| Research Abstract |
Application specific ICs (ASICs) play an important role in present-day DSP technology. However, the ASIC technology requires significant time and cost for development, and is lacking the flexibility in design modification. The FPGA (Field-Programmable Gate Array) technology, on the other hand, attracts much attention recently due to its rapid prototyping capability, but it is difficult to apply FPGAs to high-performance systems. This research project is to develop configurable signal processors --- DSP-oriented FPGA architectures employing redundant arithmetic alogorithms for achieving both high-performance and programmabiligy. Listed below are major results of this project:
1. A configurable signal processor IC, that can implement high-frequency FIR filters with 10--100 MHz sampling rate, was developed. The test chip integrates redundant SW adders and can implement FIR filters of order up to 11. Assuming the use of state-of-the-art CMOS technology, the proposed architecture could be applied to QAM modulators/demodulators and format converters for digital TV applications, for example.
2. A configurable signal processor IC (for FIR filtering) using current-mode multiple-valued logic technology was developed, where 5-level bi-directional current signals are used for intra-chip communication. The use of multiple-valued logic makes possible to reduce the chip area by 50% and power consumption by 4-40%.
3. Various redundant arithmetic algorithms for DSP applications are systematically investigated. Also, new CAD techniques were developed for designing/verifying/synthesizing signal processors using redundant arithmetic algorithms.
4. A new fault-tolerant FPGA architecture that can detect hardware faults and reconfigure its architecture to recover the correct function was proposed for use in mission-critical applications.
|
