Hi All,

I am simulating a two phase Switch-Cap sample and hold amplifier, and trying to compare the performance between using the ideal switches (G-element) and the real switches (transistors). When ideal switches are used, equivalent capacitance is also added to each node to model the junction capacitance of the real switches. Then I print out the node volatges of the output, and do fft in matlab. However, it turns out that the SNDR and SFDR of the output is better when using real switches. Did anyone experience this before?

In Hspice, if the convergence method=gear is used, the performance will be better with the ideal switches compare in two cases. But the noise floor has risen dramatically (10dB).

So I am wondering which result should I use or trust? Did anyone experience this situation before? Any help would be greatly appreciated.