Variable step size per default?

[miller]

Hi.

Can somebody help me to explain the following:

I have a section with one segment with total membrane capacity of 1pF and a point process in it, that injects NONSPECIFIC_CURRENT i in the way i=gv.
I set v_init to 10mV, g is very large, let's say 1S and dt=0.1ms.

I would expect that after 0.1 ms v would reach the value of -1MV, because it would assume v to be constant over the time intervall of dt.

But instead NEURON solves correctly the equation c*v'(t)=-g*v(t) and after 0.1ms v=0mV.

How does it work? Does NEURON use variable step size per default?

Thanks

[ted]

Very good question, Konstantin. NEURON's default integration method is
implicit ("backward") Euler, which has the distinct advantage that when dt is
much larger than the largest time constant in a passive system, the states of
that system reach their "infinite time" values in a single integration step. Neat, huh?

Numerical integration, as it pertains to computational modeling of neurons, is
discussed in some detail in chapter 4 of the NEURON Book
http://www.neuron.yale.edu/ftp/ted/book ... xedref.pdf
where you'll also learn why NEURON does not use the explicit/i] ("forward)
Euler method at all--unlike some other neural modeling software, which use
forward Euler by default.