raja.cedt wrote on Sep 14th, 2009, 1:23am:hi,
if you fix gm for noise reduction how you will get desired bandwidth?
I can't tell if this was in response to my comment so let me expand. In an MDAC there are 2 major sources of noise: 1) Thermal noise from the opamp and 2) thermal noise from the sampling switch (i.e. kT/C noise). To decrease opamp thermal noise you need to increase the current. To decrease the kT/C noise you increase the capacitance. But by doing this you ALSO have to increase the opamp current so that it can drive the additional load capacitance.
So regardless of the noise source you have to increase the opamp current.
With MDACs it has been my experience that kT/C noise dominates. This means the design flow is: 1) Pick sample cap size based on kT/C noise. 2) pick opamp current based on need to drive the sample cap.
So, to optimize you should choose the caps sizes, which determines the current. Not vice-versa.
FWIW Chiu/Gray's 2004 JSSC work touches on optimizing MDAC scaling on down the pipeline.