Far-field inlet fan noise can usually be measured under anechoic conditions in some model scale fan rig test facilities but far-field aft fan noise measurements are often not possible because the bypass flow is ducted away through a throttle into an exhaust stack. A beamformer-based technique has been developed for processing measurements taken with an in-duct axial ('phased') array in the bypass duct, which with certain modelling assumptions, enables the fan broadband noise level and directivity to be predicted in the far-field. Validation with a realistic fan noise source has been partially achieved by using experimental data from the NASA ANCF low-speed fan rig with flow up to Mach 0.17. The modal transfer functions are computed using a 'plug' flow exhaust model based upon a well-established Wiener-Hopf farfield technique ('GXMunt') but since the measured far-field data is located at three diameters, modal transfer functions are also computed at that distance and with a spreading jet model using a linearised Euler code ('FLESTURN'). Both models yield predictions that agree reasonably well with measured data but the latter is more accurate at small angles to the jet axis. To complete the validation of this method, data is required at the higher Mach numbers corresponding to the aircraft approach, cut-back and sideline conditions, with realistic bypass nozzle geometry.