In this work, growth-rate curves of the 2 f1-f2 distortion product otoacoustic emission (DPOAE) are analyzed in a population of 30 noise exposed subjects, including both normal-hearing and hearing impaired subjects. A particular embedded limit-cycle oscillator equation is used to model the cochlear resonant response at the cochlear places of the primary and secondary tone frequencies (f2 and 2 f1-f2). The parameters of the oscillator equation can be directly interpreted in terms of effectiveness of the cochlear feedback mechanisms associated with the active filter amplification. A two-sources paradigm is included in the model, in agreement with experimental evidence and with the assumptions of more detailed full cochlear models based on the transmission line formalism. According to this paradigm, DPOAEs are nonlinearly generated at the cochlear place that is resonant at frequency f2, and coherently reflected at the 2 f1-f2 place. The analysis shows that the model, which had been previously used to describe the relaxation dynamics of transient evoked otoacoustic emissions (TEOAEs), also correctly predicts the observed growth rate of the DPOAE response as a function of the primary tones amplitude. A significant difference is observed between normal and impaired ears. The comparison between the growth rate curves at different frequencies provides information about the dependence of cochlear tuning on frequency.