Sunday, October 2, 1:30 pm — 3:00 pm (Rm 409B)
Jon Boley, GN Hearing - Chicago, IL, USA
P26-1 Discrete-Time Implementation of Arbitrary Delay-Free Feedback Networks—Dave Berners, Universal Audio; CCRMA, Stanford University - Stanford, CA, USA; Jonathan S. Abel, Stanford University - Stanford, CA, USA
The delay-free feedback loop can be directly implemented in discrete time by separately discretizing the forward and backward transfer functions and simultaneously solving the resulting linear system for the nodes connecting the filters within the loop. The ability to form the solutions rests upon the fact that, at sample n, the output of a discrete-time linear system is a linear function of the input to the system at sample n. This technique allows for relatively simple calculation of coefficients for certain time-varying feedback systems, and allows for inclusion of memoryless nonlinearities inside feedback loops. We show that the technique can be generalized to discretize an arbitrary network of LTI systems arranged in multiple-loop feedback networks. Two examples are presented: one time-varying system and one nonlinear system.
Convention Paper 9685 (Purchase now)
P26-2 The Time-Varying Bilinear Transform—Jonathan S. Abel, Stanford University - Stanford, CA, USA; Dave Berners, Universal Audio; CCRMA, Stanford University - Stanford, CA, USA
The discretization of continuous-time systems is considered, and an extension of the bilinear transform to the case of time-varying systems is introduced. Termed the “time-varying bilinear transform,” the transform generates a sequence of digital filter coefficients in response to continuous-time system changes that keeps the digital filter state compatible with the changing digital filter coefficients. Accordingly, transients in the digital filter output that don’t appear in the continuous system output are avoided. For an Nth-order continuous-time system, a step change in the system produces a sequence of N intermediate sets of digital filter coefficients, bracketed by what would be generated by the bilinear transform applied to the initial and final systems. Sequences are tabulated for direct and transpose canonical forms and first-order and second-order systems, and examples of first-order and second-order analog filters with time-varying components are presented.
Convention Paper 9686 (Purchase now)
P26-3 Active Equalization for Loudspeaker Protection—Christopher Painter, Marvell Semiconductor, Inc. - Santa Clara, CA, USA; Kapil Jain, Marvell Technology Group Ltd. - Santa Clara, CA, USA
We present a time-varying linear equalization algorithm whose purpose is to protect a loudspeaker from damage under high drive conditions. It is suitable for implementation on a low-cost digital signal processor, often integrated on the same die as a high-performance audio codec. A typical application is in a portable wireless (e.g., Bluetooth) loudspeaker. For a given driver and enclosure design, the algorithm allows the power output of the loudspeaker to be maximized while introducing only minimal coloration or distortion. During the loudspeaker design phase, the parameters of the algorithm can be easily tuned by the designer, further optimizing the overall design for power output, robustness and low distortion.
Convention Paper 9687 (Purchase now)