Dear Raghunathan,

unfortunately, filtering will not only remove edges, but also exact shape of overall curve:

Assume you want perform a rising linear edge, changing actual value by 1 bit (50msec) every 50msec, from voltage of 0V on. If you then are interested in, that curve after 2sec gives a voltage of exact 2V, then overall shape of curve must not be influenced by filtering.

If you put an RC filter at output you will see the following:



The red curve shows the performance of a RC filter, where R and C are choosen in such the way, that overall curve remains unchanged (5 x R x C = 50msec). You see, the effect on smoothing is rather poor.
If you try to increase RC time constant, in order to improve smoothing (blue curve), then overall curve is relevantly changed, means after 2sec the actual output voltage will be much less than 2V!

This performance cannot be improved by the use of higher order filters. This comes from the fact, that your overall curve, means the linear ramp, contains lots of harmonics. These harmonics are necessary to build the linear ramp. If you provide relevant filtering to smooth the little edges, then you will also remove the harmonics of linear ramp and you will change the overall curve!

By the way, if you would use Butterworth or Chebycheff filters, then you will end up with ringing, whenever an edge occurs! The result will be a little overshot with every edge. Such a high order filter will not contribute to any smoothing, but will even worsen situation!!

There's another option: If you do not want to have changes with your overall curve of greater than 50mV per 50msec, then you could add a little integrator, like the following:



I have the time contants already designed for your application. The effect on output voltage would be the green curve of:



Kai