search

cedricomarchando / NB_LDPC_FB

c code for Non-Binary LDCP decoder
BSD 2-Clause "Simplified" License
18 stars 8 forks source link

The NB_LDPC_FB decoder over GF(32) #2

Closed Liubusy closed 10 months ago

Liubusy commented 11 months ago

Hi, I am very interested in the EMS algorithm for NB-LDPC codes. I have tried to use the NB_LDPC_FB decoder to decode the regular (837, 726) NB-LDPC code over GF(32) . However, there is about 0.5 dB coding gain loss at the BER of 10-4. Does it need some mofications to adapt to the regular (837, 726) NB-LDPC code over GF(32)? If yes, could you tell us where should be modifed? Thank you so much！ I am looking forward to hearing from you.

cedricomarchando commented 11 months ago

Hi, the NB-LDPC code were optimized for GF64. you should update the code for GF64. from the code designed for GF(64) you may update the GF coefficient that are greater than 32 by a random coefficient lower than 32 Best regards, cédric

De: "Liubusy" @.> À: "cedricomarchando/NB_LDPC_FB" @.> Cc: "Subscribed" @.***> Envoyé: Vendredi 20 Octobre 2023 11:50:44 Objet: [cedricomarchando/NB_LDPC_FB] The NB_LDPC_FB decoder over GF(32) (Issue #2)

Hi, I am very interested in the EMS algorithm for NB-LDPC codes. I have tried to use the NB_LDPC_FB decoder to decode the regular (837, 726) NB-LDPC code over GF(32) . However, there is about 0.5 dB coding gain loss at the BER of 10-4. Does it need some mofications to adapt to the regular (837, 726) NB-LDPC code over GF(32)? If yes, could you tell us where should be modifed? Thank you so much！ I am looking forward to hearing from you.

— Reply to this email directly, [ https://github.com/cedricomarchando/NB_LDPC_FB/issues/2 | view it on GitHub ] , or [ https://github.com/notifications/unsubscribe-auth/ANYBEK7IHE2XCG6DSQ7O6ELYAJCPJAVCNFSM6AAAAAA6IYKF5GVHI2DSMVQWIX3LMV43ASLTON2WKOZRHE2TGOJTGAZTQOA | unsubscribe ] . You are receiving this because you are subscribed to this thread. Message ID: <cedricomarchando/NB_LDPC_FB/issues/2 @ github . com>

Liubusy commented 10 months ago

Thank you so much for your timely reply. After reading your message, I am sorry for not understanding how to modify the NB_LDPC_FB decoder. Since I want to do some comparison with the GF(32) code , it is better to modify the the NB_LDPC_FB decoder opened by you to decode the GF(32) NB-LDPC codes correcttly. In fact, the NB-LDPC decoder can decode GF(32) NB-LDPC codes but will introduce performance loss. I do not know why. So, could you expain how to modify the code in detail ? Thank you so much.

cedricomarchando commented 10 months ago

a was talking about the matrix that define the LDPC code, my guess is that you use an ldpc matrice (regular (837, 726) NB-LDPC code over GF(32) ) with coefficient that are not in GF(64), (GF values higher than 32)

for example for the smalest code [ http://labsticc.univ-ubs.fr/nb_ldpc/MatricesDir/GF64/K18/merge.png | K=18 bits (3 symbols) ] defined in the NB-LDPC webpage [ http://www-labsticc.univ-ubs.fr/nb_ldpc/ | http://www-labsticc.univ-ubs.fr/nb_ldpc/ ] the AlistSimu.txt give

9 6 64 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 1 15 4 0 7 41 2 15 5 0 8 41 3 0 6 41 9 15
1 0 5 15 9 41 2 41 6 15 7 0
3 15 4 41 8 0

the lines after line 3 define the LDPC connections and the GF coefficients. the GF coefficients are given every odd column. i highlighted the GF coefficient higher than 32 and i update it here after

9 6 64 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 1 15 4 0 7 21 2 15 5 0 8 8 3 0 6 11 9 15
1 0 5 15 9 27 2 21 6 15 7 0
3 15 4 4 8 0

i hope it helps Cédric

Hi, the NB-LDPC code were optimized for GF64. you should update the code for GF64. from the code designed for GF(64) you may update the GF coefficient that are greater than 32 by a random coefficient lower than 32 Best regards, cédric

De: "Liubusy" < @. > À: "cedricomarchando/NB_LDPC_FB" < @. > Cc: "Subscribed" < @.*** > Envoyé: Vendredi 20 Octobre 2023 11:50:44 Objet: [cedricomarchando/NB_LDPC_FB] The NB_LDPC_FB decoder over GF(32) (Issue #2)

Hi, I am very interested in the EMS algorithm for NB-LDPC codes. I have tried to use the NB_LDPC_FB decoder to decode the regular (837, 726) NB-LDPC code over GF(32) . However, there is about 0.5 dB coding gain loss at the BER of 10-4. Does it need some mofications to adapt to the regular (837, 726) NB-LDPC code over GF(32)? If yes, could you tell us where should be modifed? Thank you so much！ I am looking forward to hearing from you.

— Reply to this email directly, [ https://github.com/cedricomarchando/NB_LDPC_FB/issues/2 | view it on GitHub ] , or [ https://github.com/notifications/unsubscribe-auth/ANYBEK7IHE2XCG6DSQ7O6ELYAJCPJAVCNFSM6AAAAAA6IYKF5GVHI2DSMVQWIX3LMV43ASLTON2WKOZRHE2TGOJTGAZTQOA | unsubscribe ] . You are receiving this because you are subscribed to this thread.

De: "Liubusy" @.> À: "cedricomarchando/NB_LDPC_FB" @.> Cc: "Cedric Marchand" @.>, "Comment" @.> Envoyé: Vendredi 20 Octobre 2023 16:15:24 Objet: Re: [cedricomarchando/NB_LDPC_FB] The NB_LDPC_FB decoder over GF(32) (Issue #2)

Thank you so much for your timely reply. After reading your message, I am sorry for not understanding how to modify the NB_LDPC_FB decoder. Since I want to do some comparison with the GF(32) code , it is better to modify the the NB_LDPC_FB decoder opened by you to decode the GF(32) NB-LDPC codes correcttly. In fact, the NB-LDPC decoder can decode GF(32) NB-LDPC codes but will introduce performance loss. I do not know why. So, could you expain how to modify the code in detail ? Thank you so much.

— Reply to this email directly, [ https://github.com/cedricomarchando/NB_LDPC_FB/issues/2#issuecomment-1772821508 | view it on GitHub ] , or [ https://github.com/notifications/unsubscribe-auth/ANYBEK3T2WS2OMI5DI72NFDYAKBPZAVCNFSM6AAAAAA6IYKF5GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONZSHAZDCNJQHA | unsubscribe ] . You are receiving this because you commented. Message ID: <cedricomarchando/NB_LDPC_FB/issues/2/1772821508 @ github . com>

Liubusy commented 10 months ago

I try to modify the realted parts. Thank you so much !