Opamp specification in reconfigurable pipelined ADC

[predator89]

Hello,

I am designing a reconfigurable pipelined ADC with varing sampling rate and resolution.
Can anyone tell me how the opamp specifiation of individual pipeline stage would vary with resolution and sampling rate of ADC?

Cheers

 

[dick_freebird]

Offset voltage and gain error terms need to support
LSB accuracy.

Bandwidth / settling time need to be well better than
sampling rate, and settling time is to a very low
residue error.

You would have to design the amplifier for the worst
case (max bits, max bit rate) or cases (if your scheme
varies them independently or according to some built
in trade-logic). Probably for worst-worst, and shed
power smartly as you back off along either axis.

 

[predator89]

As per the theory in most of the papers i found that by scaling sampling capacitor for the lower stages of pipeline we can get power reduction.
But as I am measuring static power consumption in my ADC, I did'nt find any change in power by scaling capacitors.

Please guide me where I am going wrong with my understanding.

 

[dick_freebird]

"Static" power (as in, nothing at all is switching)
will push no displacement current through any
capacitors, so no impact.

If you were doing a transient simulation with the
sampling etc. active, there might be some difference.
But if the input value remains constant then the
capacitors' charge may as well, again, no difference.

A simulation where the analog input varies between
two widely separated values, might be the best at
showing sampling-capacitor-value sensitivity.

In any case the Idd is probably dominated by other
things such as rollup Cdg*Vdd*f from all of the
digital "stuff", and at advanced nodes perhaps the
baseline drain current leakage along with explicit
static analog bias currents. Do a calc of the sum
of sample capacitors, C*Vdd*f, as a worst case
contribution of these caps to the net supply current.
Reality would be less except in the extreme-repetitive-
input-variation case.

"Reduction" there may be. That's science, academic.
How much, and what else it costs you, that's
engineering.