subgold wrote on Nov 12th, 2008, 4:18am: We run a pss+pxf and get one transfer function, but there should be different transfer functions for each periodic state.
Don't confuse the individual states that the amp will be in from the discrete time transfer function. Take the example of an simple integrator. During the sample time, the amp just holds its previous value and is disconnected from the input, i.e. H=0. During the transfer phase, it looks just like a standard amplifier, i.e. H=C
s/C
i (sample cap/integrate cap). However, from the charge equations the 'real' discrete time transfer function is H=C
s/C
i*1/(1+z
-1). That is very different from the individual state transfer functions, which are useless in and of themselves. The PSS and all of the other P-analyzes return information about the combined discrete time equations.
Quote:Since the amplifier only properly operates in one clock phase (the amplifier is disconnected from the CMFB in the other phase)
That's not true since the hold caps are still in the circuit. Even a SC CMFB must maintain some type of negative feedback through all phases of operation.
Quote:How to obtain the information only from the amplification phase?
You will most likely need to use an ideal track and hold that is discussed in Frank's reference.
Quote:how to simulate the small-signal gain of an amplifier which uses a typical switched capacitor CMFB?
Another thread with the same question:
http://www.designers-guide.org/Forum/YaBB.pl?num=1223636443