MULTIPLE-ERROR-CORRECTING CODES FOR IMPROVING THE PERFORMANCE OF OPTICAL MATRIX-VECTOR PROCESSORS

I examine the use of Reed-Solomon multiple-error-correcting codes for enhancing the performance of optical matrix-vector processors. An opti mal code rate of 0.75 is found, and n = 127 block-length codes are see n to increase the optical matrix dimension achievable by a factor of 2 .0 for a required system bit-error rate of 10(-15). The optimal codes required for various matrix dimensions are determined. I show that sin gle code word implementations are more efficient than those utilizing multiple code words.