The clever designs of natural transducers are a great source of inspiration for man-made systems. At small length scales, there are many transducers in nature that we are now beginning to understand and learn from. Here, we present an example of such a transducer that is used by field crickets to produce their characteristic song. This transducer uses two distinct components—a file of discrete teeth and a plectrum that engages intermittently to produce a series of impulses forming the loading, and an approximately triangular membrane, called the harp, that acts as a resonator and vibrates in response to the impulse-train loading. The file-and-plectrum act as a frequency multiplier taking the low wing beat frequency as the input and converting it into an impulse-train of sufficiently high frequency close to the resonant frequency of the harp. The forced vibration response results in beats producing the characteristic sound of the cricket song. With careful measurements of the harp geometry and experimental measurements of its mechanical properties (Young's modulus determined from nanoindentation tests), we construct a finite element (FE) model of the harp and carry out modal analysis to determine its natural frequency. We fine tune the model with appropriate elastic boundary conditions to match the natural frequency of the harp of a particular species—Gryllus bimaculatus. We model impulsive loading based on a loading scheme reported in literature and predict the transient response of the harp. We show that the harp indeed produces beats and its frequency content matches closely that of the recorded song. Subsequently, we use our FE model to show that the natural design is quite robust to perturbations in the file. The characteristic song frequency produced is unaffected by variations in the spacing of file-teeth and even by larger gaps. Based on the understanding of how this natural transducer works, one can design and fabricate efficient microscale acoustic devices such as microelectromechanical systems (MEMS) loudspeakers.