Contents 1. INTRODUCTION
A rank order filter (ROF) is a generalization of several common nonlinear and linear filtering operations which include the min, max, median and average operators among others [1] – [11]. High-speed analog ROFs can find applications in various fields such as real-time nonlinear image processing, speech processing, data compression, communications, neural and fuzzy networks, etc. In an ROF filter with N input signals, the k-th rank order signal is defined as the k-th signal in increasing order of amplitude. For an N-input ROF, the min case corresponds to , the max case to and the median case to . Three basic approaches have been proposed for the analog implementation of ROFs: a) feedforward systems based on multistage sorting min and max circuits that order signals in increasing order of magnitude [6] – [8]. In this approach, the number of stages increases with the number of input signals N b) Continuous-time negative feedback approaches based on high gain transconductors with symmetrical output current saturation characteristics [9]. This approach provides the kth rank order signal in just one step and c) Sample data approaches based on capacitive threshold logic [10] – [11]. The hardware complexity of this approach grows rapidly with N and it has an inherent speed limitation due to sample data operation. Two stage transconductance amplifier: (a) Internal structure. (b) Symbol (c) DC Transconductance characteristic.
