Turbo encoder/decoder Class. More...

#include <itpp/comm/turbo.h>

Inheritance diagram for itpp::Turbo_Codec:

Public Member Functions
	Turbo_Codec (void)
	Class constructor.

virtual	~Turbo_Codec (void)
	Class destructor.

void	set_parameters (ivec gen1, ivec gen2, int constraint_length, const ivec &interleaver_sequence, int in_iterations=8, const std::string &in_metric="LOGMAX", double in_logmax_scale_factor=1.0, bool in_adaptive_stop=false, LLR_calc_unit lcalc=LLR_calc_unit())
	Set parameters for the turbo encoder/decoder.

void	set_interleaver (const ivec &interleaver_sequence)
	Set a new internal interleaver sequence for the turbo encoder/decoder.

void	set_metric (std::string in_metric="LOGMAX", double in_logmax_scale_factor=1.0, LLR_calc_unit lcalc=LLR_calc_unit())
	Set the decoder metric.

void	set_iterations (int in_iterations=8)
	Sets the number of decoding iterations. Default value is 8.

void	set_adaptive_stop (bool in_adaptive_stop=true)
	Use and adaptive number of iterations.

void	set_awgn_channel_parameters (double in_Ec, double in_N0)
	Set parameters for decoding on an AWGN channel.

void	set_scaling_factor (double in_Lc)
	Set scaling factor for decoding on e.g. Rayleigh fading channels.

void	encode (const bvec &input, bvec &output)
	Encoder function.

virtual void	decode (const vec &received_signal, bvec &decoded_bits, const bvec &true_bits="0")
	Decoder function.

virtual void	decode (const vec &received_signal, bvec &decoded_bits, ivec &nrof_used_iterations, const bvec &true_bits="0")
	Decoder function.

void	encode_block (const bvec &input, bvec &in1, bvec &in2, bmat &parity1, bmat &parity2)
	Encode a single block.

virtual void	decode_block (const vec &rec_syst1, const vec &rec_syst2, const mat &rec_parity1, const mat &rec_parity2, bmat &decoded_bits_i, int &nrof_used_iterations_i, const bvec &true_bits="0")
	Decode a single block.

int	get_Ncoded () const
	Get number of coded bits.

int	get_Nuncoded () const
	Get number of uncoded bits.

Protected Member Functions
void	decode_n3 (const vec &received_signal, bvec &decoded_bits, ivec &nrof_used_iterations, const bvec &true_bits="0")
	Special decoder function for R = 1/3 i.e. two parity bits for each systematic bit.

Static Protected Member Functions
static std::string	string_from_metric (const Metric &m)

Protected Attributes
int	interleaver_size

int	Ncoded

int	Nuncoded

int	m_tail

int	n1

int	n2

int	n_tot

int	iterations

double	Ec

double	N0

double	Lc

double	R

double	logmax_scale_factor

bool	adaptive_stop

Metric	metric

bvec	decoded_bits_previous_iteration

Rec_Syst_Conv_Code	rscc1

Rec_Syst_Conv_Code	rscc2

Sequence_Interleaver< bin >	bit_interleaver

Sequence_Interleaver< double >	float_interleaver

Related Functions
(Note that these are not member functions.)
ITPP_EXPORT ivec	wcdma_turbo_interleaver_sequence (int interleaver_size)
	Generates the interleaver sequence for the internal turbo encoder interleaver used in WCDMA.

ITPP_EXPORT ivec	lte_turbo_interleaver_sequence (int interleaver_size)
	Generates the interleaver sequence for the internal turbo encoder interleaver used in LTE.

Detailed Description

Turbo encoder/decoder Class.

Author: Pal Frenger

To set up the turbo encoder used in e.g. WCDMA the following code can be used (assuming a code block size of 320 bits):

Turbo_Codec turbo;
ivec gen(2);
gen(0) = 013; gen(1) = 015;
int constraint_length = 4;
ivec interleaver_sequence = wcdma_turbo_interleaver_sequence( 320 );
turbo.set_parameters(gen, gen, constraint_length, interleaver_sequence);

Definition at line 66 of file turbo.h.

Member Function Documentation

void itpp::Turbo_Codec::set_parameters	(	ivec	gen1,
		ivec	gen2,
		int	constraint_length,
		const ivec &	interleaver_sequence,
		int	in_iterations = `8`,
		const std::string &	in_metric = `"LOGMAX"`,
		double	in_logmax_scale_factor = `1.0`,
		bool	in_adaptive_stop = `false`,
		LLR_calc_unit	lcalc = `LLR_calc_unit()`
	)

Set parameters for the turbo encoder/decoder.

Parameters

gen1	A vector with n1 elements containing the generator polynomials for the first constituent encoder
gen2	A vector with n2 elements containing the generator polynomials for the second constituent encoder
constraint_length	The constraint length of the two constituent encoders
interleaver_sequence	An ivec defining the internal turbo interleaver.
in_iterations	The number of decoding iterations. Default value is 8.
in_metric	Determines the decoder metric: "MAP", LOGMAP", "LOGMAX", or "TABLE". The default is "LOGMAX".
in_logmax_scale_factor	The extrinsic information from each constituent decoder is to optimistic when LOGMAX decoding is used. This parameter allows for a down-scaling of the extrinsic information that will be passed on to the next decoder. The default value is 1.0. This parameter is ignored for other metrics than "LOGMAX".
in_adaptive_stop	If this parameter is true, then the iterations will stop if the decoding results after one full iteration equals the previous iteration. Default value is false.
lcalc	This parameter can be used to provide a specific `LLR_calc_unit` which defines the resolution in the table-lookup if decoding with the metric "TABLE" is used.

Note: For issues relating to the accuracy of LLR computations, please see the documentation of LLR_calc_unit

Definition at line 55 of file turbo.cpp.

References it_assert, it_error, itpp::LOGMAP, itpp::Rec_Syst_Conv_Code::set_generator_polynomials(), itpp::Sequence_Interleaver< T >::set_interleaver_depth(), itpp::Sequence_Interleaver< T >::set_interleaver_sequence(), and itpp::Rec_Syst_Conv_Code::set_llrcalc().

Referenced by itpp::Punctured_Turbo_Codec::set_parameters().

void itpp::Turbo_Codec::set_interleaver ( const ivec & interleaver_sequence )

Set a new internal interleaver sequence for the turbo encoder/decoder.

By changing the interleaver sequence it is possible to change the code word size of the turbo codec. Note that you still must use the set_parameters function first to set the other parameters of the turbo codec.

Definition at line 115 of file turbo.cpp.

References itpp::Sequence_Interleaver< T >::set_interleaver_depth(), and itpp::Sequence_Interleaver< T >::set_interleaver_sequence().

void itpp::Turbo_Codec::set_metric	(	std::string	in_metric = `"LOGMAX"`,
		double	in_logmax_scale_factor = `1.0`,
		LLR_calc_unit	lcalc = `LLR_calc_unit()`
	)

Set the decoder metric.

Parameters

in_metric	Determines the decoder metric: "MAP", LOGMAP", "LOGMAX", or "TABLE". The default is "LOGMAX".
in_logmax_scale_factor	The extrinsic information from each constituent decoder is to optimistic when LOGMAX decoding is used. This parameter allows for a down-scaling of the extrinsic information that will be passed on to the next decoder. The default value is 1.0. This parameter is ignored for other metrics than "LOGMAX".
lcalc	This parameter can be used to provide a specific `LLR_calc_unit` which defines the resolution in the table-lookup if decoding with the metric "TABLE" is used.

Definition at line 130 of file turbo.cpp.

References it_error, itpp::LOGMAP, and itpp::Rec_Syst_Conv_Code::set_llrcalc().

void itpp::Turbo_Codec::set_adaptive_stop ( bool in_adaptive_stop = true )

Use and adaptive number of iterations.

When the adaptive stop criterion is used the iterations will stop before the maximum number of iterations is executed if the decoding results after one full iteration equals the previous iteration. Default value is true.

Definition at line 160 of file turbo.cpp.

void itpp::Turbo_Codec::set_awgn_channel_parameters	(	double	in_Ec,
		double	in_N0
	)

Set parameters for decoding on an AWGN channel.

Parameters

in_Ec	The received energy per channel symbol (i.e. per soft input bit)
in_N0	The single sided spectral density of the AWGN noise

Definition at line 165 of file turbo.cpp.

References itpp::sqrt().

void itpp::Turbo_Codec::set_scaling_factor ( double in_Lc )

Set scaling factor for decoding on e.g. Rayleigh fading channels.

Setting the correct value of the channel reliability function is important for MAP decoder algorithms. However, if the Log-MAX decoding algorithm is used, then the value of Lc is not important. Assuming that the received soft values $r_k$ from the channel equal

$r_k = h_k c_k + w_k$

where $h_k$ is the (complex valued) channel gain, $c_k$ is the transmitted symbol equal to $\{-\sqrt{E_c},+\sqrt{E_c}\}$ , and $w_k$ is the AWGN (complex valued) noise with total variance of the real plus imaginary part equal to $N_0$ . The input to the turbo decoder shall then be

$z_k = \hat{h}_k^{*} r_k$

where $\hat{h}_k^{*}$ is the conjugate of the channel estimate. Assuming that the channel estimate is perfect, the channel reliability factor shall be set to

$L_c = 4\sqrt{E_c} / {N_0}$

Parameters

in_Lc the channel reliability factor of the channel.

Definition at line 172 of file turbo.cpp.

void itpp::Turbo_Codec::encode	(	const bvec &	input,
		bvec &	output
	)

Encoder function.

This function can encode several consecutive coding blocks. The output is organized as follows:

$s(1), p_{1,1}(1), p_{1,2}(1), \ldots , p_{1,n_1}(1), p_{2,1}(1), p_{2,2}(1), \ldots , p_{2,n_2}(1), s(2), \ldots$

In the above expression $s(n)$ is the n-th systematic bit and $p_{l,k}(n)$ is the n-th bit from the k-th encoder polynomial of the l-th constituent encoder. A tail of both systematic and parity bits is added after each coding block to force both encoder to the zero state. The tail of each encoder is structured as follows (using encoder one as an example):

$t_1(1), pt_{1,1}(1), pt_{1,2}(1), \ldots , pt_{1,n_1}(1), \ldots pt_{1,n_1}(m)$

The tailbits from the first encoder are placed before the tailbits from the second encoder.

Parameters

input	The input bits to the encoder. Must contain an integer number of code blocks
output	The encoded bits including two tail, one from each constituent encoder, after each coding block.

Definition at line 178 of file turbo.cpp.

References encode_block().

Referenced by itpp::Punctured_Turbo_Codec::encode().

void itpp::Turbo_Codec::decode	(	const vec &	received_signal,
		bvec &	decoded_bits,
		const bvec &	true_bits = `"0"`
	)

virtual

Decoder function.

This function can decode several consecutive coding blocks that were encoded with the encode member function

Parameters

received_signal	The vector of received bits
decoded_bits	A vector of decoded bits
true_bits	If this input vector is provided then the iterations will stop as soon as the decoded bits equals the `true_bits`. Note that this feature can not be used if the `in_adaptive_stop` parameter in the setup function is set to `true`.

Reimplemented in itpp::Punctured_Turbo_Codec.

Definition at line 226 of file turbo.cpp.

Referenced by itpp::Punctured_Turbo_Codec::decode().

void itpp::Turbo_Codec::decode	(	const vec &	received_signal,
		bvec &	decoded_bits,
		ivec &	nrof_used_iterations,
		const bvec &	true_bits = `"0"`
	)

virtual

Decoder function.

This function can decode several consecutive coding blocks that were encoded with the encode member function

Parameters

received_signal	The vector of received bits
decoded_bits	A vector of decoded bits
nrof_used_iterations	Returns the number of used iterations for each code block.
true_bits	If this input vector is provided then the iterations will stop as soon as the decoded bits equals the `true_bits`. Note that this feature can not be used if the `in_adaptive_stop` parameter in the setup function is set to `true`.

Reimplemented in itpp::Punctured_Turbo_Codec.

Definition at line 232 of file turbo.cpp.

References decode_block(), decode_n3(), and it_assert.

void itpp::Turbo_Codec::encode_block	(	const bvec &	input,
		bvec &	in1,
		bvec &	in2,
		bmat &	parity1,
		bmat &	parity2
	)

Encode a single block.

This function is useful if rate matching is to be applied after the turbo encoder. The size of in1 and in2 equals the size of input plus the tail bits. Tailbits are appended ate the end of in1 and in2. The number of rows in parity1 and parity2 equals the lengths of in1 and in2 respectively. The number of columns of parity1 and parity2 equals the number of parity bits from the first and the second constituent encoders respectively.

Parameters

input	The input bits to the encoder. Must contain a single code block
in1	The input bits to the first constituent encoder with a tail added at the end
in2	The input bits to the second constituent encoder (i.e. the interleaved data bits) with a tail added at the end
parity1	The parity bits from the first constituent encoder (including parity bits for the first tail)
parity2	The parity bits from the second constituent encoder (including parity bits for the second tail)

Definition at line 326 of file turbo.cpp.

References itpp::concat(), itpp::Rec_Syst_Conv_Code::encode_tail(), itpp::Sequence_Interleaver< T >::interleave(), and it_assert.

Referenced by encode().

void itpp::Turbo_Codec::decode_block	(	const vec &	rec_syst1,
		const vec &	rec_syst2,
		const mat &	rec_parity1,
		const mat &	rec_parity2,
		bmat &	decoded_bits_i,
		int &	nrof_used_iterations_i,
		const bvec &	true_bits = `"0"`
	)

virtual

Decode a single block.

This function can decode a single coding blocks that was encoded with the encode_block member function. In order to speed up the decoding process it is possible to input the correct data bits. If this information is provided the decoder can stop iterating as soon as the decoded bits match the correct data bits. This simulation trick can greatly speed up the simulation time for high SNR cases when only a few iterations are required. If errors still exist after the maximum number of iterations have been performed, the decoding process stops.

The matrix decoded_bits_i contains the result from decoding iteration i on row i. Even though both rec_syst1 and rec_syst2 are given as inputs, the systematic bits in rec_syst2 will in most cases be punctured and only the tailbits at the end of the vector rec_syst2 will have values different from zero.

Note: This decoding function assumes that the input is scaled as +-2*SNR + noise. This means that the channel reliability factor Lc must be equal to 1.0. No additional scaling is performed by this function.

Parameters

rec_syst1	The received input bits to the first constituent decoder with a tail added at the end
rec_syst2	The received input bits to the second constituent decoder with a tail added at the end
rec_parity1	The received parity bits for the first constituent decoder (including parity bits for the first tail)
rec_parity2	The received parity bits for the second constituent decoder (including parity bits for the second tail)
decoded_bits_i	Contains the result from decoding iteration i on row i.
nrof_used_iterations_i	Returns the number of iterations used for decoding of this block.
true_bits	Optional input parameter. If given, the iterations will stop as soon as the decoded bits match the true bits.

Definition at line 361 of file turbo.cpp.

References itpp::concat(), itpp::Sequence_Interleaver< T >::deinterleave(), itpp::Sequence_Interleaver< T >::interleave(), it_assert, it_error, itpp::Vec< Num_T >::left(), itpp::Rec_Syst_Conv_Code::log_decode(), itpp::LOGMAP, itpp::Rec_Syst_Conv_Code::map_decode(), itpp::Vec< Num_T >::right(), and itpp::zeros().

Referenced by decode().

Friends And Related Function Documentation

ITPP_EXPORT ivec lte_turbo_interleaver_sequence ( int interleaver_size )

Author: qdelfin and Stephan Ludwig

Definition at line 1135 of file turbo.cpp.

The documentation for this class was generated from the following files:

itpp/comm/turbo.h
itpp/comm/turbo.cpp

Public Member Functions

Protected Member Functions

Static Protected Member Functions

Protected Attributes

Related Functions

Detailed Description

Member Function Documentation

Friends And Related Function Documentation