|
|
calculating phase noise of pll / sampling loop - RF Cafe Forums
|
mac Post subject: calculating phase noise of pll / sampling loop
Posted: Fri Nov 02, 2007 3:52 am
Captain
Joined:
Fri Nov 02, 2007 3:45 am Posts: 8 Location: germany consider
I want to phase lock a 2-3 ghz VCO to a 20mhz reference ( -120dBc/Hz)
and i am thinking about how to get the best phase noise performance-
if i do it with a single loop pll i can roughly calculate phase
noise to be 20log(N) (with n= 150) for 3ghz output frequency at
-76dBc/Hz inside the loop bandwith with a perfect phase detector
found some publications using mixing techniques including comb
generators and pretuning and i wondered how they are superior to
the single loop solution with that division ratio.
if i use
a sampling phase detector with wide bandwith as a clean up loop
in a configuration of : comb generator driven at 20mhz and mix 3ghz
vco output with the 149th harmonic ( 2980mhz) what phase noise can
i expect ? - how strong is the influence of the phase noise of the
149 harmonic of 20mhz onto overall system noise ?? i calculate phase
noise of 2980mhz mixing signal to be -120dBc/hz -(43.5dB)=76.5dBc/Hz
am i right to say inside the loop bandwith i can expect phase
noise of my loop to be average of reference phase noise and harmonic
phase noise , calculating sqrt( -120dbc*76.5dBc)= 95.8dBc/hz
thus would result in an improvement of nearly 20dB comparing to the
single loop solution ?
does someone has got experience with phase
locked loops using harmonic locking ? how do i calculate the phase noise
i can expect if i use mixing / harmonic locking techniques ?
mny thanks for comments
_________________ regards
mac
Top
Tony Kurlovich Post subject: Posted:
Thu Nov 08, 2007 5:10 pm
Captain
Joined: Thu Oct
19, 2006 6:02 pm Posts: 7 The answer to your last question is
no. For this application, you can't average dbs. If you could, a large
PN could be reduced by adding a lesser PN.
If the phase noise
components to be added are not correlated, it would be the sum of the
noise powers. If the components are correlated (this case), it is the
sum of the noise voltages. This will yield the same answer as 20log(150).
Which is only for small deviation PM.
If your output is 20MHz*150,
0.1 Degree at 20MHz is worth 15 Degrees at 3000Mhz. You don't get one
without the other.
Posted
11/12/2012
|
|
|