RU2082988C1 - Process of optimal detection of pulse signals with unmodulated carrier frequency - Google Patents

Abstract

FIELD: radiolocation. SUBSTANCE: process consists in reception of signal, in optimal or quasi-optimal filtration of it, in isolation of envelope, in comparison with threshold level. Signal is detected by excess of threshold. Before quasi-optimal filtration excessive transmission band ΔF of receiver is provided which is compared with width of main spectrum of pulse signal of duration τ_p, where ΔF is bigger or equal to 10/τ_p and signal is delayed for time τ_к=τ_p with realization of condition τ_к=d 1/2, where d is integral number; f_c is carrier frequency. Undelayed and delayed signals are summed up with subsequent filtration in quasi-optimal filter which transmission band satisfies condition ΔF_к much less than i/τ_p and central frequency f_p= f_c. In this case operations of delay and summing are repeated gamma times. EFFECT: improved authenticity of process. 3 dwg

Description

 The invention relates to radio engineering, can be used in radar and in some other radio communication systems.

 Known methods for optimal detection of signals based on the operations of optimal linear filtering (OLF) or one or another correlation processing. As a result of these operations, the highest signal-to-noise ratio (SNR) is provided, and from here the best signal detection quality according to one or another probabilistic criterion or message reception (with the smallest error). With a higher SNR, higher accuracy (less error) of measurement of some signal parameters is provided.

Здесь S(iω), S^*(iω) соответственно спектр и комплексно-сопряженный спектр сигнала;

двухсторонняя спектральная плотность (СП) входного шума; t₀ момент возникновения максимального напряжения на выходе ОФ. Выражения (1) и (2) справедливы и для более сложных видов сигнала со случайным временем запаздывания, случайной амплитудой. В последнем случае в (1) и (2) учитывается среднее значение амплитуды сигнала; величина ОСШ усредняется

. В случае белого шума с СП

(N₀ односторонняя СП) ОФ с АЧХ является согласованным фильтром (СФ) со спектром сигнала. При этом ОСШ определяется как параметр обнаружения, равный

где E энергия сигнала. Приведенные выражения (1) и (2) распространяются на любую флюктуационную стационарную помеху, в том числе и пассивную, которая в определенных ограниченных интервалах времени может считаться условно стационарной.According to the theory of OLF, the frequency response of the optimal filter (OF) for a fully known signal against a background of stationary noise and the SNR at the output of the OF is determined by the relations

Here S (iω), S ^* (iω) respectively the spectrum and the complex conjugate spectrum of the signal;

two-way spectral density (SP) of input noise; t ₀ moment of occurrence of the maximum voltage at the output of the OF. Expressions (1) and (2) are also valid for more complex types of signal with a random delay time, a random amplitude. In the latter case, in (1) and (2), the average signal amplitude is taken into account; SNR value is averaged

. In case of white noise with SP

(N ₀ one-way SP) OF with the frequency response is a matched filter (SF) with the spectrum of the signal. In this case, the SNR is defined as a detection parameter equal to

where E is the signal energy. The above expressions (1) and (2) apply to any fluctuation stationary interference, including passive, which can be considered conditionally stationary in certain limited time intervals.

In modern conditions of ever-increasing levels of interference and requirements for increasing the range, it is increasingly difficult to provide the required value of q ² . In fact, in a number of communications applications, technical capabilities are already limited due to the impossibility of creating transmitters and antenna-feeder devices for a large level of energy and, all the more, the peak power of the emitted signal, mass-dimensional limitations, and so on. In the case of passive interference, there are also restrictions on signal isolation due to a number of technical (hardware) reasons. Therefore, increasing noise immunity (PU) due to more efficient signal processing when they are detected (received) is appropriate and relevant.

In this method of optimal detection of pulsed signals, based on the correct signal processing operations when they are received, including well-known operations, prior to the classical OLF operation, additional filtering (DF) is carried out, which converts the shape of the envelope of the input signal, turning it into a more complex signal the same carrier frequency (f _c ), consisting of two or more private operating signals practically separated in time. In this case, the DF provides a dip-noise "well" in the noise spectral density (SP) n _ξ (ω) around the frequency f _c . Therefore, the subsequent traditional optimal filter (OF), made on the basis of the classical OLF, thanks to the “pit” in the noise SP (at the input) provides at its output a sharp increase in the signal-to-noise ratio (SNR) compared with that of the traditional OLF, i.e. . without using the specified DF. The use of DF can be multiple, which deepens and expands the noise "hole".

Such changes form the envelope of the signal and noise SP (via DF) are achieved by conducting the initial signal delay operations rectangular envelope duration t _and the amount _to τ = τ _and summation and subsequent undelayed and delayed signals. The device with which such operations are carried out is a special case of the so-called comb filter (HF), the frequency response of which has a known comb shape. Those. due to the use of an additional filter (DF; before the OF), which in this particular case is the ideal GF without amplitude and phase distortions in the delay device (LZ) and the GF adder. In this case, the condition
f _c • τ _k = d 1/2 (4)
where τ _to = τ _{and the} delay in the LZ;
d is an integer.

The bandwidth of the receiver and the HF circuits is chosen to be of excess order ΔF ≥ 10 / τ _and that does not worsen the final result due to the use then as a quasi-optimal filter with a bandwidth ΔF _to ≪ 1 / τ _and .

As a result, the signal input through DF-GF at its output (at the adder) summed input signal of duration τ _and he is delayed by an interval τ _a = τ _u but with the initial phase, wherein the phase of the end of the first (un-delayed) signal ψ = π. Those. at the junction of both signals there is a phase jump of 180 ^o , which makes them the most incoherent - anticoherent. This position follows from condition (4). Due to this condition, the frequency response of the GF has a zero “pit” at the signal frequency. When ensuring the equality of the variances of the summed noise in the GF (and SP) SP noise at the output will be
n _η (ω) = 4n _ξ (ω) cos ² ωτ _k / 2 (5),
n _ξ (ω) is a one-sided input noise SP. Denoting w ′ = ω-ω _c , expression (5) can also be written in the form
n _η (f) = 4n _ξ (f) sin ² πf′τ _to (5a)
The HF output signals and noise are then subjected to conventional OLF. Because both anti-coherent signals at the DF-GF output are separated by the interval τ _k = τ _k , and the stationary noise SP with a “pit” acts continuously in time and, of course, the subsequent signal detection circuit “does not know” that the first signal will arrive at it or if the second signal is not received (in this case, anticoherent with a phase jump of ψ = π), the detection process is determined by the first separate signal and noise (at the output of the HF) with the SP n _η (f ′), which corresponds to the causal principle of signal detection during time. Naturally, this begins with the operation of optimally filtering said signal and noise. However, the traditional theory of OLF does not take into account the possibility of such a separation of signals; it takes into account the totality of the converted DF signal, i.e. the spectrum of the entire signal at the output of the DF, in which, naturally, there will be the same spectral "well" as in the noise SP (5a).

Здесь

двусторонняя СП шума, равная

в данном случае (ГФ) спектры S_p(iω)=S(iω) т. е. спектры первого сигнала на выходе ГФ и входного сигнала (так как первый и является входным сигналом незадержанным; S $\genfrac{}{}{0ex}{}{\text{*}}{\text{p}}$ (iω) - комплексно-сопряженный спектр раздельного сигнала. Напомним, что в классической теории ОЛФ для определения К_ОФ(i ω ) и ОСШ (ρ_т) в (6) и (7) учитываются спектр входного сигнала S(i ω) и СП входного шума

.So, to determine the frequency response of the OF _p and the SNR r at the output of the optimal filter (OF _p ), according to the traditional method, it is necessary to take into account the spectrum of the separate signal S _p (iw) and the noise SP after DF-1F at the input of the OF _p

Here

two-way noise equal to

in this case, (GF) spectra S _p (iω) = S (iω), i.e., the spectra of the first signal at the output of the GF and the input signal (since the first is an undelayed input signal; S $\genfrac{}{}{0ex}{}{\text{*}}{\text{p}}$ (iω) is the complex conjugate spectrum of a separate signal. Recall that in the classical theory of OLF to determine K _OF (i ω) and SNR (ρ _t ) in (6) and (7), the spectrum of the input signal S (i ω) and the input noise SP are taken into account

.

где x=πf′τ_и, f′=f-f_c
Для квазиоптимального фильтра (КФ) с прямоугольной формой АЧХ доказано, что ОСШ

где

параметр обнаружения,
E энергия входного сигнала;
β=ΔF_к•τ_и, ΔF_к полоса КФ. При β → 0, ρ_п → ∞..For a signal with a rectangular envelope of duration t _and and input noise with a SP n _η (ω) = N _o / 2 = const, we obtained:

where x = πf′τ _and , f ′ = ff _c
For a quasi-optimal filter (CF) with a rectangular shape of the frequency response, it is proved that the SNR

Where

detection parameter
E is the energy of the input signal;
β = ΔF _to • τ _and , ΔF _to the CF band. As β → 0, ρ _p → ∞ ..

Так, при b= 0.01 g=3,4, т.е. ρ_п ≫ ρ_c=q². Разумеется, при значительном росте величины ОСШ будут соответственно улучшены качественные (вероятностные) показатели обнаружения сигналов и точность измерения их некоторых параметров.Those. the gain in SNR is equal to

So, for b = 0.01 g = 3.4, i.e. ρ _n ≫ ρ _c = q ² . Of course, with a significant increase in the SNR value, the qualitative (probabilistic) indicators of signal detection and the accuracy of measuring some of their parameters will be accordingly improved.

Although in this version of the method it is possible to obtain a very high SNR for real KF, including according to (10), however, this will require very narrow-band KF ΔF _k = β / τ _and , β ≪ 1. At ordinary intermediate frequencies, this will require the use of high-quality quartz resonators. However, for small values of τ _and in the circuits of a strip amplifier and detector, at the same intermediate frequency (f _c ), there should be a sufficiently wide band for passing a sawtooth front of the working signal of duration τ _and , i.e. ΔF _у ≈ 1 / τ _and . It is easier to provide such a band with the ratio f _c : ΔF _y ≳ 10. So, if τ _u = 1 μs, then ΔF _y ≃ 1 MHz and f _c ≥10 MHz, i.e. at the same time, within the limits of τ _and = 1 μs = 1 μs, at least 10 signal waves are placed at f _c = 10 MHz, which is sufficient to generate a working signal in the IFA and to accurately estimate its spectrum using standard formulas.

We note that quartz resonators at the first harmonic are technologically advanced up to the frequency f _c = 10 MHz. For example, at β = 0.0001, the CF band will be DF _к = β / τ _and = 100 Hz. Therefore, the equivalent Q factor of the CF loop will be Q = f _c • ΔF _k = 10 ⁷ : 10 ² = 10 ⁵ , which is practically unattainable for quartz as well. It would have to be limited to at least Q-factor Q = 10 ⁴ and, accordingly, ΔF _k = 1 KHZ and β = 0.001. Apparently, at very small b, tuning the HF circuits to accurately equalize the SP of the summed noises in the adder and to form zero in the noise "well" will be practically problematic. The same in conditions of not very strictly stationary interference, which most likely corresponds to practice. Therefore, in general, it is desirable to achieve a high SNR (i.e., high noise immunity of the PU) with relatively not very small values of b.

In this method, due to the narrow bandwidth of the CF (after the GF) after passing a pair of working anti-coherent signals of duration t _{and the} resulting free vibrations in the CF chains last for a very long time in accordance with the equivalent CF time constant equal to τ _kf ≈ 1 / ΔF _k = 1: β / τ _and / β ≫ τ _and . This means that the throughput, range resolution (RSD) will deteriorate sharply. This is a serious drawback, and measures must be taken to ensure the taxiway at least at a level close to τ _and .

The aim of the invention is, firstly, to provide increased detection efficiency of signals, increasing SNR with an acceptable not very narrow band of CF, i.e. and solving technological difficulties to ensure high efficiency. Secondly, the aim of the invention is to provide RSD detectable signals, in magnitude, close to the duration of τ _and .

The goals are achieved by the introduction of additional (distinctive) operations:
a) the operation of delaying the initial signal (τ _and ) when the condition f _c • τ _k = d 1/2 is met and summing it with an uncontrolled signal, carried out using an additional filter (DF-GF), is carried out sequentially (cascade) γ> 1 time ; these cascade operations at a selected value of b sharply increase the SNR;
b) after the triggering device is triggered when any of the RF pulses passed through the CF and the subsequent bandpass filter (PF) are detected, video pulses of duration t _OB = (1 ... γ) τ are formed _and then fed to control (key) circuits with low output resistance in the working (open) state, with the help of these small resistances to which the KF and PF circuits are connected, the latter are shunted (“zeroed”) during τ _OB , after which the entire signal detection circuit is again ready to detect the signal from the output of the adder last gth cascade of GF.

 The following additional operations are discussed in detail.

Тогда СП шума на выходе ДФ-ГФ равна (5) или (5а), что вытекает из известного соотношения для линейных фильтров:

Из изложенного выше принципа учет раздельных во времени сигналов для определения АЧХ ОФ_p (после ДФ-ГФ) и ОСШ ρ на выходе такого ОФ_p следует, что для определения этих величин надо воспользоваться соотношениями (6) и (7), подставляя в них известные функции

(f')-(5_a) и S_p(if'). Последняя для сигнала длительностью t_и с прямоугольной огибающей известна: в данном случае, когда первый раздельный сигнал на выходе ГФ совпадает с исходным сигналом длительностью τ_и

Для однократного выполнения операций п.а. или одного каскада АЧХ ОФ_p (ее модуль) и ОСШ ρ определены выражениями (8) и (9). Определены эти величины для произвольного числа каскадов g. При этом напомним, что первый импульс совокупности ВЧ антикогерентных импульсов на выходе последнего (g-го) каскада ГФ остается исходным незадержанным сигналом со спектром (13).The implementation of the above operations, as already noted, is equivalent to the use of an additional filter (DF) of the ideal GF with the parameter V = 2 equal to the number of terms in the GF adder, and therefore, the total expansion of the output signal along the envelope T = Uτ _and = 2τ _and . The frequency response of such a filter is determined by a relatively simple direct calculation and is generally known. In particular, based on the general expression for an arbitrary V [1], we have for V = 2

Then the noise SP at the output of the DF-GF is (5) or (5a), which follows from the well-known relation for linear filters:

From the above principle, taking into account time-separated signals for determining the frequency response of the OF _p (after DF-GF) and the SNR ρ at the output of such OF _p, it follows that to determine these quantities, we must use relations (6) and (7), substituting the known the functions

(f ') - (5 _a ) and S _p (if'). The latter is known for a signal of duration t _and with a rectangular envelope: in this case, when the first separate signal at the output of the HF coincides with the original signal of duration τ _and

For a single operation or one cascade of the frequency response of the OF _p (its module) and the SNR ρ are defined by expressions (8) and (9). These quantities are determined for an arbitrary number of cascades g. At the same time, we recall that the first pulse of the set of HF anticoherent pulses at the output of the last (gth) GF cascade remains the initial uncontrolled signal with spectrum (13).

в этом случае с учетом логики (12) и выражения (5а) будет равна

Тогда в общем виде

Применительно к рассматриваемому сигналу со спектром (13) при СП

АЧХ ОФ_p при γ>1 (15а) имеет тот же характер, что и при g 1: пики до бесконечности в точках

, но более острые (узкие), не столь глубокие впадины кривой АЧХ между этими пиками. Такой ОФ_p, как и при γ= 1 строго нереализуем из-за необходимости иметь "бесконечные" коэффициенты передачи.SP noise

in this case, taking into account logic (12) and expression (5a) it will be equal to

Then in general form

As applied to the signal under consideration with spectrum (13) in SP

The frequency response of the OF _p for γ> 1 (15a) has the same character as for g 1: peaks to infinity at the points

, but sharper (narrower), not so deep valleys of the frequency response curve between these peaks. Such an OF _p , as with γ = 1, is strictly unrealizable because of the need to have "infinite" transmission coefficients.

а наличие в знаменателе интеграла множителя sin^2(γ-1)x увеличивает подинтегральную функцию в (16а) при любом X (кроме

) особенно в точках, где

. Величина ρ стремится к бесконечности быстрее, чем при g=1, причем тем быстрее, чем g>1.The magnitude of the SNR r = ∞ for any value of γ = 1.2. since already with g = 1

and the presence in the denominator of the integral of the factor sin ^{2 (γ-1)} x increases the integrand in (16a) for any X (except

) especially at points where

. The quantity ρ tends to infinity faster than for g = 1, and the faster, than g> 1.

Consider the traditional CF with a rectangular shape of the frequency response. Its central tuning frequency coincides with the signal frequency f _c (and the central frequency of the noise band KF f '= 0), which satisfies (4). In general, g≥1. SP of noise at the input of CF with a one-sided SP of input real noise N ₀ = Const is equal to (14)
n _ηγ (f ′) = N _o • (2sinπf′τ _u ) ^2γ (14a)
The right branch of this joint venture with normalization N 1 using the example γ = 2 in the region of small values X = πf′τ _and up to X _n = π 0.005 is shown in FIG. 1. The dotted line marks the right boundary of the CF band —X _n = pb / 2 at β = 0.01.

Let us consider the formation of a signal at the output of the gth cascade of the GF in a conventional digital diagram, where the input (initial) signal of unit amplitude is designated through the symbol "I". When the GF cascade is delayed by the interval t _k = τ _and condition (4) is observed, when summing the phase of the signals at their junction, they jump abruptly to ψ = π. But when summing undetected and delayed signals in each stage, the complete coincidence of the phases of the simultaneously summed elementary signals of duration t _and starting with the cascade γ≥2 is observed. The meaning of what has been said about summation is illustrated by a digital diagram (17), in which it manifests itself in the sums 1 + 1 = 2, 2 + 1 = 3, 1 + 2 = 3, and the like.

При этом скачки фазы на ψ=π на стыках сигналов в сумматорах сохраняются. Например, при g=3, после первого сигнала длительностью t_и происходит скачок фазы на π, затем после второго на p, затем после третьего на p. Это легко проследить графически, укладывая на интервале t_и=τ_к d 1/2 волн частоты f_c.

In this case, phase jumps by ψ = π at the junction of the signals in the adders are stored. For example, if g = 3, the first signal duration t _and a phase jump happens to π, then after the second to p, then after the third on p. This is easy to trace graphically, laying on the interval t _and = τ _to d 1/2 waves of frequency f _c .

Thus, the total duration at the output of the γ-th GF cascade is equal to T = (γ + 1) τ _and , moreover, through the intervals τ _{and the} phases of the signal jump abruptly by ψ = π. The maximum amplitude of anticoherent pulses is U _min = g U _m (for j≅ 3 see the diagram). However, to detect a separate signal, the largest difference amplitude U _mp should be taken into account. It can be shown that it at the junction of the first and subsequent (antiphase) pulse of the signal at the output of the GF is
U _Mr = U _m (γ-1); γ≥ 2 (18)
and is equal to U _mp = U _m at γ = 1 (the amplitude of the first pulse at the output of the GF). In the case of g 3 U _mp = 2U _m , which can be seen in diagram (17). This is explained by the fact that, as the analysis shows (see below), for optimal detection, it is necessary to use CF with a narrow band DF _k , at which the time constant of the CP circuits is τ _kf ≈ 1 / ΔF _k ≫ τ _and , which leads to a sharp delay time of free oscillations after passing the previous pulse through the CF. At the moment of the appearance of the next pulse at the input of the CF with the opposite phase, the total jump in the amplitude of the signal at the input of the CF (this is reflected above by the difference formula 18) is compressed, the indicated opposite phasing of such signals is proved for single LC circuits.

So, to determine the SNR, one must take into account the amplitude of the separate signal U _mp for γ ≥3 or U _m for g = 1; 2.

Здесь максимальная амплитуда U_mk на выходе КФ с полосой ΔF_к зависит от эффективной входной амплитуды U_mp раздельного сигнала (на выходе ГФ; см. выше), от величин ΔF_к и τ_и, поскольку параметр Z определяется как
Z=πΔF_кτ_и=πβ ΔF_к=β/τ_и (20)
В (19) СП n_2γ(f′), определяемая согласно (14а), оценивается как средняя мощность шума на 1 Гц полосы. Из физических соображений и смысла вида АЧХ ОФ_p наибольшие значения ОСШ будут в области DF_к _→ 0. Для малых аргументов Z справедливо

. Поэтому с учетом (14а) и (20)

Здесь учтено, что U $\genfrac{}{}{0ex}{}{\text{2}}{\text{m}}$ τ_и=2E, E энергия сигнала, а также (3). Т.е. выигрыш равен

Естественно, для получения большей эффективности роста ρ_п и g, следует выбирать выше величину γ. При этом решающую роль имеет величина интеграла в (21) и (22) в области малых значений b<1. Заметим, что при малых значениях b и больших g потребуется весьма высокая точность расчета интеграла (непринципиальное техническое обстоятельство). Однако, чтобы и это учесть и не очень усложнять схему остановимся на конкретном примере g=2 (2 каскада ГФ), при котором U_mp=U_m. Тогда, находя интеграл в (22) имеем

При β=0,01 g₂≈50000 раз или ≈47 дБ. Для сравнения укажем, что при g1 g₁= 30,4 или 14,8 дБ. Таким образом r_п=g•ρ_c ≫ ρ_c, причем при сравнительно не очень малых β (0,01). При g=1 и b → 0 ρ_п→ ∞. Это тем более относится и к случаям γ>1, т. к. делитель в (22) при g>1 в раскрытом виде представляет собой разность малых величин, которая еще быстрее стремится к нулю, чем при g=1, когда

что легко сравнить в (22).For the considered type of CF with a rectangular frequency response

Here, the maximum amplitude U _mk at the output of the CF with the band ΔF _k depends on the effective input amplitude U _{mp of the} separate signal (at the output of the GF; see above), on the quantities ΔF _k and τ _and , since the parameter Z is defined as
Z = πΔF _to τ _and = πβ ΔF _to = β / τ _and (20)
In (19), the SP n _2γ (f ′), determined according to (14a), is estimated as the average noise power per 1 Hz band. From physical considerations and meaning of the form of the frequency response of the OF _{p, the} highest SNR values will be in the region DF _to _ → 0. For small arguments Z is valid

. Therefore, taking into account (14a) and (20)

It is taken into account that U $\genfrac{}{}{0ex}{}{\text{2}}{\text{m}}$ τ _and = 2E, E is the signal energy, as well as (3). Those. the gain is equal

Naturally, in order to obtain a greater growth efficiency of ρ _p and g, one should choose the higher γ value. The decisive role is played by the integral in (21) and (22) in the region of small values of b <1. Note that for small values of b and large g, a very high accuracy in the calculation of the integral will be required (unprincipled technical circumstance). However, in order to take this into account and not to complicate the scheme very much, let us dwell on a concrete example g = 2 (2 GF cascades), at which U _mp = U _m . Then, finding the integral in (22) we have

At β = 0.01 g ₂ ≈50000 times or ≈47 dB. For comparison, we indicate that for g1 g ₁ = 30.4 or 14.8 dB. Thus, r _p = g • ρ _c ≫ ρ _c , and with relatively not very small β (0.01). For g = 1 and b → 0 ρ _n → ∞. This is even more true for the cases γ> 1, since the divisor in (22) for g> 1 in the expanded form is the difference of small quantities, which tends to zero even faster than for g = 1, when

which is easy to compare in (22).

It should be noted that such a very high g ₂ result is achievable, of course, with strict stationary input noise and accurate and stable tuning of the GF cascade circuits, when a "nominal well" in the SP at the output is provided. Practically due to the non-observance of the strict equality of the total noise noise, the result (23) will be slightly lower, but guaranteed to be high compared to one cascade of GFs.

We note that for small β = ΔF _, τ _{and the} maximum voltage at the output of the CF is almost proportional to the amplitude of the input voltage U _m (in general, U _mp ) only for limited types of CF, for example, for the considered ideal one- and two-stage resonant LC circuits, single-cascade on coupled circuits with bumps with a relative level of the depression of 0.7, with a critical connection, which follows from the calculated curves of establishment at the output of the CF. In the cases of some real QFs with their multistage and small β output levels of the signal (and noise) nonlinearly depend on the input levels (in the establishment interval), and the question of decreasing the SNR requires special studies, in particular, the relationship between the noted nonlinearity and the group delay time of the signal with multi-loop CF.

Since relatively narrow-band filters are used as KFs, the question naturally arises of increased frequency selectivity (ES). In this case, the mode of small values b = ΔF _to τ _{and is used} . But the ES of filters is fully manifested only at β ≳ 0.5.

Since the frequency response of a simple quartz filter is almost equivalent to the frequency response of a single LC circuit, using this KF, one can expect to achieve a relatively large gain at the resonance frequency f _p = f _c and the proximity of the KF frequency response to the main region of the AF frequency response _p at a frequency f '= 0 at f _c corresponding to condition (4). This will result in high SNR. In individual applications, naturally, CFs of the type of resonant LC loops can be used.

In order to solve the problem of approximating the RSD to the value of τ _and using this method of detecting signals with narrow-band CFs having an equivalent time constant τ _kf ≫ τ _and , it is advisable to generate a special pulse signal with a duration of τ _OB , "overlapping" after detecting the signal of the first separate (or subsequent) signal "in time, the passage of the remaining separate signals at the output of the last γ-th GF cascade. During the interval t _OB, the narrow-band CF and subsequent PF circuits are shunted by the low-impedance output resistance of the key circuit, as a result of which the free vibrations (“zero”) in these circuits are completely damped, and the whole circuit is again ready to detect the next (in time) input signal of duration τ _and . In accordance with the digital diagram (17) in the light of the foregoing, it is advisable to have τ _OB = γτ _and , although in some cases the variants τ _OB <γτ _{and are} possible. In general, Τ _OB = (1 ... γ) τ _and . Then the RSD will be determined by the total duration of all separate signals at the output of the γ-th GF cascade
T _rsd = T = (γ + 1) τ _and
So, for γ = 2 T, _ssd = 3τ _and . But at γ = 3, the operation of the solver can occur for weak signals after the second separate signal (U _mp = 2U _m ; see diagram (17) and (18)).

In FIG. Figure 2 shows an example of a simplified block diagram of the receiver circuits (after the IFA), in which the basic operations (distinguishing features) of the proposed method for the optimal detection of pulsed signals using two stages of additional GF filters are implemented. The following notation is used in the figure and in the text: 1 OPC receiver; 2 delay device for the interval t _to = τ _and ; 3 amplifier to compensate for losses in the delay device; 4 adder; 5 decoupling amplifier; 6 amplifier loaded on a quasi-optimal filter; 7 amplifier loaded on a band-pass filter; 8 detector; 9 decisive device; 10 shaper control "zeroing"pulses; 11 device to "zero" the output circuits of amplifiers 6 and 7.

From the output of the amplifier of the receiver 1, the passband ΔF of which (and the subsequent circuits of two GFs 2, 3, 4 and amplifier 5) corresponds to the condition ΔF ≳ 10 / Δτ _and , the working pulse signal of duration τ _and at an intermediate frequency f _{c is} fed to the first stage of the GF , namely, the delay device (US) 2 'and the adder 4'. The ultrasound contains an ultrasonic delay line (USL) and a direct and inverse transducer of electrical vibrations into mechanical (ultrasonic) ones. Ultrasound, practically without distorting the signal, delays it for a time τ _k = τ _and , which satisfies condition (4). In this case, the delayed RF pulse arriving at the adder 4 'through the non-inverting amplifier 3' differs in initial phase from the phase of the end of the uncontrolled pulse by ψ = π.

. Изменения частоты Δf_c=100 Гц и частоты центра "ямы" на

весьма незначительны по отношению к полосе КФ ΔF_к, что практически не повлияет на эффективность - достижимую величину ОСШ. Величина 10^-5 легко достижима при кварцевой стабилизации частоты f_c (на всех этапах) и при применении в УЗ УЗЛ на кварце.The stability of the delay and frequency f _c determines the accuracy of combining the frequency of the signal with the center of the noise "well" of the SP noise (14). So, if τ _and = τ _k = 10 ^-6 s, then the width between the main ridges of this joint venture is ΔF _g = 1 / τ _k = 10 ⁶ Hz. When β = ΔF _to τ _and = 0.01, i.e., when ΔF _to = 10 ⁴ Hz, the stability and accuracy of the delay τ _to and frequencies f _c 10 ^-5 correspond

. Changes in the frequency Δf _c = 100 Hz and the frequency of the center of the "well" on

are very insignificant with respect to the CF band ΔF _k , which practically does not affect the efficiency — the attainable SNR value. The value of 10 ^{-5 is} easily achievable with quartz stabilization of the frequency f _c (at all stages) and when used in ultrasound ultrasonic scanning on quartz.

In the summing circuit, the non-phase inverting amplifier 3 'is designed to compensate for the loss of the delayed signal in the ultrasound, i.e. to equalize the noise intensities ξ (t) and ξ (t-τ _к ) during their summation, in order to correctly (with zero on f _c ) form the “hole” of the joint venture (14).

From the output of the adder 4 ', a pair of anticoherent pulses is fed to the decoupling amplifier 5, loaded on the second stage of the HF as part of devices 2 ", 3" and 4 ", identical to those of the first cascade of the HF. At the output of the adder 4", 33 anti-coherent pulses with amplitudes according to digital diagram (17) U _m , 2U _m , U _m , each of duration τ _and . This sequence of pulses arrives at amplifier 6 loaded on CF with a band ΔF _to ≪ 1 / τ _and . We only note that it is precisely in the narrow-band CF that the problem of obtaining a high SNR ρ ≫ ρ _t is solved precisely.

From the output of the CF stage 6, the signal with a sawtooth front of duration τ _and (with increased SNR) is fed to an amplifier 7 loaded on a band-pass filter (PF), which ensures coordinated operation with the detector 8. The passband of the PF in order to pass the front of the working signal of duration τ _and , elimination of the effect on the resulting CF band ΔF _k , additional suppression of the nearest main ridges of the SP noise (GF) at the input of the CF, which are ± 1/2 τ _and / (x = π / 2) away from f _c and with the aim of realizing the PF on LC- contours is ΔF _y ≲ 1 / τ _and .

. После окончания фронта рабочего сигнала (τ_и) из-за малого эквивалентного коэффициента затухания контура КФ 6 α=πΔF_к "хвост" свободных колебаний после прохождения через него рабочих антикогерентных импульсов будет относительно затянут по закону e^-αt. Из-за антикогерентности (скачков фазы на π на стыках между импульсами на входе КФ) и малого ослабления колебаний за интервалы t_и, 2τ_и и так далее после окончания всех γ+1 импульсов на входе КФ колебания "хвостов" будут в значительной степени скомпенсированы взаимно подавленными парциальными противофазными составляющими. Из цифровой диаграммы (17) видно, что сумма амплитуд с условной начальной фазой 0, определяемой первым (исходным) сигналом на выходе сумматора последнего g-го каскада ГФ, и сумма амплитуд сигналов с начальной фазой, отличающейся от первого сигнала на Jg, равны друг другу. Например, при g= 1 это соответствует 1=1; при g=2 1+1=2; при g3 1+3=3+1 и так далее. Физическая картина автокомпенсации (подавления) нежелательных "хвостов" свободных колебаний иллюстрируется на фиг. 3, где на примере первого каскада 1ф (g1) условно показаны огибающие входного (незадержанного) сигнала с обозначением условной начальной фазы "0" (фиг. 3,а), задержанного сигнала (фаза p, фиг. 3, б), их сумма (фиг. 3,в). С использованием принципа суперпозиции показаны огибающие выходных сигналов КФ с обозначенными начальными фазами "0" и "p" (фиг. 3,г,д) и их сумма (фиг. 3,е) на выходе КФ практически без "хвоста" в виде треугольной огибающей длительностью 2τ_и, т.е. почти как в обычном КФ с полосой ΔF_к ≈ 1/τ_и. В этом построении учтено то, что начальная фаза свободных колебаний частоты f_p после окончания входного сигнала частоты f_c= f_p для КФ типа LC-контура из-за его инерционности равна фазе колебаний конца входного импульса. При таком виде выходного сигнала КФ создаются обычные условия для оценки амплитуды и запаздывания сигнала, точность которой повышается с ростом ОСШ. Разумеется, при γ>1 форма сигналов усложняется; огибающая сигнала на выходе КФ строится аналогичным образом.Accordingly, the band of video chains of the load of the detector 8 is determined by the duration of the front

. After the end of the front of the working signal (τ _and ), due to the small equivalent coefficient of attenuation of the KF 6 circuit, α = πΔF _{to the} “tail” of free oscillations after passing through it of working anticoherent pulses will be relatively tightened according to the law e ^-αt . Due antikogerentnosti (phase jumps to π at the joints between the pulses at the input KF) and small attenuation of oscillations for intervals t _and, 2τ _and and so on after closure of all γ + 1 pulses inlet CF fluctuation "tails" will be largely compensated mutually suppressed partial antiphase components. It can be seen from the digital diagram (17) that the sum of the amplitudes with the conditional initial phase 0 determined by the first (initial) signal at the output of the adder of the last gth GF cascade, and the sum of the amplitudes of the signals with the initial phase different from the first signal by Jg are equal to to a friend. For example, with g = 1 this corresponds to 1 = 1; when g = 2 1 + 1 = 2; with g3 1 + 3 = 3 + 1 and so on. The physical picture of auto-compensation (suppression) of undesirable “tails” of free vibrations is illustrated in FIG. 3, where, as an example of the first stage 1ph (g1), the envelopes of the input (uncontrolled) signal are conventionally shown with the conditional initial phase “0” (Fig. 3a), delayed signal (phase p, Fig. 3b), their sum (Fig. 3, c). Using the principle of superposition, the envelopes of the output signals of the CF are shown with the initial phases "0" and "p" indicated (Fig. 3, d, e) and their sum (Fig. 3, f) at the output of the CF practically without a "tail" in the form of a triangular envelope of duration 2τ _and , i.e. almost as in a conventional CF with a band ΔF _to ≈ 1 / τ _and . In this construction, it is taken into account that the initial phase of free oscillations of frequency f _p after the end of the input signal of frequency f _c = f _p for a CF of the LC-circuit type due to its inertia is equal to the phase of oscillations of the end of the input pulse. With this type of output of the CF signal, the usual conditions are created for estimating the amplitude and delay of the signal, the accuracy of which increases with increasing SNR. Of course, for γ> 1, the waveform is complicated; the envelope of the signal at the output of the CF is constructed in a similar way.

Nevertheless, due to inaccuracies in the configuration of the KF circuit or its features (multi-circuit), this antiphase may in the strict sense be violated. Nevertheless, partial free vibrations in the CF begin at intervals of t _and , 2τ _, and so on, already therefore absolute suppression of the “tails” may not occur. Consequently, RSD can sharply worsen, especially “near” (strong) signals on the passage of “distant” (weaker) ones will especially affect. Those. measures must be taken to suppress the “tails” of free vibrations.

 From detector 8, the video signal is supplied to a decisive device (RU), tuned, for example, to the desired probability of false alarm by the Neumann-Pearson criterion.

 When γ = 1, there is one peak in the envelope of the input RU signal (Fig. 3f). Therefore, the RU reacts to the input (initial) signal according to a simple binary scheme: there is a signal, there is no signal. In the case of g> 1, there may be several such peaks (if measures are not taken to ensure reliable RSD; see below), which should be taken into account when determining the probabilities of false alarm and signal skipping.

From the output of RU 9, the standard pulse generated by it of the required amplitude and duration goes to the channel for processing the received information (target for specific communication systems) and to the driver shaper for “zeroing” the pulses 10. The latter generates control pulses of the desired polarity, amplitude and duration, which, using the devices zeroing "(controlled keys) connected by their low output resistances (on) to the circuits KF 6 and PF 7 damp free vibrations in them during the duration of the control pulses hs. After that, the implementation scheme of the signal detection method becomes ready again for processing the next input working signal (of duration t _and ). The time of "zeroing" in principle can vary τ _OB = (1 ... γ) τ _and . The simplest option is τ _OB = γτ _and , when the process of “zeroing” begins immediately after the first separate signal is detected at the output of the γ-th GF cascade and ends approximately at the end of the last separate signal, which follows from diagram (17). In this case, the RSD is of the order of T _RSD = (γ + 1) τ _and . Note that the “zeroing” mode can be switched off if necessary, which, when the input signals are not very different in level, generally saves the indicated RSD in the light of the auto-compensation of free oscillations of a narrow-band CF considered above. Note that instead of the usual envelope detector, one or another correlation signal processing can be used. In this case, narrow-band CF is used in the output video chains of the correlator.

 Thus, in this method, with a sharp increase in PU (SNR) RSD worsens slightly.

Using the proposed method for detecting pulsed signals will dramatically increase the noise immunity (PU) to fluctuation interference of any origin, including those acting simultaneously, and, of course, the sensitivity of the receiving equipment. In this case, the input noise can have a SP with an arbitrary frequency spectrum and have an arbitrary probability distribution (which, however, after a narrow-band CF still tends to normal if it was abnormal at the input). We especially note that the growth of controllers is achieved in one time cycle of the communication system, which leads to an increase in the throughput (PS) of the system, although if necessary for the same purpose of the growth of controllers, loop-by-loop signal processing (with loss of controllers) can be used. In addition to the growth of the PU to stationary fluctuation interference, with respect to which this method is mainly aimed (see above), the growth of the PU to harmonic and narrow-band interference (both "tuned" to the signal frequency f _c and very "upset"), due to the “pit” in the frequency response of the GF at f _c and the condition ΔF _to ≪ 1 / τ _and .

The growth of the PN also relates to a large extent to non-stationary broadband intermittent interference with a duration of at least τ _p ≳ 2τ _and , since the spectrum begins to form at the output of the DF-GF after the interval τ _and in accordance with the frequency response of the GF (with a “pit” at f _c ) having physical meaning and with one implementation of the interference. Moreover, such a noise should be ahead of the signal by an interval of at least τ _k = τ _and be conditionally stationary within τ _p for the formation of a noise “well” with “zero” in the frequency zone f _c . At the output of the GF cascades in the first interval with a duration of τ _k = τ _and at the same interval after the end of the intermittent interference, it passes with the same SP as at the input of the first GF (due to the peculiarities of the operation of the GF with its LZ “edge effect”) , i.e. without the effect of the “pit” in the frequency zone f _c .

 The growth of PU leads to an increase in the throughput of the communication system, which may be an important target in some applications. At the same time, the well-known Shannon limit on the transmission rate of binary information, limited by the passband (“ether” band) ΔF, average signal power, and noise power in this band, increases due to noise reduction.

 The indicated technical efficiency of the method also provides the improvement of such characteristics as mass-dimensional (reduction), operational (reduction of energy consumption, fuel), sanitary (environmental) due to the possibility of using lower emitting capacities to achieve the same effect in the communication line. All this reduces the cost of equipment and its operation.

 The proposed method and its justification are of great theoretical value. The unconventional approach found in the theory of OLF as applied to the main goal of OLF to detect pulsed signals develops the existing theory, putting into it not only mathematical justification, but also a larger physical content based on cause-effect processes in time, largely circumvented in the current theory.

 This particular method, together with other methods previously announced by the author, in our opinion, constitutes a definite scientific basis for the refined theory of OLF and detection of pulsed signals.

Claims (1)

A method for optimal detection of pulsed signals with an unmodulated carrier frequency, at which a signal is received, optimally or quasi-optimally filtered, an envelope is extracted, compared with a threshold level and, when the threshold level is exceeded, a signal is distinguished that provides an excess ΔF receiver bandwidth compared to quasi-optimal filtering compared to with the width of the main pulse signal spectrum _and duration τ, where ΔF ≥ 10 / τ, _and the received signal delayed by _a time τ = _{τ, and} when the moment loviya f _c • τ d _a = 1/2, where d is an integer; f _{c is the} carrier frequency of the signal, summarize the delayed and delayed signals, followed by filtering in a quasi-optimal filter, the passband ΔF _to which satisfies the condition ΔF _to ≪ 1 / τ _{and the} center frequency f _p f _c , while the delay and sum operations are repeated γ times where g>11> e

Images