Abstract

The ability to detect errors in a transmitted packet is currently available in the data-link layer by use of a Cyclic Redundancy Check (CRC). We analyze the selection of CRC polynomials in the context of error detection when transmitting packets of bits with rare, memoryless errors. We simulate errors in transmitted packets and show why standard CRC polynomials should not be used. Based on a realistic simulation, we show that equivalent-size primitive polynomials provide over one trillion times better protection against undetected errors than the current international standards.