Despite a significant number of new structures in the past few years, MEMS loudspeaker still are not competitive in terms of performance compared to non-MEMS loudspeakers for free field applications. For industrial perspectives, a high sound pressure level on a wide frequency band is required, as well as a low total harmonic distortion. To widen the frequency range of MEMS loudspeakers, we propose to separate the actuating element from the radiating one, in order to separate design constraints to reach an optimal structure. In this paper, the lumped element model of the loudspeaker in presented, as well as the innovative manufacturing process. Finally, the computed frequency response is compared to the measured one. At the resonance, pressures as high as 110 $dBSPL$ at 1 kHz and at 10 mm are reported for an active surface of 36 mm$^2$, which is above the known state of the art for a loudspeaker with similar dimensions. Also, the flatness of the radiated sound pressure in a wide frequency range is closer to the ideal frequency response of loudspeakers than other MEMS loudspeakers, due to the piston mode of the moving rigid plate of the loudspeaker. The total harmonic distortion, mainly due to the nonlinearity of the piezoelectric transduction, is below 5% for reasonable sound pressure levels in the usable frequency band. The use of digital signal processing and of a dedicated packaging will allow our loudspeaker to advantageously replace the main or secondary one in smartphones.