Error and failure-control proceedure for a large-size bubble memory

This paper describes a new dual-mode error-coding scheme for a large-size bubble memory. The scheme is designed to preserve the nonvolatility property of bubbles against random single-bit errors and occasional massive data loss caused by failure of a module or a card. The objective in this design was to obtain reasonable amount of fault tolerance in a cost-effective manner without substantial increase in normal ECC (error correcting code) redundancy. In this memory, blocks of data are accessed sequentially in the form of m-byte words, which corresponds to interleaved or parallel operation of m bubble modules. Each word contains one check byte. A group of N words forms a block with an additional end-of-block check byte in the Nth word. The word ECC, in its primary mode, routinely corrects random single-bit errors. In the presence of larger errors, caused by an oncoming module failure, the interaction between the word ECC and the end-of-block check provides detection and isolation of the failing module. The word ECC, then in its secondary mode, provides recovery of a large amount of data lost in the failing module. The bubble memory is then reconfigured, bypassing the failed module before restoring the data.