| Description: |
test_spike_generator checks the spike_generator device and its
consistency with the nest simulation kernel.
/start specifies the earliest possible time stamp of a spike event to
be emitted. /stop specifies the earliest time stamp of a potential
spike event that is not emitted.
spike_generator is a device emitting spike events restricted to the
time grid spanned by the computation step size h. Nevertheless, the
time stamps of the spike events to be emitted are specified by device
property /spike_times as double floats in milliseconds (ms). The
spike_generator automatically assigns the appropriate time stamps on
the simulation grid. For an arbitrary spike time s this is t=i*h with
s in the interval (t-h, t], the corresponding event with time stamp t
is emitted in the simulation step t-h -> t. See test_iaf_i0_refractory
for consistency with spike generation in integrate-and-fire type neuron
models and test_iaf_dc_aligned_stop.sli for consistency with the
generation of grid based dc currents.
A a consequence of the restriction of spike events to the simulation
grid, multiple spike events with identical time stamps may be emitted.
The results at different computation step sizes (resolutions) require
some further discussion. Let us assume that /start and /stop are at
joined grid positions of all resolutions tested. If all spike events
are specified on joined grid positions, the simulation results are
inependent of the computation step size. However, if spikes occur on
non-joined grid positions, spike times are shifted to the appropriate
position on the current grid and simulation results will differ. Also,
the spike count cannot be preserved because spikes at the end of the
interval may be shifted to grid position /stop which is not a valid
time stamp of a spike event anymore.
Note that the number of emitted spike events can always be made
independent of the computation step size by setting /stop to an
appropriately late joined grid position.
The usage of values for /start and /stop that are not simultaneous
grid positions of the computation step sizes to be used should be
avoided because the set of spike events delivered in the simulation
will vary in complex ways.
The expected output is documented at the end of the script.
|