A technique of extending a correction limit defined by an ECC is described. According to one aspect of the present invention, remaining errors that cannot be corrected by the ECCs in a data array is first identified and then formed in form of matrix with defined size. These remaining errors are flipped in value, namely from “1” to “0” or “0’ to “1” if the number of the errors are within a range or additional ECCs are applied to correct the errors in flipped data bits.