Neurons of the avian nucleus laminaris (NL) compute the interaural time difference (ITD) by detecting coincident arrivals of binaural signals with submillisecond accuracy. The cellular mechanisms for this temporal precision have long been studied theoretically and experimentally. The myelinated axon initial segment in the owl's NL neuron and small somatic spikes observed in auditory coincidence detector neurons of various animals suggest that spikes in the NL neuron are generated at the first node of Ranvier and that the soma passively receives back-propagating spikes. To investigate the significance of the "passive soma" structure, we constructed a two-compartment NL neuron model, consisting of a cell body and a first node, and systematically changed the excitability of each compartment. Here, we show that a neuron with a less active soma achieves higher ITD sensitivity and higher noise tolerance with lower energy costs. We also investigate the biophysical mechanism of the computational advantage of the "passive soma" structure by performing sub- and suprathreshold analyses. Setting a spike initiation site with high sodium conductance, not in the large soma but in the small node, serves to amplify high-frequency input signals and to reduce the impact and the energy cost of spike generation. Our results indicate that the owl's NL neuron uses a "passive soma" design for computational and metabolic reasons.