Yes, it seems to rely on a low error probability, to get high throughput. But maybe the FT8 CRC+ LDPC could be replace by CRC only and be decoded by GRAND with less CPU-cycles?
I do not if it is even possible to use GRAND - if it is possible to get a good license agreement for the WSJT software? - References: https://physics.princeton.edu/pulsar/k1jt/FT4_FT8_QEX.pdf <https://physics.princeton.edu/pulsar/k1jt/FT4_FT8_QEX.pdf> Backup: https://web.archive.org/web/20210618182741/https://physics.princeton.edu/pulsar/k1jt/FT4_FT8_QEX.pdf Quote: “… 3. Error Detection and Error Correction A 14-bit cyclic redundancy check (CRC) is appended to each 77-bit information packet to create a 91-bit message-plus-CRC word. The CRC is calculated on the source-encoded message, zero-extended from 77 to 82 bits. The CRC algorithm uses the polynomial 0x6757 (hexadecimal) and an initial value of zero (see [13, 14] for further details). Another 83 bits are appended for forward error correction, creating a 174-bit codeword. Forward error correction is accomplished using a (174, 91) LDPC code designed specifi cally for FT8 and FT4. …” https://www.granddecoder.mit.edu/ <https://www.granddecoder.mit.edu/> Backup: https://web.archive.org/web/20210918073847/https://www.granddecoder.mit.edu/ Quote: "... For further licensing information, please contact Myron Kassaraba ([email protected]) at MIT's Technology Licensing Office. ..." Glenn, OZ1HFT - > Den 21. sep. 2021 kl. 22.50 skrev Philip Gladstone > <[email protected]>: > > My understanding of GRAND is that it relies on the fact that the SNR is high > -- which is very much not the case for FT8. > > Philip
_______________________________________________ wsjt-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/wsjt-devel
