CN106324313B - The seamless measuring system of transient signal based on approximate entropy - Google Patents

Abstract

The invention discloses a kind of seamless measuring systems of the transient signal based on approximate entropy,ADC module is sampled to obtain sampled signal and be stored to following characteristics memory to analog signal to be measured,Characteristic signal coarse sizing module carries out coarse sizing to sampled signal using sample mode and stores to screening signal storage,Threshold setting module is calculated characteristic signal postsearch screening according to sampled signal based on approximate entropy algorithm and compares threshold value,The approximate entropy of characteristic signal postsearch screening module calculating sifting signal,Compare threshold value according to characteristic signal postsearch screening to judge screening signal,It is then stored the sampled signal of following characteristics memory as characteristic signal into characteristic signal memory if it is characteristic signal,Data processing and waveform mapping block read characteristic signal and are handled and mapped,It is shown by display mapping ejected wave shape when the display cycle arrives.The present invention realizes the seamless measurement of transient state non-stationary signal time-domain information with the complexity of approximate entropy quantitative description signal.

Description

Transient Signal Seamless Measurement System Based on Approximate Entropy

technical field

The invention belongs to the technical field of transient signal measurement, and more specifically relates to a transient signal seamless measurement system based on approximate entropy.

Background technique

Modern electronic signals are becoming more and more complex and diverse, the frequency range of the signal is constantly expanding, and the instantaneous and non-stationary growth of the signal is extremely rapid. How to effectively extract the information carried by the signal and realize the efficient acquisition and real-time analysis of the transient non-stationary signal has brought challenges to modern signal measurement. On the one hand, we pursue higher-speed and higher-precision acquisitions to capture as many signal details as possible; on the other hand, the massive data acquired by high-sampling rate and high-resolution acquisitions bring new opportunities for real-time signal processing and analysis. Comes the hard way, affecting system response.

The digital oscilloscope using real-time sampling and real-time processing technology is a representative of modern time-domain measurement instruments, and the research on the seamless measurement capability of digital oscilloscopes has become a hot topic in recent years. A complete measurement process of a digital oscilloscope includes signal acquisition, data processing, and waveform display. Its signal acquisition can be regarded as a kind of intermittent sampling, and the processing and display time is separated between two adjacent acquisitions, which is called dead time or measurement gap. All signals occurring in the measurement gap cannot be effectively collected. It can be seen that the size of the measurement gap directly affects the signal measurement capability of the oscilloscope. The smaller the measurement gap, the higher the ratio of valid samples to the total measurement time, and the greater the probability that the oscilloscope will capture the transient signal. Therefore, it is very important for modern signal measurement to reduce the measurement gap until seamless measurement is realized.

Analyze the data processing process of the digital oscilloscope, including waveform image processing and some common digital signal analysis, such as interpolation, averaging, mathematical operations, FFT, digital filtering, etc. The display process is to present the waveform image and related analysis results on the display. On the premise that the sampling rate is determined, to reduce the measurement gap, it is necessary to shorten the time spent on processing and display. However, in order to completely eliminate measurement gaps and realize seamless measurement, processing and display must be completed at the same time as sampling, that is to say, the speed of processing and display must completely keep up with the speed of sampling, and the difficulty can be imagined. First of all, there is a huge gap between the computing speed (1-2GHz) of existing CPU, DSP and other processing devices and the sampling rate of ADC (up to 1-100GSPS), and the more types of signal analysis are included, the longer the processing time ; Secondly, limited by the display mechanism and the refresh rate (usually 50Hz) of the liquid crystal display, the waveform display process is even more time-consuming. Therefore, under the existing architecture and device level of the oscilloscope, it is almost impossible to realize completely seamless measurement.

At present, certain research results have been obtained on seamless measurement in the industry. Literature H.Zeng, H.Q.Pan and W.H.Huang, Key technology design of 6 GSPS high-speed digital storage oscilloscope, Proceedings of 2013 IEEE 11th International Conference on Electronic Measurement & Instruments, 385-391(2013) and H.Zeng, P.Ye, H.J.Wang and CH.Y.Xiang, Research on waveform mapping technology of digital three-dimensional oscilloscope, Chinese Journal of Scientific Instrument, 30(11), 2399-2404(2009) used a parallel processing architecture for data processing, using multiple The waveform statistical superposition method is used to display, which compresses the processing time per unit time and increases the number of waveform displays, thereby reducing the measurement gap to a certain extent, but still cannot achieve seamless measurement. On this basis, the document K.J.Yang, S.L.Tian, H.Zeng, L.Qiuand L.P.Guo, A seamless acquisition digital storage oscilloscope with three-dimensional waveform display, Review of Scientific Instruments, 85, 045102 (2014) uses DDR deep memory Segmented storage and ping-pong operation are constructed to realize the seamless acquisition oscilloscope under the condition of a certain amount of data within a certain period of time. However, the research on the above-mentioned seamless measurement is still limited to the measurement of slow signals in some specific applications, and there is generally a disconnection between acquisition and display (cannot be displayed during acquisition, cannot be acquired during display), poor real-time performance, and cannot be continuously seamless The problem.

In fact, it is neither possible nor necessary to achieve completely seamless measurement of the signal under test over a long period of time. Because the information concerned by different application fields is completely different, and it is unacceptable for people to acquire too much information or display too many signal waveforms at the same time. The famous scientist Shannon once pointed out that there is redundancy in any information, and the size of the redundancy is related to the probability or uncertainty of each symbol in the information. The information entropy is the dimension that characterizes the degree of signal uncertainty. From the perspective of information entropy, it is not difficult to find that there are always some abnormal components superimposed in the unsteady signal, which makes the information entropy of the signal different. Seamless measurement of transient signals is very important. Therefore, if in the process of signal acquisition, based on the real-time control of information entropy, the features contained in the signal can be extracted, the key or useful information can be retained, and the redundant or useless information can be discarded, which can not only greatly reduce the burden of processing and display, but also realize the characteristic signal seamless measurement and make access to information truly usable by humans. However, the information entropy (Shannon entropy) is the same as the entropy in the physical category. Due to its chaotic phenomenon, the calculation process is very complicated, and it is difficult to meet the real-time requirements of the seamless measurement system.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art, provide a transient signal seamless measurement system based on approximate entropy, quantitatively describe the complexity of the signal with the approximate entropy value, and guide the acquisition and processing of the signal based on this, Realize seamless measurement of transient non-stationary signal time domain information.

In order to achieve the purpose of the above invention, the present invention based on approximate entropy transient signal seamless measurement system includes ADC module, characteristic signal rough screening module, threshold value setting module, detailed signal storage, screening signal storage, characteristic signal secondary screening module, data processing and waveform mapping modules and displays, where:

The ADC module samples the analog signal x(t) to be tested to obtain the sampled signal X(l), which is sent to the feature signal coarse screening module, threshold setting module and detailed signal memory respectively;

The feature signal coarse screening module adopts sampling mode to carry out feature value rough screening to the sampling signal X (l) to obtain the screening signal Y (n), and the screening signal Y (n) is stored in the screening signal memory;

The threshold setting module calculates the secondary screening comparison threshold G of the characteristic signal according to the sampling signal X(l) and sends it to the secondary screening module of the characteristic signal. The modules are the same, the sampling signal Y'(n) is obtained, the approximate entropy ApEn' of Y'(n) is calculated by the approximate entropy algorithm, and then the characteristic signal secondary screening comparison threshold G=P×ApEn' is calculated, and P represents the preset ratio coefficient;

The detailed signal memory is used to store the sampled signal X(l);

The filter signal memory is used to store the filter signal Y(n);

The characteristic signal secondary screening module reads the screening signal Y(n) from the screening signal memory, and uses the approximate entropy algorithm to calculate the approximate entropy ApEn of Y(n). If ApEn>G, send the dump data signal to the detailed signal memory, otherwise do not any operation;

Characteristic signal memory reads the sampling signal X (1) in the detailed signal memory and stores as characteristic signal X' (1) after receiving the dump data signal;

The data processing and waveform mapping module monitors the characteristic signal memory, and reads the characteristic signal X′(l) from the characteristic signal memory every time its data is updated, performs data processing and real-time waveform mapping, and uses three-dimensional waveform in waveform mapping Mapping; when the display period arrives, the data processing and waveform mapping module sends the mapped waveform to the display;

The display is used to display the mapped waveform sent by the data processing and waveform mapping module.

The present invention is a transient signal seamless measurement system based on approximate entropy. The ADC module samples the analog signal to be measured and stores the sampled signal in the detailed signal storage. The characteristic signal rough screening module uses sampling to roughly screen the sampled signal and stores it in the screening signal storage. , the threshold setting module is based on the approximate entropy algorithm and calculates the threshold value of the secondary screening comparison of the characteristic signal according to the sampling signal. The secondary screening module of the characteristic signal calculates the approximate entropy of the screening signal, and judges the screening signal according to the secondary screening comparison threshold of the characteristic signal. For the characteristic signal, the sampling signal of the detailed signal memory is stored as a characteristic signal in the characteristic signal memory. The data processing and waveform mapping module reads the characteristic signal for processing and mapping. When the display period arrives, the display will display the mapped waveform.

The present invention quantitatively describes the complexity (that is, the amount of information) of the sampling signal with the approximate entropy value, and based on the real-time control of the approximate entropy, adaptively captures the characteristic signal, extracts key or useful information, and discards redundant or useless information, thereby reducing The burden of data processing and waveform display is realized, and the seamless measurement of time domain information of transient signals is realized.

Description of drawings

Fig. 1 is the flow chart of approximate entropy algorithm;

Fig. 2 is the specific embodiment structural diagram of the transient signal seamless measurement system based on approximate entropy of the present invention;

Fig. 3 is the schematic diagram of peak sampling in the present embodiment;

Fig. 4 is a threshold setting flowchart;

Fig. 5 is a schematic diagram of mean value sampling;

Figure 6 is a flow chart of data processing and waveform mapping module processing of the multi-stage pipeline processing mechanism;

Fig. 7 is the sampling signal waveform figure of the first group of analog signals to be tested;

Fig. 8 is the screening signal waveform diagram of the first group of analog signals to be tested;

Fig. 9 is the sampling signal waveform diagram of the second group of analog signals to be tested;

Fig. 10 is the screening signal waveform diagram of the second group of analog signals to be tested;

Fig. 11 is the sampling signal waveform diagram of the third group of analog signals to be tested;

Fig. 12 is the screening signal waveform diagram of the third group of analog signals to be tested;

Fig. 13 is the sampling signal waveform diagram of the fourth group of analog signals to be tested;

Fig. 14 is the screening signal waveform diagram of the fourth group of analog signals to be tested;

Fig. 15 is the display result of the first group of analog signals to be tested;

Fig. 16 is the display result of the second group of analog signals to be tested;

Fig. 17 is the display result of the third group of analog signals to be tested;

Fig. 18 is the display result of the fourth group of analog signals to be tested.

Detailed ways

Specific embodiments of the present invention will be described below in conjunction with the accompanying drawings, so that those skilled in the art can better understand the present invention. It should be noted that in the following description, when detailed descriptions of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

In order to better illustrate the technical content of the present invention, the approximate entropy is briefly introduced first.

The concept of approximate entropy was proposed by Steven M.Pincus in 1991 from the perspective of measuring the complexity of time series, and it is used to measure the probability of generating new patterns in signals. The greater the probability of a signal generating a new pattern, the greater the complexity of the sequence and the greater the corresponding entropy. That is, the approximate entropy algorithm uses a non-negative number to quantitatively represent the complexity and irregularity of the signal. Figure 1 is a flowchart of the approximate entropy algorithm. As shown in Figure 1, suppose the original signal time series is x(n)=x(1), x(2),...,x(N), with a total of N sample points, the specific steps of the approximate entropy algorithm are as follows:

S101: Form the signal x(n) into a set of m-dimensional vectors in consecutive order of sequence numbers:

X _i =[x(i),x(i+1),...,x(i+m-1)], i=1～N-m+1 (1)

Define the distance d _ij between the vectors X _i and X _j as the one with the largest difference among the corresponding elements of the two, that is

And for each value of i, calculate the distance between the vector X _i and other vectors X _j (j=1˜N-m+1, j≠i).

S102: Set the threshold value r (r>0), for each value of i, count the number n _ij (r) of d _ij <r, and record the ratio of n _ij (r) to the total number of distances N-m+1 as which is:

S103: will Take the logarithm, and calculate its average value for all i, denoted as φ ^m (r), that is

S104: Add 1 to the number of dimensions, construct the signal x(n) to obtain an m+1-dimensional vector, and calculate the distance d _ij between the vector X _i and other vectors X _j .

S105: Similarly, count the number n _ij (r) of d _ij <r, and calculate

S106: Calculate φ ^m+1 (r):

S107: Calculate the theoretical approximate entropy of this signal x(n) as:

Generally speaking, this extreme value exists with probability 1. In practice, N cannot be ∞. When N is a finite value, the estimated value of the approximate entropy ApEn when the sequence length is N is obtained according to the above steps, denoted as

ApEn(m,r,N)＝ ^φm (r)-φm ⁺¹ (r) (7)

The value of ApEn is obviously related to the values of m and r. According to practice, Pincus suggests m=2, r=0.1~0.2SD (SD is the original data x(i), i=1~n standard deviation (standard deviation)). The selection of the dimension m has been discussed in detail by technical personnel in the industry. Therefore, the values of m and r can be set according to actual needs.

Through the above calculation steps, it is not difficult to find that the physical essence of approximate entropy is to measure the mean value of the logarithmic conditional probability of the emergence of new patterns in the signal sequence when the dimension changes. Therefore, theoretically approximate entropy has significance in characterizing the irregularity and complexity of signal sequences.

According to the analysis of the physical nature of approximate entropy and the comprehensive discussion of related literature, the main characteristics of approximate entropy algorithm suitable for signal analysis in the field of electronic measurement can be obtained:

1) Due to the constraints of the tolerance threshold r, the approximate entropy algorithm has better anti-noise ability. For electronic measurement, the measured signal often contains high-frequency noise interference, so the anti-noise ability of the feature extraction method is very important.

2) The approximate entropy algorithm has nothing to do with the amplitude of the signal sequence, but only with the complexity of the sequence. When the electronic measurement is faced with small signals, the characteristic that the approximate entropy has nothing to do with the amplitude is very important.

3) As a nonlinear dynamic parameter, the approximate entropy algorithm is applicable to both stochastic and deterministic processes, while the signals faced by electronic measurement are often complex signals containing both deterministic and random components, so this feature is also very important. important.

4) The approximate entropy algorithm can obtain a reasonable and robust estimate with only short data, which enables it to extract the characteristic information contained in the signal sequence in a short period of time, so it is suitable for electronic measurement with high real-time requirements field.

5) The analysis effect of approximate entropy algorithm is better than statistical parameters such as mean value, variance, standard deviation, etc., so that it can extract characteristic signals more accurately and effectively.

In summary, the approximate entropy algorithm is suitable for signal analysis in the field of electronic measurement, and it can provide a quantifiable new method for extracting characteristic signals. Based on this, the present invention guides signal acquisition and data processing, and proposes an instantaneous The state signal seamless measurement system realizes the seamless measurement of the transient signal.

Fig. 2 is a structural diagram of a specific embodiment of the transient signal seamless measurement system based on approximate entropy of the present invention. As shown in Figure 2, the transient signal seamless measurement system based on approximate entropy of the present invention comprises ADC (Analog-to-Digital Converter, analog-to-digital converter) module 1, feature signal coarse screening module 2, threshold value setting module 3, detailed signal memory 4 , filter signal memory 5 , characteristic signal secondary screening module 6 , characteristic signal memory 7 , data processing and waveform mapping module 8 and display 9 .

The ADC module 1 samples the analog signal x(t) to be tested to obtain the sampled signal X(l), which is sent to the feature signal coarse screening module 2, the threshold setting module 3 and the detailed signal memory 4 respectively.

The feature signal rough screening module 2 performs feature value rough screening on the sampling signal X(l) in a sampling manner to obtain a screening signal Y(n), and stores the screening signal Y(n) in the screening signal memory 5.

In the present invention, the seamless measurement is realized based on the calculation and comparison of the approximate entropy of the signal, and the calculation amount of the approximate entropy algorithm doubles with the increase of the signal sequence length (data volume). For long data, calculating the approximate entropy directly will be very time-consuming. However, it has been proved by research that the approximate entropy algorithm actually only needs shorter data (100-1000) to obtain a reasonable and robust estimate. Therefore, before calculating approximate entropy, it is necessary to perform eigenvalue sampling processing on long data, that is, to perform rough screening of eigensignals.

Specifically in a digital oscilloscope, the amount of data (sample points) collected at one time is determined by its storage depth. At present, according to different performances, oscilloscopes generally have a variable storage depth in the range of 1kpts ~ 1Gpts, that is, the amount of data collected at a time is 10 ³ ~ 10 ⁹ . Therefore, for different storage depths, different sampling intervals can be selected, the sampled signal can be roughly screened by the sampling method, and the data volume of the signal after rough screening can be controlled between 100 and 1000, which can greatly reduce the subsequent feature signal. Calculation time of approximate entropy for secondary screening.

The design of feature signal coarse screening module 2 in the present embodiment is based on the peak value detection function of digital oscilloscope, and its purpose is to screen out the peak-to-peak value in a certain interval time (sampling interval k) in real time from the massive data of ADC sampling signal X(l) data, forming the screening signal Y(n). Since each group of peak-to-peak data contains two data, the maximum value and the minimum value, it is also necessary to sort according to the characteristics of the signal itself to determine the order of the maximum value and the minimum value. After peak sampling, the data volume of Y(n) is reduced to 2/k of X(l), that is, the data length of Y(n) is N=2L/k, and L represents the length of the sampled signal X(l). Fig. 3 is a schematic diagram of peak sampling in this embodiment. According to Figure 3, peak sampling can be expressed by the following formula:

Threshold value setting module 3 calculates the characteristic signal secondary screening comparison threshold G according to the sampled signal X(l). Figure 4 is a flow chart of threshold setting. As shown in Figure 4, the specific method of threshold setting is:

After entering the threshold value setting module 3 and receiving the sampling signal X(l), the sampling signal X(l) is first sampled, and the sampling rate is the same as that of the characteristic signal coarse screening module, so as to obtain the sampling signal Y'(n), obviously the sampling signal Y' (n) is the same length as the screening signal Y(n). In the present embodiment, the design of the threshold setting module 3 is based on the high-resolution sampling function of the digital oscilloscope, so-called high-resolution sampling, its essence is to seek the average value of the sampling signal in a certain interval, that is, to sample the signal X (l) Mean value sampling, because feature signal coarse screening module 2 adopts peak-to-peak sampling in the present embodiment, extract two peak values in each sampling interval k, so sampling signal X (l) carries out the sampling interval k'=k/ of mean value sampling 2, then the average sampling signal Y'(n) equivalent to the data volume of the screening signal Y(n) can be obtained. Fig. 5 is a schematic diagram of mean value sampling. According to Figure 5, the principle of average value sampling can be expressed in the following formula:

Then, the approximate entropy ApEn' of Y'(n) is calculated by using the approximate entropy algorithm, and ApEn' is associated with the user-preset proportional coefficient P, and the comparison threshold G of the secondary screening of the characteristic signal can be calculated by the following formula:

G=P×ApEn' (10)

Wherein, the value range of P is P>1.

In consideration of the high real-time requirement of seamless measurement, in this embodiment, a computing architecture of parallel and pipeline processing is chosen to be constructed in FPGA to realize fast calculation of approximate entropy. In this embodiment, the standard deviation SD is used to set the threshold. According to the process shown in Figure 1, the difficulty of implementing the approximate entropy algorithm with FPGA is to calculate a large number of complex operations such as distance d _ij , standard deviation SD, and logarithm and square root. For the calculation of d _ij and SD, since the original data participating in the calculation in the present invention is a time series obtained by continuous sampling and screening, it can be realized by using parallel plus pipeline technology, that is, at each clock beat, new sample points are calculated in parallel The distance d _ij from all old sample points and the sum of the new sample point and all old sample points; for complex operations such as logarithm and square root, it can be completed by calling the integrated CORDIC algorithm IP core inside the FPGA.

The detailed signal memory 4 is used to store the sampled signal X(l).

The filter signal memory 5 is used to store the filter signal Y(n).

Characteristic signal secondary screening module 6 reads screening signal Y(n) from screening signal memory 5, calculates approximate entropy ApEn, the comparison result of ApEn and threshold value G is used as the control signal of characteristic signal memory 7, if ApEn＞G, then judge screening The signal Y(n) is the profile of the characteristic signal, the sampling signal X(l) is the detail of the characteristic signal, and the dumped data signal is sent to the detailed signal memory, otherwise it is determined that the screening signal Y(n) is the profile of the non-characteristic signal, and the sampling signal X(l) is ) are non-characteristic signal details, and do not perform any operations.

After receiving the dumped data signal, the characteristic signal memory 7 reads the sampling signal X(l) in the detailed signal memory 4 and stores it as the characteristic signal X'(l).

The data processing and waveform mapping module 8 monitors the characteristic signal memory 7, and reads the characteristic signal X'(1) from the characteristic signal memory 7 whenever its data is updated, and performs various data processing and real-time waveform mapping. Three-dimensional waveform mapping is used in waveform mapping, that is, multiple waveforms are statistically superimposed. When the display period comes, the data processing and waveform mapping module 8 sends the mapped waveform to the display 9 .

The display 9 is used to display the mapped waveform sent by the data processing and waveform mapping module 8 .

In the present invention, since the data processing and waveform mapping module 8 needs to complete various data processing of the characteristic signal X'(l) and real-time mapping from the sampling data to the waveform image, even if the characteristic after secondary screening per unit time The total number of signals X'(l) (number of waveforms) has been greatly reduced compared with the original signal X(l), but this module still undertakes heavy data processing tasks, so it is also the main source of measurement gaps. Therefore, in order to reduce the measurement gap as much as possible, the data processing and waveform mapping modules of the multi-stage pipeline processing mechanism are preferably designed.

Figure 6 is a flow chart of the data processing and waveform mapping module of the multi-stage pipeline processing mechanism. As shown in Figure 6, for each data to be processed, each stage in the pipeline processing line completes a processing task (such as interpolation, averaging, mathematical operations, FFT, digital filtering, waveform mapping, etc.). If there are Q processing tasks, a Q-level pipeline processing line is required.

Waveform mapping is set at the last stage (Q stage) of the pipeline processing line. Its main task is to map the characteristic signal processed by the previous Q-1 stage into a waveform image, and realize the statistical superposition display of several waveform images. The statistical superposition process of waveforms is based on the continuous updating of the constructed waveform database. For each pixel in the waveform image, there is an independent storage unit corresponding to it in the waveform database. Whenever a certain data in the characteristic signal relates to the unit, the internal counter of the unit will be increased by 1, if not involved, it will not be increased (the initial value of the counter is 0). Finally, when the refresh period comes, the entire waveform database stores and displays the statistical superimposition results of all waveform images in the refresh period, thus saving the time spent on refreshing the waveform images sequentially.

Because what the present invention adopts is three-dimensional waveform mapping, then mathematically, waveform database can be regarded as a two-dimensional matrix A, and the size of matrix element a _ij represents the same sampling point (time and amplitude are all the same) in S waveforms The number of occurrences, i.e. the number of hits

Each column vector of the matrix A reflects the distribution of the values of the sampling points at that moment, and 1≤j≤L, S represents the number of characteristic signals processed in the refresh period.

The so-called waveform mapping process is to mark each element in the matrix A according to each data of the characteristic signal X'(l). After S times of mapping, the matrix A contains the value frequency of each sampling point at each moment in the time period. The display may display different brightnesses according to the values of matrix elements a _ij .

In order to illustrate the effectiveness of the present invention, the following theoretical derivation is made on the conditions that need to be met to realize the seamless measurement by using the present invention.

According to the above process, the main factors affecting the realization of seamless measurement are: the sampling rate f _s of the analog-to-digital converter, the length L of the sampling signal X(l), the threshold setting time t _gate , the twice screening time t _screen , data processing and Single waveform mapping time t _process and display refresh time t _lcd etc. Among them, since the coarse screening of the characteristic signal can be carried out synchronously and in real time during the signal sampling process, and does not require additional time, the two screening time t _screen is approximately equal to the calculation time t _ApEn of the approximate entropy during the second screening, namely

t _screen ≈t _ApEn (12)

Similarly, the threshold setting time t _gate is also approximately equal to the approximate entropy calculation time t _ApEn , so there is

t _screen ＝t _gate ≈t _ApEn (13)

The calculation time of the two is equal, and the results can be obtained at the same time through parallel processing. In addition, since the present embodiment adopts the display mechanism of statistical superimposition of multiple waveforms, new waveform mapping can be performed simultaneously within the refresh time t _lcd of the display without wasting time. Therefore, the total signal processing time T _process per unit time only depends on f _s , L, t _ApEn , t _process and the number D' of characteristic signals.

As mentioned above, to achieve seamless measurement, the total processing time T _process must be less than or equal to the total acquisition time T _acq , namely

T _process ≤ _{T acq} (14)

In unit time, let T _acq =1s, then must satisfy

T _process ＝t _ApEn ×D+t _process ×D'≤1 (15)

Among them, D is the quantity (number of waveforms) of sampling signal x(l) in unit time (1s), which can be obtained by sampling rate f _s and length L, namely

D＝f _s /L (16)

And D' is the quantity of the characteristic signal X'(l) after secondary screening. Putting formula (16) into formula (15), we can get:

Then from equations (16) and (17) we can get

Equations (17) and (18) are the necessary conditions to realize the seamless measurement of transient signals, that is, within a unit time (1s), when the quantity D' (absolute quantity) of the characteristic signal X'(l) satisfies the equation (17 ), or the percentage Z (relative quantity) of the quantity D' of the characteristic signal X'(l) and the quantity D of the sampling signal x(l) satisfies the formula (18), the seamless measurement of the transient signal can be realized. However, in the present invention, the determination of the characteristic signal is realized through the threshold G, so the number of characteristic signals can be controlled by controlling G, so as to achieve the purpose of seamless measurement.

Example

In order to verify the effectiveness of the present invention in screening transient signals of different complexity, an oscilloscope acquisition system is constructed using the 8Bit ADC model (Ideal_8_Bit.adc) provided by ADI. It is assumed that the sampling rate of the oscilloscope is f _s =1GSa/s, the storage depth L=1Mpts (that is, the data volume of each sampling signal X(l) is 10 ⁶ ), and the internal system clock f _c =250MHz. Take N=200 data to participate in the secondary screening of characteristic signals, so the peak sampling interval of the rough screening module k=2L/N=2×10 ⁶ /200=10000, and the average sampling interval of the threshold setting module k'=k/2 =5000. In the approximate entropy algorithm, m=2, r=0.2SD. In the following, four sets of sample data are used for simulation.

Record the first group of analog signals to be tested x ₁ (t)=sin(2πf ₀ t), and its frequency f ₀ =1kHz. Fig. 7 is a waveform diagram of sampling signals of the first group of analog signals to be tested. Fig. 8 is a waveform diagram of the filtered signal of the first group of analog signals to be tested. After calculation, the approximate entropy ApEn ₁ of the screening signal (peak sampling signal) Y ₁ (n) at this time = 0.0846, and the approximate entropy ApEn ₁ ′ of the sampling signal (average sampling signal) Y ₁ ′ (n) = 0.0858.

Note that the second group of analog signals to be tested is x ₂ (t)=sin(2πf ₀ t), and its frequency f ₀ =1kHz, but in order to simulate transient phenomena such as occasional noise interference and AD quantization errors, random Added distorted samples (glitches). Fig. 9 is a waveform diagram of sampling signals of the second group of analog signals to be tested. Fig. 10 is a waveform diagram of the filtered signal of the second group of analog signals to be tested. After calculation, the approximate entropy ApEn ₂ of the screening signal (peak sampling signal) Y ₂ (n) at this time = 0.1040, and the approximate entropy ApEn ₂ ′ of the sampling signal (average sampling signal) Y ₂ ′(n) = 0.0858.

Record the third group of analog signals to be tested x ₃ (t)＝0.25×[sin(2πf ₀ t)+sin(6πf ₀ t)+sin(10πf ₀ t)+sin(14πf ₀ t)], its frequency f ₀ = 1kHz. It can be seen that in order to simulate harmonic distortion, the 3rd, 5th and 7th harmonics are superimposed in the sinusoidal signal with frequency f ₀ =1 kHz. FIG. 11 is a waveform diagram of a sampled signal of a third group of analog signals to be tested. Fig. 12 is a waveform diagram of the filtered signal of the third group of analog signals to be tested. After calculation, the approximate entropy ApEn ₃ of the screening signal (peak sampling signal) Y ₃ (n) at this time is 0.4537, and the approximate entropy ApEn 3 ′ _of the sampling signal (average sampling signal) Y ₃ ′(n) is 0.3706.

Note that the fourth group of analog signals to be tested is x ₄ (t)=0.5×sin(2πf ₀ t)+Noise, and its frequency f ₀ =1kHz. It can be seen that, in order to simulate severe noise interference, uniform white noise Noise with mean value 0 and variance 1 is superimposed on the sinusoidal signal with frequency f ₀ =1 kHz. Fig. 13 is a waveform diagram of the sampled signals of the fourth group of analog signals to be tested. Fig. 14 is a waveform diagram of the filtered signal of the fourth group of analog signals to be tested. After calculation, the approximate entropy ApEn ₄ of the screening signal (peak sampling signal) Y ₄ (n) at this time = 0.6901, and the approximate entropy ApEn ₄ ′ of the sampling signal (average sampling signal) Y ₄ ′ (n) = 0.0861.

Define R as the ratio of the approximate entropy ApEn of the screening signal Y(n) to the approximate entropy ApEn' of the sampling signal Y'(n):

R=ApEn/ApEn' (19)

Table 1 is the approximate entropy calculation results of the four sets of sample data.

serial number ApEn ApEn' R First group ApEn ₁ =0.0846 ApEn ₁ '=0.0858 R ₁ =0.9861 Second Group ApEn ₂ =0.1040 ApEn ₂ '=0.0858 R ₂ =1.2115 The third group ApEn ₃ =0.4537 ApEn ₃ '=0.3706 R ₃ =1.2243 Fourth group ApEn ₄ =0.6901 ApEn ₄ '=0.0861 R ₄ =8.0151

Table 1

It can be seen from Table 1 that as the complexity of the sampled signal X(l) increases, the approximate entropy ApEn of the screened signal Y(n) increases gradually; The ratio R of the approximate entropy ApEn' of n) also increases gradually. This result further shows that:

1. Approximate entropy is a quantitative indicator of signal complexity. The more complex the signal sequence, the greater the approximate entropy;

2. Use the peak sampling method to roughly screen the signal, which can retain the complexity and characteristic information of the signal;

3. The ratio of the approximate entropy of the screening signal to the sampling signal can be used to guide the secondary screening of the characteristic signal.

Therefore, for the secondary screening of the characteristic signals of the above four groups of samples, the user can set the proportionality factor P=1.2. Table 2 is the comparison threshold G of the secondary screening of the characteristic signal and the results of the secondary screening.

serial number ApEn G Comparing results filter results First group ApEn ₁ =0.0846 G ₁ =0.1030 ApEn ₁ <G ₁ non-characteristic signal Second Group ApEn ₂ =0.1040 G ₂ =0.1030 ApEn ₂ >G ₂ characteristic signal The third group ApEn ₃ =0.4537 G ₃ =0.4447 ApEn ₃ >G ₃ characteristic signal Fourth group ApEn ₄ =0.6901 G ₄ =0.1033 ApEn ₄ >G ₄ characteristic signal

Table 2

It can be seen from Table 2 that when the user sets the proportional coefficient P=1.2, the transient signals including noise interference, AD quantization error and harmonic distortion can be effectively screened. Further, the user can realize effective screening of transient signals of different complexity by setting different proportional coefficients P. The larger the P value, the stricter the screening conditions, the smaller the number D' of the characteristic signal X'(l) after secondary screening, the smaller the percentage Z of D' and the number D of the sampled signal X(l), the closer the system is to seamless.

Using the transient signal seamless measurement system based on approximate entropy of the present invention to construct a digital oscilloscope, its real-time sampling rate f _s =1GSa/s, storage depth L=1Mpts, internal system clock frequency f _c =250MHz, the calculation of approximate entropy is about takes 200 system clock cycles, ie

t _ApEn ≈200/f _c ＝8×10 ^-7 s (20)

L data for data processing and waveform mapping time

t _process ≈l/f _c ＝4×10 ^-3 s (21)

Therefore, according to formulas (17) and (18), within a unit time (1s), when the number D' of the characteristic signal X'(l)≤249 or the number D' of the characteristic signal X'(l) is the same as the sampling signal X (1) When the percentage Z of the quantity D≤24.98%, the seamless measurement of the transient signal can be realized.

In the actual test, use the Tektronix arbitrary waveform generator AWG5014B to generate the above four groups of measured analog signals of x ₁ (t), x ₂ (t), x ₃ (t) and x ₄ (t) and input them to the oscilloscope, and set the scale factor P=1.2. Fig. 15 is the display result of the first group of analog signals to be tested. Fig. 16 is the display result of the second group of analog signals to be tested. Fig. 17 is the display result of the third group of analog signals to be tested. Fig. 18 is the display result of the fourth group of analog signals to be tested. In Fig. 15, since all sampling signals X ₁ (l) are judged to be non-characteristic signals and discarded during the refresh cycle, the oscilloscope displays no waveform, and the system is always in a state of waiting for a trigger; in Fig. 16 and Fig. 17, all The sampling signals X ₂ (l) and X ₃ (l) of glitches and harmonics of each order are judged as characteristic signals and reserved, and the oscilloscope performs waveform mapping and display on them; in Fig. 18, all superimposed white noise The sampled signal X ₄ (l) was judged to be a characteristic signal and kept, but serious noise affected the trigger system of the oscilloscope, resulting in edge trigger errors, so the waveform display appeared shaking and double-edge phenomena. The actual test results of the above four analog signals in the oscilloscope further verify the effectiveness of the present invention for screening transient signals of different complexity.

Although the illustrative specific embodiments of the present invention have been described above, so that those skilled in the art can understand the present invention, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, As long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions and creations using the concept of the present invention are included in the protection list.

Claims (4)

1. A transient signal seamless measurement system based on approximate entropy, characterized in that it comprises an ADC module, a characteristic signal coarse screening module, a threshold setting module, a detailed signal memory, a screening signal memory, a characteristic signal secondary screening module, and data processing and waveform mapping modules and displays, where: The ADC module samples the analog signal x(t) to be tested to obtain the sampled signal X(l), which is sent to the feature signal coarse screening module, threshold setting module and detailed signal memory respectively; The feature signal coarse screening module adopts sampling mode to carry out feature value rough screening to the sampling signal X (l) to obtain the screening signal Y (n), and the screening signal Y (n) is stored in the screening signal memory; The threshold setting module calculates the secondary screening comparison threshold G of the characteristic signal according to the sampling signal X(l) and sends it to the secondary screening module of the characteristic signal. The modules are the same, the sampling signal Y'(n) is obtained, the approximate entropy ApEn' of Y'(n) is calculated by the approximate entropy algorithm, and then the characteristic signal secondary screening comparison threshold G=P×ApEn' is calculated, and P represents the preset ratio coefficient; The detailed signal memory is used to store the sampled signal X(l); The filter signal memory is used to store the filter signal Y(n); The characteristic signal secondary screening module reads the screening signal Y(n) from the screening signal memory, and uses the approximate entropy algorithm to calculate the approximate entropy ApEn of Y(n). If ApEn>G, send the dump data signal to the detailed signal memory, otherwise do not any operation; Characteristic signal memory reads the sampling signal X (1) in the detailed signal memory and stores as characteristic signal X' (1) after receiving the dump data signal; The data processing and waveform mapping module monitors the characteristic signal memory, and reads the characteristic signal X′(l) from the characteristic signal memory every time its data is updated, performs data processing and real-time waveform mapping, and uses three-dimensional waveform in waveform mapping Mapping; when the display period arrives, the data processing and waveform mapping module sends the mapped waveform to the display; The display is used to display the mapped waveform sent by the data processing and waveform mapping module. 2. transient signal seamless measurement system according to claim 1, is characterized in that, described characteristic signal coarse screening module adopts peak sampling mode, and its expression is: Among them, k represents the sampling interval, and L represents the length of the sampling signal X(l). 3. The transient signal seamless measurement system according to claim 1, characterized in that, the threshold value setting module adopts average value sampling for the sampling of the sampling signal X(1). 4. The transient signal seamless measurement system according to claim 1, wherein the data processing and waveform mapping module adopts a multi-stage pipeline processing mechanism, and for each data to be processed, each pipeline processing line level to complete a processing task.