PLL phase noise improvement by using a crystal filter in the feedback loop

[kurtulmehtap]

Hi All,
I have designed a PLL with a 10 MHz Reference and a 2.2-2.6 GHz VCO (using discrete PFD and Frequency divider). I wonder if can improve the phase noise/spurious by placing a 10 MHz Crystal filter in front of the feedback loop. For a divider value of 220 we get 2200 Mhz /200 =10 Mhz and then applying a 10 Mhz filter.
Any Ideas if this will work?
Thank you

 

[biff44]

So, you want to place a very tight bandwidth 10 MHz crystal filter in the divider chain?

Step away from the lab bench and think of what u just said! You have a control loop, where the feedback signal is used to improve the phase noise (or jitter) of a free running oscillator at 2 GHz. So you WANT all that feedback signal to come thru to the phase detector. If you passed only 10 MHz +/- 10 KHz of the feedback signal....then the phase noise of the total circuit will be horrible at +/- 100 KHz offset, since the PLL is NO LONGER cleaning up the free running VCO.

You might get some improvement putting a 10 MHz crystal filter in series with your 10 MHz reference clock signal.

 

[mtwieg]

biff44 is correct, filtering in the feedback path will just degrade the performance outside the filter's passband, and probably won't help at all in the passband. Putting a filter on the input or output of the whole system may help, though.

The only situation in which such a filter might help in the feedback path is if you are getting interference there from neighboring systems or circuits. And if that's the case then you should work on trying to eliminate the interference or improve isolation, rather than filtering the feedback.

 

[kurtulmehtap]

Thank you for you advice. I still wonder if I can remove the spurs caused by the 10 MHz reference using a LPF in the divider chain.

 

[0dBc]

As the name implies reference spurs are caused by the reference signal itself. Some portion of the reference signal leaks to the VCO input and modulates the VCO's control voltage. If you want to minimize the reference feedthru, you can add some extra RC filtering next to your loop filter, not to the divider path.

Thank you for you advice. I still wonder if I can remove the spurs caused by the 10 MHz reference using a LPF in the divider chain.