Adaptive fault tolerant digital filters with single and multiple bit errors in fixed-point arithmetic

Fault Tolerant Signal Processing for Nano-scale VLSI Circuit Technology

“…Reference [8] extended the capability of adaptive fault tolerance to cases of single and multiple stuck-at bit errors occurring in one or more of the adaptive filter coefficients for fixed-point number representation. It was shown that in an over-parameterized adaptive filter with a large number of equivalent MSE solutions, the occurrence of a single stuck-at bit error reduces the number of available solutions, although multiple minimum MSE solutions still exist.…”

“…If the power normalized LMS algorithm is used to update the coefficients of the TDAF, then the relevant equations are e(n) = d(n) -y(n) (7) C(n+1) = C(n) + e(n) fV(n), (8) where Al is a diagonal matrix of time-varying step size parameters.…”

“…Adaptive Fault Tolerance (AFT) takes advantage of non-canonical adaptive filter architectures that use adaptive principles to achieve error masking. It has been shown that AFT methods are capable of masking and recovering from single and multiple fixed [7,8] and transient faults [9]. In this paper we explore the possibility of effectively using AFT principles, residue number system (RNS) arithmetic [10], and stochastic learning algorithms [12,13] to detect and correct soft errors in highly scaled adaptive VLSI processors.…”

Learning characteristics of adaptive fault tolerant filters in the presence of transient errors

“…More recent results have been published that demonstrate the capability of adaptive fault tolerance in cases of single and multiple stuck-at bit errors occurring in one or more of the adaptive filter coefficients, for both fixed-point [4] and floating-point [5] number representation. It was shown that in an over-parameterized fixed-point adaptive filter with a large number of equivalent MSE solutions, the occurrence of a single stuck-at bit error reduces the number of available solutions, although multiple minimum MSE solutions still remain [4].…”

“…It was shown that in an over-parameterized fixed-point adaptive filter with a large number of equivalent MSE solutions, the occurrence of a single stuck-at bit error reduces the number of available solutions, although multiple minimum MSE solutions still remain [4].…”