Designing stimulation signals for neuroprosthetic devices can be cast as an optimal control problem. Rectangular Lilly-type stimulation waveforms have been used extensively in these devices. In this paper, we rigorously formulate the charge optimization problem from a control perspective and distinguish between monophasic and biphasic stimuli. We show that for a monophasic stimulus, the important factor in stimulating a neuron is the total delivered charge per unit cell membrane. This factor is a consequence of the subthreshold linear behavior of the neural membrane. On the other hand, biphasic pulses, which are ubiquitous in the neuron stimulation context, can stimulate a neuron in its non-linear range, thereby challenging the finding that total charge delivery is the only critical factor. As a result, there may be even more optimal stimulus profiles than Lilly-type rectangular waveforms for biphasic stimulation. Furthermore, solving the charge minimization problem also will reduce the risk of electrode corrosion, which is an important factor in improving the performance of neuroprosthetic devices.