Designer’s Guide Community

New Member

Offline

Posts: 3

Charge Pump Noise Simulation
Feb 23^rd, 2015, 5:40pm

Hi,
I have a question about the noise extraction of a charge pump + PFD. The published tutorial states to extract the current noise of CP when applied an offset phase to the PFD, and then convert noise current back into time Jitter by multiplying by Period/Kdet

For example, the noise is 10nA, the Period is 10ns, And the Charge Pump Delivers 50uA when turned on. That makes jitter = (10ns * 10nA)/50uA = 2ps?

And if so, this implies that if I double the period, the jitter will double, which does not make sense to me.

Thanks.

Back to top

IP Logged

Ken Kundert Global Moderator Offline Posts: 2384 Silicon Valley	Re: Charge Pump Noise Simulation Reply #1 - Feb 24^th, 2015, 11:25am What you are computing is an input referred noise. By doubling the period you are also halving the gain of the PFD. -Ken
Back to top	IP Logged

Mehdi Abderezai

New Member

Offline

Posts: 3

Re: Charge Pump Noise Simulation
Reply #2 - Feb 24^th, 2015, 11:35am

Hi,
Thanks for your reply Ken, yes I see that the gain of the PFD is halved when the period doubles, but then closed loop sim (verilog A simulation) would account for that and increase the observed jitter.

If do both, I increase the input referred PFD jitter and put it in the closed loop simulation, I feel like I would be double counting the jitter of the PFD.

-Mehdi

Back to top

IP Logged

Ken Kundert

Global Moderator

Offline

Posts: 2384
Silicon Valley

Re: Charge Pump Noise Simulation
Reply #3 - Feb 25^th, 2015, 3:56pm

The point is that jitter is easier to model than noise. The signal at the input of the PFD/CP is time whereas at the output it is current. In your loop simulations you need to choose where you are going to inject the noise for the PFD/CP. It is easier to model and simulate if you inject it at the input in the form of jitter than if you inject it at the output as a noise current.

-Ken

Back to top

IP Logged