Hi,
I am trying to simulate the phase noise of a bunch of frequency dividers. My oscillator runs at 12-17 GHz. I use frequency dividers to generate quadrature signals and cover the desired frequency bands. Currently I am using a divide-by-2 conventional CML divider followed by two divide-by-2 and divide-by-3 frequency dividers to cover two bands. So the final division ratio is 4 and 6.
Currently I am trying to simulate the divider chain only using ideal sources to find out the total noise floor dominated by the dividers. The divide-by-2 frequency divider that is placed after the first CML divider is a dynamic divider with a much lower noise floor compared to the CML divider and is almost negligible. So I expect to see 3db improvment in the noise floor after the second division compared to the output of the first divider, which I don't. The same applies to the divide-by-3 frequency divider where I expect to see close to 5 db improvement and not only don't I see that but also I see noise floor degredation.
I did a lot of simulations and now I am worried about my simulation setup in cadence. There are two parameters in setup simulation which is not clear to me: Number of harmonics in the PSS window and Maximum sidebands in Pnoise window. I set both of them as 50. I know that the maximum sidebands should be chosen in a way that the results converge but I am not sure about the relation between these two numbers.
The other thing is the modulated noise. is PM noise the real PHASE noise? or should I avoid modulated noise and choose "sources" as the Noise Type?