RU2451962C1 - Method to measure time interval - Google Patents

Abstract

FIELD: instrument making.

SUBSTANCE: when measuring a time interval T, a sequence of clock pulses is generated with a period τ, as synchronised with the start of the measured interval, in the end of the interval a delayed signal pulse is generated with an available shape, the number N of clock pulses starts counting between the start of the interval and the moment of registration of the delayed pulse, a correction ΔN is generated, and the interval T is determined using the formula T=τ (N+ΔN). At the same time duration T_p of a signal pulse is established as more than two clock periods τ, a non-delayed signal pulse is previously digitised with the period x, and results of digitisation are saved in W>3 arrays{Su}_w, where i=1 ,2,…, I - a sequence number of a sample in an array, w - a sequence number of an array, l>2 - number of samples, in process of digitisation in each array {S_1i}_w, samples of the non-delayed pulse are taken into moments of time counted from its start t_iw=(w-l) τ/W+(i-1)-τ, and start of its digitisation is set so that the sample with the maximum value has the same sequence number in all arrays, besides, a sample with the maximum value is included into the set of sample values {S_1i}_w, as well as at least one sample at the left and the right from it, in process of measurement of a time interval, a delayed signal pulse is digitised with binding to a clock sequence, forming an array {S_2i}, at the same time arrays {S_1i}_w and {S_2i} are formed with identical number of samples I in each array, so that the maximum sample in the array {Si;} is in the same position, as in the arrays {S_1i}_w, afterwards W estimates R_w=R({S_1i}_w, {S_2i}) are generated, which characterise proximity of arrays {S_1i}_w and {S_2i}, for instance, in the form of a total absolute deviation

a sequence number w* is determined, at the same time the estimate R_w to a largest extent characterises the proximity of arrays {S_1i}_w and {S_2i}, and a correction ΔN is generated using the formula ΔN=(w*-1)/W. Arrays {S_1i}_w and {S_2i} may be normalized so that their maximum samples have the same value. To even more increase accuracy, it is possible to generate an additional correction ΔN_C, determined previously by measurement of a reference interval of time.

EFFECT: increased accuracy of measurement of a time interval with standard equipment.

4 cl, 4 dwg

Description

The invention relates to electronic instrumentation and can be used in accurate meters of time intervals, in pulsed locations, experimental physics, etc.

The simplest method for measuring the time interval by converting it into a numerical code is known, according to the principle of counting the number of clock pulses of a stable frequency, filling the measured time interval specified by the “start” and “stop” signal pulses [1]. The extreme accuracy of this “direct counting” method is limited by the discrete nature of clock pulses.

Known ways to improve the accuracy of such a measurement without increasing the clock frequency due to the introduction of interpolation correction using auxiliary clock sequences. The closest in technical essence to the proposed method is a method [2] measuring the time interval T by forming a sequence of clock pulses with a period τ synchronized with the beginning of the measured interval, forming at the end of the interval a delayed signal pulse of a known shape, counting the number N of clock pulses between the beginning of the interval and the moment of registration of the delayed pulse, the formation of the correction ΔN and the determination of the interval T by the formula T = τ (N + ΔN). To determine the correction ΔN in this method, an additional sequence of clock pulses is used, shifted relative to the main sequence in frequency and synchronized with the delayed signal pulse.

With this method, information processing and determination of the ΔN correction are performed in real time, which requires complex high-speed equipment, which is also exposed to the risk of gross errors due to electromagnetic interference and interference and other adverse factors inherent in analog information processing.

The objective of the invention is to improve the accuracy and noise immunity of measuring the time interval with standard hardware.

, определяют порядковый номер w*, при котором оценка R_w в наибольшей степени характеризует близость массивов {S_1i}_w и {S_2i}, и формируют поправку ΔN по формуле ΔN=(w*-l)/W.This problem is solved due to the fact that in the known method of measuring the time interval T by forming a sequence of clock pulses with a period τ synchronized with the beginning of the measured interval, forming at the end of the interval a delayed signal pulse of a known shape, counting the number N of clock pulses between the beginning of the interval and the moment registration delayed pulse forming correction ΔN and determine the interval T from the formula T = τ (N + ΔN), the duration _{T, and} signal pulse set bole two clock periods τ, the undelayed signal pulse previously digitized with a period τ digitizing and storing the results in W≥3 arrays {S _{1i} w,} where i = 1, 2, ..., I - sequential sample number in the array, w - ref array number , I> 2 is the number of samples, in the process of digitization in each array {S _1i } _w , samples of the uncontrolled pulse are taken at the time instants t _iw = (w-1) τ / W + (i-1) -r, and the beginning of its digitization is set so that the sample with the maximum value has the same serial number in all arrays, and in the set of sample values of the arrays {S _1i } _w include the sample with the maximum value and at least one sample to the left and to the right of it, during the measurement of the time interval, the delayed signal pulse is digitized with reference to the clock sequence, forming the array {S _2i }, the arrays {S _1i } _w and {S _2i } are formed with the same number of samples I in each array, so that the maximum sample in the array {S _2i } is in the same position as in the arrays {S _1i } _w , then w is formed count _{R w = R ({S 1i} } w, {S 2i}), characterized x proximity array {S _{1i} w} and {S _2i}, e.g., as the sum of absolute deviation

, determine the serial number w * at which the estimate R _w characterizes the closeness of the arrays {S _1i } _w and {S _2i } to the greatest degree, and form the correction ΔN by the formula ΔN = (w * -l) / W.

To ensure maximum sensitivity of the method to the position of the delayed signal pulse, the arrays {S _1i } _w and {S _2i } are normalized so that their maximum samples have the same value.

To eliminate the unaccounted for systematic error in determining the time interval T, an additional correction ΔN _k is introduced, determined previously by measuring the reference time interval T _e , calculating the correction ΔN _k by the formula ΔN _k = (T _e -T) / τ, and the measurement T of the measured interval by the formula T = τ (N + ΔN + ΔN _k ).

. Показаны зависимости вида

, для значений сдвига сигнала S₂(t), равного 0 и 0,4 (W=10). На фиг.4а) и б) представлены реализации R_w при воздействии аддитивного шума при отношении сигнал/шум соответственно 100 и 30.Figure 1 presents a timing diagram of the process of measuring the time interval T, its binding to the clock frequency and the formation of arrays {S _1i } and {S _2i }. Figure 2 shows the signal pulse S ₂ (t) at zero offset relative to the clock pulse (figa) and with a shift of 0.4 clock period (fig.2b). Figure 3 illustrates the formation of functions

. View dependencies are shown.

, for the values of the signal shift S ₂ (t) equal to 0 and 0.4 (W = 10). On figa) and b) presents the implementation of R _w when exposed to additive noise with a signal to noise ratio of 100 and 30, respectively.

At the beginning of the measured time interval T, a signal pulse S ₁ (t) 1 is formed, and at the end, a pulse S ₂ (t) 2 is formed. A sequence of clock pulses 3 is tied to pulse 1 in such a way that the temporary position of pulse 1 corresponds to an array of sample values {S _1i } ₁ . The array {S _1i } ₁ belongs to the set of arrays {S _1i } _w 4, which are preliminarily determined by digitizing the first signal pulse. When digitizing the pulse S ₁ (t), W arrays of its sample values are formed. The samples in these arrays are determined with a period of the clock sequence τ. Digitization is performed so that the first sample in each array is delayed relative to the conditional start of the pulse S ₁ (t) by the time interval (w-1) τ / W, where w = 1, 2, ..., W is the serial number of the array {S _1i } _w ; W is the number of arrays.

Calculations and experimentally showed that the number W of arrays {S _1i } _w should be at least three. The set of sample values of arrays {S _1i } _w includes a sample with a maximum value and at least one sample to the left and to the right of it. The number of samples I in all arrays should be the same. The conditional beginning of the pulse S ₁ (t) is chosen so that, when digitizing in all arrays {S _1i } _{w, the} sample with the maximum value is at the same position i = i _max .

The resulting set of arrays {S _1i } _w (templates) is stored in the system memory.

In the process of measuring the time interval T, the second signal pulse S ₂ (t) 2 is digitized at the moments of arrival of the clock pulses. Based on the results of digitization, an array {S _2i } 5 is formed in such a way that it contains a sample with the maximum value in the same position as in the arrays {S _1i } _w , and the number of samples in this array is I.

или среднего абсолютного отклонения

. Затем определяют порядковый номер w=w* того шаблона, который по выбранному критерию ближе к массиву {S_2i}, и определяют поправку ΔN по формуле ΔN=(w*-l)/W.After the formation of the array {S _2i }, it is compared with each of the patterns {S _1i } _w , by forming the coefficients R _w = R ({S _1i } _w , {S _2i }) characterizing the proximity of the arrays {S _1i } _w and { S _2i }, for example, in the form of a correlation coefficient

or mean absolute deviation

. Then determine the serial number w = w * of the template, which is closer to the array {S _2i } by the selected criterion, and determine the correction ΔN by the formula ΔN = (w * -l) / W.

Example.

The clock sequence in a conventional time scale t has a period τ = 1.

Signal pulses on this scale are described by the expression S (t) = At ² exp (-bt). The type of signal pulses at A = 184 and b = 1 is shown in Fig.2. As can be seen from the graph, the duration of the signal pulse is 8 periods of the clock frequency, and the amplitude on a conditional scale is 100. The beginning of the signal pulse 1 of Fig.2a) is conditionally attached to the second clock pulse, and the signal pulse in Fig.2b) is shifted relative to this position by 0.4 clock periods.

As a result of digitization of the first signal pulse, ten arrays {S _1i } _{w of} normalized sample values of the signal S ₁ (t) are shown in Table 1.

Table 1 Arrays {S _1i } _w , sample values of the signal S ₁ (t) = S (t) = At ² exp (-bt) Array (pattern) Sample No. i = 1 i = 2 i = 3 i = 4 i = 5 {S _1i } ₁ 0.3823 1,0000 0.9714 0.6851 0.4115 {S _1i } ₂ 0.4609 1,0000 0.9280 0.6411 0.3806 {S _1i } ₃ 0.5369 1,0000 0.8902 0.6032 0.3540 {S _1i } ₄ 0.6099 1,0000 0.8570 0.5702 0.3311 {S _1i } ₅ 0.6796 1,0000 0.8277 0.5413 0.3112 {S _1i } ₆ 0.7458 1,0000 0.8017 0.5159 0.2936 {S _1i } ₇ 0.8087 1,0000 0.7783 0.4932 0.2781 {S _1i } ₈ 0.8684 1,0000 0.7573 0.4730 0.2644 {S _1i } ₉ 0.9250 1,0000 0.7383 0.4549 0.2520 {S _1i } ₁₀ 0.9786 1,0000 0.7210 0.4385 0.2410

The sampling frequency is equal to the clock period τ = 1. The initial digitization shift of the wth pattern is (w-1) / 10, that is, the first pattern has a zero shift, and the shift of the 10th pattern is 0.9 periods of the clock sequence. When digitizing, the beginning of the signal S ₁ (t) is chosen so that the maximum sample falls at one position in all patterns (j = 2). Samples with values S _1i > 0.2 were left in each template. In this case, the number of samples in all the patterns is I = 5.

As a result of digitization, normalized signal arrays S ₂ (t) with a zero shift and a shift of 0.4 clock periods are obtained. The results of digitization {S _2i } at two values of the signal shift S ₂ (t), shown in figure 2, are shown in table 2. As in the templates, the number of samples is I = 5, and the maximum sample is in the second position.

table 2 Digitization results {S _2i } of the signal pulse S ₂ (t) when it is shifted by 0 and 0.4 clock periods signal shift Sample No. i = 1 i = 2 i = 3 i = 4 i = 5 0 0.3823 1,0000 0.9714 0.6851 0.4115 0.4 0.6796 1,0000 0.8277 0.5413 0.3112

. Результаты сравнения приведены в Табл. 3 и на графиках фиг.3а) (нулевой сдвиг сигнала) и фиг.3б) (сдвиг сигнала равен 0,4).Comparison of arrays {S _2i } with patterns {S _1i } _{w was} carried out according to the criterion of mean absolute deviation

. The comparison results are shown in Table. 3 and in the graphs of FIG. 3a) (zero signal shift) and FIG. 3b) (signal shift is 0.4).

Table 3 The average absolute deviation R _{w of the} array {S _2i } from the patterns {S _1i } _w signal shift Pattern No. one 2 3 four 5 6 7 8 9 10 0 0 0.1969 0.3752 0.5373 0.6851 0.8203 0.9448 1.0594 1,1655 1.2638 0.4 0.6851 0.4882 0.3099 0.1478 0 0.1352 0.2594 0.3743 1,1655 0.4804

The minimum value of R _w 6 corresponds to a signal shift of 0.4 clock periods. By the number w * of the corresponding template, the correction ΔN is formed. In the first case, it is (1-1) / 10 = 0, and in the second (5-1) / 10 = 0.4, which corresponds to a real shift.

The results are obtained in the absence of noise distortion.

Table 4 shows the results of data processing by the proposed method when exposed to several implementations of additive noise in the frequency band corresponding to the spectral characteristic of the signal.

Table 4 The average absolute deviation R _{w of the} array {S _2i } from the patterns {S _1i } _w when exposed to additive noise. Minimum values are underlined. signal to noise ratio Pattern No. one 2 3 four 5 6 7 8 9 10 one hundred 0.6779 0.481 0.3027 0.1406 0.017 0.1424 0.2669 0.3815 0.4876 0.5859 one hundred 0.661 0.4641 0.2858 0.1237 0,0241 0.1593 0.2838 0.3984 0.5045 0.6028 one hundred 0.7044 0.5075 0.3292 0.1671 0,0317 0.1159 0.2404 0.355 0.4611 0.5594 one hundred 0.6898 0.4929 0.3146 0.1525 0,0373 0.1305 0.255 0.3696 0.4757 0.574 thirty 0.5743 0.3774 0.2041 0.108 0.128 0.246 0.3705 0.4851 0.5912 0.6895 thirty 0.6689 0.472 0.2937 0.1316 0,1502 0.1994 0.2759 0.3905 0.4966 0.5949 thirty 0.6955 0.4986 0.3203 0.1582 0,048 0.148 0.2493 0.3639 0.47 0.5683 thirty 0.81 0.6131 0.4348 0.2727 0.1249 0,0537 0.1348 0.2494 0.3555 0.4538 10 0.8918 0.6949 0.5166 0.3895 0.3811 0.3783 0.4088 0.454 0.4973 0.539

From the table and graphs of Figure 4 it can be seen that when the signal to noise ratio = 100, the dependence of R _w 6 on w practically does not differ from the ideal one. With a signal-to-noise ratio = 30, the minimum position of the function can be shifted to an adjacent position. In this case, the error in measuring the time interval reaches 0.1τ. With a further decrease in the signal-to-noise ratio, the peak of the dependence R _w becomes less pronounced, and this can lead to an increase in the error. It should be noted that under such conditions and under given restrictions, the error inherent in the known technical solution [2] and other similar solutions may exceed the error provided by the proposed method, since the known methods are not designed for measurements in the presence of noise and are not optimized for this factor. The error in them can reach the duration of the signal pulse front [1, p. 77].

The greatest uniqueness and noise immunity of measurements is ensured by normalizing the arrays {S _1i } _w and {S _2i } so that their maximum samples have the same value. In the above example, the maximum sample size of all arrays is 1.

In order for the maximum sample to be in the same position when forming the patterns {S _1i } _w in each of them, the beginning of the pulse S ₁ (t), by which it is tied to the clock sequence, is selected conditionally. In this case, a systematic shift in the measurement evaluation can occur, which is excluded by preliminary calibration with the reference time interval T _e and determining the corresponding correction ΔN _k = (T _e -T _e ^* ) / τ, where T _e is the measured value of the reference interval. In the measurement process, an estimate of T of the measured interval is determined by the formula T = τ (N + ΔN + ΔN _k ).

The proposed method is implemented in an experimental sample of a laser rangefinder with the following characteristics. Clock frequency F _T = 25 MHz (clock period T = 40 ns). The number of arrays {S _0i } W = 100. The maximum samples of all arrays {S _0i } _w are in the 2nd position, and the total number of samples is K = 5. The standard error of the measurement of the interval T does not exceed 0.4 ns (this corresponds to a measurement error of the range of 0.06 m). In known devices, the measurement error is tens of times greater than this value.

The proposed method for measuring the time interval in comparison with the known methods provides a significantly more accurate estimate of the position of the second signal pulse with respect to the beginning of the measured interval both in the presence of noise and in their absence.

As a result of using the method, a significant increase in the accuracy and noise immunity of measuring the time interval by standard hardware is provided.

Information sources

1. E.A. Meleshko "Integrated Circuits in Nanosecond Nuclear Electronics". M .: Atomizdat, 1977, p. 139.

2. RF patent No. 2133053, 1999. E.I. Gurin. "Method of accelerated nonius interpolation of time intervals" - prototype.

Claims (3)

1. The method of measuring the time interval T by forming a sequence of clock pulses with a period τ synchronized with the beginning of the measured interval, forming at the end of the interval a delayed signal pulse of a known shape, counting the number N of clock pulses between the beginning of the interval and the moment of registration of the delayed pulse, generating the correction ΔN and determining an interval T using the formula T = τ (N + ΔN) , characterized in that the duration T _and the pulse signal is set for more than two clock periods τ, nezade zhanny previously digitized pulse signal with a period τ and stores the result in the digitizing arrays W≥3
{S _1i } _w , where i = 1, 2, ..., I is the serial number of the sample in the array, w is the serial number of the array, I> 2 is the number of samples, in the process of digitization in each array {S _1i } _w take samples of undelayed pulse at the time moments t _iw = (w-1) τ / W + (i-1) · τ counted from its beginning, and the beginning of its digitization is set so that the sample with the maximum value has the same serial number in all arrays, and in a set of sample values of arrays {S _1i } _w include a sample with a maximum value and at least one sample to the left and to the right of it, in the process All time interval measurements, the delayed signal pulse is digitized with reference to the clock sequence, forming the array {S _2i }, while the arrays {S _1i } _w and {S _2i } are formed with the same number of samples I in each array so that the maximum sample in the array {S _2i } was in the same position as in the arrays {S _1i } _w , after which form W estimates R _w = R ({S _1i } _w , {S _2i }), characterizing the proximity of the arrays {S _1i } _w and {S _2i }, for example, in the form of a total absolute deviation

, determine the serial number w *, in which the estimate R _w characterizes the closeness of the arrays {S _1i } _w and {S _2i } to the greatest extent, and form the correction ΔN by the formula ΔN = (w * -1) / W. 2. The method according to claim 1, characterized in that the arrays {S _1i } _w and {S _2i } are normalized so that their maximum samples have the same value. 3. The method according to claim 1, characterized in that when determining the time interval T, an additional correction ΔN _k is introduced, determined previously by measuring the reference time interval T _e , calculating the correction ΔN _k by the formula ΔN _k = (T _e -T _e ^* ) / τ, where T _e ^* is the measured value of the interval T _e , and during the measurement process, an estimate of T of the measured interval is determined by the formula T = τ (N + ΔN + ΔN _k ).

Images