yvkrishna wrote on Apr 25^th, 2009, 1:20am:


Why do you expect an "exactly exponential" response ?
This is the case for a first order system only. Of course, all other poles influence the step response. When they come closer to the dominant pole, the step response will exhibit overshoot and ringing. That´s pure system theory.

For my opinion, the response as shown in the figure looks rather good, don´t you think so ?
There is no overshoot ; the settling time to reach the final value may come from an pole-zero-doublet which is not exactly matched.

yvkrishna wrote on Apr 25^th, 2009, 3:46am:


What do you really need ? 4th order filter design ?
It´s not clear to me. Sorry.

Vamshi,

  Have you run Monte Carlo analysis on your circuit? You are
relying on precise matching of the poles and zeroes to cancel
each other. Small misses in the cancellation result in doublets
that degrade settling.

                                              Best Regards,

                                                  Sheldon

yvkrishna wrote on Apr 25^th, 2009, 1:20am:


This would depend on what kind of ADC you are making and what inputs you are giving it.

For a pipelined or cyclic or other form of switched-capacitor-based ADC, the above response may be unacceptable because the slow settling of the doublet would either degrade performance too much, or force you to operate at lower speed and/or higher current levels.

For a continuous-time delta-sigma modulator, the above response may not be an issue, provided that you have some pre-filtering that prevents your modulator's STEP RESPONSE from being excited. As long as the input consists of a combination of finite sine wave inputs and the overall signal level is small (all very diffuse terms as you notice), your amp may be useful. As soon as the step response of the amp comes into play or your amp slews, you are in trouble.

Regards,

Vivek

Hi Vamshi,

A single-pole settling can be achieved with higher phase margin provided there are no doublets in the unity gain band. Doublets may not affect stability very much but they do impact the step response appreciably, as you have found out.

Typically, doublets and nondominant poles etc. can be minimized by making the opamp as simple as possible (folded cascode/telescopic). I would recommend this course. Also, would be better off trying to find where these pole/zero pairs come from and pushing them out of band. This would be better than trying to analytically account for the effects of incomplete settling as this may be very dependent on process, supply and signal levels, and you will likely find it difficult to achieve consistent performance.

Regards,

Vivek