Under Low Voltage Probability Characteristics Analysis And The Application Of Digital Circuit

Today，the digital products has penetrated into every area of the world. The energyefficient is becoming more and more serious issue as more and more functions areadded into the digital systems.In order toachieveexcellentperformance and low powerconsumption, the feature size of semiconductorprocess has been reduced. But thecontinuous reduction of the feature size, we have topay a considerableprice and thefunction of digital becomes unreliable. Therefore, how toreducethedigitalICpowerconsumptionwith guaranteed performance is becoming newresearch subject in academic and industry area for digital system design.Power dissipation of DigitalICismainly composed ofstaticand dynamic powerconsumption; they are both proportional to the squareofthevoltage.Thus, the reductiontheoperating voltageofthe circuitcanreducepower consumption significantly. But the lowvoltage can also lead to probabilistic characteristic of IC. On the other hand, thetraditional digital signal processing(DSP)requires the correct calculations. It is notneeded for the DSP in communications, radar and multimedia. Therefore, it is a feasiblemethod to achieve low power consumption by reducing voltages.In this dissertation, the factors, which cause the soft error for deepsubmicron(DSM) device with low supply voltage, are introduced, and the probabilisticmodels for the Complementary Metal Oxide Semiconductor (CMOS)logic gatesarededuced based on the structure of devices. A novel state-transfer methodis proposedto establish the probabilistic models for the digital circuit modules.In order to verify therationality of this method, we establish the hardware platform. To study the practicalapplication of the probabilistic characteristic, we proposed the design method with thecombination of the Reduced Precision Redundancy (RPR)algorithm, Triple-ModularRedundancy (TMR) algorithm and Redundant Residue Number Systems (RRNS)algorithm. A25-taps FIR filter is appliedas an example in this paper, two kinds ofprobability FIR filter structures,named RPR-TMR and RPR-RRNS, are designed andanalyzed. From the case study, we find that the proposed methods can provide a newkind of feasible design ideas for the applications of probability device.