Philip Koopman is a professor at Carnegie Mellon University, with research interests in the areas of software robustness, embedded networking, dependable embedded computer systems, and autonomous vehicle safety. Some key research findings that are discussed include: a well-chosen CRC is usually dramatically better than a checksum for relatively little additional computational cost you can usually do a lot better than “standard” CRC (especially CRC-32) Hamming Distance at the target payload length is the predominant selection criterion of interest and it is important to avoid bit encoding approaches that undermine CRC effectiveness.ĭr. This talk will covers the following areas: checksum and CRC theory with an emphasis on intuitive understanding rather than heavy math why using a standard or widely used CRC can be suboptimal (or worse) how to pick a good checksum/CRC the key parameters that affect the error detection capability of a checksum/CRC CRC pitfalls illustrated via examples from Controller Area Network and ARINC-825 an example CRC selection process for achieving a required level of functional criticality and a “seven deadly sins” list for CRC/checksum use. However, recent work has been able to exhaustively explore the CRC design space and identify optimal selection criteria based on key system characteristics. More than 50 years since the invention of the CRC, the proper use of these error detection codes is still hampered by a combination of misleading folklore, sub-optimality of standard approaches, general inaccessibility of research results, and the occasional typographical error in key reference materials. ![]() This talk includes both a tutorial and explanation of research results on the proper use of Cyclic Redundancy codes (CRCs) and checksums in an aviation context. ![]() ![]() Philip Koopman, Carnegie Mellon University
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |