Nonlinear Mixing System

This thesis presents a novel massively mutual compression system for use as a nonlinear mixing tool. In this interdisciplinary work, the system is explained and explored from the perspectives of both audio engineering and music production aesthetics. First, the traditional method for presenting the time-frequency energy distribution of musical audio signals, the short-time Fourier transform, is assessed. Following that, an alternative, the constant-Q transform, is suggested as a better approach to illustrating variations in time-frequency content of audio signals. By exploring the architecture of digital compressors, a method to create a bilateral cross-adaptive architecture is proposed and created in the Max/MSP visual programming language. This platform is chosen because of its routing flexibility and performance. In this thesis, it was found that commercially available digital audio workstation applications are not able to support the feedback cross-modulation which is crucial for this architecture. The behaviour of the system, in particular a ‘double knee’ curve, is characterised, verified, and explained by a series of tests. Then, two original audio examples are presented which are mixed in different ways to highlight how the system developed in this research can be utilised. These examples highlight some unique properties of this system. The first is the lack of one perpetually dominant element in the mix. Secondly, constant changes are introduced into the tonal balance of the mix. The research also found that the ‘top down’ approach to mixing is the most intuitive approach to using the system. Finally, user tests were conducted involving 15 participants. The experiment found that users will most likely use the architecture as a mixing tool to reduce masking. Furthermore, most of the participants agree that there are advantages of utilising this tool in ther production. Additionally, a majority of the respondents have discovered mixing technique that are unique to the nonlinear mixing system.