Download in PDF Format

For a PDF-Viewer click here ...

Multiple-input min/max circuits, median filters, and rank based filters are finding increasing applications in nonlinear signal processing (particularly speech and image data processing), data acquisition, fault tolerant systems, and power electronics. Many of these applications would previously have used a DSP core. However, sorting is a computational expensive operation, and a large area and power reduction may be possible with simpler analog implementations.

This work focuses on continuous-time analog rank extractors and their applications, with an emphasis on MOS circuits. Sources of error in previous analog implementations are analyzed, and improved multiple-input min/max circuit confgurations are proposed. A general rank extractor is discussed as well as its limitations on the number of inputs and rank, together with a general approach to larger analog sorting networks. Several extensions to sampled data systems are also presented, among which is a multiple-input strobed comparator and its generalization to other rank extractors.

Applications discussed in this work include a high-speed median filter, fast precision rectifiers based on min/max circuits, and the concept of an operational rank extractor. This versatile building block can implement a variety of nonlinear transfer functions such as a dead-zone amplifer, a limiter, a full-wave rectifer, and a tri-state comparator (including hysteretic behavior).
 

Opris, I. E., "Analog Rank Extractors and Sorting Networks," Ph.D. Thesis, Stanford University, CA, 1996.