The noise looks like a small ceramic disk, say .01uF or .1uF might help.
Larger caps will affect your goal to get to transfer function as the cap
will contribute to system dynamic behavior as its an energy storage
element.
An alternative way is, if scope is DSO, and has ability to export capture
data to post process it in excel or MatLab. With a LPF or running average
filter.
Or use DSO to average several step response cycles....
Regards, Dana.