JP6409802B2 - Motor noise cancellation device - Google Patents

Description

  The present invention relates to an apparatus for detecting noise generated when a motor is driven.

In general, a motor generates a rotational force using a magnetic attractive force / repulsive force between a coil and a coil, or a ferromagnetic metal such as a coil and iron, or a coil and a permanent magnet. The magnetic attraction / repulsive force that generates rotational force by the motor also generates mechanical deformation force in addition to rotational force.
http://eprints.whiterose.ac.uk/891/1/zhuzq22.pdf
As shown in FIG. 2, this mechanical deformation force induces mechanical resonance and generates a large noise.

  In addition, since the magnetic attractive force / repulsive force is not constant with respect to the rotational direction, a vibration phenomenon in the rotational direction of the rotor called cogging or torque ripple also occurs, which also causes vibration and noise. As a countermeasure against this, as shown in Non-Patent Document 1, a structure for suppressing the occurrence of mechanical resonance has been devised.

  Further, Non-Patent Document 2 suggests that a rectangular wave as a drive signal for a coil has a large noise in order to reduce the vibration of the casing, but the noise is reduced by reducing its harmonics. Similarly, in Non-Patent Document 3, harmonics are intentionally injected to reduce cogging.

Nau, Sebastiao Lauro, and Hugo Gustavo Gomez Mello. "Acoustic noise in induction motors: causes and solutions." Petroleum and Chemical Industry Conference, 2000. Record of Conference Papers. Industry Applications Society 47th Annual. IEEE, 2000. (Paper A) Cameron, Derrick E., Jeffrey H. Lang, and Stephen D. Umans. "The origin and reduction of acoustic noise in doubly salient variable-reluctance motors." Industry Applications, IEEE Transactions on 28.6 (1992): 1250 -1255. Stephenson, J. M., A. Hughes, and R. Mann. "Torque ripple minimization in a switched reluctance motor by optimum harmonic current injection." Electric Power Applications, IEE Proceedings-. Vol. 148. No. 4. IET, 2001.

  However, when structural measures are taken as in Non-Patent Document 1, there is a problem that the degree of freedom in housing design is lowered. Non-Patent Document 2 does not disclose a specific method for reducing harmonics. In Non-Patent Document 3, since the amplitude and phase of the harmonics to be injected are determined by simulation, it is difficult to sufficiently reduce noise unless there is a highly accurate simulation model. In addition, when noise fluctuates due to individual differences, temperature fluctuations, rotation speed dependence, etc., there is a problem that they cannot be handled unless they are modeled in advance.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a motor noise canceling device that can reduce noise generated when a motor is driven by an easier method.

  According to the motor noise canceling apparatus of the first aspect, the basis vector output unit generates and outputs a set of harmonics with respect to the fundamental frequency of the drive signal output to the motor as a basis vector. The coefficient generator updates the coefficient corresponding to each base vector by performing error minimization using the detection signal obtained from the noise sensor that detects noise generated when the motor is driven and the step size parameter μ. While generating. Here, “noise generated when the motor is driven” includes “vibration” and the like. Further, the “error minimization calculation” includes an LMS algorithm using the least square method, a Normalized LMS algorithm, a sign-sign LMS algorithm, and the like, which are applications thereof.

  Then, when the product-sum operation unit performs a product-sum operation on the basis vector and the coefficient corresponding to each basis vector, the signal obtained as the operation result is added to or subtracted from the drive signal, thereby driving the motor. Cancel the generated noise. That is, the detection signal obtained from the noise sensor becomes an error signal as a result of the noise contained in the motor drive system. On the other hand, the coefficient generator performs an error minimization operation and updates the coefficient corresponding to each base vector, so that the error signal approaches zero level, and noise generated when the motor is driven is reduced. The Therefore, it is not necessary to prepare a high-precision simulation model in advance or to modify the model in consideration of individual noise fluctuation factors in advance, and noise generated when the motor is driven can be reduced by real-time processing. .

Functional block diagram showing the configuration of the noise canceling apparatus according to the first embodiment The figure which illustrates the attachment state of the noise sensor Diagram showing fluctuations in motor rotation speed Partial equivalent diagram of FIG. 1 showing the arrangement of the differentiation circuit FIG. The figure which shows the relationship between the transmission wave currently disclosed by 2 and a reception wave FIG. FIG. 3 is a diagram showing an echo canceller disclosed in FIG. Diagram showing the relationship between the transmitted wave and its reflection The figure explaining the product-sum calculation performed simulating a FIR filter Diagram showing time waveform and frequency spectrum of motor rotation frequency The figure which shows what applied the product-sum operation shown in FIG. 8 to the rotational frequency of the motor and its harmonic components Functional block diagram showing the configuration of the noise cancellation device according to the second embodiment The figure which shows what applied the sum of products operation shown in FIG. Diagram showing motor drive signal and noise waveform used for simulation (Condition 1) Diagram showing waveform and noise waveform generated by simulation (Condition 1) Diagram showing coefficients obtained by simulation (Condition 1) Diagram showing motor drive signal and noise waveform used for simulation (Condition 2) Diagram showing waveform and noise waveform generated by simulation (Condition 2) Diagram showing coefficients obtained by simulation (Condition 2) Functional block diagram showing the configuration of the noise canceling apparatus according to the third embodiment Diagram showing product-sum operation using square wave basis vectors Diagram showing motor drive signal and noise waveform used for simulation Diagram showing waveforms generated by simulation and noise waveforms Diagram showing coefficients obtained by simulation Functional block diagram showing a part of the configuration of the noise cancellation device according to the fourth embodiment Functional block diagram showing the configuration of the noise canceling apparatus according to the fifth embodiment Functional block diagram showing the configuration of the noise canceling apparatus according to the sixth embodiment Diagram showing motor drive signal and noise waveform used for simulation Diagram showing waveforms generated by simulation and noise waveforms Diagram showing coefficients obtained by simulation Functional block diagram showing a part of the configuration of the noise cancellation device according to the seventh embodiment Waveform diagram showing a drive signal and its fourth harmonic component in the case of driving a three-phase motor in the eighth embodiment Waveform diagram showing the drive signal and its third harmonic component Functional block diagram showing the configuration of the noise cancellation device

(First embodiment)
First, the principle of this embodiment will be described. In general, as a noise reduction technique by digital signal processing, there is an echo canceller used in full-duplex communication as shown in the following paper A, for example.
https://www.researchgate.net/profile/Hector_PerezMeana/publication/4186057_Adaptive_echo_canceller_using_a_modified_LMS_algorithm/links/
54b74bb50cf24eb34f6e9ebf.pdf

  As shown in FIG. 5, it is desired to receive only the Tx2 signal as the received signal Rx signal, but there are cases where the Tx1 signal *, which is a reflection component of the transmission signal Tx1 signal, is mixed as noise. In order to cancel this noise component, the echo canceller generates a correction signal y (T) with an Echo Canceller block configured by a FIR (Finite Impulse Response) filter, as shown in FIG. The coefficient of the Echo Canceller block is adjusted so that e (T) that is a difference between y (T) and d (T) that is a noise component becomes zero.

As a method for adjusting the coefficient, the above-mentioned paper A discloses using an LMS (Least Mean Square) algorithm as shown in Equation (1). In this technical field, such an LMS algorithm and its modified algorithm are generally used. Using the LMS algorithm, when a Tx1 signal is input, the coefficient of the FIR filter having the same characteristics as the echo can be determined (Principle 1).
w (T + 1) = w (T) + μx (T) e (T) (1)
w (T + 1): Next filter coefficient
w (T): Current filter coefficient
x (T): Input signal
e (T): Filter error signal
μ: Step size parameter

  Hereinafter, the principle of canceling noise generated when the motor is driven by applying the present technology will be described. Although the waveform of Tx1 signal * shown in FIG. 7 is different from that of Tx1 signal, the waveform of Tx1 signal * can be shown as the sum of delay components of Tx1 signal. FIG. 7A shows a Tx1 signal and its echo Tx1 signal *. Although the waveforms are different from each other, as shown in FIG. 7B, when the waveform is decomposed and an echo for one isolated pulse is seen, a delayed signal and a signal subjected to amplitude fluctuation at a certain rate are obtained. It turns out that it is addition. Then, as shown in FIG. 7A, the waveform whose amplitude changes randomly is obtained by adding a plurality of such isolated pulses.

  As can be seen from FIG. 7B, the characteristic of the sum of the delayed pulse multiplied by a certain coefficient can be simulated using an FIR filter. By applying the FIR filter, echoes can be reproduced even for randomly changing waveforms. In the echo canceller simulating this FIR filter, as shown in FIG. 8, it is only necessary to obtain the optimum amplitude fluctuations (w_1 to w_n) for signals delayed by various times.

  Here, the plurality of input signals shown in FIG. 8 are shifted in the time direction, but there is no correlation at the same time, that is, they can be regarded as independent basis vectors in the time direction. Therefore, it can be considered that the LMS algorithm is to obtain a coefficient that is the sum of the basis vectors. In other words, it can be said that the LMS algorithm is a mechanism for obtaining the coefficient of the basis vector having the same characteristics as the echo when the sum is taken (principle 2).

  When the motor is operated by applying a drive signal having a rotational frequency, noise having a higher-order component of the rotational frequency is generated as described above. Since the signal that simulates the noise does not become a delay wave of the drive signal, a delay by the FIR filter cannot be made. Therefore, the same method as the conventional echo canceller based on the principle 1 cannot be applied. However, if the noise signal is reproduced by synthesizing independent basis vectors based on the principle 2, the coefficient of amplitude can be obtained by the LMS algorithm.

  In the present embodiment, the waveform of each frequency is generated as the basis vector in the same manner as the drive signal, not the waveform generated by the delay. There are various generation methods, such as those based on numerical calculations and those referring to tables. In FIG. 9, the case where the noise frequency f2 is produced | generated with respect to the waveform of the rotation frequency f1 is considered. As shown in FIG. 10, coefficients (w_1 to w_n) that are optimal for simulating a noise waveform are generated for each basis vector using the LMS algorithm. Let these product-sum results be y (T).

  The basis vectors do not necessarily have to be prepared for all high-order harmonics. If the order in which noise is likely to occur is known by prior analysis or the like, it is only necessary to generate the basis vectors. Even when a harmonic component whose order is a non-integer multiple is included, a basis vector corresponding to the frequency may be added similarly.

  As shown in FIG. 1, the noise cancellation apparatus 1 of the present embodiment includes a base vector generation block 2, a coefficient generation block 3, and a product-sum block 4, and is configured by, for example, a DSP (Digital Signal Processor). The basis vector generation block 2 generates basis vectors x_1 to x_n, which are a set of fundamental frequencies of the motor drive signal and harmonic components corresponding to the fundamental frequencies, and outputs them to the product-sum block 4. Note that each frequency of the base vectors x_1 to x_n also changes as the motor driving frequency f1 changes. A signal detected by a noise sensor disposed in the motor is input to the coefficient generation block 3.

  As shown in FIG. 2, the motor 6 includes a rotation shaft 8 and a rotation angle measuring device 9 attached to the rotation shaft 8 inside the housing 7. The rotation angle measuring device 9 is, for example, a rotary encoder or a resolver. When the detection target is a sound as the noise sensor, for example, in the vicinity of the motor 6, a microphone 11 </ b> A is attached to a fixed part 10 such as a wall that is not directly connected to the motor 6, and noise that is vibration waves of air As the sound is detected directly. Further, in the case where there is no fixing portion 10, if the microphone 11B is attached to the motor fixing jig 12, the microphone 11B side vibrates, so that vibration sound can be detected relatively. Similarly, the vibration can be directly detected by attaching the acceleration sensor 13 to the motor housing 7 or the motor fixing jig 12.

In the coefficient generation block 3, the coefficients w_1 to w_n are updated by the LMS algorithm which is an error minimizing calculation algorithm, and these are output to the product-sum block 4. Here, each term on the right side of the equation (1) corresponds as follows.
w (T): Current coefficient w_1 to w_n
x (T): basis vectors x_1 to x_n
e (T): Sensor detection signal

  The product-sum block 4 performs a product-sum operation on each of the base vectors x_1 to x_n and the corresponding coefficients w_1 to w_n, and outputs a signal generated as the operation result to the subtracter 5. Then, the subtracter 5 subtracts the signal input from the product-sum block 4 from the noise signal generated when the motor is driven, and the noise sensor detects the subtraction result. That is, when the coefficient generation block 3 updates the coefficients w_1 to w_n by the LMS algorithm, the error between the generated waveform and the noise waveform becomes sufficiently small. That is, the noise cancellation apparatus 1 generates a waveform similar to the noise waveform.

  Here, the subtractor 5 is shown as a signal model, and actually, it is determined whether to perform subtraction or addition according to the action in the “noise generation process” in the figure. That is, the signal {x_1 (T) + y (T)} on which the product-sum result y (T) generated in FIG. 10 is superimposed is used for driving the motor. Then, y (T) becomes z (T) in the noise generation process, and when this z (T) is subtracted from the noise waveform in the same process, the error between the z (T) generation waveform and the noise waveform becomes sufficiently small. That is, a waveform similar to the noise waveform is generated. In addition, in the noise generation process, z (T) may be added to the noise waveform, that is, the adder 5 may be used instead of the subtractor 5, and in this case, x_1 (T) -y ( T)} may be a motor drive signal.

In general, since noise is composed of harmonic components, the fundamental wave component x_1 is not used for waveform generation in the product-sum block 4. However, since a fundamental wave is required for generating the motor drive signal, the base vector generation block 2 can be easily integrated with the drive signal generation circuit. In the present embodiment, the LMS algorithm has been described. However, the feature is that a harmonic of the motor rotational speed is used as the basis vector, and the error minimizing calculation algorithm for determining the coefficient is not limited to this. For example,
http://www.cs.tut.fi/~tabus/course/ASP/SGN2206LectureNew5.pdf
As described in, algorithms such as sign-sign LMS and Normalized LMS derived from LMS are also applicable.

  Also, by setting the convergence factor; step size parameter μ used in the LMS algorithm to a value expressed by a power of 2, the multiplication performed inside the LSI can be replaced with a bit shift operation, and the circuit scale can be reduced. .

In addition, as shown in FIG. 2, by using the rotation angle measuring device 9, it is possible to detect a swing in the rotation direction around the rotation shaft 8 of the motor 6. Between the rotation angle Φ [rad] and the rotation speed ω [rad / sec], ω = dφ / dt
There is a relationship. The rotational speed ω is desirably constant with no error with respect to a specified value, but may vary with time as shown in FIG. Such fluctuations in the rotational speed are undesirable because they cause vibrations and the like, but by applying this embodiment, the fluctuations from the indicated value of the rotational speed can also be cancelled. In this case, as shown in FIG. 4, if the rotation angle is input to the differentiation circuit 14, the rotation speed can be obtained.

  As described above, according to the present embodiment, the basis vector generation block 2 generates and outputs a set of harmonics with respect to the fundamental frequency of the drive signal output to the motor 6 as the basis vectors x_1 to x_n. The coefficient generation block 3 performs an error minimizing calculation using a detection signal obtained from a noise sensor such as a microphone 11 or an acceleration sensor 13 and a step size parameter μ, which detects noise generated when the motor 6 is driven. The coefficients w_1 to w_n corresponding to the respective base vectors x_1 to x_n are generated while being updated. Here, “noise generated when the motor is driven” includes “vibration” and the like.

  Then, when the product-sum block 4 performs the product-sum operation on the basis vectors x_1 to x_n and the corresponding coefficients w_1 to w_n, the signal obtained as the operation result and the drive signal are added or subtracted, whereby the motor 6 Cancels noise generated during driving. That is, the detection signal obtained from the noise sensor becomes an error signal as a result of including noise in the drive system of the motor 6. On the other hand, when the coefficient generation block 3 performs the error minimization calculation and updates the coefficients w_1 to w_n, the error signal comes close to the zero level, and noise generated when the motor 6 is driven is reduced.

  Therefore, it is not necessary to prepare a high-precision simulation model in advance or to modify the model by considering individual noise fluctuation factors in advance, and reduce the noise generated when the motor 6 is driven by real-time processing. it can. In addition, by applying the driving signal to a sine wave, noise generated when the motor 6 is driven in a sine wave can be reduced.

(Second Embodiment)
Hereinafter, the same parts as those in the first embodiment are denoted by the same reference numerals, description thereof will be omitted, and different parts will be described. The phase of the actually generated noise is rotated even at the same frequency, and it does not always coincide with the phase of the drive signal and can take various phases. Therefore, it is desirable that an arbitrary phase can be given to a waveform generated for noise removal.

  Therefore, in the noise cancellation device 21 of the second embodiment shown in FIG. 11, a base vector having two orthogonal phases is prepared for a base vector for a harmonic of a certain frequency, and the amplitudes thereof can be adjusted independently. And By combining these two basis vectors, an arbitrary phase can be given to the basis vectors for each harmonic.

  The basis vector generation block 22 outputs two basis vectors x_1 to n_I (T) and x_1 to n_Q (T) having phases different from each other by 90 ° for each basis vector to the product-sum block 24. In the coefficient generation block 23, the coefficients w_1 to n_I and w_1 to n_Q are updated by the LMS algorithm and output to the product-sum block 24. The product-sum block 24 performs a product-sum operation on each of the basis vectors x_1 to n_I (T), x_1 to n_Q (T) and the corresponding coefficients w_1 to n_I, w_1 to n_Q, and is generated as an operation result. The signal is output to the subtracter 5.

  As shown in an example in FIG. 12, there are two basis vectors x_2_I (T) and x_2_Q (T) having a phase difference of 90 ° with respect to the second harmonic having the same frequency f2. By adjusting these amplitudes independently by the coefficients w_2_I and w_2_Q, the combined waveform has an arbitrary phase at the same frequency f2.

  Since x_n_I and x_n_Q are also orthogonal, each becomes an independent basis vector. Therefore, the number of basis vectors increases as shown above, but the other circuit configurations are exactly the same. Note that the fundamental waves x_1_I (T) and x_1_Q (T) are not generated for noise reduction as in the first embodiment, but are necessary for the motor drive signal and integrated with the drive signal generation circuit. Easy to do.

  Further, although the phase difference between x_n_I and x_n_Q is 90 °, the case of 90 ° is the most independent. As long as the phase difference is not 0 ° or 180 °, it is possible to generate a signal having an arbitrary phase by vector addition. Therefore, the phase difference is not necessarily limited to 90 °.

Next, the result of having confirmed the effect of 2nd Embodiment by simulation is shown. The simulation condition 1 is as follows, and FIG. 13 shows a time waveform corresponding to the condition 1. Note that μ is a step size parameter used in the LMS algorithm.
<Condition 1>
Drive signal frequency: F1 [Hz]
Noise component frequency: 3F1 [Hz]
In-phase: 45 °
Same amplitude: 20% of drive signal
μ: 0.003
Of the LMS algorithm for the control period T
Update frequency: F1 / 720 [Hz]

As shown in FIG. 14, it can be seen that the generated waveform converges to the same waveform as the noise waveform over time. Further, the finally converged coefficient value shown in FIG. 15 corresponds to the noise component frequency being 3F1 [Hz], and the component is obtained. Further, w_3_I and w_3_Q have the same amplitude so as to correspond to the phase 45 ° of the noise component, and a phase of 45 ° is generated by their vector sum.
<Condition 2>
Drive signal frequency: F1 [Hz]
Noise component frequency: 5F1 [Hz]
In-phase: 60 °
Same amplitude: 15% of drive signal
μ: 0.003
Update frequency of LMS algorithm: F1 / 720 [Hz]

  Also in this case, it can be seen that the generated waveform converges to the noise waveform as shown in FIG. Further, from the finally converged coefficient shown in FIG. 18, the same 5 × F1 [Hz] coefficient as the noise component is increased, and the amplitude ratio between w — 5 — I and w — 5 — Q is adjusted corresponding to the noise phase of 60 °.

  If an approximate value that can be expressed by a power of 2 is used instead of the value 0.003 of the step size parameter μ, the value is 0.00390625 or 0.001953125. The simulation results using these values are almost the same as when μ = 0.003.

  As described above, according to the second embodiment, the basis vector generation block 22 generates two types of vectors x_1 to n_I (T) and x_1 to n_Q (T) that are different in phase by 90 ° as the basis vectors. I tried to do it. By combining these two orthogonal vectors, noise having an arbitrary phase component can also be suppressed.

(Third embodiment)
In the first and second embodiments, the waveform of each base vector is a sine wave. However, since the circuit for generating the sine wave is large in scale and needs to be generated for all candidate harmonics, it causes an increase in cost. Therefore, as shown in FIGS. 19 and 20, in the basis vector generation block 32 of the noise cancellation device 31, for example, a rectangular wave r_n having the same fundamental frequency is used instead of x_n (T) as a basis vector corresponding to the nth harmonic. (T) is used. In this case, since the rectangular wave r_n (T) includes harmonic components other than x_n (T), even if the n-th harmonic noise is removed by r_n (T), an extra r_n (T), for example, 3n Noise due to the second harmonic remains.

  However, the 3n-order harmonic of r_n (T) is removed by the basis vector of r_3n (T) having this frequency as a fundamental wave. Similarly, the 9n-th harmonic, which is a further harmonic of r_3n (T), is removed by r_9n (T). Thereafter, the harmonic component is recursively converted into a higher-order harmonic having a small amplitude, and finally becomes a frequency lower than the detection limit or higher than the audible range. For example, the Fourier transform is a technique for expressing a rectangular wave or the like as a set of sine waves, but the third embodiment can be said to be a technique for expressing a sine wave as a set of rectangular waves.

  According to the third embodiment, it is not necessary to generate a sine wave that increases the circuit scale, and only a +1 and −1, that is, a rectangular wave that can be expressed by 1 bit, may be generated. Can be reduced. It is also easy to implement the third embodiment in combination with the configuration of adding the I and Q vectors of the second embodiment.

Next, a simulation result to which the third embodiment is applied is shown. In the coefficient generation block, a sign-sign LMS algorithm is used.
<Condition 2>
Drive signal frequency: F1 [Hz]
Noise component frequency: 2F1 [Hz]
In-phase: 15 °
Same amplitude: 30% of drive signal
μ: 0.0002
Algorithm update frequency: F1 / 720 [Hz]

  Here, the motor drive signal shown in FIG. 21 is a sine wave x_1 (T) similar to that of the first embodiment. The base vector generation block 32 shown in FIG. 19 outputs a square wave base vector r_1 (T). However, it may be converted into a sine wave, or a sine wave base vector x_1 (T) may be converted in advance. It may be output.

  As shown in FIG. 22, since there is an upper limit to the frequency that can be prepared as a basis vector, the remaining noise component is larger than that in the second embodiment, but at least the second harmonic component can be synthesized almost correctly. Since the frequency region in which the basis vector cannot be prepared is a frequency that does not cause a problem because it exceeds the audible range in the first place, the remaining noise component does not cause a problem as well. Further, as a feature different from the second embodiment, as shown in a portion surrounded by a broken line in FIG. 23, the 6th, 10th and 14th harmonic components are seen in addition to the 2nd harmonic. As described above, this is a result of generating a sine wave of the second harmonic by combining these order rectangular waves.

  If an approximate value that can be expressed by a power of 2 is used instead of the value 0.0002 of the step size parameter μ, 0.000244140625 or 0.0001220703625 is obtained. The simulation results using these values are almost the same as when μ = 0.0002.

  As described above, according to the third embodiment, the basis vector generation block 32 outputs the rectangular wave signals r_n (1) to r_n (T) having the corresponding harmonics as fundamental waves as basis vectors. The circuit scale can be further reduced. The motor 6 may be driven using a rectangular wave motor drive signal r_1 (T).

(Fourth embodiment)
For example, in a brushless DC motor, the rotor is provided with a permanent magnet, and a rotating magnetic field with an angular velocity ω1 is generated from the rotor rotating at the angular velocity ω1. On the other hand, when a harmonic signal of the angular velocity ω2 is superimposed on the stator coil in order to reduce noise, the magnetic field generated by the coil modulates the angular velocity ω2. In this case, the attractive force / repulsive force determined by the product of the rotor magnetic field and the stator magnetic field is, for example, in a certain direction component,
(Rotating force) ∝ (Rotating magnetic field) x (Stator magnetic field)
∝sin (ω1 · t) × sin (ω2 · t + θc)
∝cos {(ω1 + ω2) · t + θc} −cos {(ω1−ω2) · t−θc}
... (2)
It becomes. Where θc is the phase angle difference between the rotor and the stator at time 0.

  That is, even if the harmonic wave ω2 is injected, the modulation of the rotational force at the angular frequency ω2 cannot be obtained, and the frequency conversion that results in the modulation at (ω1 + ω2) or (ω1-ω2) occurs. Therefore, when noise of the angular frequency ωn is generated, it is necessary not to superimpose the harmonic ωn but to superimpose a waveform that has the angular frequency ωn after frequency conversion.

  In FIG. 24, the signal sin (ω1 · t) multiplied by the generated waveform by the multiplier 33 means frequency conversion, and there is no direct corresponding multiplier. In this case, the sensor detection signal is also in a state where sin (ω1 · t) is multiplied by the generated waveform. On the other hand, in the LMS algorithm, it is assumed that the sensor detection signal includes only an error between the generated waveform and the noise waveform. Therefore, it cannot be applied as it is to the relationship between the sensor detection signal multiplied by the signal sin (ω1 · t) and the generated waveform. Therefore, it is conceivable that the relationship between the sensor detection signal and the generated waveform is restored by dividing the sensor detection signal by sin (ω1 · t).

  However, the range of sin (ω1 · t) is in the range of −1 to +1. When the value becomes 0, sin (ωn · t) / sin (ω1 · t) diverges infinitely. In addition, even if it does not completely become zero, the value of sin (ωn · t) / sin (ω1 · t) becomes very large in the vicinity thereof, so that calculation becomes difficult. Therefore, in the fourth embodiment, the sensor division signal also uses the multiplier 34 as shown in FIG.

Thus, if the generated waveform sin (ω2 · t + θc) is considered excluding the noise waveform, the sensor detection waveform after conversion is sin (ω2 · t + θc) × sin (ω1 · t) × sin (ω1 · t).
= Sin (ω2 · t + θc) × 1/2 {1-cos (2 · ω1 · t)}
= 1 / 2sin (ω2 ・ t + θc)
-1 / 2sin (ω2 · t + θc) · cos (2 · ω1 · t) (3)
It becomes. That is, the first term on the right side of equation (3) is multiplied by ½, but the converted sensor detection waveform has the same frequency component as the generated waveform. Therefore, the conventional LMS algorithm can be applied as it is.

  The second term on the right side means that the noise frequency is converted to another frequency, but as described above, that frequency is recursively canceled using another harmonic. Finally, it is converted to a frequency that has no effect. Note that ω1 is an angular frequency corresponding to the rotor rotational speed, and x_1_I is a signal applied to the stator. However, when the motor is rotating at a constant speed, the frequency of x_1_I and the rotational frequency of the rotor coincide with each other. Desired characteristics can be obtained by this method.

  As described above, according to the fourth embodiment, the multiplier 34 that inputs the result of multiplying the sensor detection signal by the drive signal sin (ω1 · t) to the coefficient generation block 3 is provided, so that noise is removed after frequency conversion. The applied harmonics before the frequency conversion can be generated so that the signal becomes the same.

(Fifth embodiment)
As shown in FIG. 25, the noise canceling device 41 of the fifth embodiment is obtained by applying the configuration of the fourth embodiment to the noise canceling device 21 of the second embodiment, and performs multiplication on the input side of the coefficient generation block 23. The device 34 multiplies the sensor detection signal and x_1_I (T). Here, when only the sign-sign LMS algorithm is used in the coefficient generation block 23, for example, only the sign of the multiplication result of the multiplier 34 is used. Therefore, instead of multiplying by x_1_I (T), the same result can be obtained by multiplying by the square wave base vector r_1_I (T) as in the sixth embodiment described below. Thereby, the circuit scale can be reduced.

(Sixth embodiment)
As shown in FIG. 26, the noise cancellation apparatus 51 of 6th Embodiment applies the structure of 4th Embodiment to the noise cancellation apparatus of 2nd and 3rd Embodiment as mentioned above. In this case, the base vector generation block 22R generates an IQ orthogonal vector for each of the square wave base vectors r_1 to r_n. Further, the motor drive signal multiplied by the sensor detection signal is not limited to r_1_I (T), but may be x_1_I (T). Even in this case, x_1_I (T) is originally used for driving the motor and is not newly generated, so the circuit scale is not increased.

The simulation conditions shown in FIGS. 27 and 28 are as follows.
Drive signal frequency: F1 [Hz]
Noise component frequency: 8F1 [Hz]
In-phase: 45 °
Same amplitude: 20% of drive signal
μ: 0.001
Algorithm update frequency: F1 / 720 [Hz]

The purpose is to obtain a signal waveform close to the noise waveform shown in FIGS. 27, 28B, 28D, etc. after frequency conversion. As shown in FIG. 28B, it can be seen that the product of the waveform generated in the sixth embodiment and the motor drive signal is asymptotic to the noise waveform. In the sixth embodiment, as shown in FIG. 29, the waveform after frequency conversion needs to be a sine wave, and the waveform superimposed on the drive signal is not a single sine wave as in the second embodiment. , Generated by combining multiple sine waves.
If an approximate value that can be expressed by a power of 2 is used instead of the value 0.001 of the step size parameter μ, 0.0009765625 is obtained.

(Seventh embodiment)
As stated in the literature below,
http://toshiba.semicon-storage.com/en/designsupport/e-learning/brushless_motor/chap3/1274521.html
In general, driving a motor with a sine wave causes lower vibration and lower noise than rectangular wave driving, but there is a problem that the circuit scale increases as a price.

  However, since the vibration and noise can be automatically reduced by applying the technique of the present invention, the original drive signal does not necessarily need to be a sine wave. Even in the case of the rectangular wave drive, it finally converges to a state where the detection noise is the smallest (close to a sine wave but a detection noise lower than that of the sine wave). Therefore, low noise driving can be realized while achieving a significant circuit reduction.

In each of the above embodiments, for the sake of simplicity, the angular frequency of the rotating magnetic field generated when the rotational angular velocity of the rotor is ω1 is also ω1.
http://toshiba.semicon-storage.com/jp/design-support/e-learning/brushless_motor/chap2/1274507.html
However, as explained, in a normal brushless DC motor, the rotor has a plurality of poles, and the stator has a plurality of slots. This term and its definition differ depending on the manufacturer, etc., but here, the case of 4 poles shown in the above document is the number of pole pairs p = 2.

In the case of the number of pole pairs p, if the rotational angular frequency of the rotor is ω1, the rotating magnetic field is p · ω1. The number of pole pairs p is a fixed value that does not change depending on the operating state. When the number of pole pairs p is introduced, for example, as shown in FIG. 30, the frequency conversion parameter sin (ω1 · t) of the sensor detection signal in the fourth embodiment is set. It may be replaced with the parameter sin (p · ω1 · t).
As described above, according to the seventh embodiment, the frequency of the motor drive signal is multiplied by the number of pole pairs of the motor 6 and supplied to the multiplier 34, so that the present invention can be applied to the motor 6 having the number of pole pairs p.

(Eighth embodiment)
In each of the above-described embodiments, it is a principle that base vectors corresponding to all harmonic components are prepared. However, in the case of the three-phase drive most used in a brushless DC motor, even if a harmonic multiple of 3 is applied, the same waveform is applied to the three-phase coil, so that the rotational force is not affected. Even so, the contribution is very low.

  In FIG. 31, with respect to the motor driving voltages Vu, Vv, and Vw for three phases, the quadruple harmonic components are indicated by V4u, V4v, and V4w. In this case, the 4th harmonics V4u, V4v, and V4w are clearly different waveforms, while the 3rd harmonics V3u, V3v, and V3w shown in FIG. 32 overlap each other.

  In general, many brushless DC motors employ star connection. In this case, even if the same waveform is given as the third harmonics V3u, V3v, and V3w, the neutral point is also at the potential, so there is no effect on the flowing current. In other words, these components can be very large in operation because they have no effect. Therefore, as in the noise canceling device 61 shown in FIG. 33, if the harmonics corresponding to the triple order are deleted from the basis vectors in the basis vector generation block 62, the remaining basis vectors will cause noise. This is suitable because it is controlled to an optimum state for reduction.

  In practice, the three-phase physical structure of the motor is not completely symmetric, and the application of the third harmonic may be necessary due to its asymmetry. In that case, even if the corresponding basis vector is not completely deleted as in the noise canceling device 61, the upper limit of the triple base vector is set so that the value becomes too large and the amplitude of the synthesized wave is not saturated. Set the coefficient value corresponding to the higher-order third-order harmonic in the coefficient generation block 3 to be smaller, or set the value of the step size parameter μ to be slower than other basis vectors. Setting a smaller value is also effective.

  As described above, according to the eighth embodiment, the basis vector output generation block 62 stops outputting the basis vector corresponding to the third harmonic with respect to the fundamental frequency when the motor has a three-phase configuration. Therefore, it is possible to prevent the harmonic component of the third order from becoming excessive when the noise canceling operation is performed.

  Further, in the coefficient generation block 3, the upper limit of the coefficient corresponding to the basis vector whose degree is a multiple of 3 is set lower than the coefficients corresponding to the other basis vectors, or the value of the corresponding step size parameter μ is set to The same effect can be obtained by setting the value smaller than the values corresponding to other basis vectors.

  The present invention is not limited to the embodiments described above or shown in the drawings, and can be appropriately modified or expanded without departing from the scope of the invention.

  1 Noise canceling device, 2 basis vector generation block, 3 coefficient generation block, 4 product-sum block, 5 subtractor.

Claims (16)

A basis vector output unit (2, 22, 32, 62) for generating and outputting a set of harmonics with respect to the fundamental frequency of the drive signal output to the motor (6) as a basis vector;
By performing an error minimizing operation using a detection signal obtained from a noise sensor (9, 11, 12) for detecting noise generated when the motor is driven and a step size parameter μ, each of the base vectors is handled. A coefficient generator (3, 23) that generates the coefficients while updating them;
A product-sum operation unit (4, 24) for performing a product-sum operation on the basis vector and a coefficient corresponding to each basis vector;
A motor noise canceling device that cancels the noise by adding a signal obtained as a result of the calculation to the driving signal or subtracting the signal from the driving signal.   The motor noise canceling device according to claim 1, wherein the basis vector output unit (32) outputs a rectangular wave signal having a harmonic corresponding to each as a fundamental wave as the basis vector.   The motor noise canceling device according to claim 1, wherein the driving signal is applied to a sinusoidal signal.   The motor noise canceling device according to claim 1, wherein the driving signal is applied to a rectangular wave.   The motor noise canceling device according to any one of claims 1 to 4, wherein the basis vector output unit (22) generates at least two types of vectors having the same frequency and different phases as each basis vector.   The motor noise canceling device according to claim 5, wherein a phase difference between the two vectors is 90 °.   The motor noise canceling device according to any one of claims 1 to 6, further comprising a multiplier (33) that inputs a result obtained by multiplying the detection signal by the drive signal to the coefficient generation unit.   The motor noise canceling device according to claim 7, wherein the frequency of the drive signal is multiplied by the number of pole pairs of the motor and supplied to the multiplier.   8. The motor noise canceling device according to claim 7, wherein the multiplication is simplified by indicating at least one of the detection signal and each of the basis vectors with a sign.   10. The basis vector output unit (62) according to any one of claims 1 to 9, wherein when the motor has a three-phase configuration, the basis vector output unit (62) stops outputting a basis vector corresponding to a third harmonic with respect to the fundamental frequency. The motor noise canceling device described.   When the motor has a three-phase configuration, the coefficient generation unit (3) sets an upper limit of a coefficient corresponding to a base vector whose degree is a multiple of 3 lower than coefficients corresponding to other base vectors. The motor noise cancellation apparatus as described in any one of Claim 1 to 9.   When the motor has a three-phase configuration, the coefficient generation unit (3) sets the value of the step size parameter μ corresponding to the basis vector whose order is a multiple of 3 to the value corresponding to the other basis vectors. The motor noise canceling device according to any one of claims 1 to 9, wherein the motor noise canceling device is set to be smaller.   The motor noise canceling device according to any one of claims 1 to 12, wherein the step size parameter μ is set to a value represented by a power of two.   The motor noise canceling device according to any one of claims 1 to 13, wherein the noise sensor is a microphone (11).   The motor noise canceling device according to any one of claims 1 to 13, wherein the noise sensor is an acceleration sensor (12). When the noise sensor is a rotational speed sensor (9), a differentiator (14) for differentiating a detection signal obtained from the rotational speed sensor,
The motor noise canceling device according to any one of claims 1 to 13, wherein a differentiation result obtained by the differentiator is input to the coefficient generation unit.

Images