Real-time processing of linear and nonlinear systems using CROSSCUT computer memory.

摘要

CROSSCUT (CC) computer memory architecture is presented, which has potential applications in real-time signal processing of linear and nonlinear systems. It can easily be added to any computer system bus and gives the resident processor the ability to realize bit manipulating/table lookup algorithms at reasonably high frame rates, especially if the processor's addressing modes have enough power. Algorithms for which the CC memory applies are quickly and inexpensively implemented using simple machine language software programs. The CC memory system consists of a large block of memory which is partitioned into multiple state registers of word-length w, a DEPTH register that points to a set of all the active state register locations, and a sliding window which allows the processor to write or read the {dollar}isp{lcub}th{rcub}{dollar} bits of those state register locations pointed by the DEPTH register. The term "CROSSCUT" is used since memory can be accessed either in a normal read/write mode or in a bit-slice read/write mode that allows the access to all {dollar}isp{lcub}th{rcub}{dollar} bits from the active state register locations. All {dollar}isp{lcub}th{rcub}{dollar} bits can be assessed with a single memory access cycle.; Assuming a single processor environment, distributed arithmetic concepts are used to derive a number of CC memory algorithms for real-time processing of selected linear and nonlinear systems. Using CC memory, software algorithms are presented for implementation of digital filters, two-dimensional digital filters, multiple input multiple output systems, quadratic filters, and nonlinear systems. Measures of frame times for each of the above categories are investigated by writing assembly language software programs. Although execution times do not compete with complex special-purpose computer hardware systems, CC memory methods greatly reduce the long computation times and greatly simplify the software programs needed to solve such systems using traditional sequential processor techniques. Utilizing off-the-shelf ICs, a CC memory hardware design with 9 state registers, a depth of 1K, and 32-bit word-length is developed. Conclusions and further investigation regarding the application of CC memory are given.