Stability measurement control method

By monitoring transfer characteristics and adjusting gain and phase margins using phase rotation and bandpass filters, the stability of closed-loop control systems for environmental sounds is managed, addressing the challenges of stability and ease of use in existing technologies, achieving effective noise rejection and howling suppression.

WO2024236831A1PCT designated stage expired Publication Date: 2024-11-21KAKUMOTO JUNICHI

Patent Information

Application Number
PCT/JP2023/032785
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-17
Filing Date
2023-09-08
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

Existing control systems for environmental sounds, such as active noise cancellers and howling suppressors, face challenges in stability and ease of use due to complex acoustic environments and the trade-off between performance and howling generation, making it difficult for engineers to achieve satisfactory noise rejection ratios without advanced theoretical knowledge.

Method used

A method for measuring and controlling the stability of closed-loop control systems by monitoring transfer characteristics and adjusting gain and phase margins using phase rotation and bandpass filters, allowing for intuitive programming and maintenance without requiring advanced adaptive control theory.

Benefits of technology

This approach enables the construction of control systems with simple algorithms that directly correspond to actual operations, providing stable performance and increased control freedom while avoiding self-oscillation and noise generation, thus effectively managing howling and noise rejection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2023032785_21112024_PF_FP_ABST
    Figure JP2023032785_21112024_PF_FP_ABST
Patent Text Reader

Abstract

[Problem] To meet the need for a control method for measuring and controlling the stability of a control system having a closed loop including acoustic coupling during the actual operation of the system, the control method requiring no advanced theoretical knowledge, having an intuitive algorithm, enabling simple programming and easy maintenance, and making it possible to ascertain the state of the closed loop by monitoring signals of the system during actual operation. [Solution] A closed loop of a control system is divided into a plurality of bands. Gains and phases of the respective bands are changed to such a degree that a control object is not affected. A signal of a location at which the operation state can be monitored within the closed loop of the control system is detected. The stability of the system is identified by deriving a change applied to the closed loop gain and a change in the signal at a monitoring point. The gains and phases are adjusted so that a desired state is reached.
Need to check novelty before this filing date? Find Prior Art

Description

Stability measurement and control method

[0001] Adaptive filter Robustness of adaptive control Observer-based estimation of state variables Environmental noise control Filter Impulse response Simple adaptive control (SAC)

[0002] Howling suppression, environmental noise canceller, public address system, PA system

[0003] The terms and symbols defined in the claims shall also be used in the specification. Considerations in the development process of this proposal: Part 1. The design of environmental sound control systems, such as active noise cancellers, feedback cancellers, and open-speaker hearing aids, presents a high barrier to entry for acoustic engineers. Adaptive control theory is difficult for average engineers to understand, and even commercially available application programs often fail to achieve satisfactory performance. Furthermore, when problems arise, they are practically infeasible to address. Meanwhile, there is a strong, both latent and apparent, need for this technology in the production market, and there is a need for even more stable, high-performance, and easy-to-integrate systems.

[0004] Considerations from the development process of this proposal, part 2. The biggest problem that occurs during actual operation when controlling environmental noise is feedback. As long as the control system is a closed loop, this problem always exists, with a delicate trade-off between improving performance and the issue of feedback. To alleviate this trade-off, a robust control method that is less susceptible to the influence of the complexity of the acoustic environment, as represented by keywords such as three-dimensional acoustic space, sound wave propagation and reflection (which is energy), multiple resonance, group delay, and nonlinearity, is required. While numerous documents explain simple control methods, they still pose a high hurdle for average engineers. Creating practical software, let alone an algorithm, is nearly impossible.

[0005] Consideration during the development process of this proposal, part 3. As an example, in the case of an active road noise canceller for passenger cars, the highest performance model installed in commercially available passenger cars has a maximum of 10dB around 100Hz and a maximum of about 3dB around 1kHz. At 10dB, the reduction difference is noticeable, but at 3dB it is at the limit of whether it is noticeable or not. For example, if users require a noise rejection ratio, a noise rejection ratio of at least 10dB is required for a small passenger car, and at least 20dB for a light truck, in the band around 1kHz.

[0006] Consideration during the development process of this proposal, No. 4. Closed loops that include normal acoustic coupling have sharp frequency gain characteristics at many points. Feedback often occurs at such points, and conversely, there are many points with sharp frequency gain characteristics just before feedback occurs. Even an average engineer can understand that, no matter how complex the system, problems arise at feedback points due to the relationship between phase and gain. Acoustic systems in particular involve group delay, so frequency and phase rotation are proportional, and phases that satisfy the oscillation conditions appear everywhere. Furthermore, from the system designer's perspective, acoustic systems are difficult to measure, with fluctuation factors for both gain and phase being scattered throughout, such as multiple resonances in physical vibration systems and standing waves due to reflections in transmission systems.

[0007] Consideration during the development process of this proposal, No. 5. When feedback is occurring, the simplest way to suppress it is to lower the loop gain. Conversely, even if feedback is not occurring, the loop gain is higher at points where feedback is more likely to occur than at other points. Of course, there is also the issue of phase, but we can first focus on gain. However, there is no way to directly measure the loop gain of a control system in operation. If we could simply measure the loop gain of a control system in operation, or if we could measure some kind of index corresponding to the loop gain, it would be possible to take measures to move away from the danger zone of feedback, although there would be a trade-off with control performance.

[0008] Considerations during the development process of this proposal, part 6. Figure 1(a) shows the simplest model of a control system including a closed loop. The loop gain is G*F*Y*K, and the transfer characteristic is F*Y*K / (1-G*F*Y*K). Figure 1(b) shows the numerical value of the change in gain of the transfer characteristic when K of this system is changed. As K increases, the gain margin of the closed loop decreases. This example shows a case where the phase characteristic of the closed loop is 0. The gain margin is one indicator of the stability of a system, and the gain margin is 6dB near K=0.505 in No. 20. For the measurement signal P in this vicinity, if K is changed by 5%, the rate of change in the intensity of P for column No. 19 is 10.1% (0.8dB). In an acoustic system, a 10% change in the controlled object is perceived as an extremely small change by human hearing, so by changing K to the extent that switching noise does not occur and measuring the strength of P, it is possible to grasp the state of the gain margin of the control system. This value is just one example, but it is possible to measure changes in P for any system, regardless of the state the system is in. Therefore, regardless of the closed-loop model, it is possible to know the state of the system's stability. As a result, it is possible to change the system's operation to the target state.

[0009] Considerations from the development process of this proposal, No. 7. Let Wx be the second-order filter, and let Y in Figure 1(a) be Wx*Yx. Applying a phase transformation Yx to the output signal of Wx can further change the state of the system. For example, by applying a phase angle change ΔAx to Yx and measuring the change in the transmission gain and the second-order change in the transmission gain, the phase state of the system can be estimated. Figures 6(a) and 6(b) show how to apply a phase transformation to Yx, and Figure 6(c) shows its principle. By applying differentiation or integration to the output signal of Wx to create a signal that is π / 2 the original signal and then combining it with the original signal, the phase of the output signal of Wx can be naturally rotated by one full rotation. Second-order digital filters generally have five parameters, which determine the filter's center frequency, selectivity, and gain. The sharper the filter characteristics, the greater the risk of self-oscillation in a second-order digital filter, depending on the initial values. Furthermore, adjustments to parameters during operation must be performed using procedures and methods that prevent self-oscillation and switching noise. However, applying phase rotation by combining signals with a π / 2 phase difference eliminates the possibility of self-oscillation, eliminating the need to worry about safety and noise issues when adjusting parameters during operation. Figure 6(d) shows a graph of the relationship between phase conversion and the gain of the control system's transfer characteristics. By measuring and evaluating the change in transfer gain due to phase correction, we can determine the impact of that phase correction on system stability. This allows adjustments to the target phase state. This cut-and-try technique allows us to determine the closed-loop gain and phase state and drive them toward the target state suitable for control.

[0010] Consideration during the development process of this proposal, No. 8: If further correction of the bandpass filter Wx is required even after applying the optimal phase shift through phase rotation, correction can be made to the frequency and band of Wx, but the characteristics of Wx are determined depending on the application, and this is not the essence of this proposal.

[0011] Consideration during the development process of this proposal, No. 9: Figure 3 is a modification of the control system model in Figure 1(a). By dividing the F*Y*K portion of Figure 1(a) into multiple bandpass filters and combining the characteristics of each bandpass filter, Wj*Yj*Kj, the control system model in Figure 3 can be practically approximated to the model in Figure 1. Because each of Wj, Yj, and Kj is linear, the model in Figure 3 can be considered to be multiple parallel models of Figure 1(a), each operating independently. In other words, by measuring and controlling the stability of each Wj*Yj*Kj without being affected by the mutual influence of different bandpass filters, it is possible to control the stability of the entire system, even if there are multiple unstable factors.

[0012] There are many books and documents, but as it is difficult to determine which documents need to be listed in relation to this case, I will omit listing them here.

[0013] Japanese Patent Application No. 2009-512075: Automated Tuning of Closed-Loop Control; Japanese Patent Application No. 2004-103763: Optical Disk Device; Japanese Patent Application No. 2012-256043: Non-adaptive Active Noise Control; Japanese Patent Application No. 06-137334: Acoustic Echo Canceller; Japanese Patent Application No. 05-057261: Noise Reducer; Japanese Patent Application No. 04-006280: Noise Control Device; Japanese Patent Application No. 63-049447: Hands-Free Telephone. None of these have content in common with the present invention. Japanese Patent Application No. 2011-48010: Active Feedback Control Device and Program; This relates to gain correction when the transfer characteristics of the system are known in advance. Japanese Patent Application No. 2009-512075: Automated Tuning of Closed-Loop Control; This relates to a method of adjusting closed-loop control that requires a step of measuring the transfer characteristics of the system.

[0014] This is a method for measuring and controlling the stability of a control system with a closed loop including acoustic coupling during the actual operation of the system, and does not require advanced theoretical knowledge, has an intuitive algorithm, is easy to program, is easy to maintain, and allows for grasping of the closed loop state by monitoring the signals of the system during actual operation. For details, see Considerations in the development process of this proposal, Parts 1 to 5.

[0015]

[0016] Method 1: Measure the gain margin of the closed loop during operation. For details, see "Considerations in the development process of this proposal, Part 6."

[0017] Means 2: Adjust the phase of the closed loop during operation to the required optimum state. For details, see Part 7 of the Considerations in the Development Process of this Plan.

[0018] Measure 3: Incorporate multiple bandpass filters into the signal processing system. Because these bandpass filters are independent of the other filters, they are not affected by the state of the other filters, and the gain and phase of each closed loop can be adjusted to the optimum state required for each. For details, see Considerations in the Development Process of This Proposal, Part 9.

[0019] Effect 1: By using the means 1 and 2, a control system can be configured with a simple algorithm and signal processing program in which the actual operation and parameters correspond one-to-one, without requiring theoretical knowledge about adaptive control.

[0020] Effect 2: The means 3 increases the degree of freedom of control.

[0021] (a) Model diagram of the simplest control system including a closed loop (b) Table of calculated values ​​of the state of the system when the closed loop gain in Figure 1(a) is changed (c) Graphical representation of the closed loop gain and its rate of change in the table in Figure 1(b). Schematic diagram of an example of the algorithm of the present invention Schematic diagram of an algorithm for dividing Figure 1(a) into multiple bandpass filters Schematic diagram of an example of an algorithm for specifying and adjusting the gain margin of an operating control system Schematic diagram of an example of an algorithm for phase adjustment by phase conversion of an operating control system (a) Schematic diagram of an algorithm for phase conversion using a differential function (b) Schematic diagram of an algorithm for phase conversion using an integral function (c) Explanatory diagram of the principle of applying phase correction (d) Explanatory diagram of the relationship between phase correction, transmission gain, and its rate of change

[0022] First, application to the control program of an active road noise canceller. Second, application to the control program of a feedback canceller.

[0023] First, application software for applied products. Second, applied products.

[0024] Figure 1(a) is a model diagram of the simplest control system including a closed loop. This model diagram explains the principle of determining the gain margin of a control system, which cannot be directly measured in actual operation and is the essence of this proposal. Like loop gain, the state of the loop phase cannot be directly measured, so measurement of the transfer gain is used to determine this state. Si is the input signal to the controlled object, F*Y*K is the transfer characteristic of the signal processing system, and P is its output, which is the monitoring signal for the actual operation state corresponding to the controlled object. G is the transfer characteristic of the feedback path, consisting of acoustic coupling and the signal processing transfer system. The output of G is the feedback signal and is added to Si. The transfer characteristic from Si to P in this control system can be expressed as P=Si*F*Y*K / (1-G*F*Y*K). When (1-G*F*Y*K) is near zero, the system is in an unstable region. Very roughly, the strength of (1-G*F*Y*K) is 0.5, i.e., the gain margin is 6 dB, which is one guideline for determining whether the system is stable or not. The specific boundary value is determined by the specifications of the system to which it is applied.

[0025] Figure 1(b) is a table of calculated values ​​of the transmission gain and the rate of change of the transmission gain when the gain of the closed loop in Figure 1(a) is changed, i.e., when the value of K is changed. In relation to the values ​​shown in the table in Figure 1(b), this is a table of calculated values ​​to explain that although it is theoretically impossible to measure the closed loop gain of a control system in operation, it is possible to estimate the closed loop gain by detecting the change in the transmission gain when the closed loop gain is changed slightly. R21 Reference Table (an example) by 5% step of ΔK indicates a correspondence table of calculated values ​​of R21, which is the rate of change of the transmission gain of the system, when K is changed in 5% steps. This is a simple example in which G, F, and Y all do not have imaginary parts. No. indicates the column number, the Loop Gain=K(1+ΔK) column indicates the change in K when ΔK=K*0.05 is sequentially corrected, the 1 / ((1 / K)–1) column indicates the system's transmission gain when G*Y*F is set to 1, and the R21=P2 / P1 column indicates the ratio of the value in that column to the value in the column above it for 1 / (1–GFYK), with the % and dB columns indicating numerical values ​​in the respective units. For example, since P2 is the transmission gain of No. 20, 1.022, and P1 is 0.928, the value of R21=P2 / P1 is 10.1% 0.8dB. Empirically, even in complex control systems where the Loop Gain is unknown, the relationship between K and R21 tends to be fairly similar near the unstable region. In this table, for rows in columns No. 19 and No. 20, when ΔK increases by 5% for a 6dB gain margin, P will increase by 10.1% or 0.8dB. Therefore, although it may not necessarily be a perfect match, if you correct K by 5%, and the change in P is less than 10.1% 0.8dB, you can roughly estimate that the system has a gain margin of 6dB or more. Conversely, if the gain margin exceeds 6dB, you can conclude that some kind of correction is necessary to ensure control stability.While a 5% forced correction to K may be inconvenient for general control systems, it is practically acceptable when controlling an acoustic system. While the table shows K increasing from bottom to top, it is safer to decrease K when making corrections to an operating control system rather than increasing it. Unstable indicates the direction in which control becomes unstable, while Stable indicates the direction in which control becomes stable. The boundary setting is determined by the control system's application, but generally, closed-loop control controls something for a specific purpose, so it is desirable for the closed-loop gain margin to be as small as possible. The constant trade-off between control and instability is a common issue in all control systems. The 6dB Gain Margin column indicates a 6dB gain margin. Beyond 6dB, the rate of change of System Gain increases dramatically. The 12dB Gain Margin column indicates a 12dB gain margin.

[0026] Figure 1(c) is a graph comparing the system transmission gain ( ) and the change in the intensity of the signal P at the monitoring point of the system state in the numerical table of Figure 1(b). System Gain & R21 / 5% Step indicates that this graph shows the transmission gain and the rate of change of the transmission gain, R21. The horizontal axis is the value of the coefficient K, which changes in 5% steps. The left vertical axis, System Gain, represents the system transmission gain, and the right vertical axis, R21%, represents the rate of change of the transmission gain. Since the transmission gain depends on the system state, evaluating the value itself is meaningless. However, the rate of change of the transmission gain when K is slightly changed is the ratio of the change to the original intensity, so it is a value that can be used to evaluate the system state regardless of the system state. P changes rapidly with changes in K as K approaches 1. While closed-loop gain cannot be measured directly, we can estimate the closed-loop gain of the system from the resulting change in P by slightly changing the closed-loop gain and determine the stability level. For example, since it is possible to calculate the transmission gain near a K value where the gain margin is 6 dB, it is possible to estimate the closed-loop gain margin by actually measuring the rate of change of the transmission gain and then measuring the value of R21 corresponding to that value. In this way, the closed-loop gain margin can be driven to the target value by adjusting the value of K. R21max is an example of an allowable limit value for R21 determined by design. Stable indicates a stable region further to the left, and un-Stable indicates an unstable region further to the right. Not Controlled indicates a region with weaker control to the left, and Controlled indicates a region with stronger control to the right.

[0027] Figure 2 is a schematic diagram of an example of an algorithm that embodies this proposal. Stability Conditioning indicates that the diagram is an explanatory diagram of stability adjustment. SQ20 is a preparation process. Build Wx,Yx,Kx with F indicates incorporating Wx, Yx, and Kx for stability measurement and adjustment after the transfer characteristic F of a part of the closed loop. Start up L is the process of actually operating the control system, and L is Running indicates that L is in the actual operating state. SQ21 is the process of adjusting the closed-loop gain and phase. for x = a,b,,,,,j,,,,,n is the process of sequentially executing x from a to n, Gain Conditioning Kx is the process of gain correction of Kx, and Phase Conditioning Yx is the process of phase correction of Yx. SQ22 is the process of determining whether to continue gain and phase adjustment. Continue? is a process to determine whether further adjustment is necessary, and if necessary, continue and execute SQ21. If not necessary, this process is completed with End. This process is incorporated as part of the closed loop as a program.

[0028] Figure 3 is a schematic diagram of an example of an algorithm for controlling stability by dividing Figure 1(a) into multiple band-pass filters. Details are as described in claim 2. The outline is as explained in Part 9 of the Considerations in the Development Process of the Proposal. Divided F indicates that the diagram is an example of a band-divided control system. To make the model in Figure 1(a) equivalent to the model in Figure 3, the transfer characteristic of F*K in Figure 1(a) is replaced with the product of {F} and {multiple band-pass filters Wx, phase conversion means Yx, and coefficient Kx}. "Equivalent" means characteristics that are close enough to be practically acceptable in acoustic signal processing. G and F are transfer characteristics specific to the system. Wx*Yx*Kx x=a,b,,,,,j,,,,n is the product of Wx, Yx, and Kx from a to n. Px x=a,b,,,,,j,,,,n is the output signal of Wx*Yx*Kx. Σ is the sum of signals from Pa to Pn. P is its output and is equivalent to P in Figure 1(a). Wx is set in advance in the band where it is expected that the closed loop will become unstable. A typical design example is Wa=Ya=Ka=1. This allows the entire band to pass, and the other Wx, Yx, Kx can be responsible for controlling the unstable bands. As explained in the discussion, since Wx*Yx*Kx is linear, controlling one Wx*Yx*Kx does not affect the state of the other Wx, Yx, Kx, nor is it affected by the others. By controlling the stability of each band independently, the overall stability can be controlled.

[0029] Figure 4 is a schematic diagram of an example of an algorithm for measuring and adjusting the gain margin of a control system during operation. Since it is not possible to directly measure the state of the closed-loop gain margin during operation, the closed-loop gain is slightly changed to measure the ratio between the state before and after the monitoring point, thereby indirectly estimating the state of the closed-loop stability. This is as described in the closed-loop gain correction means of claim 1. Gain Conditioning indicates that the diagram is a loop gain adjustment function. Initial Kx=Kx sets the initial value of Kx to Kx. SQ40 begins execution from here. SQ41 is the measurement process for R21. Prms1=Measured Prms sets the measured Prms to Prms1. Kx=Kx*(1-ΔKx) reduces Kx by ΔKx. Since increasing the gain may cause the system to become unstable, it is reduced and the change in transmission gain is measured. Prms2=Measured Prms sets the measured Prms to Prms2. R21=Prms2 / Prms1 sets the rate of change of transmission gain R21 to Prms2 / Prms1. SQ42 is the process to confirm the range of the measured R21. R21>R21max? determines whether the R21 obtained as a result of the measurement is larger than the design value R21max. If it is larger, return to the process to further lower the closed-loop gain. Prms1=Prms2 is used to omit the measurement of Prms1 in the next process. SQ43 is the end process of this sequence. Kx=Kx+ΔKx is returned to the current Kx, and this process ends with End. By repeating the above operations while the system is operating, it is possible to drive Kx into the target gain margin.

[0030] Figure 5 is a schematic diagram of an example of the algorithm for the phase correction method of the phase conversion means Yx. Since the state of the phase margin of the closed loop during actual operation cannot be directly measured, by adding a little phase correction, the change in the transfer gain becomes steeper at the frequencies where the phase characteristics of the closed loop are in phase, and becomes gentler on the frequency axis where it is in antiphase. After confirming the change in the transfer gain and determining the direction of adjusting the phase characteristics, the phase correction in that direction is continuously repeated until it can be judged as an appropriate phase state. Phase Conditioning indicates that the figure is the algorithm for phase correction. Initial Phase(Yx)=Ax sets the initial value of the phase conversion of Yx as Ax. SQ50 is the start of this sequence. SQ51 is the step of obtaining the change in the transfer gain at the current phase. Prms3=Measure Prms measures Prms and sets its value as Prms3. Ax=Ax+ΔAx changes Ax by ΔAx. Prms4=Measure Prms measures Prms and sets its value as Prms4. R43=(Prms4-Prms3) / Prms4 is the change rate of the transfer gain at Prms4. SQ52 is the step of checking whether the measured R43 is within the allowable range between R43max and R43min. If R43>R43max, then Ax=Ax-2*ΔAx and the phase is changed by -2*ΔAx. If R43<R43min, no correction is made to Ax, and the execution of SQ51 and SQ52 is repeated. SQ53 is the step of determining whether it is in the stable region by checking the slope of R43 even if R43 is within the allowable range. Ax=Ax+ΔAx further changes Ax by ΔAx. Prms5=Measure Prms measures Prms and sets its value as Prms5. R54=(Prms5-Prms4) / Prms5 is the change rate of the transfer gain at Prms5. abs(R54-R43)>ΔR54max? is the judgment of whether the absolute value (R54-R43) is greater than the designed ΔR54max.If the slope is greater than the allowable value and it is determined that there is an unstable region, set Ax = Ax - π - ΔAx and return to step SQ51. -π shifts from the in-phase region to the out-of-phase region, and -ΔAx returns Ax to its original value and returns to step SQ51. SQ54 is the end step. Ax = Ax - 2 * ΔAx returns Ax, which was changed for measurement, to its current value. End ends the process. Generally, the closer the closed-loop phase approaches the unstable region, the more rapidly the signal strength P at the monitoring point increases. Therefore, it is possible to determine whether the system is in a safe region using the rate of change value of R43. If the measurement shows that it is in the unstable region, the phase is further displaced until it reaches the stable region. By repeating the above steps while the system is operating, even if the phase characteristics of the original control system's closed loop are unknown, phase conversion can be applied to align the phase to the target within that band, i.e., the band Wx. Measurements clearly show that when the closed-loop gain is at its minimum, G*F*Wx*Yx are in anti-phase, and when it is at its maximum, G*F*Wx*Yx are in phase. There is a range in between where the phase is 90 degrees out of phase. The target phase is determined by design according to the specifications of the control system.

[0031] Figure 6(a) is a schematic diagram of an example algorithm for phase correction using a differentiation function. The Phase Generation Model by Differentiation for High Frequency Range (Phase Generation Model by Differentiation for High Frequency Range) is a diagram of phase correction using a differentiation function and is suitable for phase conversion in the high frequency band. Si is the input signal to the control system, and So is the signal corresponding to the controlled object of the control system. Wx is the bandpass filter, Ss is the output signal of the bandpass filter, d() / dt is the differentiation function, St is the output signal of the differentiation function, and Su is the signal whose phase is π / 2 ahead of Ss after being corrected to have the same intensity as Ss. If the phase angle for which stability is to be measured is A, then So=Ss*cos(A)+Su*sin(A) is the signal that leads Ss at a phase angle A. RMS is a function for obtaining signal intensity, Mss is the intensity of signal Ss, and Mst is the intensity of signal St. Mss / Mst is the ratio of the intensity of signals Ss and St. By setting Su=St*(Mss / Mst), the intensities of Ss and Su are equal. Because the bandwidth of Wx has a certain width, this algorithm cannot perform strictly accurate phase conversion within the effective bandwidth of Wx. However, the goal is to adjust the gain margin, which is known (albeit an estimate), to the target value by changing the phase. Therefore, strictly accurate phase conversion is not necessary. As long as abs(R54-R43)>ΔR54max in Figure 5 is not met, the state satisfies the phase condition. Gain adjustment is constantly repeated, as shown in Figure 2, so the target state can be achieved. Programming is possible using standard, intuitive sequences without requiring knowledge of complex adaptive control theory. Signal monitoring for each component is also possible, making maintenance extremely easy.

[0032] Figure 6(b) is a schematic diagram of an example algorithm for phase correction using an integration function. The Phase Generation Model by Integration for Low Frequency Range is a diagram of the phase correction using an integration function, and is suitable for phase conversion in the low frequency range. The difference from Figure 6(a) is that the integration function creates a signal with a phase of -π / 2. -∫()dt is the integration function.

[0033] Figure 6(c) is an explanatory diagram of how phase correction is applied to the output signal of an operating bandpass filter. It shows that a phase correction of 2*π can be applied using two signals with a phase difference of π / 2. The Phase Generation Model shows that the diagram illustrates the principle of phase conversion. Since the intensities of Ss, Su, and So are the same, Ss*cos(A)+Su*sin(A) moves around a circle by an angle A. Even if there are slight differences in phase or intensity, the circular orbit becomes an ellipse, and it still moves around the ellipse. This is a sufficient means of phase conversion. A=Phase(So / Ss) indicates that the phase angle of So relative to Ss is A. abs(So)=abs(Ss)=abs(Su) indicates that the intensities of So, Ss, and Su are the same.

[0034] Figure 6 (d) is an explanatory diagram of the relationship between the phase correction of the phase conversion means Yx and the closed-loop gain. System Gain & ΔGain / Phase Shift indicates that the diagram is a graph showing the transmission gain of the control system and its rate of change with respect to the phase shift. System Gain=0.5 / (1.25-cos(A))^0.5 is the formula for calculating the change in the transmission gain of the control system by changing the phase of the phase conversion means. The horizontal axis is the correction phase angle, which is the output of Yx, and the vertical axis is the gain. System Gain on the left vertical axis is the transmission gain due to the phase correction of Yx, and R43% on the right vertical axis is the rate of change in the transmission gain. ΔGain R43 indicates the rate of change in the transmission gain. While the relationship with the original transmission gain due to phase correction is unknown, the rate of change of the transmission gain due to the phase correction of Yx, R43%, is known through calculation. Therefore, although the phase characteristics of the closed loop of the control system are unknown, the rate of change of the gain with respect to the phase correction can be measured. Therefore, the state of the closed loop phase can be estimated from the measured value of the gain change rate R43. By rotating the phase rotation Ax at most once, it is possible to grasp the full extent of the change in R21 or whether it is within the tolerance limits. When the phase of the entire closed loop including Yx is π, the transmission gain is minimum, and conversely, when it is 0, it is maximum. Therefore, by rotating Ax at most once, the state of the wavy transmission gain can be obtained. The state of the system can be determined by the rate of change of the transmission gain and can be driven to the desired phase state. This utilizes the fact that the slope of R43 is steep near the unstable region and gentle in the stable region. The allowable range of R43 is defined as a maximum value R43max and a minimum value R43min. Furthermore, since the steep slope of R43 is an unstable region, a slope threshold of R54=abs(ΔR43)max is defined as the allowable range of the slope.

[0035] Supplementary Explanation of Claim 1, Part 1. Regarding the outline of claim 1, this is a stability measurement control means for a control system that forms a closed loop including acoustic coupling. Stability evaluation is based on the gain margin and phase margin of the control system. There is no simple method for directly measuring the gain margin and phase margin of a control system during operation. Figure 1(b) shows calculated values ​​for a representative example of the relationship between the gain margin and transmission gain of a control system. This table of calculated values ​​shows how the transmission gain is affected by the loop gain. Generally, problems with control systems with closed loops occur more or less near the balance between the gain margin and system stability. Therefore, a gain margin of approximately 6 dB can be considered important as a guideline. Near a gain margin of 6 dB, the transmission gain tends to increase rapidly toward a region where the gain margin is zero. By measuring the change in transmission gain caused by slightly changing the loop gain of a control system during operation, the approximate gain margin can be estimated. As a result, the gain margin can also be driven to a target level. Regarding the phase margin, Figure 1(b) shows the calculated values ​​of a typical example of the relationship between the phase rotation of a closed loop in a control system and the transmission gain. This table of calculated values ​​shows how the transmission gain is affected by phase rotation. Generally, problems with control systems with closed loops occur more or less near the balance between the phase margin and system stability. Therefore, a phase margin of π / 2 to π / 6 serves as a guideline for the phase margin near the phase margin. This exploits the fact that when the phase margin is near 0, the transmission gain tends to increase rapidly toward the region where the phase margin is zero. Conversely, when the phase margin is near π, the change in the transmission gain is most gradual. Generally, a control system is stable when the phase is π. However, when the phase π is balanced between 0 and π, instability and stability are side by side. In such cases, a π / 2 phase is desirable. For example, if the transmission system has a notch filter, the phase will be discontinuously reversed near the singular point, so unless the characteristics of the transmission system are fully understood, 0 phase or π phase must be avoided. In any case, the phase conversion by the means of claim 3 allows for free rotation of 2*π.By measuring the change in transmission gain caused by slightly changing the phase rotation of an operating control system, the approximate gain margin can be estimated. As a result, the gain margin can be driven to the target level. Figures 1(a), (b), (c), 2, 4, 5, and 6(d) are explanatory diagrams of claim 1. 2. Regarding the transfer characteristic F of a part of the control system: For this proposal, the transfer characteristic F, which is part of the loop of a given control system, does not need to be known. Whether F is a property inherent in the original control system or includes transfer characteristics added for the purpose of stabilizing the control system is irrelevant to this proposal. The essence of this proposal is to add Wx, Yx, and Kx to F, making F F*Wx*Yx*K. 3. A stability evaluation table is a table of the change in transmission gain when the loop gain is changed, prepared in advance by calculation or measurement. Information is also prepared that allows for determining from this table which rate of change in transmission gain corresponds to the target value or limit value of stability. 4. Regarding the stability evaluation value, the stability evaluation value in the stability evaluation table is a criterion for determining the optimum gain correction based on the universal fact that the smaller the gain margin becomes when the closed-loop gain is changed, the greater the rate of change of the transmission gain, as shown in the numerical table in Figure 1(b) and the representative example in Figure 1(c), which is a graph of the numerical table.

[0036] Supplementary Explanation of Claim 2, Part 1. Regarding the outline of Claim 2, this is a stability measurement control means for a control system that forms a closed loop including acoustic coupling, and is a practical configuration that applies the basic configuration of Claim 1. Figure 1 shows the basic configuration of this proposal, and Figure 3 shows a practical configuration. This is a modified version of the F*Y*K portion of the closed loop in Figure 1. The Y*K portion is defined as Wx*Yx*Kx and divided into multiple bands using a bandpass filter called Wx. Alternatively, it can be divided into multiple bands where problems are likely to occur. In the most practical configuration, Wa*Ya*Ka is set to 1, and the remaining Wj*Yj*Kj are assigned to bands where problems are likely to occur. Bands outside the scope of the scope are handled by Wa*Ya*Ka=1. In other words, the bands are divided. Generally, the number of unstable regions tends to increase as the loop gain in a closed-loop control system is increased, so the number of unstable regions varies depending on the individual control system. Furthermore, depending on the control system, instability factors may fluctuate depending on the control environment, so dynamic control of Wj, Yj, and Kj may be necessary. Part 2: Bandpass filter Wx is a bandpass filter installed in a band requiring particular attention to stability. It may be necessary to install it uniformly and densely throughout the entire band, or if there is a particularly unstable band and detailed stability control is required for that band, a bandpass filter may be installed only in that band. Whether or not Wx has group delay characteristics is also determined by design. Regarding the closed-loop division method, band division may also be determined dynamically during the actual operation of the control system. In this case, the band is roughly divided, for example, into two, and which band is furthest from stability is identified. Then, that band is further roughly divided to identify which band is most unstable. By repeating this process, the bands requiring the greatest stability improvement are driven toward the target stability level. This method of dynamically determining band division is also determined by design. Alternatively, additional bands can be established within a single band. By layering the bands, control performance can be improved.Point 3. Regarding band division, band division does not mean dividing the entire band, but rather assigning filters to only those bands that require particular gain and phase control.

[0037] Supplementary Explanation of Claim 3: Part 1. Regarding the outline of claim 3, this relates to a stability measurement control means for a control system that forms a closed loop including acoustic coupling, and relates to phase control of the bands described in claims 1 and 2. Phase control is required when the band contains instability factors related to phase and gain. There are many methods for controlling phase, but excluding phase due to group delay, this is angle control that can be expressed by the horizontal and vertical axes on a Cartesian coordinate system, as shown in Figure 6(c). In an actual system, the control system signal is treated as a periodic signal, and its phase is controlled as a periodic signal. A signal with a phase of π / 2 relative to the original signal can be created and combined with the original signal to obtain a signal with any phase relative to the original signal. Part 2. Regarding phase conversion: There are many methods for converting phase. The term "phase" only applies if the signal is a steady sine wave. It is generally known that noise and signals handled in human communication and listening environments have a high degree of redundancy. While it is not possible to explain whether the concept of phase applies transiently when noise and signals are constantly changing, the fact that sounds in our living environment at least have a high degree of redundancy makes it clear that the technique of phase control is meaningful. On the other hand, since oscillation phenomena such as feedback are theoretically supported by the two basic factors of phase and gain, it is quite meaningful to control the stability of control systems through phase conversion under conditions of high signal redundancy. 3. Regarding differentiation and integration functions, differentiation and integration functions are essential for any filter to obtain a phase difference of π / 2.

[0038] Si: Input signal of the control system +-* / : Addition, subtraction, multiplication and division operators F: Part of the transfer loop of the control system G: Part of the transfer loop of the control system Y: Phase transformation for control stabilization, which is the essence of this proposal K: Coefficient for control stabilization, which is the essence of this proposal P = Si*F*Y*K / (1-G*F*Y*K) Calculation formula for the transfer characteristics of the control system in Figure 1(a)

[0039] R21 Reference Table / an example / by 5% step of ΔK List of R21 values ​​for a 5% change in ΔK No. Column number Stimulus A change is applied to the system for measurement 0.05 ΔK is 5% K=K(1+ΔK) K is the ΔK correction System Gain The transmission gain of the control system is GFY=1,FY=1 G*F*Y=1. Assume F*Y=1 1 / (1–GFK) Assume G*F*Y=1, F*Y=1, value of 1 / (1 – GFK) R21=P2 / P1 Ratio of 1 / (1 – GFK) value in that column and the column above % % value of that ratio dB dB value of that ratio 6dB Gain Margin Column with gain margin of 6dB 12dB Gain Margin Column with gain margin of 12dB Un-Stable Direction in which control becomes unstable table Direction in which control becomes stable

[0040] System Gain & R21 / 5% Step Graph of transmission gain and rate of change of transmission gain when K is changed in 5% steps 5% Step (K(1+ΔK) Rate of change of K is 5% Gain Transmission gain R21 % Rate of change of transmission gain 6dB Gain Margin Line where gain margin is 6dB 12dB Gain Margin Line where gain margin is 12dB System Gain Graph of transmission gain R21 % Rate of change of transmission gain Graph of R21 Stable Direction of more stable control system Un-Stable Direction of more unstable control system Not Controlled Direction of looser control of control system Controlled Direction of stronger control of control system

[0041] Stability Conditioning The diagram shows the stability adjustment process SQ20 Preparation process Build Wx,Yx,Kx with F Incorporate Wx,Yx,Kx into the loop transfer characteristics Start up L Operate the control system L is Running State where L is in the actual operating state SQ21 Kx adjustment process for x = a,b,,,,,j,,,,nx from a to n Gain Conditioning Kx Gain adjustment Phase Conditioning Yx Phase adjustment SQ23 End process Continue ? The adjustment process has finished, but is further adjustment required? End End

[0042] Divided F Diagram of how to divide a closed loop into bands Si Input signal of the control system F Part of the transfer loop of the control system G Part of the transfer loop of the control system Wx*Yx*Kx Bandpass filter and phase transformation and coefficients from x=a,b,,,,j,,,,na to n Px x=a,b,,,,j,,,,n Output signal of Wx*Kx Σ Addition of signals from Pa to Pn P Output of the addition result of signals from Pa to Pn

[0043] Gain Conditioning Gain adjustment sequence Initial Kx=Kx The initial value of Kx is Kx SQ40 Start sequence Start Start SQ41 Measurement process for the rate of change of transmission gain Prms1=Measure Prms Measure Prms and set it as Prms1 Kx=Kx*(1-ΔKx) Apply ΔKx correction to Kx Prms2=Measure Prms Measure Prms and set it as Prms1 R21=Prms1 / Prms2 Calculate the rate of change of transmission gain R21 SQ42 Confirmation process for whether R21 is within the allowable range R21 > R21max Is the measured R21 greater than R21max? Prms1=Prms2 Set the value of Prms2 to Prms1 SQ43 End process Kx=Kx*(1+ΔKx) Set Kx to the current value

[0044] Phase Conditioning The figure shows the sequence of phase conditioning. Initial Phase(Yx)=Ax The initial value of the phase of Yx is Ax. SQ50 Start sequence Start Start SQ51 Change the phase and measure the change in transfer gain. Prms3=Measure Prms Measure Prms and assign it to Prms3. Ax=Ax+ΔAx Change Ax by ΔAx. Prms4=Measure Prms Measure Prms and assign it to Prms4. R43=(Prms4-Prms3) / Prms4 Calculate the change rate of transfer gain, R43. SQ52 Evaluation process of R43. R43>R43max? Is R43 greater than R43max? R43<R43min? Is R43 less than R43min? Ax=Ax-2*ΔAx Change Ax by another -2*ΔAx and go to the SQ51 process. SQ53 Measurement and judgment process of the slope of the change rate of transfer gain. Ax=Ax+ΔAx Change Ax by another ΔAx. Prms5=Measure Prms Measure Prms and assign it to Prms5. R54=(Prms5-Prms4) / Prms5 Calculate the transfer gain change rate, R54. abs(R54-R43)>ΔR54max? Is the absolute value of the slope of the transfer gain change rate greater than ΔR54max? Ax=Ax-π-ΔAx Change Ax by –π-ΔAx and go to the SQ51 sequence. SQ54 End process. Ax=Ax-2*ΔAx Restore Ax to the current value. End End

[0045] Phase Generation Model by Differentiation for High Frequency Range Model diagram of phase transformation using differentiation Phase Generation Model by Integration for Low Frequency Range Model diagram of phase transformation using integration Siw Input signal of bandpass filter Wx So Output signal of phase transformation Wx Bandpass filter Ss Output signal of bandpass filter d() / dt Differentiation function -∫()dt Integration function St Output signal of differentiation or integration function Su Ss with coefficient corrected to be the same as the intensity of Ss Function to obtain RMS signal intensity Mss Intensity of Ss Mst Intensity of St Mss / Mst Ratio of intensity of Ss and St A Phase transformation angle So=Ss*cos(A)+Su*sin(A) So is a signal that leads Ss by phase angle A Su=St*(Mss / Mst) Equalizes the intensity of Su to Ss

[0046] Phase Generation Model Phase transformation diagram Ss*cos(A)+Su*sin(A) So moves along the circumference by angle A A=Phase(So / Ss) The phase angle of So relative to Ss is phase transformation A abs(So)=abs(Ss)=abs(Su) The intensities of So, Ss and Su are the same

[0047] System Gain & ΔGain / Phase Shift Example of a graph showing the gain change rate and gain change rate slope due to phase shift Phase Horizontal axis: Phase angle of phase shift System Gain Vertical axis: Transfer gain R43 % Vertical axis: Gain change rate System Gain Change in transfer gain R43 % Gain change rate R43max Maximum allowable value of R43 R43min Minimum allowable value of R43 R54=abs(ΔR43 %)max Maximum allowable value of gain change rate slope

Claims

1. The stability control function of an acoustic control system having a closed loop is called the stability control means, a specific acoustic signal band is called B, and a specific acoustic signal band is taken to mean from the entire band to any partial band in terms of human hearing, the closed loop of the control system including acoustic coupling in band B is called L, the loop transfer characteristic of L is called Z, and the transfer characteristic Z is specific to the closed loop L and is the dynamic transfer characteristic of L which changes from moment to moment, Wx, Yx, and Kx are called the transfer characteristic correction means which are added to Z for the purpose of correcting the stability of the control system during actual operation, and the structure and characteristics of the transfer characteristic correction means are determined by design, Wx is a band filter, Yx is a phase conversion means, and Kx is a proportional coefficient, i.e., Z*Wx*Yx*Kx is added to Z to control the stability of the original control system, and L when the control system is in an actual operating state Let P be the controlled object or a signal corresponding to the controlled object, the quantity representing the strength of the signal such as absolute value, average value, effective value, maximum value, etc. be Strength, the strength of P be Prms, for a band Wx, let Prms1 be the value of P when the coefficient is Kx, and Prms2 be the value of P when the coefficient Kx is changed in the direction decreasing it by ΔK %, that is, when Kx = Kx (1 - ΔKx), let R21 = Prms2 / Prms1 be the stability evaluation value when a change of -ΔKx is applied to Kx, and the stability evaluation value is closely related to the gain margin, which is an evaluation value of the stability of the control system, and utilizes the fact that it is a value with which the stability of the control system in the actual operating state can be evaluated, let a list of R21 values ​​corresponding to changes in Kx be the stability evaluation table, and the result of evaluating stability using the stability evaluation shown in the stability evaluation table corresponding to the value of R21, A function for correcting the closed loop gain as necessary is defined as the closed loop gain correction means, the stability evaluation table and evaluation criteria are determined by design, the phase converted by the phase conversion means Yx is defined as Ax, the rate of change of the transmission gain with respect to the change in phase Ax is defined as the phase influence rate, the slope of the phase influence rate with respect to the change in phase Ax is defined as the phase influence rate slope, the allowable range of the phase influence rate is defined as the phase influence rate allowable range,a phase influence factor slope allowable range; the phase influence factor allowable range and the phase influence factor slope allowable range are determined by design; Ax is adjusted so that both the phase influence factor and the phase influence factor slope are within the allowable ranges; the adjustment function of Ax is a closed loop phase correction means; the phase obtained by the closed loop phase correction means is an optimum phase; the closed loop phase correction means functions to adjust the converted phase amount of Yx to the optimum phase; a transfer characteristic correction means is a first transfer characteristic correction means; a stability evaluation table is a second transfer characteristic correction means; a closed loop gain correction means is a third transfer characteristic correction means; and a closed loop phase correction means is a fourth transfer characteristic correction means; 2. Wj*Yj*Kj, where j = a,b,,,,,j,,,n, is incorporated into the transfer characteristic Z of the closed loop including the acoustic coupling of claim 1 so that Z*{Wa*Ya*Ka + Wb*Yb*Kb + ,,,,,+ Wj*Yj*Kj + ,,,,,+ Wn*Yn*Kn}, where j = a,b,,,,,j,,,,n is the type of W, Y, K having different parameters from a to n, and Wj*Yj*Kj is incorporated into the closed loop for the purpose of improving the stability and performance of the control system having Z as the closed loop transfer characteristic, where Wj is a bandpass filter, Yj is a phase conversion means, and Kx is a coefficient, and Wx*Yx*Kx of claim 1 is expressed as Wx*Yx*Kx = Σ(Wj*Yj*Kj) j = a,b,,,,,j,,,n is defined as "closed loop division", and an example of closed loop division is also taken to include the case where Wa=Ka=1, in which case Wx*Kx = 1+Σ(Wj+Kj) j = b,c,,,,,j,,,n, and the function of dividing the closed loop and applying the stability measurement control of claim 1 to each individual closed loop is defined as "closed loop division means", and "closed loop division" means a plurality of bands requiring control, and the specific method of dividing the bands is determined by design, and each divided closed loop has the first, second, third and fourth of claim 1, and the closed loop division means is the fifth, and a stability measurement control method for a control system forming a closed loop including acoustic coupling having the fifth.

3. In the phase correction means described in claim 1, a method for adjusting phase characteristics, in which Yx is provided for the purpose of phase conversion in the output signal of bandpass filter Wx, Si is the input signal of Wx, Ss is the output of Wx, the differentiation function of Ss is d() / dt, the integration function of Ss is -∫()dt, the output of both the differentiation function and the integration function is St, St is obtained from either the differentiation function or the integration function, a signal whose intensity of St is equal to the intensity of Ss, or a signal that can be considered to be equal in practical terms is Su, that is, abs() is the function for obtaining the intensity, the intensity of St is abs(St), the intensity of Ss is abs(Ss), Su=St*abs(Ss) / abs(St), a signal Ss*cos(A)+Su*sin(A) is generated from Ss and Su, and this signal is Let So be, that is, So=Yx(Ss)=Ss*cos(A)+Su*sin(A), and utilize the fact that So rotates 2*π, or once, relative to Ss by varying A from 0 to 2*π; when the band of Wx is in the mid to treble range, the differentiated signal of Ss is used for signal processing, and when the band of Wx is in the bass range, the integrated signal of Ss is used for signal processing, and in this case the bass range and treble range are determined by design in light of the object to be controlled; a function is applied to the output Ss of Wx to generate Su, either a differentiation function or an integration function, and a function is defined as generating a signal So having a desired phase difference from Ss from a signal Su having a phase difference of π / 2 with Ss, and the phase generation means is a sixth phase generation means, and the first, second, third and fourth of claim 1 and the fifth and sixth of claim 2 are included. A method for measuring and controlling the stability of a control system that forms a closed loop including acoustic coupling.

Citation Information

Patent Citations

  • noise cancellation system

    JP1994503897A

  • Noise cancellation headphone and its variation adjusting method

    JP2007108522A

  • Active vibration noise control device and transfer characteristic measuring method

    JP2010188871A

  • Noise spectrum distribution detection method and noise volume sound quality control method

    JP2018182480A

Cited By

  • Ground surface deformation monitoring method and system based on distributed millimeter wave radar

    CN122110020A