Ineffective simulation with small duration current injection

NMODL and the Channel Builder.
Post Reply
doctorxue
Posts: 2
Joined: Tue Feb 27, 2018 10:54 pm

Ineffective simulation with small duration current injection

Post by doctorxue » Fri Aug 24, 2018 12:03 pm

hi everyone:

I'm a freshman in NEURON and recently encountered a problem.

When I was doing current injection(with h.IClamp method) into a 99-divided 20um diameter spherical soma inserted with 'pas' membrane mechanism, if duration is set to like 0.1ms and amplitude is 1nA, I got observable voltage response. But if the duration is 0.01ms and amplitude is 1000nA, no observbable response, the voltage does not change even 1 mV with time.

I would like to know whether this is limitation of time-step simulation, or some real biophysical phenomenon.

Thank you so much!

ramcdougal
Posts: 138
Joined: Fri Nov 28, 2008 3:38 pm
Location: Yale School of Medicine

Re: Ineffective simulation with small duration current injection

Post by ramcdougal » Fri Aug 24, 2018 12:38 pm

Are you using a fixed time step or variable? I assume variable step will choose time points at the beginning and end of each current change... And thus you would see the current injection having an effect.

If fixed, what is dt?

Note that for fixed step, the default timestep (0.025 ms) is longer than your current pulse, so you could easily have a situation with fixed step where no current should be injected at any of the calculated time points.

[Edit: fixed typos.]
Last edited by ramcdougal on Fri Aug 24, 2018 5:28 pm, edited 1 time in total.

ted
Site Admin
Posts: 5351
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: Ineffective simulation with small duration current injection

Post by ted » Fri Aug 24, 2018 3:07 pm

Whoa, ramcdougal, disable auto spellcheck!

doctorxue, what you observed is exactly what happens in a fixed time step simulation when the stimulus duration is shorter than dt. If you were using adaptive integration, NEURON's computational engine would have automatically changed dt to accurately capture the time course of the stimulus current. With fixed time step integration, any attempted perturbation must be at least as long as the time step, or it will have no effect.

Post Reply