Multiple bit error correcting architectures over finite fields

This thesis proposes techniques to mitigate multiple bit errors in GF arithmetic circuits. As GF arithmetic circuits such as multipliers constitute the complex and important functional unit of a crypto-processor, making them fault tolerant will improve the reliability of circuits that are employed in safety applications and the errors may cause catastrophe if not mitigated. Firstly, a thorough literature review has been carried out. The merits of efficient schemes are carefully analyzed to study the space for improvement in error correction, area and power consumption. Proposed error correction schemes include bit parallel ones using optimized BCH codes that are useful in applications where power and area are not prime concerns. The scheme is also extended to dynamically correcting scheme to reduce decoder delay. Other method that suits low power and area applications such as RFIDs and smart cards using cross parity codes is also proposed. The experimental evaluation shows that the proposed techniques can mitigate single and multiple bit errors with wider error coverage compared to existing methods with lesser area and power consumption. The proposed scheme is used to mask the errors appearing at the output of the circuit irrespective of their cause. This thesis also investigates the error mitigation schemes in emerging technologies (QCA, CNTFET) to compare area, power and delay with existing CMOS equivalent. Though the proposed novel multiple error correcting techniques can not ensure 100% error mitigation, inclusion of these techniques to actual design can improve the reliability of the circuits or increase the difficulty in hacking crypto-devices. Proposed schemes can also be extended to non GF digital circuits