ballstick.ses is a model cell that is a ball & stick approximation 
to the charging properties of the pyramidal cell that comes with 
NEURON's demo program. Its specification is:
soma { nseg=1  L=60.6576  Ra=160
	diam=60.6576
	cm=1
	insert hh { gnabar_hh=0.12 gkbar_hh=0.036 gl_hh=0.0003 el_hh=-54.3}
}
dend { nseg=37  L=1632  Ra=160
  diam = 4.3
	soma connect dend (0), 1
	insert pas { g_pas=0.001 e_pas=-65}
	cm=1
}

init_rc_comp3.hoc loads ballstick.ses and rc_comp3.ses,
in which the electrode is represented by an equivalent T 
(a "better" electrode model).
Adjust the gain of A2 to compensate for series resistance, 
and Af's gain to compensate for electrode capacitance.
The potentials at the ungrounded ends of Ce and Cf must be properly 
initialized, or spurious current will be injected at t = 0.
The correct initializations are 
Vx    v_init
Ve    v_init
Vy    5*v_init (i.e. gain of A1 * gain of Af * Ve)
Vo    v_init
In this implementation, Cf is 1/10 of Ce, and Rf is set so that 
the product Cf Rf equals the electrode time constant Ce Re/4.
Other values of Cf can be used, but the time constants must match
or else performance is degraded--ringing ensues, or the observed 
pulse looks "good" while the spike looks "bad" etc..
This circuit illustrates the impossibility of eliminating all 
artifact even if the amplifiers are perfect.
Try adjusting the gain of A2 and see what happens to the 
artifact at the start and end of the current pulse as you 
null out the resistance.  Also try changing the frequency 
response of A1, Af, or A2.

