A Fault Tolerance Method for Encoding-Based Parallel Digital Channelizers

Abstract

The present invention relates to a fault tolerance method for a coded parallel digital channelizer, including the following steps: According to the number N of digital channelizers to be protected, construct two redundant digital channelizers, the input of which is the weighted sum of the inputs of N basic digital channelizers; perform weighted subtraction on the outputs of the redundant digital channelizers and the outputs of the basic digital channelizers to calculate the difference between one redundant channelizer and the basic channelizer and the difference pair between two redundant channelizers and the basic channelizer, perform threshold detection on them, count the number of those exceeding the corresponding thresholds in the difference pair, and perform a primary fault detection operation based on the statistical value; the error correction module generates a recovery signal and replaces the original faulty output.

Description

Technical Field

[0001] The present invention relates to a fault tolerance method for a parallel digital channelizer based on coding. Background Art

[0002] Digital channelizers are widely used in satellite communications. Their main function is to extract narrow sub-channels from broadband signals to achieve large-scale switching between wireless sub-channels. In particular, large communication satellite platforms often support multiple beams and high system capacity based on parallel digital channelizers. In recent years, the field programmable gate array (FPGA), a new type of custom circuit device, has developed rapidly. The SRAM-based FPGA (SRAM-FPGA) has the advantages of low cost, high density, and reconfigurability while having good programmability, which makes it possible to efficiently and economically implement large-scale parallel digital channelizers on satellites.

[0003] However, the space where satellites are located is filled with a large number of space radiation particles. Spaceborne semiconductor devices are extremely vulnerable to their influence and may malfunction, resulting in errors in the operation of the on-chip signal processing system. The most common fault is the single event upset (SEU) fault. In addition, as the signal processing system becomes more complex and the chip scale increases, the possibility of being affected by SEUs also gradually increases. Therefore, it is necessary to study solutions for the reliability problem of large-scale parallel digital channelizers on satellites.

[0004] The traditional SEU fault tolerance solution is the triple modular redundancy scheme, which has 3 identical original system instances and completes fault tolerance through a majority selector at the output end. It has the advantages of simple implementation logic and fast calculation speed. However, the resource overhead of this scheme is more than 3 times that of the module to be protected, and it is not suitable for spaceborne communication platforms with limited resources. An effective method is to construct a dedicated fault tolerance scheme according to the arithmetic characteristics of the module to be protected. The present invention designs an efficient fault tolerance method by using the arithmetic characteristics of the parallel digital channelizer. Summary of the Invention

[0005] Aiming at the above problems, the purpose of the present invention is to provide a fault tolerance method for a parallel digital channelizer based on coding. Based on the arithmetic characteristics of the channelizer and the analysis results of faults caused by SEUs, corresponding redundant channelizers and error detection and correction modules are designed to achieve the detection and correction of SEUs. Experiments have proved that the present invention can effectively tolerate SEU faults, and its resource overhead is significantly less than that of triple modular redundancy. To achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A fault tolerance method for a parallel digital channelizer based on coding, comprising the following steps:

[0007] 1) According to the number N of digital channelizers to be protected, construct 2-channel redundant digital channelizers, the input of which is the weighted sum of the inputs of N basic digital channelizers:

[0008] The input of the first-channel redundant digital channelizer is through the following formula:

[0009] The input of the second-channel redundant digital channelizer is through the following formula: where x i is the input of the basic digital channelizer, and the weighting coefficients a i = 1, b i = i, i = 1, 2,..., N;

[0010] 2) Subtract the outputs of the redundant digital channelizers from the outputs of the basic digital channelizers with weights, and calculate the differences between the first-channel redundant channelizer and the basic channelizer and the differences between the second-channel redundant channelizer and the basic channelizer The method is as follows:

[0011] Assume the number of output branches of the channelizer is D, The calculation results of and

[0012] are obtained through the following formula: where is the output of the basic digital channelizer, are the outputs of the first-channel and second-channel redundant channelizers at branch d respectively, is the fault signal caused by SEU in the d-th branch of the n-th channelizer. When no fault occurs in the system, and are the fixed-point quantization noises of the first-channel and second-channel redundant channelizers at branch d respectively, d = 1,..., D;

[0013] 3) Perform threshold detection on and Count the number of values exceeding the corresponding thresholds in and , and perform a fault detection operation based on the statistical values. The method is as follows:

[0014] Preset and are the thresholds of and respectively. After threshold detection, according to and the number of values exceeding the corresponding thresholds in Statistical value The fault detection situation is divided into three types:

[0015] · And It is considered that the outputs of the basic channelizer and the redundant channelizer maintain a linear relationship, and there is no fault signal There is no need to recover the output of the basic channelizer, and return to step 2);

[0016] · And And satisfy Or At this time, it is judged that the output result of the basic channelizer is correct, and the redundant channelizer fails. There is no need to recover the output of the basic channelizer, and return to step 2);

[0017] · And And do not satisfy And At this time, it is judged that the output result of one of the basic channelizers is incorrect. To achieve fault location and recovery, perform the operation in step 4);

[0018] 4) Set a time period T, and perform spatio-temporal averaging on the And that simultaneously exceed the threshold under each channel within the time period, and respectively obtain And The spatio-temporally averaged means Δ1 and Δ2, and estimate the fault channelizer label based on the averaged Δ1 and Δ2. The method is as follows:

[0019] The values of Δ1 and Δ2 are obtained through the following formula, and the formula is: Among them, the set ε(j,k) represents all the And that simultaneously exceed the threshold in the buffer matrix with a length of 2D on the spatial scale and a width of T on the temporal scale; the corresponding coordinates

[0020] The estimated value of the fault channelizer label is obtained through the following formula:

[0021] 5) The majority voting module continuously captures M estimated values of the fault channelizer label calculated in step 4 And perform majority voting on these M estimated values And generate a recovery signal;

[0022] 6) The error correction module generates a recovery signal and replaces the original faulty output.

[0023] Further, in step 5), among the M estimated values, if the number of occurrences of the estimated value of a certain faulty channelizer label is greater than half of the total number, then is identified as the channelizer label where the fault lies.

[0024] Due to the above technical solutions adopted by the present invention, it has the following advantages:

[0025] (1) The present invention protects the entire parallel digital channelizer implemented based on SRAM-FPGA, replacing the previous fault tolerance for partial structures of the channelizer such as filters and DFTs, and has a high implementation efficiency.

[0026] (2) The present invention adopts a fault tolerance method based on linear coding, utilizes the characteristic that a linear relationship can be maintained between the multiple inputs and outputs of the parallel digital channelizer for fault tolerance protection, and optimizes the fault tolerance algorithm for the quantization error of the fixed-point system.

[0027] (3) The present invention can provide efficient fault tolerance protection, and can ensure that the output signal-to-noise ratio of the protected channelizer is higher than 20 dB, meeting the requirements of most signal processing systems.

[0028] (4) The present invention has a small fault tolerance overhead, and the fault tolerance efficiency increases with the increase in the number of parallel digital channelizers. Description of the Drawings

[0029] Figure 1 is a flowchart of a fault tolerance method for a parallel digital channelizer system based on coding according to an embodiment of the present invention;

[0030] Figure 2 is the output signal SNR distribution of a parallel digital channelizer without protection and under coding protection affected by SEU according to an embodiment of the present invention Detailed Embodiments

[0031] Figure 1 is a flowchart of a fault tolerance method for a parallel digital channelizer system based on coding according to an embodiment of the present invention. The fault tolerance method for a parallel digital channelizer system based on coding according to an embodiment of the present invention includes the following steps:

[0032] 1) According to the number N of digital channelizers to be protected, construct 2-channel redundant digital channelizers, and their inputs are the weighted sums of the inputs of N basic digital channelizers.

[0033] In an embodiment of the present invention, the number N of digital channelizers to be protected is 8, the number D of output branches of the channelizer is 8, and the input of the first-channel redundant digital channelizer can be obtained through the following formula. The formula is The input of the second redundant digital channelizer can be obtained through the following formula: where x i is the input of the basic digital channelizer, and the weighting coefficients a i = 1, b i = i, i = 1, 2, …, N.

[0034] 2) Subtract the output of the redundant digital channelizer from the output of the basic digital channelizer after weighting to calculate the difference between the first redundant channelizer and the basic channelizer and the difference between the second redundant channelizer and the basic channelizer

[0035] Specifically, The calculation results of and

[0036] can be obtained through the following formula: is the output of the basic digital channelizer, are the outputs of the first and second redundant channelizers in branch D respectively, is the fault signal caused by SEU in the d-th branch of the n-th channelizer. When the system has no fault, and are the fixed-point quantization noises of the first and second redundant channelizers in branch D respectively.

[0037] 3) Perform threshold detection on and , count the number of values exceeding the corresponding thresholds in and , and perform a fault detection operation based on the statistical value

[0038] Specifically, set and to be the thresholds of and respectively. In an embodiment of the present invention, their values are preset based on the distribution characteristics of and when the system is running normally. After threshold detection, according to and the statistical values of the values exceeding the corresponding thresholds the fault detection situation can be divided into three types:

[0039] · It can be considered that the outputs of the basic channelizer and the redundant channelizer maintain a linear relationship and there is no fault signal The system will not recover the output of the basic channelizer and return to step 2.

[0040] · and satisfy or At this time, the output results of the 8-channel basic channelizer are correct, and the redundant channelizer fails. Since the redundant channelizer does not affect the final output result, error correction may not be performed. Therefore, the system will not recover the output of the basic channelizer and return to step 2.

[0041] · and satisfy At this time, the output result of one of the basic channelizers is incorrect. To achieve fault location and recovery, the system will perform the next operation.

[0042] 4) Spatially and temporally average the and that exceed the threshold value under each channel within a period of time to obtain and The mean values Δ1 and Δ2 after spatio-temporal averaging, and estimate the label of the faulty channelizer based on the averaged Δ1 and Δ2.

[0043] The values of Δ1 and Δ2 can be obtained through the following formula. The formula is where the set ε(j,k) represents all and that simultaneously exceed the threshold value in the buffer matrix with a length of 2D in the spatial scale and a width of T in the temporal scale. In an embodiment of the present invention, D = 8 and T = 3. The estimated value of the label of the faulty channelizer can be obtained through the following formula. The formula is

[0044] 5) The majority voting module continuously captures M estimated values of the label of the faulty channelizer calculated in step 4 and performs majority voting on these M estimated values to generate a recovery signal.

[0045] Specifically, among these M estimated values, if the occurrence times of a certain is greater than half of the total number, then finally this is determined as the label of the faulty channelizer. In an embodiment of the present invention, M = 15.

[0046] 6) The error correction module generates a recovery signal and replaces the original faulty output.

[0047] In an embodiment of the present invention, the recovery signal corresponding to the faulty digital channelizer can be obtained through the following formula. The formula is Among them, n is the subscript of the channelizer where a fault actually occurs. After generating the recovery signal, the error correction module bypasses the output of the faulty channelizer and connects the corresponding output port to the recovery signal output port. Thus, the fault tolerance scheme for the parallel digital channelizer based on coding is completed.

[0048] In order to verify the effect of the present invention, the following experiment was carried out. Systems with and without protection were designed and implemented when the number N of parallel digital channelizers was 8 or 16. The channelizer includes an 8-point FFT module and a 24th-order FIR filter module, and its input data and coefficients are quantized to 16 bits. The above fault tolerance method was implemented in HDL and run on a Xilinx FPGA. First, the resource overhead of the present invention was evaluated, as shown in Table 1 below. The percentages in parentheses in Table 1 are the overhead multiples relative to the unprotected parallel digital channelizer. It can be seen from Table 1 that the overhead of fault tolerance protection for the embodiment with N = 8 according to the present invention is less than 1.7 times that of the parallel digital channelizer to be protected. When N = 16, the fault tolerance overhead is further reduced to less than 1.4 times. In contrast, at least 3 times more resource overhead is required for triple modular redundancy. Thus, it can be seen that the present invention can effectively reduce the resource overhead of the system, and the fault tolerance efficiency increases with the increase in the number of parallel digital channelizers.

[0049]

[0050]

[0051] Table 1 shows the comparison of the resource consumption of the parallel digital channelizer without protection and with coding-based protection in two embodiments where the number N of parallel digital channelizers is 8 and N = 16.

[0052] Then, the effectiveness of the present invention was evaluated. The SNR distributions of the output signals of the parallel digital channelizer without protection and with coding-based protection under the influence of SEU are as Figure 2 shown. Without protection, the minimum SNR of the sub-channel output of the channelizer affected by SEU can reach -20 dB. After protection using the method of the present invention, the SNR of its sub-channel output can be increased to more than 20 dB. Thus, it can be seen that the present invention can provide strong fault tolerance for the parallel digital channelizer.

Claims

1. A fault tolerance method for a coded-based parallel digital channelizer, comprising the following steps: 1) According to the number N of digital channelizers to be protected, construct 2-channel redundant digital channelizers, whose input is the weighted sum of the inputs of N basic digital channelizers: The input of the first redundant digital channelizer is through the following formula: The input of the second redundant digital channelizer is given by the following formula: where x i is the input of the basic digital channelizer, and the weighting coefficient a i = 1, b i = i, i = 1, 2, …, N; 2) Weightedly subtract the output of the redundant digital channelizer from the output of the basic digital channelizer to calculate the difference between one redundant channelizer and the basic channelizer and the difference between two redundant channelizers and the basic channelizer The method is as follows: Let the number of output branches of the channelizer be D, and The calculation result of is obtained through the following formula: wherein is the output of the basic digital channelizer, are respectively the outputs of the first and second redundant channelizers on branch d, is the fault signal caused by SEU on the d-th branch of the n-th channelizer. When no fault occurs in the system, and are respectively the fixed-point quantization noises of the first and second redundant channelizers on branch d, where d = 1, …, D; 3) For and perform threshold detection, count and the number exceeding the corresponding threshold, and perform a primary fault detection operation based on the statistical value, the method is as follows: Preset and are respectively and Thresholds of. After threshold detection, according to and The statistical values that exceed the corresponding thresholds in are used to divide the fault detection situation into three types: ​ · and it is considered that the output of the basic channelizer and the redundant channelizer maintains a linear relationship and there is no fault signal There is no need to recover the output of the basic channelizer, and return to step 2); · and and satisfy or At this time, it is determined that the output result of the basic channelizer is correct, and the redundant channelizer fails. There is no need to recover the output of the basic channelizer, and return to step 2); · and and does not meet with At this time, it is determined that the output result of one of the basic channelizers is incorrect. To achieve fault location and recovery, the operation in step 4) is performed; 4) Set a time period T, and perform spatio-temporal averaging on the and that simultaneously exceed the threshold under each channel within the time period, and respectively obtain and the mean values Δ1 and Δ2 after spatio-temporal averaging. Then estimate the labels of the faulty channelizers based on the averaged Δ1 and Δ2. The method is as follows: The values of Δ1 and Δ2 are obtained through the following formula: where the set ε(j,k) represents all the and corresponding coordinates that simultaneously exceed the threshold in the buffer matrix with a length of 2D in the spatial scale and a width of T in the temporal scale; Estimated value of the faulty channelizer label Obtained by the following formula: 5) The majority voting module continuously captures M estimated values of the faulty channelizer labels obtained through step 4) and for these M estimated values performs majority voting and generates a recovery signal; 6) The error correction module generates a recovery signal and replaces the original faulty output.

2. The fault tolerance method of the parallel digital channelizer according to claim 1, wherein In step 5), among the M estimated values, if the number of occurrences of the estimated value of a certain faulty channelizer label is greater than half of the total number, then is identified as the channelizer label where the fault lies.

Images