3.2.4.7.1. Polar PAC Encoder parameters

3.2.4.7.1.1. --enc-type

Type:

text

Allowed values:

POLAR_PAC AZCW COSET USER

Default:

POLAR_PAC

Examples:

--enc-type AZCW

Select the encoder type.

Description of the allowed values:

Value

Description

POLAR_PAC

Select the standard Polar PAC encoder. The --enc-conv argument is required.

AZCW

See the common --enc-type parameter.

COSET

See the common --enc-type parameter.

USER

See the common --enc-type parameter.

Note

Only non-systematic encoding is supported as defined in [Pace-Ark19].

3.2.4.7.1.2. --enc-conv

Type:

text

Default:

NO

Examples:

--enc-conv 0o33

Define the convolutionnal code (ex.: “0o33”).

Warning

It is required to specify this argument with the POLAR_PAC encoder.

Note

Currently, we only support the octal format for the impulse response of the convolutional transform. For instance, consider an impulse response represented as [1, 0, 1, 1, 1, 0, 1]. When converting this to octal format, we will ignore the first 1 and convert the remaining vector, since AFF3CT will append the 1 at the start.

Thus, the vector [0, 1, 1, 1, 0, 1] will be converted to the octal value 35. Consequently, the input for AFF3CT will be 0o35.

3.2.4.7.1.3. --enc-fb-gen-method

Type:

text

Allowed values:

RM FILE GA TV BEC 5G

Examples:

--enc-fb-gen-method FILE

Select the frozen bits generation method.

Description of the allowed values:

Value

Description

RM

This is the preferred method to generate short length PAC codes.

GA

Select the GA method from [Plre-Tri12].

TV

Select the TV method which is based on Density Evolution (DE) approach from [Plre-TV13], to use with the --enc-fb-awgn-path parameter.

FILE

Read the best channels from an external file, to use with the --enc-fb-awgn-path parameter.

BEC

Generate frozen bits for the BEC channel from [Plrd-Ari09].

5G

Generate the frozen bits as described in the 5G standard [Plre-3GPP].

Note

The RM method is typically good for short block lengths and with PAC codes [Pace-Ark19], [Pacd-YFV20]. This is the default generator on PAC codes.

Note

By default, when using the GA or the TV method, the frozen bits are optimized for each SNR point. To override this behavior you can use the --enc-fb-noise parameter.

Note

When using the FILE method, the frozen bits are always the same regardless of the SNR value.

Note

When using the BEC method, the frozen bits are optimized for each erasure probability.

Note

When using the 5G method, the codeword size must be inferior to 1024.

3.2.4.7.1.4. --enc-fb-awgn-path

Type:

path

Rights:

read only

Examples:

--enc-fb-awgn-path example/path/to/the/right/place/

Set the path to a file or a directory containing the best channels to select the frozen bits.

An ASCII file is expected, for instance, the following file describes the most reliable channels optimized for a codeword of size \(N = 8\) and for an AWGN channel where the noise variance is \(\sigma = 0.435999\):

8
awgn
0.435999
7 6 5 3 4 2 1 0

Given the previous file, if we suppose a Polar code of size \(N = 8\) with \(K = 4\) information bits, the frozen bits are at the 0, 1, 2, 4 positions in the codeword. The strategy is to freeze the less reliable channels.

Warning

The FILE frozen bits generator expects a file and not a directory.

Warning

The TV frozen bits generator expects a directory and not a file. AFF3CT comes with input configuration files, a part of those configuration files are a set of best channels pre-generated with the TV method (see conf/cde/awgn_polar_codes/TV/).

3.2.4.7.1.5. --enc-fb-dump-path

Type:

folder

Rights:

write only

Examples:

--enc-fb-dump-path example/path/to/the/right/place/

Set the path to store the best channels.

Note

Works only for the GA and BEC frozen bits generation methods.

3.2.4.7.1.6. --enc-fb-noise

Type:

real number

Examples:

--enc-fb-noise 1.0

Select the noise for which the frozen bits will be optimized.

Can be a gaussian noise variance \(\sigma\) for GA and TV generation methods, or an event probability for the BEC generation method. All the noise points in the simulation will use the same frozen bits configuration.

3.2.4.7.1.7. References

[Pace-Ark19] (1,2)

Erdal Arıkan. From sequential decoding to channel polarization and back again. 2019. URL: https://arxiv.org/abs/1908.09594, arXiv:1908.09594.