Berti wrote on Aug 28th, 2009, 5:05am:Ron needs to be small enough to satisfy the settling/linearity requirements. You can do hand calculations to get an idea.
The settling would be determined by the total Ron, but I think he is looking for the optimal
relative sizes of the two switches.
I hadn't seriously analyzed it, but I would think you would want the top switch resistance to be as small as possible (i.e. make it a wide switch). This is because its resistance is signal dependent so you want to minimize its effect. In addition, charge injection effects are minimized because the bottom switch is open so there is little penalty to a large switch.
On the other hand, the bottom switch contributes to the charge injection. Most is common mode and signal independent, but some isn't. You would want to minimize its effects by making the switch as small as possible... that is, you would want the bottom switch resistance to dominate the settling time.
I'm just rattling this off the top of my head, but the above seems to make sense. It doesn't seem to be consistent with the plot. Ideas?
rg