 0ps steps when you programmed 20ps. )
-> So my question is: why is the quoted accuracy of 20 ps while it seems to perform much better? what is the limitation (temperature effects, ageing, ... ? )? Is there a Silabs alternative which would achieve 3ps?
According to the formulas of phase step in Si5338 RM, and Fvco range is 2.2G to 2.84G. The minimum step size is about 1/(128*2.2G)=3.55ps.
The step size is about 2.75ps when VCO is 2.84G.
So it is reasonable if you measure step <4ps.
The spec in datasheet is just a maximum step accuracy over PVT with some margin.
Because the step depends on VCO frequency, what's you output frequency for step=3ps requirement?