NR LDPC Decoder

Decode LDPC code according to 5G NR standard

Since R2020a

• 

×

Libraries:
Wireless HDL Toolbox / Error Detection and Correction

Description

The NR LDPC Decoder block implements a low-density parity-check (LDPC) decoder with hardware-friendly control signals. The block accepts punctured log-likelihood ratio (LLR) values, a stream of control signals, a base graph number, and lifting sizes. The block outputs decoded bits, a stream of control signals, lifting sizes, and a signal that indicates when the block is ready to accept new inputs.

This block provides an option to implement layered belief propagation with either the normalized min-sum approximation algorithm or the min-sum approximation algorithm. This implementation matches that of the function nrLDPCDecode (5G Toolbox). You can use this block for channel coding of downlink and uplink shared channels and paging channel according to 5G new radio (NR) standard TS 38.212 [1].

The NR LDPC Decoder block supports scalar and 64-element column vector inputs. The block supports the early termination feature to help improve decoding performance and faster convergence speeds at high signal noise ratio (SNR) conditions. The block enables decoding of multiple code rates to help achieve high throughput efficiency with a high degree of code rate flexibility. The block provides an architecture suitable for HDL code generation and hardware deployment. For more information, see Algorithms.

Examples

LDPC Encode and Decode of 5G NR Streaming Data

Simulate NR LDPC Encoder and NR LDPC Decoder blocks and compare their hardware-optimized results with results from 5G Toolbox™ functions.

• Open Script

LDPC Decode 5G NR Streaming Data for Multiple Code Rates with Early Termination

Use multiple code rates and early termination criteria features in NR LDPC Decoder block.

• Open Script

Ports

Input

data — Input LLR values

scalar | vector

Input log-likelihood ratio (LLR) values, specified as a scalar or a column vector of size 64.

The data type of this input must be a signed fixed-point with a word length from 4 to 16 bits. For more information on how to specify vector input data, see Specifying Vector Input.

Data Types: int8 | int16 | fixed point

ctrl — Control signals accompanying sample stream

samplecontrol bus

Control signals accompanying the sample stream, specified as a samplecontrol bus. The bus includes the start, end, and valid control signals, which indicate the boundaries of the frame and the validity of the samples.

• start — Indicates the start of the input frame

• end — Indicates the end of the input frame

• valid — Indicates that the data on the input data port is valid

For more details, see Sample Control Bus.

Data Types: bus

bgn — Base graph number

scalar

Base graph number, specified as a scalar. When this value is 0, the block applies bgn 1. When this value is 1, the block applies bgn 2. For more information about bgn 1 and bgn 2, see section 5.3.2, of TS 38.212 [1].

Data Types: Boolean

liftingSize — Input lifting size

scalar

Input lifting size, specified as a scalar.

For an invalid liftingSize value, the block discards the current frame and waits for the new frame. For more information about the supported lifting size values, see section 5.3.2, of TS 38.212 [1].

Data Types: uint16

iter — Number of iterations

scalar

Number of iterations, specified as a integer in the range from 1 to 63.

If you specify iter as a value greater than 63, the block automatically sets the iter value to 8 and performs the decoding operation.

Dependencies

To enable this port, set the Source for number of iterations parameter to Input port.

Data Types: uint8

numRows — Number of rows

scalar

Number of rows, specified as a scalar.

When you set the bgn value to 0 the block supports the number of rows in the range from 4 to 46. When you set the bgn value to 1, the block supports the number of rows in the range from 4 to 42.

Dependencies

To enable this port, select the Enable multiple code rates parameter.

Data Types: fixdt(0,6,0)

Output

data — Decoded output data bits

scalar | vector

Decoded output data bits, returned as a scalar or a column vector of size 64.

The block outputs data bits in a similar format as the input LLR values. Extract these output data bits in a similar format for further processing.

Data Types: Boolean

ctrl — Control signals accompanying sample stream

samplecontrol bus

Control signals accompanying the sample stream, returned as a samplecontrol bus. The bus includes the start, end, and valid control signals, which indicate the boundaries of the frame and the validity of the samples.

• start — Indicates the start of the output frame

• end — Indicates the end of the output frame

• valid — Indicates that the data on the output data port is valid

For more details, see Sample Control Bus.

Data Types: bus

liftingSize — Output lifting size

scalar

Output lifting size, returned as a scalar.

Data Types: uint16

nextFrame — Block ready indicator

Boolean scalar

Block ready indicator, returned as a Boolean scalar.

The block sets this signal to 1 (true) when the block is ready to accept the start of the next frame. If the block receives an input ctrl.start signal while nextFrame is 0 (false), the block discards the frame in progress and begins processing the new data.

Data Types: Boolean

actIter — Actual number of iterations

scalar

Actual number of iterations the block takes to decode the output, returned as a scalar.

Dependencies

To enable this port, set the Decoding termination criteria parameter to Early.

Data Types: uint8

parityCheck — Parity check status indicator

scalar

Parity check status indicator, returned as a Boolean scalar. The port indicates the status of the parity check after the decoding operation.

• 0 — Indicates that the parity check failed

• 1 — Indicates that the parity check passed

Dependencies

To enable this port, select the Enable parity check output port parameter.

Data Types: Boolean

Parameters

Algorithm — Type of algorithm

Min-sum (default) | Normalized min-sum

Select the type of algorithm. For more information, see Algorithm (5G Toolbox).

Scaling factor — Scaling factor

0.75 (default) | values in the range from 0.5 to 1, incremented by 0.0625

Specify the scaling factor.

Dependencies

To enable this parameter, set the Algorithm parameter to Normalized min-sum.

Decoding termination criteria — Termination criteria

Max (default) | Early

Select the decoding termination criteria.

• Max — Terminates decoding when the block reaches the number of iterations specified through the Number of iterations parameter or through the iter input port

• Early — Terminates decoding when all of the parity checks are met or when the block reaches the maximum number of iterations specified through the Maximum number of iterations parameter or through the iter input port

Source for number of iterations — Source selection for number of iterations

Property (default) | Input port

Select the source for specifying the number of iterations.

You can set the number of iterations by using either an input port or a parameter.

• Select Property to enable either the Number of iterations parameter or the Maximum number of iterations parameter.

• Select Input port to enable the iter port.

Number of iterations — Number of iterations

8 (default) | integer in the range from 1 to 63

Specify the number of iterations.

Dependencies

To enable this parameter, set the Decoding termination criteria parameter to Max and the Source for number of iterations parameter to Property.

Maximum number of iterations — Maximum number of iterations

8 (default) | integer in the range from 1 to 63

Specify the maximum number of iterations.

Dependencies

To enable this parameter, set the Decoding termination criteria parameter to Early and the Source for number of iterations parameter to Property.

Enable multiple code rates — Multiple code rates

off (default) | on

Select this parameter to enable the numRows input port to support multiple code rates. For more information about multiple code rates, see Multiple Code Rates.

Enable parity check output port — Parity check status

off (default) | on

Select this parameter to enable the parityCheck output port to view the status of the parity check.

More About

Specifying Vector Input

Vector input data for the block must be specified as a column vector of size 64. You must provide inputs as an integer number of ceil(liftingSize/64) clock cycles.

The total number of clock cycles that the block requires to receive a frame of LLR values for decoding is equal to n x ceil(liftingSize/64), where n is the number of columns in the parity check matrix. n depends on the base graph number, specified by the bgn input port. When the bgn port value is 0, the block sets n to 66. When the bgn port value is 1, the block sets n to 50.

These sections show how the block accepts input LLR values based on the liftingSize and bgn port values.

liftingSize input value is less than 64 and bgn value is 0

For a liftingSize input value of 2 and bgn input value of 0, the block can accept 132 LLRs. In this case, the block accepts the first two LLR input bits in each clock cycle and ignores the remaining 62 elements in that clock cycle. The total number of clock cycles the block requires to receive a frame of LLR values is 66.

The L_n elements represent LLR bits, and the X elements represent ignored values.

Input LLR Values	Number of Clock Cycles
	1 Clock Cycle	2 Clock Cycles	3 Clock Cycles	4 Clock Cycles	...	66 Clock Cycles
data[0]	L0	L2	L4	L6	...	L130
data[1]	L1	L3	L5	L7	...	L131
...	X	X	X	X	X	X
data[63]	X	X	X	X	X	X

liftingSize input value is greater than 64 and bgn value is 0

For a liftingSize input value of 104 and bgn input value of 0, the block can accept 6,864 LLRs. In this case, the block accepts 104 LLR values in two clock cycles: 64 LLRs in the first clock cycle and 40 LLRs in the second clock cycle. The block ignores the remaining 24 elements in the second clock cycle. The total number of clock cycles the block requires to receive input LLR values is 132.

The L_n elements represent LLR bits, and the X elements represent ignored values.

Input LLR Values	Number of Clock Cycles
	1 Clock Cycle	2 Clock Cycles	3 Clock Cycles	4 Clock Cycles	...	...	131 Clock Cycles	132 Clock Cycles
data[0]	L0	L64	L104	L168	...	...	L6760	L6824
data[1]	L1	L65	L105	L169	...	...	L6761	L6825
	...	...	...		...	...	...	...
...	...	L103	...	L207	...	...	...	L6863
...	...	X	...	X	...	...	...	X
data[63]	L63	X	L167	X	...	...	L6823	X

Multiple Code Rates

NR LDPC codes can support flexible code rates based on the parity check matrix (PCM) extension to achieve high throughputs and meet low latency requirements. The block supports multiple code rates by varying the number of rows of the parity check matrix.

For LDPC codes, the base parity check matrix (H_b) is a product of the number of rows (m_b) and the number of columns (n_b) of the matrix. The output (K) of the block is calculated as k_b x Z, where Z is the expansion factor or lifting size that can be in the range from 2 to 384, and k_b is equal to 22 for bgn value 0 and 10 for bgn value 1 as defined in the standard [1].

The input size (N) is calculated as, n_b x Z, where n_b is equal to m_b + k_b.

This figure shows a parity check matrix marked with a specified number of rows and columns, which you can use to calculate the code rates of the block. The code rate R is calculated as, k_b / (k_b – 2 + m_b) for the specified bgn value. In this figure, the values n_b1, n_b2, and n_b3 indicate the number of columns for the specified bgn value and values m_b1, m_b2, and m_b3 indicate the number of rows for the specified bgn value.

Algorithms

This figure shows the architecture block diagram of the NR LDPC Decoder block. The Controller block controls the layer and iteration count of the decoding process. The Variable node RAM block stores the variable node (VN) messages, and Check node RAM block stores the check node messages (CN). The Functional Unit block calculates the variable node (VN) messages and check node (CN) messages based on the layered belief propagation and either the normalized min-sum approximation algorithm or the min-sum approximation algorithm. The Termination/Parity check status block calculates the parity checks and provides the parity check status after each iteration. For more information about decoding algorithms, see Algorithm (5G Toolbox).

The implementation of the block matches the performance of the function nrLDPCDecode (5G Toolbox).

This plot shows the performance of the block for a 4 bit BPSK-modulated LLR input, when you set the Algorithm parameter to Min-sum, the bgn parameter to 0, and the liftingSize parameter to 384.

This plot shows the performance of the block for a 4 bit BPSK-modulated LLR input, when you set the Algorithm parameter to Normalized min-sum, the bgn parameter to 0, and the liftingSize parameter to 384.

This plot shows the average number of iterations taken to decode the data per EbNo for a 4-bit BPSK-modulated LLR input, when you set the Algorithm parameter to Min-sum and the Decoding termination criteria parameter to Early.

Latency

The latency of the block varies based on the values of the bgn, liftingSize, and numRows input ports and the number of iterations. Because the latency varies, use the nextFrame control signal output port to determine when the block is ready for a new input frame.

Scalar Input

The latency of the block is equal to r x (t + (m x 8) x ceil(liftingSize/64) + t + m x (7 – ceil(liftingSize/64))) + (n x liftingSize) + 18. In this calculation, r is the number of iterations, n is the number of columns in the parity check matrix, t is twice the total number of non –1 elements in the parity check matrix, m is the number of rows in the parity check matrix, and d is the pipeline delays. When you select the Enable multiple code rates parameter, d is 26. Otherwise, d is 18.

This figure shows a Logic Analyzer waveform of the sample output of the NR LDPC Decoder block with latency. In this case, the bgn and liftingSize input port values are set to 0 and 128, respectively, and the Number of iterations parameter is set to 8. The latency of the block is 31,362 clock cycles.

Vector Input

For vector inputs, the latency of the block is equal to r x (t + (m x 9)) + n x (ceil(liftingSize/64)) + d. In this calculation, r is the number of iterations, n is the number of columns in the parity check matrix, t is twice the total number of non -1 elements in the parity check matrix, m is the number of rows in the parity check matrix, and d is the pipeline delays. When you select the Enable multiple code rates parameter, d is 26. Otherwise, d is 18.

This figure shows a Logic Analyzer waveform of the sample output of the NR LDPC Decoder block with latency. In this case, the bgn and liftingSize input port values are set to 0 and 384, respectively, and the Number of iterations parameter is set to 8. The latency of the block is 8,782 clock cycles.

This figure shows a Logic Analyzer waveform of the sample output of the NR LDPC Decoder block with latency. In this case, the bgn, liftingSize, and numRows input port values are set to 0, 384, and 4, respectively, and the Number of iterations parameter is set to 8. The latency of the block is 1,674 clock cycles.

Throughput

The throughput of the block is calculated as (cwLen / latency) x f_max. In this calculation:

• cwLen is the code word length which is equal to k_b x Z, where k_b is 22 for the bgn value 0 and 10 for the bgn value 1.

• latency is the latency of the block for the specified configuration

• f_max is the maximum operating frequency

For more information about the latency calculation, see Latency. For more information about the maximum operating frequency, see Performance.

This plot shows the throughput versus the number of rows specified at the block input, when you set the Algorithm parameter to Min-sum, the Number of iterations parameter to 8, and the bgn input port to 0.

Performance

The performance of the synthesized HDL code varies with your target and synthesis options. It also varies based on the type of algorithm and the word length of the input LLR values.

This table shows the resource and performance data synthesis results of the block, when you set the Algorithm parameter to Min-sum, set the Number of iterations parameter to 8, and specify the input LLR values of data type fixdt(1,4,0). The generated HDL is targeted to the AMD^® Zynq^® UltraScale+™ RFSoC evaluation board.

Input Data	Slice LUTs	Slice Registers	Block RAMs	Maximum Frequency in MHz
Scalar	45319	54356	193	299.3
Vector	68987	63458	128.5	295.2

This table shows the resource and performance data synthesis results of the block for the vector input when you set the Algorithm parameter to Min-sum, set the Decoding termination criteria parameter to Max, set the Number of iterations parameter to 8, select the Enable multiple code rates parameter, and specify the input LLR values of data type fixdt(1,4,0). The generated HDL is targeted to the AMD Zynq UltraScale+ RFSoC evaluation board.

Slice LUTs	Slice Registers	Block RAMs	Maximum Frequency in MHz
74700	64425	128.5	280.3

References

[1] 3GPP TS 38.212. “NR; Multiplexing and Channel Coding.” 3rd Generation Partnership Project; Technical Specification Group Radio Access Network.

[2] Gallager, R. “Low-Density Parity-Check Codes.” IEEE Transactions on Information Theory 8, no. 1 (January 1962): 21–28. www.doi.org/10.1109/TIT.1962.1057683.

Extended Capabilities

C/C++ Code Generation

Generate C and C++ code using Simulink® Coder™.

This block supports C/C++ code generation for Simulink^® accelerator and rapid accelerator modes and for DPI component generation.

HDL Code Generation

Generate VHDL, Verilog and SystemVerilog code for FPGA and ASIC designs using HDL Coder™.

HDL Coder™ provides additional configuration options that affect HDL implementation and synthesized logic.

HDL Architecture

This block has one default HDL architecture.

HDL Block Properties

ConstrainedOutputPipeline	Number of registers to place at the outputs by moving existing delays within your design. Distributed pipelining does not redistribute these registers. The default is 0. For more details, see ConstrainedOutputPipeline (HDL Coder).
InputPipeline	Number of input pipeline stages to insert in the generated code. Distributed pipelining and constrained output pipelining can move these registers. The default is 0. For more details, see InputPipeline (HDL Coder).
OutputPipeline	Number of output pipeline stages to insert in the generated code. Distributed pipelining and constrained output pipelining can move these registers. The default is 0. For more details, see OutputPipeline (HDL Coder).

Restrictions

You cannot generate HDL for this block inside a Resettable Synchronous Subsystem (HDL Coder).

Version History

Introduced in R2020a