3.2.4.10.2. Turbo Decoder parameters¶
3.2.4.10.2.1. --dec-type, -D¶
Type: text Allowed values: TURBOCHASEMLDefault: TURBOExamples: --dec-type CHASE
Select the decoder algorithm.
Description of the allowed values:
| Value | Description |
|---|---|
TURBO |
Select the Turbo decoder, the two sub-decoders are from the RSC code family. |
CHASE |
See the common --dec-type, -D parameter. |
ML |
See the common --dec-type, -D parameter. |
3.2.4.10.2.2. --dec-implem¶
Type: text Allowed values: STDFASTDefault: FASTExamples: --dec-implem FAST
Select the implementation of the decoder algorithm.
Description of the allowed values:
| Value | Description |
|---|---|
STD |
Select the STD implemenation. |
FAST |
Select the fast implemenation from [CTL+16]. |
3.2.4.10.2.3. --dec-sub-type, -D¶
Please refer to the RSC --dec-type, -D parameter.
3.2.4.10.2.4. --dec-sub-implem¶
Please refer to the RSC --dec-implem parameter.
3.2.4.10.2.5. --dec-sub-simd¶
Please refer to the RSC --dec-simd parameter.
3.2.4.10.2.6. --dec-crc-start¶
Type: integer Default: 2Examples: --dec-fnc-crc-ite 1
Set the first iteration to start the CRC checking.
Note
This parameter requires the Turbo code to be concatenated with a CRC to work, see the CRC parameters.
3.2.4.10.2.7. --dec-fnc¶
Enable the FNC post processing technique.
Note
The FNC post processing technique is detailed in [TLLeGal+16].
Note
This parameter requires the Turbo code to be concatenated with a CRC to work, see the CRC parameters.
3.2.4.10.2.8. --dec-fnc-ite-m¶
Type: integer Default: 3Examples: --dec-fnc-ite-m 2
Set the first iteration at which the FNC is used.
See the --dec-fnc parameter.
3.2.4.10.2.9. --dec-fnc-ite-M¶
Type: integer Default: 10Examples: --dec-fnc-ite-M 6
Set the last iteration at which the FNC is used.
See the --dec-fnc parameter.
3.2.4.10.2.10. --dec-fnc-ite-s¶
Type: integer Default: 1Examples: --dec-fnc-ite-s 2
Set the iteration step for the FNC technique.
See the --dec-fnc parameter.
3.2.4.10.2.11. --dec-fnc-q¶
Type: integer Default: 10Examples: --dec-fnc-q 6
Set the search space for the FNC technique.
See the --dec-fnc parameter.
3.2.4.10.2.12. --dec-ite, -i¶
Type: integer Default: 6Examples: --dec-ite 8
Set the maximal number of iterations in the Turbo decoder.
If the Turbo code is concatenated with a CRC and if the CRC is checked, the decoder can stop before making all the iterations.
3.2.4.10.2.13. --dec-sc¶
Enable the Self-Corrected (SC) decoder.
Note
The SC decoder is detailed in [Ton17] (in French).
Note
This parameter requires the Turbo code to be concatenated with a CRC to work, see the CRC parameters.
3.2.4.10.2.14. --dec-sf-type¶
Type: text Allowed values: ADAPTIVEARRAYCSTLTELTE_VECExamples: --dec-sf-type ADAPTIVE--dec-sf-type CST 0.5
Select a scaling factor (SF) to be applied to the extrinsic values after each half iteration.
This is especially useful with the max-log-MAP sub-decoders (BCJR with the \(\max\) approximation): the SF helps to recover a part of the decoding performance loss compare to the MAP algorithm (BCJR with the \(\max^*\) operator).
Note
The SF technique is detailed in [VF00].
Description of the allowed values:
| Value | Description |
|---|---|
ADAPTIVE |
Select the adaptive SF, for the first and second iterations a SF of 0.5 is applied, for the other iterations the SF is 0.85. |
ARRAY |
Select an hard-coded array of SFs (c.f. Table 3.3). |
CST |
Set the same SF to be applied for each iterations. |
LTE |
Select a 0.75 SF. |
LTE_VEC |
Select a 0.75 vectorized SF (faster than LTE), used in
[CTL+16]. |
| Iteration | Value |
|---|---|
| 1 | 0.15 |
| 2 | 0.25 |
| 3 | 0.30 |
| 4 | 0.40 |
| 5 | 0.70 |
| 6 | 0.80 |
| 7 | 0.90 |
| 8 | 0.95 |
3.2.4.10.2.16. References¶
| [CTL+16] | (1, 2) A. Cassagne, T. Tonnellier, C. Leroux, B. Le Gal, O. Aumage, and D. Barthou. Beyond Gbps turbo decoder on multi-core CPUs. In International Symposium on Turbo Codes and Iterative Information Processing (ISTC), 136–140. IEEE, September 2016. doi:10.1109/ISTC.2016.7593092. |
| [Ton17] | T. Tonnellier. Contribution to the Improvement of the Decoding Performance of Turbo Codes : Algorithms and Architecture. PhD thesis, Université de Bordeaux, 2017. URL: https://tel.archives-ouvertes.fr/tel-01580476. |
| [TLLeGal+16] | T. Tonnellier, C. Leroux, B. Le Gal, B. Gadat, C. Jégo, and N. Van Wambeke. Lowering the error floor of turbo codes with CRC verification. IEEE Wireless Communications Letters (WCL), 5(4):404–407, August 2016. doi:10.1109/LWC.2016.2571283. |
| [VF00] | J. Vogt and A. Finger. Improving the max-log-MAP turbo decoder. IET Electronics Letters, 36(23):1937–1939, November 2000. doi:10.1049/el:20001357. |