Run-time reconfigurable adaptive signal processing system with asynchronous dynamic pipelining: a case study of DLMS ADFE

Abstract

Most pipelined adaptive signal processing systems are inherently subject to a trade-off between throughput and signal processing performance because of the adaptation feedback loops. To mitigate this dilemma, we propose to apply an asynchronous pipeline to implement pipelined adaptive signal processing systems that can support run-time reconfigurable throughput/performance trade-offs. This can be leveraged to improve the overall system performance in many applications. In this work, we demonstrate this design approach using a delayed LMS (DLMS) adaptive decision feedback equalizer (ADFE) for the communication links with variable data rate. By adjusting the pipelining depth on-the-fly, the DLMS ADFE can dynamically track the best equalization performance allowed by different data rates. We develop the design scheme that applies a zero-delay-overhead asynchronous pipeline style to implement ADFE and support dynamic pipelining depth control. Simulation result shows a significant performance improvement compared with its synchronous counterpart.