February 3, 2010 -- Current design methodologies, whether for software or hardware, are in constant pursuit of making applications faster. For software, execution time is determined by measuring input, processing the data, and displaying or transmitting the output. Here, input and output times depend on the peripheral ports and the processing of data depends on the application implementation and processor operating speed.

For most SOC (system-on-chip) applications, signal processing consumes the majority of the execution time. Using extensive mathematical libraries for processing data is expensive in terms of both CPU time and memory usage. If the input and output are expected to be in a specified range, developers can optimize the data processing with the use of LUTs (lookup tables). The time spent to build the LUT will go into the one-time cost of design rather than CPU overhead for every execution. In scenarios where the LUT method is feasible, its use depends on the tradeoff between execution time versus design time, and accuracy verses memory usage.

The following case studies discuss the trade-offs between different methods of implementations for thermistor sensing and PID (proportional-integral-differential) processing. By Archana Yarlagadda. (Yarlagadda is an applications engineer at Cypress Semiconductor, Inc.)