Pipelined recursive filter : the recursive greedy way of design

• Main Author: Ong, Van Vinh
• Other Authors: Lim Yong Ching
• Format: Final Year Project
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Signal processing
• Online Access: http://hdl.handle.net/10356/62082
• Institution: Nanyang Technological University

In signal processing, digital filter is a small system but it is so essential that each chipset nowadays would have millions (even billions) of digital filter inside. The filter system is used to perform mathematical operations on a signal in order to enhance certain aspect of that signal. Therefore, if the speed of each filter can be increased just a bit, the computational speed of the whole system can increase significantly. That is the reason why there is always a need to design a high-speed digital filter. Pipelining is one effective technique to increase the computing speed of digital filters. The higher the degree of pipelined segmentation is, the faster the speed of processor could become. Develop further from Minimum Order Augmentation method, this FYP report will introduce and discuss a new method of converting a consequential transfer function into pipelined transfer function called recursive greedy method. For any stable digital filter, the poles of its transfer function must always be inside the unit circle. Thus, to convert the consequential transfer function into pipelined transfer function, we must do it in the way that all the poles will eventually stay inside the unit circle. We adopt a greedy algorithm in which we minimize the largest absolute value of root in every step until that largest absolute value stays inside the unit circle. As it is “greedy”, we can only receive a good pipelined transfer function but couldn’t be the best (and the result could be bad in many cases). Therefore, the greedy algorithm will be applied recursively to create possibly the best-pipelined transfer function.