Document

LoRaWAN Fragmented Data Block Transport v1.0.0 Specification ©2018 LoRa Alliance™ Page 22 of 30 The authors reserve the right to change specifications without notice. 9 Performance of the coding scheme. 542 543 As described in Gallager's thesis Parity Check codes have a non-zero statistical overhead 544 independent of the coded word length. In our case the word length used is M. The actual 545 overhead depends on the way the parity check matrix is built. To be able to reconstruct the 546 uncoded fragments the receiver must receive at least M linearly independent coded 547 fragments. Said in another way, the parity check matrix reconstructed by the receiver based 548 on the fragments received must be of rank M. 549 This condition is fulfilled ideally as soon as M coded fragments have been received. But 550 sometimes, those M received first fragments are not all independent and the matrix resulting 551 rank is 100) , because 567 adding 2 fragments on a 100 fragments session only represent 2% relative statistical 568 overhead. The overhead increases when the number of fragments is lower. This coding 569

• Cover
• 1 Conventions
• 2 Introduction
• 3 Downlink Fragmentation Transport Message Package
• 3.1 PackageVersionReq & Ans
• 3.2 FragSessionSetupReq & Ans
• 3.3 FragSessionDeleteReq & Ans
• 3.4 Downlink Data Fragment message
• 3.5 FragSessionStatusReq & Ans
• 4 File integrity check and authentication
• 5 Fragmentation algorithm
• Appendix: Data block fragmentation forward error correction code proposal
• 6 Introduction
• 7 Fragment error coding
• 8 Fragment decoding and reassembling
• 9 Performance of the coding scheme.
• 10 End-device memory requirement
• 11 Preliminary Matlab code
• 12 Glossary
• 13 Bibliography
• 13.1 References
• 14 NOTICE OF USE AND DISCLOSURE