Issue with current injection via ACSource in Brian

[appukuttan-shailesh]

There is a problem with the output of current injection via ACSource in Brian. The current profile is incorrect (as can be inferred from the voltage trace in below figure), and also the current injection fails to cease at t_stop.

The figure below illustrates this (for this source code):

Brian Output:

NEURON Output:

[appukuttan-shailesh]

The problems here are similar to those described in issue #437. The solution for ensuring that current injection stops at t_stop is same as that discussed there.

Brian requires time to be specified with units of seconds. Most of the required conversion is done, except for simulator.state.dt which is still specified in ms, and the frequency which is converted to 1/ms. To fix this, we need to update the following:

i) method _generate() in line https://github.com/NeuralEnsemble/PyNN/blob/master/pyNN/brian/standardmodels/electrodes.py#L114
from:

self.times = numpy.arange(self.start, self.stop, simulator.state.dt)

to:

self.times = numpy.arange(self.start, self.stop, simulator.state.dt * 1e-3)

ii) method _compute() in line https://github.com/NeuralEnsemble/PyNN/blob/master/pyNN/brian/standardmodels/electrodes.py#L117
from:

return self.amplitude * numpy.sin(time * 2 * numpy.pi * self.frequency / 1000. + 2 * numpy.pi * self.phase / 360)

to:

return self.amplitude * numpy.sin(time * 2 * numpy.pi * self.frequency + 2 * numpy.pi * self.phase / 360)

Brian output with above fixes:

[appukuttan-shailesh]

If the alternative fix for problem 1 in issue #437 is adopted, then the method '_generate' in line https://github.com/NeuralEnsemble/PyNN/blob/master/pyNN/brian/standardmodels/electrodes.py#L114 should be changed from:

def _generate(self):
        self.times = numpy.arange(self.start, self.stop, simulator.state.dt)

to

def _generate(self):
    self.times = numpy.arange(self.start, self.stop, simulator.state.dt * 1e-3)
    self.times = numpy.append(self.times, self.stop)

Note: Issue #445 would need to be fixed for this to work properly.

[appukuttan-shailesh]

Found two more issues with current injection via AC source, as seen in the below outputs:
Note: start = 47.5 ms and offset = 0.1

NEURON output:

Brian output:

The issues are:

1. Brian implementation evaluates current amplitudes with times from t = 0 ms, instead of from t = t_start
2. Brian implementation is not taking into consideration the parameter 'offset' while evaluating the amplitudes

Combined fix for all of the above

Methods _generate() and _compute() for the class ACSource in file https://github.com/NeuralEnsemble/PyNN/blob/master/pyNN/brian/standardmodels/electrodes.py
needs to be updated to:

def _generate(self):
        self.times = numpy.arange(self.start, self.stop + simulator.state.dt * 1e-3, simulator.state.dt * 1e-3)

def _compute(self, time):
        return self.offset + self.amplitude * numpy.sin((time-self.start) * 2 * numpy.pi * self.frequency + 2 * numpy.pi * self.phase / 360)

Brian output with above fix:

[apdavison]

Fixed in #467