Purpose Although heart transplantation is the gold standard for advanced heart failure, donor shortage is driving the increasing use of mechanical circulatory support such as left ventricular assist devices (LVADs). Whilst providing an effective therapy, device complications persist and diagnosis remains challenging. Improved non-invasive methods of assessing LVAD function and pathological states are needed. Methods Between 09/05/2018 to 29/08/2018, 17 HeartWare HVAD outpatient patients and those admitted with pump complications were recruited from the St Vincent's Heart and Lung Clinic. A laboratory-based mock circulatory loop was used to simulate pump speed changes, suction states, outflow graft kinks, and haematocrit changes. An electronic stethoscope was used to record audio signals of the pump in vivo and in vitro. Data was analyzed using Welch's power spectral density estimate in MATLAB where amplitude versus frequency graphs were generated, as well as with IBM SPSS. Results Acoustic signature of Heartware was characterised by up to 24 harmonic peaks [31.25-1121Hz] where latter peaks occurred at frequency multiples of the first [31.25-46.88Hz]; this is supported by previous research where all peaks were proposed to be harmonic in nature. Of all peaks, 6 were most prominent [31.25-558.6Hz] and coincided with the 6-peak pattern found in previous research. Harmonic peaks 19, 21, 22 and 23 in the predicted frequency ranges of 570-886.67Hz, 630-980Hz, 660-1026.67Hz, and 690-1073.33Hz were absent. Peak frequency values measured in vivo and in vitro correlated with each other. The frequency distribution of harmonic peaks was positively transposed along the x-axis with increasing speed (linear, R2>0.961). Increasing speed had varying effects on the amplitude of each peak. Inducing suction states increased the amplitude of most peaks except 4 peaks [121.1Hz, 199.2Hz, 238.3Hz, 320.3Hz]. Increasing outflow graft kinks increased the amplitude of all peaks. Increasing haematocrit increased the amplitude of most peaks except 6 peaks [39.06Hz, 78.12Hz, 160.2Hz, 718.8Hz, 800.8Hz, 960.9Hz]. Conclusion With a better understanding of LVAD acoustic signature, deviations can be identified early and used to predict pump malfunction or pathological states. Pump audio signals have the potential to be utilized as a non-invasive method of assessing pump function in the management of LVAD patients.