Dear all,
    I have a question about using PSS and Pnosie to simulate an auto-zeroed SC integrator in sigma-delta modulator.
I want to know the 1/f noise corner of an auto-zeroed SC integrator
I have tried three different ways.

All simulations are done with zero input and 5V clock signal.

1) SC integrator (with Cs and Cint), normal connection:
Choosing noise type to "time domain", I can only plot output noise. Thus I need to divided it by the transfer function.

Thus I simulated the PAC simulation, however I can not get
the expected transfer function of the SC integrator. I use SC CMFB.
If I change it to CT CMFB. I can get the PAC transfer function I want.


2) SC integrator (with Cs and Cint), close loop by adding an S&H to feedback the output of the integrator to the input during the next integration phase.

This is described in Ken's paper about device noise simulation in Sigma-delta.
Since it's a unity gain feedback, the output noise will equal to the input noise.
PAC transfer function = 1 at low frequency  


3) SC integrator (with Cs and Cint), normal connection, and then followed by a S&H, not close loop. This technique is described in Ken's paper about SC-filter. I only sample the point at the end of the integratotion phase.

For these three techniques, I got different 1/f noise corner, I am not sure which one is the correct way to simulate
the input inferred noise for auto-zeroed SC integrator.

Is there anyone who knows about this problem?
Thank you very much in advance and I appreciate it.

pnoise plot of SC_integrator

pnoise plot of SC integrator (close loop) with Sample & hold

Hi,

I am also simulating the auto-zeroed SC integrator, and I have come up with similar results.

To my greatest surprise is that the noise floor of open-loop simulation and closed-loop simulation are different. I also saw that in your simulation result: about 250 nV/sqrt(Hz) for the open loop simulation and about 600 nV/sqrt(Hz) for the closed loop simulation.

For me, the noise floor of the two simulations are also different: 80 nV/sqrt(Hz) for open-loop and 275 nV/sqrt(Hz) for closed-loop. I am using 10 pF Cs and sampling at 5 Mhz.

I wonder why this is the case.