JP2023040159A - Phase shift detector and phase shift detection method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable detection of a frequency and a phase shift amount using adaptive control with one detection circuit even though a nominal value of the frequency of an input signal is unknown.

SOLUTION: A phase shift detector includes phase shift calculation means which calculates a unit sample cosine value and a unit sample sine value on the basis of two discrete signals in the past that are at least consecutive to an input discrete signal, sequentially updates a first filter coefficient and a second filter coefficient by using an estimation error signal that is a difference between an output signal of a delay circuit and an estimation signal, calculates a phase difference between the input discrete signal and the output signal of the delay circuit means, on the basis of the first filter coefficient, the second filter coefficient, the unit sample cosine value, and the unit sample sine value, and calculates a phase shift amount on the basis of behaviors of a phase difference differential value between a phase difference in a certain time section and a phase difference in an immediately previous time section and of a convergence rate of an adaptive filter.

SELECTED DRAWING: Figure 1

COPYRIGHT: (C)2023,JPO&INPIT

Description

The present invention relates to a phase shift detection device and a phase shift detection method, and for example, to a phase shift detection device for detecting whether or not an input signal is a single frequency signal in real-time communication and measurement. applicable.

For example, when a communication device connected to a communication line detects an input signal, which is a single-frequency signal, through the communication line, it performs control processing according to the signal. In particular, communication devices that require real-time performance are required to detect single-frequency signals while suppressing processing delay time.

Conventionally, there are various methods for detecting whether an input signal is a single frequency signal.

For example, when the frequency of the input signal is known, in order to detect the frequency of the input signal in the time domain, a band pass filter (BPF: Band Pass Filter) that transmits a desired selected band signal by 0 dB is provided. The determination unit calculates the average power value POW_S of the selected band signal S and the average power value POW_N of the selected out-of-band signal N, and further calculates the signal-to-noise ratio SNR. and when the average power value POW_S of the selected band signal S is greater than or equal to a predefined threshold TH_POW_S and the signal-to-noise ratio SNR is greater than or equal to a predefined threshold TH_SNR, the input signal is a single frequency passable through the BPF. There is a method for determining that it is a signal.

Also, for example, when the frequency of the input signal is known, in order to detect the frequency of the input signal in the frequency domain, a discrete Fourier transform (DFT) is provided to set the input signal in the time domain in advance. Then, the DFT outputs the average power value POW_S of the frequency set in advance. The determination unit calculates the average power value POW_N for frequencies other than the preset frequency, and further calculates the signal-to-noise ratio SNR.
and when the average power value POW_S of the preset frequency is greater than or equal to a predefined threshold TH_POW_S, and the signal-to-noise ratio SNR is greater than or equal to a predefined threshold TH_SNR, the input signal is determined to be a single-frequency signal. There is a way to judge.

Furthermore, for example, in Patent Document 1, when the input signal is a discrete sine wave signal, a discrete sine wave signal that is a single frequency signal is proved from a trigonometric function formula, A method for detecting frequency is disclosed.

Conventionally, methods for detecting a phase shift of an input signal, which is a single-frequency signal, are disclosed in Patent Documents 2 to 4, for example.

JP-A-8-184618 Japanese Patent Publication No. 2000-504172 JP 2006-308809 A JP 2007-104721 A

However, the conventional method of detecting a single frequency signal can only determine the presence or absence of a signal with a frequency specified in advance, so there is a problem that it cannot be applied when the frequency of the input signal is unknown. .

When trying to detect the frequency of an input signal whose frequency is unknown using a conventional frequency detection method, it becomes necessary to arrange a plurality of BPFs, DFTs, etc., each having a different passband in parallel. There is a problem that it causes an increase in size, an increase in cost, and an increase in power consumption.

The conventional phase shift detection method also assumes that the nominal value of the frequency of the single-frequency signal that is the input signal is known, and is applied when the nominal value of the frequency of the single-frequency signal is unknown. I have a problem that I can't.

In this case as well, if a conventional phase shift detection method is used to detect a plurality of frequency candidates, a plurality of receiver circuits must be prepared in parallel, resulting in an increase in size and cost of the device. There is a problem that it causes an increase in power consumption and an increase in power consumption.

SUMMARY OF THE INVENTION Accordingly, the present invention has been made to solve the above problems. If so, the object is to provide an apparatus and method capable of detecting that frequency.

Also, the present invention provides a phase shift detection apparatus and method that can detect the frequency and phase shift amount by adaptive control with one detection circuit even if the nominal value of the frequency of the input signal is unknown. and

In order to solve this problem, the phase shift detector according to the first aspect of the present invention detects the amount of phase shift occurring in the input signal based on discrete signals obtained by sampling the input signal at a predetermined sampling period. In a phase shift detector for detecting, (1) for an input discrete signal, a unit sample cosine value and a unit sample sine value are calculated based on at least two consecutive past discrete signals. (2) delay circuit means for delaying the input discrete signal by a delay time; (3) using the output signal of the delay circuit means as a reference signal to generate an estimated signal of the output signal; (4) the first filter coefficient; calculating a phase difference between the input discrete signal and the output signal of the delay circuit means based on the second filter coefficient, the unit sample cosine value, and the unit sample sine value; phase shift calculation means for calculating a phase shift amount of the input signal based on behavior of a phase difference difference value between the phase difference and the phase difference in the immediately preceding time interval and a convergence rate of the adaptive filter. It is characterized by

A phase shift detection method according to a second aspect of the present invention is a phase shift detection method for detecting a phase shift amount generated in an input signal based on a discrete signal obtained by sampling the input signal at a predetermined sampling period, (1) A unit sample cosine/sine value calculating means calculates a unit sample cosine value and a unit sample sine value for an input discrete signal based on at least two consecutive past discrete signals; (2) delay circuit means for delaying the input discrete signal by a delay time; (3) adaptive filter for generating an estimated signal of the output signal using the output signal of the delay circuit means as a reference signal; sequentially updating the first filter coefficient and the second filter coefficient using an estimated error signal that is the difference between the signal and the estimated signal; calculating a phase difference between the input discrete signal and the output signal of the delay circuit means based on the second filter coefficient, the unit sample cosine value, and the unit sample sine value; The phase shift amount of the input signal is calculated based on the behavior of the phase difference difference value between the phase difference of and the phase difference of the immediately preceding time interval and the convergence rate of the adaptive filter.

According to the present invention, the presence or absence of an arbitrary single-frequency signal can be determined simply by preparing one detection circuit, and furthermore, if the input signal is a single-frequency signal, its frequency can be detected.
Further, according to the present invention, even if the nominal value of the frequency of the input signal is unknown, it is possible to detect the frequency and the phase shift amount by adaptive control with one detection circuit.

1 is a configuration diagram showing the configuration of a single frequency detection device according to a first embodiment; FIG. 1 is a configuration diagram showing the configuration of a conventional frequency detection device (No. 1); FIG. FIG. 2 is a configuration diagram showing the configuration of a conventional frequency detection device (No. 2); 1 is a configuration diagram showing a configuration when trying to detect the frequency of an input signal whose frequency is unknown with a conventional frequency detection device (No. 1); FIG. FIG. 10 is a configuration diagram showing a configuration when trying to detect the frequency of an input signal whose frequency is unknown with a conventional frequency detection device (No. 2); It is a state transition diagram when adopting the protection logic according to the first embodiment. FIG. 4 is a diagram showing the detection frequency and convergence rate of the single-frequency detection device when 10 types of signals are passed continuously in the first embodiment; FIG. 9 is a configuration diagram showing a configuration example of a first-order FIR filter in the second embodiment; FIG. 10 is a configuration diagram showing an architecture for realizing a method of calculating coefficients h ₀ and h ₁ in the second embodiment; FIG. 10 is a relational diagram showing the timings of the input signal x(n) of the adaptive digital filter and the output signal y(n) of the pure delay unit according to the second embodiment; FIG. 10 is a configuration diagram showing the configuration of a phase shift detector according to a second embodiment; FIG. FIG. 10 is an explanatory diagram illustrating a process of calculating a phase shift amount γ1 in a phase shift calculation circuit according to the second embodiment; FIG. 10 is a diagram showing an example of operating conditions of a phase shift detector according to the second embodiment; 14 shows an example of operation of the convergence rate CR(k) and the phase difference β(k) of the phase shift detector under the operating conditions of FIG. 13;

(A) First Embodiment Hereinafter, embodiments of a single frequency detection device and a single frequency detection method according to the present invention will be described in detail with reference to the drawings.

[Introduction]
First, three conventional single-frequency signal detection methods are illustrated.

The first method is to detect a single frequency signal in the time domain.

As shown in FIG. 2, a band pass filter (BPF) 110 is set in advance so that a desired selected band signal can be transmitted at 0 dB. Assuming that the selected band signal that can be transmitted by the BPF 110 at 0 dB is "S" and the selected out-of-band signal is "N", the input signal can be represented as "S+N" and the BPF output signal can be represented as "S".

The determiner 120 calculates the average power POW_S value of the selected band signal S and the average power POW_N value of the out-of-band signal N, and further calculates the signal-to-noise ratio SNR.

When the value of the average power POW_S of the selected band signal S is greater than or equal to a predefined threshold TH_POW_S and the signal-to-noise ratio SNR is greater than or equal to a predefined threshold TH_SNR, the input signal is a single frequency signal that can pass through the BPF. It can be determined that

This is concretely expressed by a formula.

Here, in equation (A-1-2), the operator <> represents time average, and the average value <SN> of the product SN of the single-frequency signal S and the other signal N is zero.

The second method is a single-frequency signal detection scheme in the frequency domain.

As shown in FIG. 3, a Discrete Fourier Transform (DFT) 210 transforms an input signal in the time domain into real and imaginary parts of preset complex frequency components. It is well known that the sum of the squares of the real part and the imaginary part is equivalent to the average power in the time domain according to Parseval's theorem. Therefore, the DFT 210 can output the value of the average power POW_S of the preset frequency. In other words, the DFT 210 corresponds to the BPF 110 in FIG.

Subsequent processing is the same as in the first method, and the determiner 220 determines whether the input signal is a desired single-frequency signal from the value of the average power POW_S of the signal S and the signal-to-noise ratio SNR. can be determined.

A third method is the method disclosed in Patent Document 1. A discrete sinusoidal signal sin(ωnT), which is a single-frequency signal, can be expressed by the recurrence formula (A-2). This can be proved from the trigonometric formulas.

Here, ω is the angular frequency of the signal under detection, T is the sampling period, and n is an integer representing the index of discrete time.

To simplify the expression, let x(n)=sin(ωnT) and substitute it into the equation (A-2).

Re-expressing the first term cos(ωT) on the right side of equation (A-3) yields equation (A-4).

Further, if the equation (A-4) is re-expressed with respect to the angular frequency ω using the arc cosine function, the equation (A-5) is obtained.

Here, since the angular frequency ω and the frequency f have a relationship of ω=2πf, the above equation can be rewritten with respect to the frequency f as Equation (A-6).

Note that when the denominator x(n-1) in { } of the arc cosine function of equation (A-6) is zero, the division result is undefined. To avoid this, if x(n−1)=0 then the numerator x(n)+x(n−2)=0 when x(n) is a discrete sine wave, and x Focusing on the fact that the ratio of x(n-1) and x(n)+x(n-2) is constant for any n if (n-1) ≠ 0, formula (A-6) becomes It can be modified as in formula (A-7).

In Patent Document 1, when x(n) is assumed to be a discrete sine wave, the frequency f can be calculated by equation (A-6), and the arc cosine function of equation (A-6) Even if the denominator x(n−1) in { } is zero, the adoption of the formula (A-7) is characterized in that the probability of avoiding division by zero increases. Note that in Patent Document 1, dividing by zero cannot be completely avoided only by using equation (A-7), so further processing is added to reduce the probability of encountering dividing by zero.

However, the conventional first and second methods for detecting a single frequency signal as described above can only determine the presence or absence of a signal with a pre-specified frequency, so if the frequency of the input signal is unknown, cannot be applied.

For example, if the first method is used to deal with the case where the frequency of the input signal is unknown, as illustrated in FIG. Is required.

In the example of FIG. 4, for example, m BPFs 110 are arranged in parallel. Therefore, the parallel arrangement of m BPFs results in an increase in device size, cost, and power consumption. .

Similarly, when the second method is used, it can be applied when the frequency of the input signal is specified in advance, but when the frequency of the input signal is unknown, it cannot be applied as it is. .

When the frequency of the input signal is unknown, using the second method, a plurality of DFTs 210 with different passbands are arranged in parallel, as illustrated in FIG.
FIG. 5 also illustrates the case where m DFTs 210 are arranged in parallel. Also in this case, the value of m must be increased as the frequency resolution required for the detection frequency becomes more detailed.

Therefore, the parallel arrangement of m DFTs 210 results in an increase in device size, cost, and power consumption. .

Note that if a fast Fourier transform (FFT) is used instead of the DFT 210, it is possible to suppress it to log ₂ m times instead of m times. However, even if FFT is adopted, the more the frequency resolution is refined, the more samples of the input signal are required for frequency conversion, resulting in an increase in processing delay time. In this regard, there is no difference between DFT and FFT, and this is a theoretically unavoidable problem. Since this increase in processing delay often becomes a fatal drawback in the fields of real-time communication and real-time measurement, it is common to adopt the time domain BPF method in such fields.

The third method is based on the formula (A-2) similarly to the basic principle of the first embodiment, but there is a decisive difference in its realization method. A detailed explanation of the difference will be given later.

The third method does not present a method for determining whether or not the input signal is a single-frequency signal. I don't get it. In contrast, the first embodiment adopts an adaptive filter, converges only when the input signal is a single frequency signal, and calculates the convergence rate quantitatively. , the decision of a single frequency becomes clear.

(A-1) Basic Principle A discrete sinusoidal signal sin(ωnT), which is a single-frequency signal, can be expressed by the following recurrence formula.

Here, ω is the angular frequency of the signal to be detected, T is the sampling period, and n is an integer indicating the discrete time index.

The above equation (B-1) expresses that the current sampled value of the discrete sinusoidal signal is determined from the past two sampled values. In other words, the above equation (B-1) represents the predictor.

In order to simplify the expression, the discrete sine wave signal of the above equation (B-1) is re-expressed as a general discrete signal x(n).

The above equations (B-2-1) and (B-2-2) can be interpreted in terms of static control and dynamic control (adaptive control).

[Perspective of static control]
The viewpoint of static control corresponds to the disclosure of Patent Document 1. In equation (B-2-1), the input signal x(n) is assumed to be an unknown sinusoidal signal. At this time, since h ₁ is treated as an unknown quantity, formula (B-2-1) can be re-expressed for h ₁ as formula (B-3).

The meaning of the expression (B-3) is that the input signal x(n) is used to calculate according to the above expression (B-3), and if _h1 converges to a certain value, the input signal x (n) is a single frequency signal (sine wave signal), and its frequency f is obtained from the following equation (B-4) from equation (B-2-2).

[Perspective of dynamic control]
The aspect of dynamic control is the aspect according to the first embodiment.

The right side of equation (B-2-1) represents a difference equation of a finite impulse response (FIR) filter, x(n) on the left side is the reference signal, and the right side is the reference signal x(n). It can be regarded as the prediction signal x̂(n). It should be noted that, due to restrictions on expression in the specification, it is expressed as "^" here.

That is, by rewriting formula (B-2-1) as follows, the intention will become clear.

In the above equation (B-5), among the three coefficients h ₀ , h ₁ and h ₂ on the right side, only h ₁ is the unknown (variable), and h ₀ and h ₂ are constants (h ₀ =0, h ₂ = -1).

When calculating _h1 , it is possible to use a learning identification method that is widely and commonly used in adaptive filters such as echo cancellers. The learning identification method is also called another name, Normalized Least Mean Square (NLMS).

(A-2) Configuration and Operation of First Embodiment FIG. 1 is a configuration diagram showing the configuration of a single frequency detection device according to the first embodiment. FIG. 1 is a block diagram showing how to implement the formula (B-5).

In FIG. 1 , the single frequency detection device 10 has a decision section 11 , a shift register (SFR) 12 , multipliers 13 and 14 and adders 15 and 16 .

A shift register (SFR) 12 shifts to the right each time an input signal x(n) is input, and produces three consecutive sample values x(n), x(n-1), x(n-2). to store That is, the shift register 12 stores two consecutive past discrete signals for the input signal x(n) (also referred to as an input discrete signal). The shift register (SFR) 12 outputs the signal x(n-1) to the multiplier 13 from the position where the previous first discrete signal x(n-1) is stored, and the first discrete signal x( The signal x(n-2) is output to the multiplier 14 from the location where the second discrete signal x(n-2) immediately before n-1) is stored.

The multiplier 13 multiplies the signal x(n−1) by the coefficient h ₁ and outputs the result to the adder 15 . The multiplier 14 multiplies the signal x(n−2) by “−1” and outputs the result to the adder 15 . It should be noted that it is not necessary to multiply x(n−2) by “−1” in implementation, and the adder 15 may simply perform subtraction processing.

The adder 15 outputs the sum of the output signal from the multiplier 13 and the output signal from the multiplier 14 to the adder 16 as a prediction signal x̂(n).

The adder 16 takes the difference between the prediction signal x̂(n) from the adder 15 and the input signal x(n), and outputs the difference value as a prediction error signal e(n).

The dashed line in FIG. 1 represents that the prediction error signal e(n) is fed back to update the applied filter coefficient _h1 .

The decision unit 11 monitors whether or not the adaptive filter coefficient _h1 converges to a certain value based on the prediction error signal e(n) output from the adder 16, and the adaptive filter coefficient _h1 is If it converges to a certain value, the input signal is determined to be a single frequency signal. Also, when the input signal is a single-frequency signal, the determination unit 11 calculates the frequency of the input signal that is a single-frequency signal. A detailed description of the method for calculating the frequency of the input signal will be given later.

[Operation of adaptive filter]
In FIG. 1, the prediction signal x̂(n) of the input signal x(n) is calculated according to the following equation (B-6).

A prediction error signal e(n) is obtained by the following equation (B-7).

Using this prediction error signal e(n), filter coefficient _h1 is updated according to the following equation (B-8).

Here, α is a constant called a step gain that determines the convergence speed in the above equation, and is selected within the range of 0<α<1. In addition, in the above equation, the reason _why the discrete time index n is added to the suffix of _h1 , which should be a constant, is written as "h1 _,n ". is treated as a variable with respect to n.

It should be noted that the coefficient updating process according to equation (B-8) is not always performed for each sample of the input signal x(n). This is because when the input signal x(n) is not applied or its level is very small, the second term on the right side of equation (B-8) may become indefinite due to division by zero. . In this case, h _1,n is not updated, and the previous value can be inherited as h _1,n =h _1,n−1 .

It should be noted that, in general, when a sine wave signal is input in the learning identification method, there are an infinite number of local minima, so these local minima are traversed over time, and as a result, the nonlinearity of the learning identification method increases. , in the worst case, oscillates and destroys the adaptive control.

However, according to the first embodiment, on the contrary, only when the input signal is a sinusoidal signal, based on the equation (B-2) that can guarantee that one solution exists, shown in FIG. Since it is implemented by configuration, it can be converged. This is the feature of the first embodiment.

[Input signal level check]
Next, before confirming that the input signal x(n) is a sinusoidal signal, it is necessary to confirm that the value of the power level of the input signal x(n) is above a certain threshold. This can be done by simply calculating the mean square value POW_X of the input signal x(n) according to the following equation and comparing the value of POW_X with the threshold.

Here, N is the number of samples included in the measurement interval of the mean square power, and is determined in advance. Normally, the average power value is normalized by dividing the right side of the above equation by N, but this is equivalent to multiplying the power value threshold by N times. Intentionally not normalized.

Letting TH_POW_X be the power threshold, when POW_X≧TH_POW_X, it is not known whether the input signal is a sinusoidal signal or not, but it is known that some signal is present.

Conversely, since it can be assumed that there is no input signal in a section (N sample sections) where POW_X<TH_POW_X holds, the coefficient updating process shown in Equation (B-8) is stopped for N samples.

[Convergence rate]
In order to determine whether or not the input signal x(n) is a sine wave signal, it is required that the mean square value POW_E of the prediction error signal e(n) expressed by the following equation (B-10) is sufficiently small. be.

However, since the mean square value POW_E of the prediction error signal e(n) is proportional to the mean square value POW_X of the input signal x(n), it is necessary to divide POW_E by POW_X for normalization. From the above, the convergence rate CR of the single frequency signal detection device 10 is defined by the following equation (B-11).

When the threshold value of the convergence rate CR is TH_CR, the determination unit 11 of the single-frequency signal detection device 10 determines that the input signal x(n) is a single-frequency signal when POW_X≧TH_POW_X and CR≦TH_CR. can judge.

[Average of detection frequency]
Next, when the existence of the single-frequency signal is confirmed, the determination section 11 calculates the frequency f of the input signal by the following procedure.

Here, the reason why the average value MEAN_h1 of N samples of h1 _,n is calculated is that noise is mixed in the input signal x(n). In this case, a noise component also appears in h1 _,n calculated by equation (B-8). Therefore, the determination unit 11 reduces the influence of the noise component by averaging the frequency of the input signal x(n).

Note that the arc cosine function appearing in the above equation may be replaced with a table lookup in implementation, but this is not the concern of the present invention.

[Protection logic]
The average input signal power value POW_X represented by equation (B-9), the convergence rate CR represented by equation (B-11), and the frequency f represented by equation (B-12) are each an average value of N samples. By appropriately selecting the value of N, it is possible to set from a short-term average to a long-term average.

Practically, however, it is more flexible to take N as a short-time average and to take concurrences of the respective determination results.

Therefore, as exemplified below, the determination unit 11 sets a first condition value (a condition flag c0 described later) based on the comparison result between the average input signal power value POW_X and the threshold TH_POW_X, the convergence rate CR and the threshold TH_CR. Calculates a third condition value (condition flag det, described later) that indicates a logical evaluation value with a second condition value (condition flag c1, described later) based on the comparison result of , and based on continuous matching of the third condition value to determine whether the single-frequency signal is undetected or detected.
A specific example will be described below.

Three bits c0, c1, and det are prepared as condition flags and defined as follows.
Flag c0: c0=1 if POW_X≧TH_POW_X, c0=0 if POW_X<TH_POW_X Flag c1: c1=1 if CR≦TH_CR, c1=0 if CR>TH_CR Flag det: det=c0&c1 (logical product)

Here, since POW_X and CR are calculated every N samples, the flags c0, c1 and det are also updated every N samples.

FIG. 6 is a state transition diagram when adopting the protection logic according to the first embodiment.

FIG. 6 exemplifies a case where protection logic of four stages of rear protection and two stages of front protection is employed in detecting a unit frequency signal. The number of stages of rear protection and front protection is not limited to this.

Backward protection is the protection logic from the single frequency signal undetected state (S2) to the single frequency signal detected state (S3) after a reset and an instruction to start detecting a single frequency signal (S1). For example, when the flag det1=1 in a certain sample, it is not immediately determined that a single-frequency signal is detected, but it is determined whether the flag det of the next successive sample is det=1. Then, if the flag det of the sample one step after is 1, it is determined whether the flag det of the sample two steps after is set to det=1. If the flag det=1, it is further determined whether the flag det of the sample after three consecutive stages is det=1. Thus, when det=1 in a certain sample, if det=1 continuously from the current sample to the sample three rows later, that is, if det=1 in four consecutive samples, then The single frequency signal detection state (S3) is determined. On the other hand, if det=0 in any sample from a certain sample to the sample after three consecutive stages, it is not determined as a state of single-frequency signal detection, and is set to a single-frequency signal undetected state.

Forward protection is the protection logic from the state of single frequency signal detection (S3) to the state of no single frequency signal detection (S2). For example, when the flag det1=0 in a certain sample, it is not immediately determined that the single frequency signal is not detected, but it is determined whether the flag det of the sample immediately preceding it is det=0. Then, if the flag det of the sample one stage before is 0, it is determined that the single-frequency signal has not been detected. In this way, when det=0 in a certain sample, if det=0 continuously in the sample one row before the current sample, that is, if det=0 in two successive samples, It is determined that the single frequency signal is not detected (S2). On the other hand, if det=1 in the sample one stage before, it is not determined that the single-frequency signal has not been detected.

[Example of operation]
In the following, an example of operation when various signals flowing on a fixed telephone line are received by the single frequency signal detector of the present invention will be presented.

Note that rear and front protection are intentionally omitted here to clarify the operation of the detector.

The input signals consist of 5 kinds of single-frequency signals S and 5 kinds of non-single-frequency signals NS, a total of 10 kinds, which are alternately arranged in the following order at intervals of 200 milliseconds.
S(1), NS(1), S(2), NS(2), S(3), NS(3), S(4), NS(4), S(5) NS(5)

Here, the types of each signal are shown below.
S(1): 2100 Hz single frequency signal, ITU-T V. Destination modem response signal NS(1) defined by IUT-T V.25: white noise S(2): 980 Hz single frequency signal, IUT-T V. 21 modem ch1 Mark
bit signal NS(2): voice S(3): 1650 Hz single frequency signal, ITU-T V. 21 modem ch2 Mark bit signal NS(3): hold tone S(4): 2225 Hz single frequency signal, Bell 103A modem mark bit signal NS(4): PB signal, '1', 697 Hz + 1209 Hz S(5): 1300 Hz Single frequency signal, IUT-T V. Mark bit signal NS(5) of V.23 modem: multiple line spectrum signal, ITU-T V.3. 22 modem Mark bit signal

FIG. 7 is a diagram showing the detection frequency and the convergence rate of the single-frequency detection device 1 when the 10 types of signals described above are passed continuously.

Here the rear/front protection has been removed so that the internal operation of the single frequency detector 1 can be observed.

In FIG. 7, when the single frequency signal is used, the convergence rate is close to zero, and the detection frequency at that time reflects that of the single frequency signal.

On the other hand, non-single-frequency signals have a higher convergence rate and are clearly not single-frequency signals.

Note that the convergence rate may be small in the case of voice signals. This is why forward protection and rear protection are useful.

(A-3) Effect of the First Embodiment As described above, according to the first embodiment, it is possible to detect the presence or absence of any single-frequency signal simply by preparing one detector. , is a single frequency signal, its frequency can be detected.

(B) Second Embodiment Next, embodiments of the phase shift detection device and the phase shift detection method of the present invention will be described in detail with reference to the drawings.

[Introduction]
First, prior to describing the phase shift detection device and phase shift detection method according to the second embodiment, conventional phase shift detection methods will be described with reference to Patent Documents 2 to 4. FIG.

In US Pat. Nos. 5,300,002 and 5,100,000, the nominal value of the frequency of the single frequency signal is known in advance.
A local oscillator centered at the nominal frequency is provided which is synchronizable to the frequency and phase of the input signal within a small tolerance and produces sine and cosine signals synchronized to the input signal. can be generated. Then, the product of the input signal and the sine wave signal and the product of the input signal and the cosine wave signal are obtained and added through a low-pass filter to detect the amount of phase shift applied to the input signal.

Also in US Pat. No. 6,200,000, the nominal value of the frequency of the single-frequency signal is known in advance. The phase reversal detector performs a discrete Fourier transform on the input signal and outputs nominal frequency level and phase information. Then, the phase shift amount is calculated from the difference between the current phase information and the past phase information.

As mentioned above, the conventional assumption is that the nominal value of the frequency of the single-frequency input signal is known. Therefore, when the frequency of the single frequency signal is unknown, there is a problem that the conventional phase shift detector cannot be applied.

When trying to derive the phase shift amount of a single frequency signal whose frequency is unknown using a conventional phase shift detection device, when there are multiple types of frequency candidates to be detected, the number corresponding to the number of types Since it is necessary to arrange a plurality of receiving circuits in parallel, the scale of the apparatus becomes large, and there arises a problem of an increase in cost and an increase in power consumption.

This embodiment has been made to solve the above-mentioned problems, and even if the nominal value of the frequency of the input signal is unknown, the frequency and phase shift amount can be detected by adaptive control in one receiving circuit. can do.

(B-1) Basic Principle FIG. 8 is a configuration diagram showing a configuration example of a primary FIR (Finite Impulse Response) filter in the second embodiment.

In FIG. 8, consider the output signal y(n) when a discrete sinusoidal signal sin(ωnT) is applied to the first-order FIR filter 20 . As shown in FIG. 8, the first-order FIR filter 20 has a shift register 21, two multipliers 22 and 23, and one adder 24.

In FIG. 8, the shift register (SFR) 21 shifts the data signal to the right every time the discrete sine wave signal sin(ωnT), which is the input signal, is applied. Shift register 21 stores signal x(n), which is the current sample value, and signal x(n−1), which is the previous sample value, and outputs signal x(n) to multiplier 22 . , the signal x(n−1) is output to the multiplier 23 .

The multiplier 22 multiplies the signal x(n) output from the shift register 21 by h0 and outputs the result to the adder ₂₄ , and the multiplier 23 outputs the signal x(n-1) output from the shift register 21. is multiplied by _h1 and output to the adder 24 .

Adder 24 outputs a signal obtained by adding the output signal of multiplier 22 and the output signal of multiplier 23 as output signal y(n).

Therefore, the output signal y(n) can be represented by equation (C-1).

Here, as in the following formulas (C-2-1) and (C-2-2), p=h ₀ +h ₁ ·cos(ωT) and q=−h ₁ ·sin(ωT) are set, and the formula By substituting into (C-1) and re-expressing, the formula (C-3) is obtained.

Equation (C-3) indicates that when a sinusoidal signal sin(ωnT) is applied to the first-order FIR filter 20, the filter coefficients h ₀ and h ₁ can produce a sinusoidal signal sin(ωnT+β) having an arbitrary phase β. is shown.

[Calculation of phase difference β]
Next, a method for automatically calculating this phase difference when sin(ωnT) and sin(ωnT+β) are given will be described.

When calculating the phase difference β, it is necessary to calculate p and q appearing in equation (C-4). Furthermore, when calculating p and q, it is necessary to calculate h ₀ and h ₁ and cos (ωT) and sin (ωT) shown in formulas (C-2-1) and (C-2-2). There is Therefore, a method for calculating h ₀ and h ₁ , and cos(ωT) and sin(ωT) will be described below.

[How to calculate cosωT and sinωT]
A discrete sine wave signal sin(ωnT) can be expressed by the difference equation of the following equation (C-5).

From this, cos(ωT) and sin(ωT) are obtained by equations (C-6-1) and (C-6-2).

According to the above equation (C-6-2), sin(ωT) takes a value of 0 or more, which is guaranteed by the following fact. Assuming that the frequency of the discrete sine wave signal sin(ωnT) is F _C and the sampling frequency is F _S (=1/T), F _C is guaranteed to be less than F _S /2 according to the sampling theorem. Then, since ωT=2πF _C T=2π(F _C /F _S )<π, sin(ωT)≧0 is guaranteed.

Expressions (C-6-1) and (C-6-2) are x(n), x(n-1)≠0, x(n- 2) shows that cos(ωT) and sin(ωT) are calculated from three consecutive sample values.

Zero division avoidance by x(n-1) appearing in the denominator on the right side of equation (C-6-1) can be realized by setting a lower threshold for the absolute value of x(n-1).

Furthermore, it must be ensured that the input signal x(n) is _a sinusoidal signal, which is a sinusoidal signal from the convergence rate of the adaptive filter used in the calculation of h0 and _h1 described below. It can be determined whether there is

As another solution, the single frequency detection device 1 described in the first embodiment determines whether or not the input signal is a single frequency signal, and if it is a single frequency signal, the frequency can also be detected, the single frequency detector 1 of the first embodiment may be used.

[How to calculate _h0 and _h1 ]
For example, when an input signal sin(ωnT) and a reference signal sin(ωnT+β) are applied to an adaptive digital filter (ADF) that employs a learning identification method (normalized least squares method), the adaptive digital filter After convergence, its coefficients h ₀ and h ₁ are obtained.

FIG. 9 illustrates an architecture that implements this h ₀ and h ₁ calculation method. The architecture 30 shown in FIG. 9 has an adaptive digital filter 31 employing a learning identification method (normalized least squares method), a pure delay device 32 and an adder 33 .

The pure delay unit 32 receives sin(ωnT) as an input signal, delays sin(ωnT) by time Td, and outputs the delayed signal. The fixed time (delay time) Td given to the input signal by the pure delay device 32 corresponds to the phase difference β=-δ, and the signal output from the pure delay device 32 is the output signal y(n).

The adaptive digital filter 31 is based on the first-order FIR filter 20 illustrated in FIG. 8, and has a coefficient update function based on the learning identification method. That is, the adaptive digital filter 31 updates the coefficients based on the learning identification method to estimate the output signal y(n) output from the pure delay device 32 .

The adder 33 takes the difference between the signal y(n) output from the pure delay unit 32 and the signal y^(n) output from the adaptive digital filter 31 .

Here, x(n) is an input signal and is a discrete sinusoidal signal sin(ωnT).
Let y(n) be a signal delayed by a fixed time Td with respect to x(n).
Let y^(n) be the estimated signal of y(n).
Let e(n) be the estimated error signal between y(n) and y^(n).

Then, the coefficients h ₀ and h ₁ can be calculated by the learning identification method as follows.

Here, α is called a step gain, which is a parameter that determines the convergence speed and convergence accuracy of the adaptive digital filter 31, and is a value in the range of 0<α<1. The value of the step gain α can be preselected.

The convergence judgment of the adaptive digital filter 31 can be made based on the value of the power ratio between the input signal x(n) and the estimated error signal e(n). Therefore, the convergence rate CR of the adaptive digital filter 31 is defined by the following equation (C-10).

Here, the number of samples N used to determine the convergence rate CR should be appropriately selected according to implementation. When the input signal x(n) is a sine wave signal and no phase shift occurs in x(n), the value of CR becomes a very small value. , x(n) are sinusoidal signals can be determined by monitoring the value of CR. When CR is near zero value, the coefficients _h0 and _h1 can be considered converged.

[Method for detecting γ when phase shift γ is applied to sine wave signal]
Now suppose that the discrete sinusoidal signal x(n)=sin(ωnT) is given a phase shift γ1 at some point, so that x(n)=sin(ωnT+γ1). Assume that at the next instant, a further phase shift γ2 is applied such that x(n)=sin(ωnT+γ1+γ2). Hereinafter, it is assumed that the phase shift is applied in the same manner. However, the timing of this phase shift is not limited to a constant period, and the amount of phase shift is not limited to be constant.

At this time, the architecture 30 shown in FIG. 9 can detect these phase shift amounts γ1, γ2, γ3, . This will be explained.

FIG. 10 is a relational diagram showing timings of the input signal x(n) of the adaptive digital filter and the output signal y(n) of the pure delay unit according to the second embodiment.

In FIG. 10, the phase differences β detected in sections 1, 2, 3 and 4 are -δ, -δ-γ1, -δ and -δ-γ2, respectively. Here, -δ is an amount determined by the pure delay prepared in the phase shift detector and is seen as an offset. 4 can detect −γ2. That is, it is possible to detect the amount of phase shift applied to the input signal x(n). This is the reason why the fixed time (delay time) Td of the pure delay unit 32 is provided in FIG. It should be noted that Td must be set to a value greater than the time required for the adaptive digital filter 31 to converge.

The above is the basic principle of the phase shift detection device and the phase shift detection method according to the second embodiment.

(B-2) Configuration and Operation of Embodiment Below, a configuration example of a phase shift detector for realizing the basic principle of the above-described second embodiment will be shown, and the operation of the phase shift detector will be described. .

FIG. 11 is a configuration diagram showing the configuration of a phase shift detector according to the second embodiment.

11, a phase shift detector 40 according to the second embodiment has a delay circuit 41, an adaptive filter 42, a phase shift calculation circuit 43, and a unit sample cosine/sine value calculation circuit 44. FIG.

[input signal]
Description will be made assuming that the input signal x(n) is a discrete sinusoidal signal represented by the following equation (D-1).

Here, ω is the angular frequency, n is the discrete time index (integer value), and T is the sampling period of the discrete signal x(n).

Further, the input signal x(n) is subjected to phase shifts (in units of radians) γ1, γ2, γ3, . . . at times n1, n2, n3, . This is represented by formula (D-2).

Examples of such signals include phase shift keying (PSK) modem signals and ITU-T V. There is a 2100 Hz sinusoidal signal with a phase reversal of 450 ms period specified in 25.

Phase shift detector 40 can detect the phase shifts γ1, γ2, γ3, . The operation will be described below.

[Delay circuit]
The delay circuit 41 has a function of delaying the input signal x(n) by M samples. A shift register, FIFO, memory, or the like can be used to implement the delay circuit 41 . The following equation shows the relationship between the input signal x(n) of the delay circuit 41 and y(n) when the input signal x(n) is a discrete sine wave signal.

The meaning of equation (D-3) is that y(n) is a signal obtained by delaying x(n) by M samples, so when a phase shift occurs in x(n) at a certain time n=n1, It is n=n1+M after M samples that is reflected in y(n). In other words, in this M-sample interval, only x(n) has a phase shift, and y(n) does not reflect the phase shift. provides an opportunity to calculate This is the purpose of delay circuit 41 .

Assuming that the time required for this phase shift calculation is _Tlearn , it is necessary to select the value of M in advance so as to satisfy MT> _Tlearn (D-4).

[Adaptive filter]
The adaptive filter 42 performs the FIR filter operation represented by equation (D-5) on the input signal x(n) to generate an estimated value y(n) of y(n).

Here, since the adaptive filter coefficients _h0 and _h1 are not signals related to discrete time n but are time-varying parameters, the addition of suffix n expresses the temporal relationship explicitly.

The following coefficient updating process is performed so that the estimated value ŷ(n) represented by Equation (D-5) can follow the true value y(n).

The estimated error signal e(n) between the true value y(n) and its estimated value ŷ(n) is obtained by the following equation (D-6).

This estimated error signal e(n) is fed back to update the adaptive filter coefficients h ₀ and h ₁ according to the following equations (D-7-1) to (D-7-3).

Here, α is called a step gain and is a parameter that determines the convergence speed and convergence accuracy of the adaptive filter. If the convergence speed is important, the value of the step gain α should be closer to 1, and if the convergence accuracy is important, the value of the step gain α should be closer to 0.

When the input signal is a sine wave signal, the estimated value y(n) produced by the adaptive filter 42 converges to the true value y(n) very quickly. The convergence rate CR of this adaptive filter 42 is defined by the ratio of the average power value POW_x of the input signal x(n) and the average power value POW_e of the estimated error signal e(n).

Upon implementation, a threshold value TH_CR for the convergence rate CR may be determined and determined.

By the way, the adaptive filter coefficients h ₀ and h ₁ calculated by the formulas (D-7-2) and (D-7-3) are provided to the phase shift calculation circuit 43, but the input signal x(n ) is superimposed with extraneous noise and coding quantization noise, the values of h ₀ and h ₁ fluctuate accordingly. Therefore, the average values MEAN_h0 and MEAN_h1 of _h0 and _h1 can be calculated to reduce the influence of noise. Here, a calculation example of the average values MEAN_h0 and MEAN_h1 is shown.

[Unit sample cosine/sine value calculation circuit]
Here, cos(ωT) is called a unit sample cosine value, and sin(ωT) is called a unit sample sine value.

When the input signal x(n) is a discrete sine wave signal sin(ωnT) represented by the formula (D-1), the unit sample cosine value cos(ωT) is expressed by the formula of the trigonometric function as the formula (D-10) can be obtained with

By the way, in general, the input signal x(n) is superimposed with extraneous noise or quantization noise of coding. 2) It is dangerous to take the cos(ωT) calculated using the formula (D-10) from only 2) as a true value. be able to. Therefore, the unit sample cosine/sine value calculation circuit 44 can calculate MEAN_cwt by the following equation (D-11), for example.

The above formula exemplifies the case of integrating in the range of 0≦j≦N−1 and dividing by N. For a certain j, when the denominator x(n−1−j) is zero, Division by zero may occur and the calculated value may become indefinite. Therefore, before dividing, check the value of x(n-1-j), and if the value of x(n-1-j) is near zero, take measures not to add it to the averaging process of the above formula. good.

The average value MEAN_swt of the unit sample sine values sin(ωT) is obtained by the following equation.

[Phase shift calculation circuit]
The phase shift calculation circuit 43 calculates the average power value POW_x of the input signal from the adaptive filter 42, the average values MEAN_h0 and MEAN_h1 of the coefficients _h0 and _h1 , the convergence rate CR, and the average unit from the unit sample cosine/sine value calculation circuit 44. A sample cosine value MEAN_cwt and an average unit sample sine value MEAN_swt are each acquired once every N sample periods, and based on this information, the phase shift occurring in the input signal x(n) is detected.

For the threshold value of the convergence rate CR of the adaptive filter 42, two types TH_CRH and TH_CRL are prepared to have hysteresis, and the relationship TH_CRH>TH_CRL is assumed.

When the learning of the adaptive filter 42 is completed, that is, when y(n)≈ŷ(n), CR<TH_CRL, and when a phase shift occurs in x(n) or y(n), y( Since n)≠ŷ(n), the amplitude of the estimated error signal e(n) increases, that is, the convergence rate CR deteriorates and its value increases. The adaptive filter 42 appropriately changes the values of the coefficients h0 and _h1 in order to restore y( _n )≈ŷ(n), and finally y(n)≈ŷ(n) again. Return to CR<TH_CRL.

A phase difference β between the input signal x(n) and the output signal y(n) of the delay circuit 41 is calculated by the following procedure.

Here, since the phase difference β is calculated every N samples, a new time index k is introduced with N samples as one unit time, and is denoted as β(k).

β(k) is β(k)=− ωMT.

When a phase shift γ1 occurs in x(n) and x(n)=sin(ωnT+γ1), β(k)=−ωMT−γ1. Here, since -ωMT radians corresponding to the delay time of the delay circuit 41 are superimposed on β(k) as an offset, the difference of β(k) itself is taken to cancel this. This new variable φ(k) is defined by the following equation.

Here, L has N samples as one unit, like the index k, and is set in advance to a value longer than at least the learning time of the adaptive filter 42 . Thus, φ(k)=0 when no phase shift occurs in x(n), and φ(k)=γ1 when a phase shift γ1 occurs in x(n). .

Now that the preparation for the description has been completed, the process of calculating the phase shift amount γ1 when the phase shift γ1 occurs in the input signal x(n) at a certain point will be described with reference to FIG.

In FIG. 12, when the phase shift γ1 occurs in the input signal x(n) at the time index k, the learning of the adaptive filter 42 increases the convergence rate CR to exceed the upper threshold TH_CRH. Assuming that the time required for learning is N samples or less, the lower limit threshold TH_CRL is not yet exceeded at the next time index k+1, and CR<TH_CRL is satisfied at time index k+2. The value of the phase difference differential φ(k+2) at this time is γ1.

After the phase shift γ1 in x(n) at time index k, a phase shift appears in y(n) M samples later, eg at time index k+3 in FIG. is detected at the time index k+5.

It should be ignored. FIG. 12 indicates that up to the time index k+7 should be ignored. That is, after detecting γ1 at time index k+2, the N×L+M sample interval should be ignored. Specific values are implementation-dependent.

By the way, when the input signal x(n) is a single frequency signal and it is necessary to know its frequency, the frequency f can be calculated from MEAN_cwt according to the following equation when the convergence rate CR<TH_CRL.

[Example of operation]
Next, an operation example of the phase shift detector 40 according to the second embodiment will be described.

FIG. 13 is a diagram showing an example of operating conditions of the phase shift detector 40 according to the second embodiment.

FIG. 14 shows an operation example of the convergence rate CR(k) and the phase difference β(k) of the phase shift detector 40 under the operating conditions of FIG.

As shown in FIG. 13, the operating conditions were that the input signal x(n) was a sine wave signal of 2079 Hz, and a phase shift of 205 [deg] was applied after 200 ms. The sampling frequency was 8 [kHz], the delay time was M=128 [samples] (=16 ms), and the calculation cycle of the convergence rate CR and the phase difference β was N=40 [samples] (=5 ms).

As shown in FIG. 14, the value range of the phase difference β(k) is limited to the range of 0 [deg]≦β(k)<360 [deg] for convenience.

Substituting the theoretical value -ωMT for the phase difference before the phase shift, -ωMT=-2π·2079·(128/8000)
=-209 [rad]
=-95[deg]
= 265 [deg]. In FIG. 14, the phase difference β(k) before the phase shift indicates a value around 265 [deg].

Next, when a phase shift of 205 [deg] occurs in x(n), the theoretical value is 265-205=60 [deg] because it is closer by 205 [deg] than the previous phase difference. In FIG. 14, the phase difference β(k) shows values around 60 [deg].

Furthermore, as shown in FIG. 14, when a phase shift occurs in x(n) and y(n), the value of the convergence rate CR rises significantly, and the timing of phase shift occurrence can be known.

(B-3) Effects of the Second Embodiment As described above, according to the second embodiment, even if the nominal value of the frequency of the input signal is unknown, one receiving circuit can adjust the frequency by adaptive control. and the amount of phase shift can be detected.

(C) Third Embodiment Next, another embodiment of the phase shift detection device and phase shift detection method according to the present invention will be described.

In the second embodiment described above, the average value MEAN_cwt of the unit sample cosine values of the input signal is generated from the input signal x(n).

On the other hand, in the third embodiment, a method of calculating an average value MEAN_cwt of unit sample cosine values from average values MEAN_h0 and MEAN_h1 of adaptive filter coefficients is shown.

The amplitude (p ² +q ² ) ^1/2 of the sinusoidal signal y(n) represented by the equation (C-3) shown in "(B-1) Basic principle" of the second embodiment is, for example, " 1” can be set. This corresponds to the case where the gain of the pure delay device 32 in FIG. 9 is "1".

Then, the following formula (E-1), the formula (C-2-1) and the formula (C-2-2) become the following formula (E-2-1) and changing the expression, the formula (E-2 -2).

By substituting the mean values MEAN_h0 and MEAN_h1 of _h0 and _h1 into _h0 and _h1 of the formula (E-2-2) respectively, the mean of unit sample cosine values is obtained as in the following formula (E-3): The value MEAN_cwt can be obtained.

However, in the above equation (E-3), when MEAN_h0 or MEAN_h1 is a zero value, the above equation becomes indefinite and MEAN_cwt cannot be calculated. Therefore, this constraint is shown concretely.

[When MEAN_h0=0]

Here, f is the frequency of the input signal x(n), M is the number of delay taps of the delay circuit 41 in FIG. is the sampling frequency.

[When MEAN_h1=0]

Here, f is the frequency of the input signal x(n), M is the number of delay taps of the delay circuit in FIG. 11, k is an integer greater than or equal to 0 and less than M/2, and fs (=1/T) is the sampling frequency. .

The third embodiment cannot be applied when the frequency f of the input signal x(n) satisfies the equations (E-4-3) and (E-5-3). However, if the nominal value of the frequency of the input signal is known, it can be applied by choosing the value of M such that equations (E-4-3) and (E-5-3) do not apply.

Note that the third embodiment is obtained by modifying the method of calculating the unit sample cosine value of the second embodiment as described above, and other than that, it is the same as the second embodiment. are omitted.

As described above, according to the third embodiment, the values of coefficients _h0 and _h1 (values of mean values MEAN_h0 and MEAN_h1) are used to calculate a unit sample cosine value (value of mean value MEAN_cwt). can be done. Even in this case, the same effects as in the second embodiment are obtained.

(D) Other Embodiments For example, ITU-T G. The tone disabler specified in H.168 is required to have the function of detecting the phase inversion of the 2100 Hz response signal sent by the called modem at the start of communication. A permissible deviation of 1900 [Hz] to 2350 [Hz] is specified for the nominal value of the carrier frequency of 2100 [Hz], and a permissible deviation of 110 [deg] for the nominal value of the phase inversion angle of 180 [deg]. ~250 [deg] is defined. This allowable deviation is so wide that it can be said to be unusual for a communication standard, but according to the second and third embodiments, even this allowable deviation can be detected without any problem.

10... Single frequency detector, 11... Determination unit, 12... Shift register (SFR), 13 and 14... Multipliers, 15 and 16... Adders,
40... Phase shift detector, 41... Delay circuit, 42... Adaptive filter, 43... Phase shift calculation circuit, 44... Unit sample cosine/sine value calculation circuit.

Claims (6)

A phase shift detector that detects the amount of phase shift occurring in an input signal based on a discrete signal obtained by sampling the input signal at a predetermined sampling period,
unit sample cosine/sine value calculation means for calculating a unit sample cosine value and a unit sample sine value for an input discrete signal based on at least two consecutive past discrete signals;
delay circuit means for delaying the input discrete signal by a delay time;
Using the output signal of the delay circuit means as a reference signal, generating an estimated signal of the output signal, and using an estimated error signal, which is the difference between the output signal and the estimated signal, to obtain the first filter coefficient and the second filter coefficient. an adaptive filter that sequentially updates filter coefficients;
phase difference between the input discrete signal and the output signal of the delay circuit means based on the first filter coefficient, the second filter coefficient, the unit sample cosine value, and the unit sample sine value; is calculated, and the phase shift amount of the input signal is calculated based on the behavior of the phase difference difference value between the phase difference in a certain time interval and the phase difference in the immediately preceding time interval and the convergence rate of the adaptive filter. and phase shift calculating means for calculating a phase shift. The delay circuit means delays the input discrete signal by a predetermined number of samples. 2. The phase shift detection device according to claim 1, wherein the number of samples is set. 2. The phase shift detector according to claim 1, wherein said adaptive filter averages said first filter coefficient and said second filter coefficient in a predetermined sampling interval. 2. The unit sample cosine/sine value calculating means according to claim 1, wherein the unit sample cosine value and the unit sample sine value are each averaged over a predetermined sample interval to calculate an average value. Phase shift detector. The phase shift calculation means is
determining the occurrence of the phase shift by comparing the convergence rate of the adaptive filter, a first threshold, and a second threshold larger than the first threshold;
2. The phase shift detection device according to claim 1, wherein the phase difference differential value is determined in a time index interval longer than the learning time of the adaptive filter. In a phase shift detection method for detecting a phase shift amount generated in an input signal based on a discrete signal obtained by sampling the input signal at a predetermined sampling period,
A unit sample cosine/sine value calculating means calculates a unit sample cosine value and a unit sample sine value for the input discrete signal based on at least two consecutive past discrete signals,
delay circuit means for delaying the input discrete signal with a delay time;
An adaptive filter uses the output signal of the delay circuit means as a reference signal, generates an estimated signal of the output signal, and uses an estimated error signal, which is the difference between the output signal and the estimated signal, to obtain a first filter coefficient. and the second filter coefficients are sequentially updated,
Phase shift calculating means calculates the input discrete signal and the delay circuit means based on the first filter coefficient, the second filter coefficient, the unit sample cosine value, and the unit sample sine value. calculating the phase difference from the output signal, and calculating the input based on the behavior of the phase difference difference value between the phase difference in a certain time interval and the phase difference in the immediately preceding time interval and the convergence rate of the adaptive filter; A phase shift detection method, comprising calculating a phase shift amount of a signal.

Images