Lambda rule for fast spiking axon

[vladimirov]

Hi,
Would the standard d_lambda rule still applicable for a fast spiking axon, when interspike intervals are ~1.5-2 ms? The generic formula suggests

Code: Select all

nseg = int((L/(0.1*lambda_f(100))+.9)/2)*2 + 1

is it still good for an axon which conducts trains of high-frequency spikes (>300 Hz)?
Thank you!

[ted]

Appropriate discretizations for time and space are always empirical issues. The d_lambda rule for spatial discretization generates a very good first guess, and it often produces a satisfactory result. However, it is always a good idea to verify. You might want to do a control run with the discretization produced by the d_lambda rule, then
forall nseg*=3
(which should reduce spatial error by almost an order of magnitude),
then do a test run and compare it against the control result.
This recommendation has nothing to do with whether the model in question has a fast spiking axon--such a test is advisable regardless of what kind of model is being studied.

By the way, I see that hoc code exported from a recent version of the CellBuilder uses a slightly different formula for nseg--
nseg = int((L/(0.1*lambda_f(100))+.999)/2)*2 + 1

[vladimirov]

Ok, thanks!