CN106324313A - Approximate entropy-based transient signal seamless measurement system - Google Patents

Abstract

The invention discloses an approximate entropy-based transient signal seamless measurement system. Sampled signal are obtained after analog signals to be measured are sampled via an ADC module, the sampled signals are subjected to coarse screening operation via a sampling mode through a characteristic signal coarse screening module, post-coarse-sizing signals are stored in a screened signal storage device, a characteristic signal secondary screening comparison threshold value can be obtained via calculation conducted via a threshold value setting module based on an approximation entropy algorithm according to the sampled signals, approximation entropy of the screened signals is calculated via a characteristic signal secondary screening module, and whether the screened signals are characteristic signals is determined according to the characteristic signal secondary screening comparison threshold value; if the screened signals are characteristic signals, the sampled signals of a detailed signal storage device are stored in a characteristic signal storage device as characteristic signals, a data processing and waveform mapping module is used for reading the characteristic signals and subjecting the characteristic signals to processing and mapping operation, and a mapping waveform is displayed via a display device when a display cycle comes. According to the approximate entropy-based transient signal seamless measurement system, signal complexity is described in a quantitative manner via an approximation entropy value, and transient non-stationary signal time domain information can be measured seamlessly.

Description

Measurement system that transient signal based on approximate entropy is seamless

Technical field

The invention belongs to transient signal field of measuring technique, more specifically, relate to a kind of transient state based on approximate entropy Measurement system that signal is seamless.

Background technology

Hyundai electronics signal is increasingly sophisticated various, and the frequency range of signal is constantly widened, the increasing instantaneous, non-stationary of signal Long extremely rapid.The most effectively extract the information entrained by signal, it is achieved the highly effective gathering of transient state non-stationary signal divides with real-time Analysis brings challenge to modern signal measurement.On the one hand, we pursue the collection of more speed and higher precision, with the most Signal acquisition details；On the other hand, the mass data that high sampling rate and high-resolution acquisition obtain, give again the real-time place of signal Reason and analysis bring difficulty, affect system response.

Employing real-time sampling, the digital oscilloscope of real-time processing technique are the representatives of modern time domain measurement instrument, for number The oscillographic seamless measurement capability research of word becomes focus in recent years.The measurement process that digital oscilloscope is the most complete comprises signal and adopts The links such as the process of collection, data and waveform show.Its signals collecting is considered as a kind of intermittent sampling, between adjacent twice collection between Every processing and the time of display, referred to as Dead Time or measurement gap.All generations signal in measuring gap all can not Effectively gathered.Visible, the size measuring gap directly affects oscillographic signal measurement ability.Measure gap the least, have It is the highest that effect sampling accounts for the ratio of overall measurement time, and oscillograph is the biggest to the capture probability of transient signal.Therefore, measurement seam is reduced Gap is until realizing seamless measurement, extremely important for modern signal is measured.

Analyze the data handling procedure of digital oscilloscope, contain waveform image process and some common digital signals divide Analysis, such as interpolation, average, mathematical operation, FFT, digital filtering etc..And show process, then it is that waveform image and correlation analysis are tied Fruit presents over the display.On the premise of sample rate determines, measurement gap to be reduced it is necessary to shorten flower and processing and in display Time.And thoroughly to eliminate measurement gap, it is achieved and seamless measurement, then must complete to process and display while sampling, also I other words the speed of process and display must catch up with the speed of sampling completely, difficulty is well imagined.First, existing CPU, DSP etc. Huge spread is there is between arithmetic speed (1～2GHz) and the sample rate (up to 1～100GSPS) of ADC of processing apparatus, and And the signal analysis kind that comprises is the most, process the time the longest；Secondly, by display mechanism and the refresh rate of liquid crystal display (usual 50Hz) limits, and waveform display process is the hugest.Therefore, in the existing architecture of oscillograph and device level Under, the most seamless measurement to be realized hardly possible.

In industry, seamless measurement is had been achieved for certain achievement in research at present.Document 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) uses the architectural framework of parallel processing to carry out data process, uses the mode of multiple waveform statistical stacking Show, within the unit interval, have compressed the process time, add waveform display number, thus reduce survey to a certain extent Amount gap, but still it is unable to reach seamless measurement.On this basis, document K.J.Yang, S.L.Tian, H.Zeng, L.Qiu and L.P.Guo,A seamless acquisition digital storage oscilloscope with three- dimensional waveform display,Review of Scientific Instruments,85,045102(2014) The deep memorizer of DDR is used to build fragmented storage and ping-pong operation, it is achieved that seamless in certain time, under the conditions of certain data volume Gather oscillograph.But, above-mentioned seamless survey quantifier elimination remains in some certain applications at present for slow speed signal Measure, and generally exist gather with display disconnection (can not show during collection, can not gather during display), poor real, cannot Continue seamless problem.

It is true that to realize in a long time the complete seamless measurement of measured signal was both difficult to, also it is that do not have must Want.Because the information that different application field is paid close attention to is entirely different, and obtains excessive quantity of information simultaneously or show too much Signal waveform cannot be all people all effect accept.Famous scientist Shannon once pointed out: any information all exists superfluous Remaining, probability of occurrence or the uncertainty of redundancy size symbol each with information are relevant.And comentropy to characterize signal just the most true Determine the dimension of degree.From the angle of comentropy it is seen that, unstable signal is always superimposed with some abnormal components, makes The comentropy obtaining signal is different, and this provides the foundation for feature signal extraction based on comentropy, and to finally realizing wink The seamless measurement of state signal is the most important.Therefore, if can be in signal acquisition process, control in real time based on comentropy, carry The feature contained in the number of winning the confidence, retains crucial or useful information, abandons redundancy or garbage, not only can greatly reduce process and The burden of display, it is achieved the seamless measurement of characteristic signal, and make the acquisition information can be really used by people.But, information Entropy (Shannon entropy) is as the entropy of physical world, owing to its chaos phenomenon causes calculating process extremely complex, it is difficult to meet nothing The requirement of real-time of seam measurement system.

Summary of the invention

It is an object of the invention to overcome the deficiencies in the prior art, it is provided that a kind of seamless survey of transient signal based on approximate entropy Amount system, to approximate the complexity of entropy quantitative description signal, and instructs collection and the process of signal, it is achieved wink based on this The seamless measurement of state non-stationary signal time-domain information.

For achieving the above object, the seamless measurement system of present invention transient signal based on approximate entropy include ADC, Characteristic signal coarse sizing module, threshold setting module, following characteristics memorizer, screening signal storage, characteristic signal secondary sieve Modeling block, data process and waveform mapping block and display, wherein:

Analogue signal x (t) to be measured is sampled by ADC, obtains sampled signal X (l), is sent respectively to characteristic signal Coarse sizing module, threshold setting module and following characteristics memorizer；

Sampled signal X (l) is used sample mode to carry out eigenvalue coarse sizing and is screened by characteristic signal coarse sizing module Signal Y (n), stores screening signal Y (n) to screening signal storage；

Threshold setting module compares threshold value G according to sampled signal X (l) calculating characteristic signal postsearch screening and is sent to feature letter Number postsearch screening module, its computational methods are: be sampled sampled signal X (l), sampling rate and characteristic signal coarse sizing module Identical, obtain sampled signal Y ' (n), use approximate entropy algorithm to calculate the approximate entropy ApEn' of Y ' (n), then calculate characteristic signal Threshold value G=P × ApEn' is compared in postsearch screening, and P represents default proportionality coefficient；

Following characteristics memorizer is used for storing sampled signal X (l)；

Screening signal storage is used for storing screening signal Y (n)；

Characteristic signal postsearch screening module reads screening signal Y (n) from screening signal storage, uses approximate entropy algorithm meter Calculate the approximate entropy ApEn of Y (n), if ApEn is ＞ G, sends unloading data signal to following characteristics memorizer, the most do not make any Operation；

Characteristic signal memorizer, after receiving unloading data signal, reads sampled signal X in following characteristics memorizer L () also stores as characteristic signal X ' (l)；

Data process and characteristic signal memorizer is monitored by waveform mapping block, after its data update, from spy Levy and signal storage reads characteristic signal X ' (l), carry out data and process and real-time waveform mapping, use three when waveform maps Dimension waveform maps；When the display cycle arrives, data process and mapping waveform is sent to display by waveform mapping block；

The mapping waveform that display processes for video data and waveform mapping block sends.

Measurement system that present invention transient signal based on approximate entropy is seamless, analogue signal to be measured sampling is obtained by ADC Sampled signal stores to following characteristics memorizer, and characteristic signal coarse sizing module uses sample mode that sampled signal is carried out scalping Selecting and store to screening signal storage, threshold setting module is calculated feature letter based on approximate entropy algorithm according to sampled signal Threshold value is compared in number postsearch screening, and the approximate entropy of characteristic signal postsearch screening module calculating sifting signal, according to characteristic signal secondary Screening compare threshold value to screening signal judge, if characteristic signal then using the sampled signal of following characteristics memorizer as Characteristic signal stores to characteristic signal memorizer, and data process and waveform mapping block reads characteristic signal and processes and reflect Penetrate, shown by display mapping ejected wave shape when the display cycle arrives.

The present invention is to approximate the complexity (i.e. quantity of information) of entropy quantitative description sampled signal and based on approximate entropy real-time Control, adaptive capture characteristic signal, extract crucial or useful information, abandon redundancy or garbage, thus decrease number According to the burden processed and waveform shows, it is achieved that the seamless measurement of transient signal time-domain information.

Accompanying drawing explanation

Fig. 1 is approximate entropy algorithm flow chart；

Fig. 2 is the detailed description of the invention structure chart of the present invention seamless measurement system of transient signal based on approximate entropy；

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

Fig. 4 is threshold value setting procedure figure；

Fig. 5 is average value sample schematic diagram；

The data of Fig. 6 multi-stage pipeline treatment mechanism process and waveform mapping block process chart；

Fig. 7 is the sampled signal waveform figure of first group of analogue signal to be measured；

Fig. 8 is the screening signal waveforms of first group of analogue signal to be measured；

Fig. 9 is the sampled signal waveform figure of second group of analogue signal to be measured；

Figure 10 is the screening signal waveforms of second group of analogue signal to be measured；

Figure 11 is the sampled signal waveform figure of the 3rd group of analogue signal to be measured；

Figure 12 is the screening signal waveforms of the 3rd group of analogue signal to be measured；

Figure 13 is the sampled signal waveform figure of the 4th group of analogue signal to be measured；

Figure 14 is the screening signal waveforms of the 4th group of analogue signal to be measured；

Figure 15 is the display result of first group of analogue signal to be measured；

Figure 16 is the display result of second group of analogue signal to be measured；

Figure 17 is the display result of the 3rd group of analogue signal to be measured；

Figure 18 is the display result of the 4th group of analogue signal to be measured.

Detailed description of the invention

Below in conjunction with the accompanying drawings the detailed description of the invention of the present invention is described, in order to those skilled in the art is preferably Understand the present invention.Requiring particular attention is that, in the following description, when known function and design detailed description perhaps When can desalinate the main contents of the present invention, these are described in and will be left in the basket here.

In order to the technology contents of the present invention is better described, first pairing approximation entropy is simply introduced.

The concept of approximate entropy is that Steven M.Pincus proposed from the angle of measure time sequence complexity in 1991 , in metric signal, produce the probability of new model.The probability that signal produces new model is the biggest, shows the complexity of sequence more Greatly, corresponding entropy is the biggest.I.e. approximate entropy algorithm carrys out complexity and the scrambling of quantificational expression signal by a nonnegative number. Fig. 1 is approximate entropy algorithm flow chart.As it is shown in figure 1, set primary signal time series as x (n)=x (1), x (2) ..., x (N), N number of sampling point, specifically comprising the following steps that of approximate entropy algorithm altogether

S101: signal x (n) is formed one group of m dimensional vector by sequence number consecutive order:

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

Definition vector X_iAnd X_jBetween distance d_ijFor in the two corresponding element one of difference maximum, i.e.

$d_{ij}=\underset{k=0~m-1}{max}\[|x(i+k)-x(j+k)|\]---\left(2\right)$

And to each i value, calculate vector X_iWith its complement vector X_jDistance between (j=1～N-m+1, j ≠ i).

S102: set threshold value r (r ＞ 0), to each i value, adds up d_ijThe number n of ＜ r_ij(r), and by n_ij(r) with away from Ratio from sum N-m+1 is denoted asThat is:

$C_i^m\left(r\right)=\frac{1}{N-m+1}{n}_{ij}\left(r\right)---\left(3\right)$

S103: willTake the logarithm, and seek its meansigma methods to all i, be denoted as φ^m(r), i.e.

${\mathrm{\φ}}^m\left(r\right)=\frac{1}{N-m+1}\underset{i=1}{\overset{n-m+1}{\Σ}}\ln C_i^m\left(r\right)---\left(4\right)$

S104: add 1 by dimension, builds signal x (n) and obtains m+1 dimensional vector, and calculates vector X_iWith its complement vector X_jIt Between distance d_ij。

S105: similarly, adds up d_ijThe number n of ＜ r_ijR (), calculates

S106: calculate φ^m+1(r):

${\mathrm{\φ}}^{m+1}\left(r\right)=\frac{1}{N-m}\underset{i=1}{\overset{n-m}{\Σ}}\ln C_i^{m+1}\left(r\right)---\left(5\right)$

S107: in computational theory, the approximate entropy of this signal x (n) is:

$ApEn(m,r)=\underset{N\→\mathrm{\∞}}{\lim }\[{\mathrm{\φ}}^m\left(r\right)-{\mathrm{\φ}}^{m+1}\left(r\right)\]---\left(6\right)$

It is said that in general, this ultimate value exists with probability 1.During real work, N can not be ∞.When N is finite value, by upper State that step draws is sequence length estimated value of approximate entropy ApEn when being N, is denoted as

ApEn (m, r, N)=φ^m(r)-φ^m+1(r) (7)

Obvious and m, r the value of the value of ApEn is relevant.Pincus is according to practice, it is proposed that take m=2, r=0.1～0.2SD (SD is initial data x (i), the standard deviation (standard deviation) of i=1～n).Current industry is chosen for dimension m The most also technical staff is had to have been discussed in detail.The value of therefore m, r can be configured according to actual needs.

By above calculation procedure it is seen that, the physical essence of approximate entropy weighs the signal sequence when dimension changes exactly The logarithm conditional probability average that middle new model occurs.Approximate entropy is characterizing scrambling and the complexity of signal sequence the most in theory Property aspect has meaning.

Analysis and comprehensive pertinent literature according to approximate entropy physical essence are discussed, and can obtain approximate entropy algorithm and be suitable for The main feature of electronic surveying field signal analysis:

1) owing to being retrained by tolerance threshold r, approximate entropy algorithm has preferable anti-noise ability, for electronic surveying, Measured signal often contains high-frequency noise interference, therefore the anti-noise ability of feature extracting method is very important.

2) approximate entropy algorithm is unrelated with the amplitude of signal sequence, the most relevant with sequence complexity, when faced by electronic surveying During small-signal, the approximate entropy feature unrelated with amplitude is critically important.

3) approximate entropy algorithm is as a nonlinear kinetics parameter, is all suitable for stochastic process and deterministic process, and Signal faced by electronic surveying not only comprises definitiveness composition but also comprise the sophisticated signal of random element often, therefore this feature The most critically important.

4) approximate entropy algorithm only needs relatively short data just can draw the most sane estimated value, and this feature makes it when shorter In extract the characteristic information lying in signal sequence, therefore be applicable to the electronic surveying field that requirement of real time is high.

5) analytical effect of approximate entropy algorithm is good compared with statistical parameters such as average, variance, standard deviations, makes it more accurately to have Effect ground extracts characteristic signal.

To sum up, approximate entropy algorithm is applicable to the signal analysis in electronic surveying field, and it is provided that a kind of quantifiable extraction The new method of characteristic signal, the present invention instructs signals collecting and data to process based on this, it is proposed that wink based on approximate entropy Measurement system that state signal is seamless, it is achieved the seamless measurement of transient signal.

Fig. 2 is the detailed description of the invention structure chart of the present invention seamless measurement system of transient signal based on approximate entropy.Such as Fig. 2 Shown in, present invention transient signal based on approximate entropy is seamless, and measurement system includes ADC (Analog-to-Digital Converter, analog-digital converter) module 1, characteristic signal coarse sizing module 2, threshold setting module 3, following characteristics memorizer 4, Screening signal storage 5, characteristic signal postsearch screening module 6, characteristic signal memorizer 7, data process and waveform mapping block 8 With display 9.

Analogue signal x (t) to be measured is sampled by ADC 1, obtains sampled signal X (l), is sent respectively to feature letter Number coarse sizing module 2, threshold setting module 3 and following characteristics memorizer 4.

Sampled signal X (l) is used sample mode to carry out eigenvalue coarse sizing and is screened by characteristic signal coarse sizing module 2 Signal Y (n), stores screening signal Y (n) to screening signal storage 5.

In the present invention, it is based on to the calculating of signal approximate entropy with compare and realize seamless measurement, and approximate entropy is calculated The operand of method is doubled and redoubled along with the increase of signal sequence length (data volume).For long data, directly calculate approximate entropy Can be the most time-consuming.And prove after deliberation, approximate entropy algorithm is actual only needs relatively short data (100～1000) just can draw the most steadily and surely Estimated value.Therefore, before calculating approximate entropy, need long data is done eigenvalue sample process, i.e. carry out characteristic signal Coarse sizing.

Specific in digital oscilloscope, the data volume (sampling point number) once gathered is determined by its storage depth.At present, root According to the difference of performance, oscillograph generally has the variable storage depth in the range of 1kpts～1Gpts, i.e. single acquisition data volume 10³～10⁹.Therefore, for different storage depths, different sampling interval time can be chosen, sampled signal be used and takes out Sample loading mode carries out characteristic signal coarse sizing, and the data volume of signal after coarse sizing is controlled 100～1000, after can greatly reducing The calculating time of approximate entropy during continuous characteristic signal postsearch screening.

The design peak detection function based on digital oscilloscope of characteristic signal coarse sizing module 2, its mesh in the present embodiment Be from the mass data of ADC sampled signal X (l) filter out the certain intervals time (peak-to-peak in sampling interval k) in real time Value Data, forms screening signal Y (n).Contain maximum and two data of minima owing to often organizing peak-to-peak value packet, the most also need It is ranked up according to the feature of signal own, determines the sequencing of maximum and minima.After peak value is sampled, the number of Y (n) The data length N=2L/k, L that are reduced to the 2/k of X (l), i.e. Y (n) according to amount represent the length of sampled signal X (l).Fig. 3 is this reality Execute the schematic diagram of peak value sampling in example.Equation below can be used to express peak value sampling according to Fig. 3:

$Y\left(n\right)=\left\{\begin{array}{c}\underset{i=1~k}{\min }\[X(sk+i)\],n=1,3,5,\mathrm{...},2L/k-1,X\[sk+1\]\≤X\[(s+1)k\]\\ \underset{i=1~k}{\max }\[X(sk+i)\],n=1,3,5,\mathrm{...},2L/k-1,X\[sk+1\]>X\[(s+1)k\]\\ \underset{i=1~k}{\min }\[X(sk+i)\],n=2,4,6,\mathrm{...},2L/k,X\[sk+1\]>X\[(s+1)k\]\\ \underset{i=1~k}{\max }\[X(sk+i)\],n=2,4,6,\mathrm{...},2L/k,X\[sk+1\]\≤X\[(s+1)k\]\end{array}\right.,s=0~L/k-1---\left(8\right)$

Threshold setting module 3 calculates characteristic signal postsearch screening according to sampled signal X (l) and compares threshold value G.Fig. 4 is threshold value Setting procedure figure.As shown in Figure 4, threshold value is arranged method particularly includes:

After entrance threshold setting module 3 receives sampled signal X (l), first sampled signal X (l) is sampled, sampling Rate is identical with characteristic signal coarse sizing module, obtains sampled signal Y ' (n), it is clear that sampled signal Y ' (n) and screening signal Y (n) Length identical.In the present embodiment, the design high-resolution sampling functions based on digital oscilloscope of threshold setting module 3, so-called High-resolution is sampled, and its essence is to ask for the meansigma methods of sampled signal in the certain intervals time, i.e. carries out sampled signal X (l) Average value sample, owing in the present embodiment, characteristic signal coarse sizing module 2 uses peak-to-peak value to sample, takes out in each sampling interval k Taking two peak values, therefore sampled signal X (l) is averaged the sampling interval k'=k/2 of value sampling, then available with screening signal Average value sample signal Y ' (n) that Y (n) data volume is suitable.Fig. 5 is average value sample schematic diagram.Can be by meansigma methods according to Fig. 5 The principle of sampling is expressed with equation below:

$Y^{\′}\left(n\right)=\frac{1}{k^{\′}}\underset{i=1}{\overset{k^{\′}}{\Σ}}X\[(n-1)\×k^{\′}+i\]---\left(9\right)$

Then approximate entropy algorithm is used to calculate the approximate entropy ApEn' of Y ' (n), and by the ratio system of ApEn' Yu user preset Number P are associated, and can be calculated comparison threshold value G of characteristic signal postsearch screening by following formula:

G=P × ApEn'(10)

Wherein, the span of P is P ＞ 1.

In view of the seamless measurement high request to real-time, the present embodiment selects build in FPGA and add streamline parallel The computing framework processed realizes the quick calculating of approximate entropy.The present embodiment selects arrange threshold value, root by standard deviation SD According to flow process shown in Fig. 1, then the difficult point using FPGA to realize approximate entropy algorithm is to calculate substantial amounts of distance d_ij, standard deviation SD with And the complex calculation such as logarithm, square root.For d_ijWith the calculating of SD, it is to pass through owing to the present invention participating in the initial data of calculating The time series that continuous sampling and screening obtain, realizes it is therefore possible to use add pipelining parallel, i.e. saves at each clock Clap, the new sampling point of parallel computation and distance d of had been friends in the past sampling point_ijAnd new sampling point and the sum of had been friends in the past sampling point；And for logarithm, flat The complex calculation such as root, the cordic algorithm IP kernel can being internally integrated by calling FPGA completes.

Following characteristics memorizer 4 is used for storing sampled signal X (l).

Screening signal storage 5 is used for storing screening signal Y (n).

Characteristic signal postsearch screening module 6 reads screening signal Y (n) from screening signal storage 5, calculates approximate entropy The comparative result of ApEn, ApEn and threshold value G is as the control signal of characteristic signal memorizer 7, if ApEn is ＞ G, then judges sieve Selecting signal Y (n) to be characterized signal general picture, sampled signal X (l) is characterized signal detail, sends unloading to following characteristics memorizer Data signal, otherwise judges that screening signal Y (n) is non-characteristic signal details as non-characteristic signal general picture, sampled signal X (l), no Make any operation.

Characteristic signal memorizer 7, after receiving unloading data signal, reads the sampled signal in following characteristics memorizer 4 X (l) also stores as characteristic signal X ' (l).

Data process and characteristic signal memorizer 7 is being monitored by waveform mapping block 8, after its data update, From characteristic signal memorizer 7, read characteristic signal X ' (l), carry out various data and process and real-time waveform mapping, reflect at waveform Three-dimensional waveform is used to map when penetrating, the most multiple waveform statistical stackings.When the display cycle arrives, data process and waveform maps mould Mapping waveform is sent to display 9 by block 8.

The mapping waveform that display 9 processes for video data and waveform mapping block 8 sends.

In the present invention, process due to data and waveform mapping block 8 has needed the various numbers to characteristic signal X ' (l) According to processing and by the real-time mapping of sampled data to waveform image, though the characteristic signal in the unit interval after postsearch screening Total quantity (waveform number) relatively primitive signal X (l) of X ' (l) is greatly reduced, but this module still carries heavy number According to the task of process, it the most also it is the major source producing and measuring gap.Measure gap therefore to reduce as much as possible, preferably set The data having counted multi-stage pipeline treatment mechanism process and waveform mapping block.

The data of Fig. 6 multi-stage pipeline treatment mechanism process and waveform mapping block process chart.As shown in Figure 6, pin To each pending data, the every one-level in stream treatment line complete a process task (as interpolation, average, mathematical operation, FFT, digital filtering, waveform mapping etc.).If there being Q process task, it is necessary to Q level stream treatment line.

Waveform mapping settings is at the afterbody (Q level) of stream treatment line, and its main task is by front Q-1 level Characteristic signal after reason is mapped as waveform image, and realizes showing the statistical stacking of several waveform images.The statistics of waveform Additive process is based on the continuous renewal to the waveform database built.For each pixel in waveform image, waveform number According to having in storehouse, independent memory element is the most corresponding.Certain data in characteristic signal relate to this unit, in unit Portion's enumerator just adds 1, is not involved with, and is not added with (counter initial value is 0).Finally, when the brush screen cycle arrives, whole ripple Graphic data library storage and send aobvious be the refresh cycle in the statistical stacking result of all oscillogram elephants, thus saved oscillogram As the time that brush screen is consumed successively.

Map owing to the present invention uses three-dimensional waveform, then mathematically, waveform database is considered as one two Dimension matrix A, matrix element a_ijSize represent that identical sampled point (time is the most identical with amplitude) in S waveform occurs time Number, i.e. hit-count

$A=\left(\begin{array}{c}{a}_{11},a_{12},...,a_{1L}\\ a_{21},a_{22},...,a_{2L}\\ ..................\\ a_{q1},a_{q2},...,a_{qL}\end{array}\right)---\left(11\right)$

Each column vector of matrix A then reflects the distribution situation of this instance sample point value, and1≤j≤ The characteristic signal quantity that L, S are handled in representing the refresh cycle.

It is i.e. that each data according to characteristic signal X ' (l) are to each element in matrix A that so-called waveform mapping processes It is marked.Through S time map, matrix A in just containing this time period each sampled point value in the value frequency in each moment. Display is according to matrix element a_ijValue carry out different brightness and show.

In order to effectiveness of the invention is described, next the condition using the present invention to realize seamless measurement needs satisfied is entered Row theoretical derivation.

According to above-mentioned flow process, affecting the seamless principal element realized of measuring has: the sample rate f of analog-digital converter_s, sampling letter Number length L of X (l), threshold value arrange time t_gate, twice screening time t_screen, data process and single waveform mapping time t_processAnd display brush screen time t_lcdDeng.Wherein, can synchronize during signal sampling due to characteristic signal coarse sizing Carry out in real time, be not required to take additional time, therefore twice screening time t_screenIt is approximately equal to approximate entropy during programmed screening Calculating time t_ApEn, i.e.

t_screen≈t_ApEn (12)

In like manner, threshold value arranges time t_gateIt is also approximately equal to the calculating time t of approximate entropy_ApEn, therefore have

t_screen=t_gate≈t_ApEn (13)

It is equal that both calculate the time, can try to achieve result by parallel processing simultaneously.Additionally, due to the present embodiment have employed The display mechanism of multiple waveform statistical stackings, at the brush screen time t of display_lcdIn, new waveform can be carried out simultaneously and map, and Will not lose time.Therefore, the total processing time T of signal in the unit interval_processIt is solely dependent upon f_s、L、t_ApEn、t_processAnd The quantity D of characteristic signal '.

As it was noted above, to realize seamless measurement, it is necessary to make total processing time T_processLess than or equal to when always gathering Between T_acq, i.e.

T_process≤T_acq (14)

Within the unit interval, make T_acq=1s, then must be fulfilled for

T_process=t_ApEn×D+t_process×D'≤1 (15)

Wherein, D is the quantity (waveform quantity) of unit time (1s) interior sampled signal x (l), can be by sample rate f_sAnd length L obtains, i.e.

D=f_s/L (16)

And D' is characteristic signal X'(l after postsearch screening) quantity, bring formula (16) into formula (15), can obtain:

$D^{\′}\≤\frac{1-t_{ApEn}\×f_s/L}{t_{process}}---\left(17\right)$

Can be obtained by formula (16), (17) again

$Z=\frac{D^{\′}}{D}\×100\%\≤\frac{L/f_s-t_{ApEn}}{t_{process}}\×100\%---\left(18\right)$

Formula (17) and (18) are the essential condition realizing the seamless measurement of transient signal, i.e. within the unit interval (1s), when Characteristic signal X'(l) quantity D ' (absolute magnitude) meet formula (17), or characteristic signal X'(l) quantity D ' and sampled signal x (l) The percentage ratio Z (relative quantity) of quantity D when meeting formula (18), so that it may realize the seamless measurement of transient signal.And in the present invention, The judgement of characteristic signal is realized by threshold value G, therefore can carry out the quantity of controlling feature signal by controlling G, thus Reach seamless measurement purpose.

Embodiment

In order to verify the effectiveness that the transient signal of different complexities is screened by the present invention, ADI company is utilized to provide 8Bit ADC model (Ideal_8_Bit.adc) build oscillographic acquisition system.If oscillographic sample rate f_s=1GSa/ (data volume of the most each sampled signal X (l) is 10 for s, storage depth L=1Mpts⁶), internal system time clock f_c=250MHz.Take N =200 data participate in characteristic signal postsearch screening, therefore peak value sampling interval k=2L/N=2 × 10 of coarse sizing module⁶/ 200=10000, the average value sample interval k'=k/2=5000 of threshold setting module.Approximate entropy algorithm takes m=2, r= 0.2SD.Emulate by four groups of sample datas individually below.

Remember first group of analogue signal x to be measured₁(t)=sin (2 π f₀T), its frequency f₀=1kHz.Fig. 7 is first group of mould to be measured Intend the sampled signal waveform figure of signal.Fig. 8 is the screening signal waveforms of first group of analogue signal to be measured.It is computed, now sieves Select signal (peak value sampled signal) Y₁The approximate entropy ApEn of (n)₁=0.0846, and sampled signal (average value sample signal) Y₁′ The approximate entropy ApEn of (n)₁'=0.0858.

Remember second group of analogue signal x to be measured₂(t)=sin (2 π f₀T), its frequency f₀=1kHz, but in order to simulate accidental making an uproar Sound interference, AD quantify the transient phenomenons such as mistake, add distortion sample (burr signal) at random in preferable ADC sampling model. Fig. 9 is the sampled signal waveform figure of second group of analogue signal to be measured.Figure 10 is the screening signal wave of second group of analogue signal to be measured Shape figure.It is computed, now screens signal (peak value sampled signal) Y₂The approximate entropy ApEn of (n)₂=0.1040, and sampled signal is (flat Average sampled signal) Y₂The approximate entropy ApEn of ' (n)₂'=0.0858.

Remember the 3rd group of analogue signal x to be measured₃(t)=0.25 × [sin (2 π f₀t)+sin(6πf₀t)+sin(10πf₀t)+ sin(14πf₀T)], its frequency f₀=1kHz.Visible, in order to simulate harmonic distortion, in frequency f₀In the sinusoidal signal of=1kHz, Superposition 3 times, 5 times and 7 subharmonic.Figure 11 is the sampled signal waveform figure of the 3rd group of analogue signal to be measured.Figure 12 is the 3rd group The screening signal waveforms of analogue signal to be measured.It is computed, now screens signal (peak value sampled signal) Y₃The approximate entropy of (n) ApEn₃=0.4537, and sampled signal (average value sample signal) Y₃The approximate entropy ApEn of ' (n)₃'=0.3706.

Remember the 4th group of analogue signal x to be measured₄(t)=0.5 × sin (2 π f₀T)+Noise, its frequency f₀=1kHz.It is visible, In order to simulate critical noisy interference, in frequency f₀In the sinusoidal signal of=1kHz, superposition average is 0, variance be 1 the whitest Noise Noise.Figure 13 is the sampled signal waveform figure of the 4th group of analogue signal to be measured.Figure 14 is the 4th group of analogue signal to be measured Screening signal waveforms.It is computed, now screens signal (peak value sampled signal) Y₄The approximate entropy ApEn of (n)₄=0.6901, and Sampled signal (average value sample signal) Y₄The approximate entropy ApEn of ' (n)₄'=0.0861.

Definition R is the ratio of the approximate entropy ApEn and the approximate entropy ApEn' of sampled signal Y ' (n) of screening signal Y (n):

R=ApEn/ApEn'(19)

Table 1 is the approximate entropy result of calculation of four groups of sample datas.

Sequence number	ApEn	ApEn'	R
				First group	ApEn1=0.0846	ApEn1'=0.0858	R1=0.9861
Second group	ApEn2=0.1040	ApEn2'=0.0858	R2=1.2115
				3rd group	ApEn3=0.4537	ApEn3'=0.3706	R3=1.2243
4th group	ApEn4=0.6901	ApEn4'=0.0861	R4=8.0151

Table 1

From table 1, along with the increase of sampled signal X (l) complexity, the approximate entropy ApEn of screening signal Y (n) is gradually Increase；Meanwhile, the ratio R of the approximate entropy ApEn and the approximate entropy ApEn' of sampled signal Y ' (n) of screening signal Y (n) the most gradually increases Greatly.This result further demonstrates that

1. approximate entropy is the quantizating index of signal complexity, and signal sequence is the most complicated, and approximate entropy is the biggest；

2. the mode using peak value sampling carries out coarse sizing to signal, can be with the complexity of stick signal and characteristic information；

3. the ratio of the approximate entropy of screening signal and sampled signal, can be used for instructing the postsearch screening of characteristic signal.

Therefore, for the characteristic signal postsearch screening of above-mentioned four groups of samples, user can Set scale FACTOR P=1.2.Table 2 It is comparison threshold value G and the postsearch screening result of characteristic signal postsearch screening.

Sequence number	ApEn	G	Comparative result	The selection result
					First group	ApEn1=0.0846	G1=0.1030	ApEn1＜ G1	Non-characteristic signal
Second group	ApEn2=0.1040	G2=0.1030	ApEn2＞ G2	Characteristic signal
					3rd group	ApEn3=0.4537	G3=0.4447	ApEn3＞ G3	Characteristic signal
4th group	ApEn4=0.6901	G4=0.1033	ApEn4＞ G4	Characteristic signal

Table 2

From table 2, as the proportionality coefficient P=1.2 of user setup, can to comprise noise jamming, AD quantify mistake and The transient signal of harmonic distortion effectively screens.Further, user can be by arranging different proportionality coefficient P, it is achieved right The Effective selection of the transient signal of different complexities.P value is the biggest, and screening conditions are the strictest, characteristic signal after postsearch screening The quantity D of X ' (l) ' the fewest, D' is the least with the percentage ratio Z of the quantity D of sampled signal X (l), and system is closer to seamless.

Using and use the present invention seamless measurement system building digital oscilloscope of transient signal based on approximate entropy, it is adopted in real time Sample rate f_s=1GSa/s, storage depth L=1Mpts, internal system time clock frequency f_c=250MHz, the calculating of approximate entropy about needs 200 Individual system clock cycle, i.e.

t_ApEn≈200/f_c=8 × 10^-7s (20)

L data carry out data and process and waveform mapping time

t_process≈l/f_c=4 × 10^-3s (21)

Therefore, according to formula (17) and (18), within the unit interval (1s), when characteristic signal X ' (l) quantity D '≤249 or The quantity D of characteristic signal X ' (l) ' with percentage ratio Z≤24.98% of the quantity D of sampled signal X (l) time, transient signal can be realized Seamless measurement.

During actual test, produce above-mentioned x by Tyke AWG (Arbitrary Waveform Generator) AWG5014B₁(t)、x₂(t)、x₃(t) and x₄ T () four groups of tested analogue signals are input in oscillograph, Set scale FACTOR P=1.2.Figure 15 is first group of analogue signal to be measured Display result.Figure 16 is the display result of second group of analogue signal to be measured.Figure 17 is the display of the 3rd group of analogue signal to be measured Result.Figure 18 is the display result of the 4th group of analogue signal to be measured.In Figure 15, due to sampled signals X all in the brush screen cycle₁ L () is all judged as non-characteristic signal and abandons, therefore oscilloscope display is without waveform, and system such as is in all the time at the state to be triggered； In Figure 16 and Figure 17, all sampled signals X comprising burr and each harmonic₂(l) and X₃(l) be all judged as characteristic signal and Retain, and oscillograph has carried out waveform to it and mapped and display；In Figure 18, all superpositions sampled signal X of white noise₄(l) All it is judged as characteristic signal and retains, but the oscillographic triggering system of serious influence of noise, cause edging trigger mistake, Therefore waveform shows and occurs in that and rock and the phenomenon on double edges.The aforementioned four analogue signal actual test result in oscillograph, enters One step demonstrates the present invention effectiveness to the transient signal screening of different complexities.

Although detailed description of the invention illustrative to the present invention is described above, in order to the technology of the art Personnel understand the present invention, the common skill it should be apparent that the invention is not restricted to the scope of detailed description of the invention, to the art From the point of view of art personnel, as long as various change limits and in the spirit and scope of the present invention that determine in appended claim, these Change is apparent from, and all utilize the innovation and creation of present inventive concept all at the row of protection.

Claims (4)

1. the seamless measurement system of transient signal based on approximate entropy, it is characterised in that include ADC, characteristic signal scalping At modeling block, threshold setting module, following characteristics memorizer, screening signal storage, characteristic signal postsearch screening module, data Reason and waveform mapping block and display, wherein:

Analogue signal x (t) to be measured is sampled by ADC, obtains sampled signal X (l), is sent respectively to characteristic signal scalping Modeling block, threshold setting module and following characteristics memorizer；

Characteristic signal coarse sizing module carries out eigenvalue coarse sizing to sampled signal X (l) employing sample mode and obtains screening signal Y N (), stores screening signal Y (n) to screening signal storage；

Threshold setting module compares threshold value G according to sampled signal X (l) calculating characteristic signal postsearch screening and is sent to characteristic signal two Secondary screening module, its computational methods are: be sampled sampled signal X (l), sampling rate and characteristic signal coarse sizing module phase With, obtain sampled signal Y ' (n), use approximate entropy algorithm to calculate the approximate entropy ApEn' of Y ' (n), then calculate characteristic signal two Threshold value G=P × ApEn' is compared in secondary screening, and P represents default proportionality coefficient；

Following characteristics memorizer is used for storing sampled signal X (l)；

Screening signal storage is used for storing screening signal Y (n)；

Characteristic signal postsearch screening module reads screening signal Y (n) from screening signal storage, uses approximate entropy algorithm to calculate Y N the approximate entropy ApEn of (), if ApEn is ＞ G, sends unloading data signal to following characteristics memorizer, does not the most make any behaviour Make；

Characteristic signal memorizer after receiving unloading data signal, read sampled signal X (l) in following characteristics memorizer and Store as characteristic signal X ' (l)；

Data process and following characteristics memorizer is monitored by waveform mapping block, after its data update, believe from feature Number memorizer reads characteristic signal X ' (l), carries out data and process and real-time waveform maps, the employing three-dimensional wave when waveform maps Shape maps；When the display cycle arrives, data process and mapping waveform is sent to display by waveform mapping block；

The mapping waveform that display processes for video data and waveform mapping block sends.

Measurement system that transient signal the most according to claim 1 is seamless, it is characterised in that described characteristic signal scalping modeling Block uses peak value sample mode, and its expression formula is:

$Y\left(n\right)=\left\{\begin{array}{c}\underset{i=1~k}{\min }\[X(sk+i)\],n=1,3,5,\mathrm{...},2L/k-1,X\[sk+1\]\≤X\[(s+1)k\]\\ \underset{i=1~k}{\max }\[X(sk+i)\],n=1,3,5,\mathrm{...},2L/k-1,X\[sk+1\]>X\[(s+1)k\]\\ \underset{i=1~k}{\min }\[X(sk+i)\],n=2,4,6,\mathrm{...},2L/k,X\[sk+1\]>X\[(s+1)k\]\\ \underset{i=1~k}{\max }\[X(sk+i)\],n=2,4,6,\mathrm{...},2L/k,X\[sk+1\]\≤X\[(s+1)k\]\end{array}\right.,s-0~L/k-1$

Wherein, k represents that sampling interval, L represent the length of sampled signal X (l).

Measurement system that transient signal the most according to claim 1 is seamless, it is characterised in that described threshold setting module is to adopting The sampling of sample signal X (l) uses average value sample.

Measurement system that transient signal the most according to claim 1 is seamless, it is characterised in that described data process and waveform reflects Penetrating module and use multi-stage pipeline treatment mechanism, for each pending data, the every one-level in stream treatment line completes one Individual process task.