Supersonic jets that are subject to off-design operating conditions are marked by three distinct regions in their far-field spectra: mixing noise, screech and Broadband Shock Associated Noise (BBSAN). BBSAN is conspicuous by the prominent multiple peaks. The Morris and Miller BBSAN model that is based on an acoustic analogy, offering a straightforward implementation for RANS, forms the foundation of the present work. The analogy model robustly captures the peak frequency noise, that occurs near Strouhal number of about 1, based on the nozzle exit diameter but leads to major sound under prediction for higher frequencies. In the jet mixing noise literature, it has been shown that an inclusion of frequency dependence into the characteristic length and temporal scales of the effective noise sources improves the far-field noise predictions. In the present paper, several modifications of the original Morris and Miller model are considered that incorporate the frequency dependent scales as recommended in the jet mixing noise literature. In addition to these, a new mixed scale model is proposed that incorporates a correlation scale that depends both on the mean-flow velocity gradient and the standard mixing noise-type scaling based on the dissipation of turbulent kinetic energy. In comparison with the original Morris and Miller model, the mixed scale model shows considerable improvements in the noise predictions for the benchmark axisymmetric convergent-divergent and convergent jets. Further to this validation, the new model has been applied for improved predictions for elliptic jets of various eccentricity. It has been shown that,