JPH1144539A - Signal processor for inertia sensor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a signal processor for inertia sensor which can detect a physical amount, e.g. an angle or an angular speed, with small error while shortening the waiting time. SOLUTION: A stop determining section 3 determines the stationary state of a piezoelectric signal gyro 2 after turn on power, and starting drifts δ1 , δ2 , δ3 being outputted from an A/D converting section 5 during that interval are sampled at a predictive value operating section 6. The predictive value operating section 6 operates the predictive value d(t) of a subsequent starting drift based on the sampling data δ1 , δ2 , δ3 . Since the predictive value d(t) of a starting drift is subtracted from a sensor output signal S(t) at a synthesizing section 12, an integrating section 12 receives the sensor output signal S(t) from which the starting drift is subtracted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a signal processing device for an inertial sensor. More specifically, the present invention relates to an inertial sensor using a piezoelectric body, for example, a three-dimensional data input device using a vibration gyro or an acceleration sensor, a pointing device, a signal processing device of an inertial sensor used for a car navigation system, or the like.

[0002]

[Prior art]

(Piezoelectric Vibration Gyro) An example in which a vibration gyro using a piezoelectric body (hereinafter referred to as a piezoelectric vibration gyro) among the inertial sensors will be described. Piezoelectric vibration gyro is
Used to detect changes in the angle and direction,
An angular velocity signal is output from the piezoelectric vibrating gyroscope.
The angular velocity signal output from the piezoelectric vibrating gyroscope is converted into a digital signal by an A / D (analog / digital) converter after passing through a predetermined analog circuit such as an amplifier, a buffer, and a filter. When the angular velocity signal converted into a digital signal is used as an angle signal, the microprocessor (CPU) normally performs integration processing by a software algorithm using a software algorithm to convert the signal into an angle signal.

[0003]

[Problems to be solved by the invention]

(Start-up drift) However, in such a piezoelectric vibrating gyroscope, an output fluctuation is shown immediately after the power is turned on. For example, when the power is turned on at time t = 0 and an angular velocity is applied to the piezoelectric vibrating gyroscope, the sensor output signal output from the piezoelectric vibrating gyroscope changes as shown in FIG. Here, when the power is turned on at time t = 0 in a state where the piezoelectric vibrating gyroscope is fixed so that the angular velocity is not applied to the piezoelectric vibrating gyroscope, only the output fluctuation not including the angular velocity signal is output from the piezoelectric vibrating gyroscope. The state of the sensor output signal at this time is shown in FIG. Therefore, it can be seen that, in the sensor output signal of the piezoelectric vibrating gyroscope, the output fluctuation as shown in FIG. 2 is usually superimposed on the angular velocity signal indicating the angular velocity applied to the piezoelectric vibrating gyroscope. This output fluctuation is called a start drift, and as shown in FIG. 2, power is turned on (t = 0).
Immediately after that, during the time T1, the startup drift continues.

[0004] The start-up drift is caused by fluctuations in circuit constants in various parts of the circuit due to a rise in temperature after energization, and physical properties caused by continuous stress being applied to an adhesive for attaching the piezoelectric element to the vibrating element. Changes, fluctuations in Qm (mechanical Q) of the resonator element itself, fluctuations in the piezoelectric constant, etc.
It is thought that various factors overlapped. For this reason, the manner in which the starting drift changes differs for each piezoelectric vibrating gyroscope, and the duration T1 of the starting drift also differs.

When an angular velocity signal (digital signal) is integrated with a piezoelectric vibrating gyroscope having such a start-up drift immediately after power-on, angle data and velocity data not originally generated are integrated and output. And may hinder accurate angle detection and speed detection.

In order to avoid such inconvenience, in the conventional signal processing device, the piezoelectric vibrating gyroscope is kept stationary for a certain period of time immediately after the power is turned on, and the starting drift duration T
Waiting for the start-up drift to become stable after 1 has elapsed, processing such as integration is performed by subtracting the start-up drift stable value δ _inf from the sensor output signal as an angular velocity signal.

However, in such a method in which the piezoelectric vibrating gyroscope is kept stationary for a certain period of time immediately after the power is turned on, it is necessary to wait several minutes until the starting drift is stabilized after the power is turned on. In such a case, there is a disadvantage that the angle detection sensor must be kept stationary for a few minutes during that time.

As a method of reducing the influence of the starting drift, there is a method of removing an unnecessary direct current (DC) component by inserting a high-pass filter in an output section of the piezoelectric vibrating gyroscope.

However, in the method using a high-pass filter, when a steady turning motion or a constant acceleration motion is applied, the angular velocity signal and the acceleration signal of the DC component are removed, and even the signal originally required is removed unnecessarily. Will be
This could hinder accurate angle and speed detection.

(Temperature Drift) In such a piezoelectric vibrating gyroscope, an offset variation (hereinafter referred to as an output characteristic) shown in FIG. , Temperature drift).

[0011] Even when such a temperature drift occurs, if the sensor output signal (digital signal) is integrated, erroneous angle data and speed data are integrated and output, and accurate angle detection and speed detection are performed. Had trouble.

The present invention has been made in view of the above-mentioned drawbacks of the conventional example, and has as its object to detect a physical quantity such as an angle or acceleration with a small error in a shorter standby time. Another object of the present invention is to provide a signal processing device of an inertial sensor capable of detecting an angle, a speed, and the like without impairing necessary signal components even in a case of a steady turning motion or a constant acceleration motion.

[0013]

A signal processing device for an inertial sensor according to claim 1 is a signal processing device for an inertial sensor that performs signal processing by removing unnecessary signal components from a sensor output signal containing unnecessary signal components. A prediction that measures a sensor output signal output from the inertial sensor while the stationary state is maintained after the power supply of the inertial sensor is turned on, and predicts a value of an unnecessary signal component thereafter based on the obtained measurement data. And a signal synthesizing unit for subtracting the predicted value of the unnecessary signal component predicted by the prediction calculating unit from the sensor output signal output from the inertial sensor.

In the signal processing device for an inertial sensor according to the first aspect, after the power is turned on, an unnecessary signal output from the inertial sensor is measured while the inertial sensor is kept stationary. The value of the unnecessary signal in the future can be predicted from the value of the unnecessary signal sampled during the period. Therefore, after the inertial sensor starts operating, by subtracting the predicted value of the unnecessary signal from the sensor output signal, a true sensor output signal that does not include the unnecessary signal can be obtained.

Therefore, the influence of the unnecessary signal can be reduced, and the measurement accuracy of the inertial sensor can be improved. Moreover, in the present invention, the influence of the unnecessary signal is reduced by predicting the unnecessary signal, so that it is not necessary to keep the inertial sensor stationary for a long time as in the conventional method of waiting until the start-up drift is stabilized. In addition, it is possible to immediately shift to the measurement operation after the power is turned on. Further, unlike the method using the high-pass filter, there is no possibility that even the signal that is originally required is unnecessarily removed.

The prediction of the unnecessary signal is performed by approximating a sensor output signal output from the inertial sensor kept in a stationary state by a power series, or by comparing the sensor output signal with unnecessary signal data held in data holding means. Or by approximating with a logarithmic function.

According to a second aspect of the present invention, there is provided a signal processing apparatus for an inertial sensor which performs signal processing by reducing offset fluctuations from a sensor output signal including offset fluctuations due to environmental changes. Means for extracting a low-frequency component from a sensor output signal output from the inertial sensor and comparing the sensor output signal with the low-frequency component to determine whether the inertial sensor is in a stationary state; and When it is determined that it is in a stationary state, prediction calculation means for predicting the value of the subsequent offset fluctuation based on the low frequency component of the sensor output signal,
Signal synthesis means for subtracting the predicted value of the unnecessary signal component predicted by the prediction calculation means from the sensor output signal output from the inertial sensor.

In the signal processing device for an inertial sensor according to the second aspect, it is determined whether the inertial sensor is stationary by comparing the sensor output signal with its low frequency component, and sampling is performed while the inertial sensor is stationary. The value of the offset fluctuation in the future can be predicted from the value of the offset fluctuation such as the temperature drift. Therefore, by subtracting the predicted value of the offset variation from the sensor output signal after the inertial sensor starts operating, a true sensor output signal that does not include the offset variation can be obtained.

Therefore, the influence of the offset fluctuation can be reduced, and the measurement accuracy of the inertial sensor can be improved. Moreover, in the present invention, since the influence of the offset fluctuation is reduced by predicting the offset fluctuation, there is no need to wait for a long time by keeping the inertial sensor in a stationary state as in the conventional method of waiting until the startup drift is stabilized. After the power is turned on, the operation can be shifted to the measurement operation in a short standby time. Further, unlike the method using a high-pass filter, even signals that are originally required are not unnecessarily removed.

In predicting the offset fluctuation, the sensor output signal output from the inertial sensor kept stationary is approximated by a power series, or the sensor output signal is compared with the offset fluctuation data held in the data holding means. Or by approximating with a logarithmic function.

[0021]

BEST MODE FOR CARRYING OUT THE INVENTION

(First Embodiment) FIG. 4 shows a signal processing device 1 for a piezoelectric signal gyro according to an embodiment of the present invention, wherein a sensor output signal S from a piezoelectric vibration gyro (gyro sensor) 2 is provided.
1 shows a signal processing device 1 for processing (t).
The signal processing device 1 includes an analog circuit unit 4 including an amplifier, a buffer, a filter, and the like, an A / D conversion unit 5, a stop determination unit 3,
Prediction calculation unit 6, timer 7 for measuring elapsed time after power-on,
A sampling time measuring timer 8, a memory 9, and an integration processing unit 10 are provided.

A sensor output signal S (t) output from the piezoelectric vibrating gyroscope 2 passes through an analog circuit section 4 including an amplifier, a buffer, a filter, and the like, and is converted into a digital signal by an A / D converter 5. . The sensor output signal S (t) converted to a digital signal by the A / D converter 5 [for convenience, time t
Are branched in three directions in the signal branching unit 11, the sensor output signal S (t) branched to one is guided to the prediction operation unit 6, and the sensor output signal S branched to the other. (T) is led to the stop judging unit 3, and the sensor output signal S (t) branched to the other is supplied to the combiner 12.
Led to.

The timer 7 for measuring the elapsed time after the power is turned on is
During a predetermined time T2 from when the power of the signal processing device 1 is turned on, a timer signal q is output to the stop determination unit 3 and the prediction calculation unit 6. The predetermined time period T2 is a time period after the power is turned on, in which the piezoelectric vibrating gyroscope 2 does not start moving sufficiently and is considered to be kept almost stationary. Even when the piezoelectric vibrating gyroscope 2 is turned on, the angular velocity is rarely applied immediately immediately. For example, in a pointing device such as a computer, the piezoelectric vibrating gyro 2
Is normally operated, and the power is not turned on during the operation of the piezoelectric vibrating gyroscope 2 whose power is off. When used in a car navigation system or the like, it is common to turn on the power and then start the car. Therefore, after the power of the piezoelectric vibrating gyroscope 2 is turned on, the normal angular velocity is not applied to the piezoelectric signal gyroscope 2 for a certain period of time, and the time T2 is shorter than the normal piezoelectric vibrating gyroscope stop time when the power is turned on. Can be determined.

The stop judging section 3 sets the time within this time T2.
This is to confirm that the piezoelectric vibrating gyroscope 2 is stopped and no angular velocity is applied, and it is usually based on the fact that the starting drift is a signal having a lower frequency than the angular velocity signal and a smaller signal than the angular velocity signal. And That is, for a predetermined time T2 which is measured by the power-on after the elapsed time measuring timer 7, stop determination unit 3, the sensor output signal value S ₀ immediately after power based S (t)
Is within the threshold value ± Vth [that is, -Vth ≦ S
(T) −S ₀ ≦ + Vth], the sensor output signal S (t) does not include the angular velocity signal, and it is determined that the piezoelectric vibrating gyroscope 2 is stopped. If the difference exceeds the threshold value ± Vth, [S (t) −S ₀ ≦ −Vth or + Vth ≦ S (t)
-S ₀ ], it is determined that the sensor output signal S (t) includes the angular velocity signal and the piezoelectric vibrating gyroscope 2 is not stopped. The result of the determination by the stop determination unit 3 is transmitted to the prediction calculation unit 6, and when it is determined that the piezoelectric vibrating gyroscope 2 is stopped, the prediction calculation unit 6 predicts and calculates the starting drift and performs the sensor output signal S. Correct (t). However, when it is determined that the piezoelectric signal gyro 2 is not stopped even within the predetermined time T2, the prediction calculation unit 6 does not perform the prediction calculation of the starting drift, and corrects the sensor output signal S (t). Also do not.

When the prediction calculation unit 6 receives the stop determination signal from the stop determination unit 3, the prediction calculation unit 6 receives the timer signal q from the timer 7 for measuring the elapsed time after power-on (T2 time after power-on). ), The sensor output signal S (t), that is, the startup drift δ (t) is sampled at a constant time interval ΔT. This sampling time is defined as t ₁ , t ₂ (=
Assuming that t ₁ + ΔT) and t ₃ (= t ₂ + ΔT), the prediction calculation unit 6 calculates three startup drift values δ (t ₁ ) and δ
(T ₂ ), δ (t ₃ ) are sampled, and the sampling data δ (t ₁ ), δ (t ₂ ), δ (t
Based on ₃ ), a value δ (t) of the starting drift (unnecessary signal component) after a predetermined time T2 has elapsed is calculated. The sampling time interval ΔT is measured by the sampling time measuring timer 8. On the other hand, during the reception time T2 of the timer signal q, the prediction calculation unit 6 receives the sampling data δ (t ₁ ), δ (t ₂ ),
δ (t ₃ ) is output to the synthesizer 12 as it is, and when the reception time T2 of the timer signal q elapses, the predicted start-up drift value (estimated value) d (t) is output to the synthesizer 12. Note that since the angular velocity signal ω (t) has a higher frequency than the startup drift, a low-pass filter (LPF) may be inserted before the prediction calculation unit 6 to cut the angular velocity signal ω (t).

The other sensor output signal S (t) output from the A / D conversion unit 5 and branched by the signal branching unit 11 is used as a sampling data of the starting drift output from the prediction operation unit 6 in the synthesizer 12. δ (t ₁ ), δ (t ₂ ), δ
(T ₃ ) [t ₁ , t ₂ , t ₃ ≦ T ₂ ] or the starting drift prediction value d (t) [t> T 2] is subtracted. In this way, from the sensor output signal S (t), the predicted value d of the starting drift is calculated.
Correction signal A (t) from which (t) is subtracted = S (t) -d
(T) [this is substantially equal to the angular velocity signal ω (t) as described later] is input to the integration processing unit 10 and the integration processing unit 1
At 0, it is converted to an angle signal.

The components of the signal processing apparatus 1 other than the A / D converter 5 are generally implemented as software algorithms using a microprocessor. In addition, a one-chip microcomputer (CP) including an A / D converter 5
U) may be used.

Next, the principle of removing only the starting drift by the signal processing device 1 of the present invention to obtain only the angular velocity signal and outputting highly accurate angle information will be described. In the following description, a sensor output signal or the like is expressed as an analog signal for convenience.

When the power of the signal processing device 1 is turned on, the timer 7 for measuring the elapsed time after the power is turned on is started, and the stop judging unit 3 judges whether the piezoelectric vibrating gyroscope 2 is in a stationary state. During the time T2 during which the timer signal q is being output, the stop judging unit 3 sets the piezoelectric vibration gyro 2
Is determined to be stationary, the piezoelectric vibrating gyroscope 2
It is determined that the sensor output signal S (t) of only the starting drift δ (t) that does not include the angular velocity signal is output from. At this time, the prediction calculation unit 6 samples the startup drift δ (t) every sampling time ΔT.

The starting drift δ (t) is considered to be caused by various factors overlapping as described in connection with the conventional example. The drift behaves like the curve shown in FIG. As a result of collecting start-up drift data for a large number of inertial sensors such as piezoelectric vibrating gyroscopes, it was found that the increase in the start-up drift curve tends to decrease at a substantially constant damping rate.

Now, after turning on the power of the signal processing device 1, the sampling times t ₁ , t ₂ (= t ₁ + ΔT), t
₃ (= t ₂ + ΔT) [≦ T2], the starting drift sampled by the prediction calculation unit 6 is represented by δ ₁ = δ (t ₁ ), δ ₂
= Δ (t ₂ ) and δ ₃ = δ (t ₃ ). Times t ₁ and t ₂
Slope of the start drift curve between (δ ₂ -δ ₁₎ / slope of the start drift curve between time t ₂ and t ₃ for [Delta] T ([delta] ₃ -
The attenuation rate κ of δ ₂ ) / ΔT is expressed by the following equation (1).

[0032]

(Equation 1)

The start drift [delta] (t) is also the time t ₃ or later, increases at a constant attenuation factor κ every sampling time [Delta] T is assumed to Yuku is suppressed, the time for each time [Delta] T t _m =
_{t 3 + (m-3)} ΔT [m = 4~n] and t _{m + 1} start drift [delta] _m = [delta] in _{(t m), δ m +1} = δ (t m + 1) deviation of the [delta] _{m +} the ₁ - [delta _m sequentially, is expressed by the following (2A) ~ (2N) equation. δ ₄ −δ ₃ = κ (δ ₃ −δ ₂ )… (2A) δ ₅ −δ ₄ = κ (δ ₄ −δ ₃ ) = κ ² (δ ₃ −δ ₂ )… (2B) δ ₆ −δ ₅ = κ (δ ₅ −δ ₄ ) = κ ³ (δ ₃ −δ ₂ ) (2C)…… δ _n −δ _n-1 = κ (δ _n−1 −δ _n−2 ) = κ ^{n -3} (δ ₃ −δ ₂ ) (2N) By adding these equations (2A) to (2N) for each side, the following equation (3) is obtained.

[0034]

(Equation 2)

Therefore, by substituting the above equation (1) for κ in the above equation (3), the starting drift at time t _n = t ₃ + (n−3) ΔT [where n is an integer of 4 or more] δ _n
It can be represented by a power series such as the following equation (4).

[0036]

(Equation 3)

Therefore, during the time T2 from the time when the power of the signal processing device 1 is turned on, the stop judging section 3 makes the piezoelectric vibration gyro 2
It is determined whether or not the piezoelectric vibrating gyroscope 2 is stationary.
Is in a stationary state and the starting drift δ from the A / D converter 5
If it is determined that only (t) is output, the prediction calculation unit 6 calculates the startup drift value δ during this time T2.
If (t ₁ ), δ (t ₂ ), δ (t ₃ ) are actually measured, the time t _n = t ₃ + (n−3) Δ after the time t ₃ after the power is turned on.
The value δ (t _n ) of the startup drift at T can be predicted based on the equation (4). The two-dot chain line in FIG. 5 [the starting drift is exaggerated for convenience in FIG. 5] indicates a prediction curve of the starting drift δ (t) calculated by the equation (4). . The predicted value of the start-up drift at time t _n (n ≧ 4) obtained based on the actually measured values δ (t ₁ ), δ (t ₂ ), and δ (t ₃ ) is denoted by d (t). From the above equation (4), the predicted value d (t) of the starting drift (where t> t ₃ ) is expressed by the following equation (5).

[0038]

(Equation 4)

On the other hand, when it is determined that the piezoelectric vibrating gyroscope 2 is in a stationary state, the sensor having only the starting drift δ (t) is used at times t _{1 to} t ₃ before the time T ₂ elapses from the power-on. The signal output is being output.
At time t _n (n ≧ 4) after two hours have elapsed, sensor output signal S (t) = ω (t) + δ (t) in which start drift δ (t) is superimposed on angular velocity signal ω (t). Is output. In the stationary state for T2 hours after the power is turned on,
A sensor output signal S (t) = ω (t) + output from the piezoelectric vibrating gyroscope 2 to which the angular velocity has been applied after a lapse of time T2.
An example of δ (t) is shown by a solid line in FIG.

Therefore, when the predicted value d (t) of the starting drift is subtracted from the sensor output signal S (t) = ω (t) + δ (t) by the combiner 12 of the signal processing device 1, the combiner 1
The correction signal A (t) output from 2 is as follows: A (t) = S (t) −d (t) = ω (t) + [δ (t) −d (t)]. Here, if the predicted value d (t) of the starting drift has sufficient accuracy, it is possible to satisfy δ (t) −d (t) で き る 0, so that the correction signal A (t) output from the combiner 12 is obtained. A (t) ≒ ω (t), and the correction signal A (t) represents the angular velocity signal ω (t). The correction signal A thus obtained
(T) or the angular velocity signal ω (t) is shown by a broken line in FIG.

FIG. 6 is a flow chart showing an algorithm for calculating the predicted value d (t) of the start-up drift in the prediction calculation section 6. According to this flowchart, the power of the signal processing device 1 is turned on, the prediction operation unit 6 receives the timer signal q from the elapsed time measuring timer 7 after the power is turned on, and the piezoelectric signal gyro 2 is stopped from the stop determination unit 3. When receiving a signal indicating that the operation is performed (S1), the counter is reset to N = 1 (S2). When the counter is reset, the sampling time measuring timer 8 starts (S3), and waits until a certain sampling time ΔT has elapsed (S4).

When the sampling time measuring timer 8 times out after the elapse of the sampling time ΔT, it is determined whether the value N of the counter is 3 or less (S5). If the value N of the counter is 3 or less, it is confirmed whether or not the timer signal q is received from the timer 7 for measuring the elapsed time after the power is turned on (S6). If the timer signal q is received, the output from the piezoelectric vibration gyro is performed. Starting Drift δ
(T _N ) is sampled (S7). If the timer signal q has not been received, the processing in the prediction calculation unit 6 is stopped because the time T2 has elapsed since the power was turned on (S2).
10).

When the sampling of the startup drift δ (t _N ) is completed, the value of the counter is increased by 1 (S8), the timer 8 for sampling time measurement is reset and restarted (S9), and the processing after step S4 is performed. Try again.

By repeating the above processing,
The prediction calculation unit 6 calculates the starting drifts δ (t ₁ ), δ (t ₂ ), δ
(T ₃ ) is acquired and stored in the memory 9.

Thus, the starting drifts δ (t ₁ ), δ
When the values of (t ₂ ) and δ (t ₃ ) are obtained, the value of the counter becomes N = 4, so the prediction calculation unit 6 branches in step S5 and shifts to a startup drift prediction value calculation process. (S1
1). That is, the prediction calculation unit 6 calculates the starting drifts δ (t ₁ ), δ (t ₂ ), δ already stored in the memory 9.
Using the value of (t ₃ ), the predicted value d (t _N ) of the starting drift is calculated from the above equation (5) and sent to the combiner 12 (S1).
2). When the predicted value d (t) of the start-up drift has been output, the value of the counter is increased by 1 (S13), the timer 8 for sampling time measurement is reset and restarted (S14), and the processing after step S4 is repeated. .

The prediction calculation unit 6 repeats the above-described processing to obtain the start drift predicted values d (t ₄ ) and d (t ₄ ).
(T ₅ ), d (t ₆ ),... Every predetermined time ΔT
Output to

FIG. 7 is a flowchart showing an algorithm for calculating a correction value in the synthesizer 12. Synthesizer 12
Is the sensor output signal S output from the A / D converter 5
When (t) is sampled (S21) and the start-up drift predicted value d (t) output from the prediction calculation unit 6 is received (S22), the correction signal A is calculated based on these values.
(T) = S (t) -d (t) is calculated (S23), and the calculated correction signal A (t) is output to the integration processing unit 10 as the angular velocity signal ω (t).

In the signal processing device 1 of the present invention, if the starting drift is removed from the sensor output signal by such a method, even before the starting drift is stabilized,
Since the starting drift can be predicted and the starting drift can be removed from the sensor output signal, the waiting time T2 until the angular velocity is applied to the piezoelectric vibrating gyroscope 2 can be very short, for example, about 10 seconds.

(Second Embodiment) Next, a signal processing device 1 according to another embodiment of the present invention will be described. The signal processing device 1 according to this embodiment also has a configuration as shown in FIG. However, the prediction calculation method in the prediction calculation unit 6 is different from that of the first embodiment. This embodiment, as described below, measures the actual value δ of the startup drift.
The estimated value of the starting drift is obtained by approximating the starting drift with reference data based on (t ₁ ), δ (t ₂ ), and δ (t ₃ ).

The startup drift output from an inertial sensor such as a piezoelectric vibrating gyroscope determined to be in a stationary state fluctuates as shown in FIG. 2 after the power is turned on. The curve representing the start-up drift differs for each inertial sensor, but shows a similar change tendency. Thus, for a piezoelectric vibrating gyroscope in a stationary state, the starting drift outputted from the piezoelectric vibrating gyroscope was actually measured, and the measured values were subjected to statistical processing to represent the starting drift as several patterns. This pattern is the reference data, and is stored in the memory 9
Is stored within. For example, functions of eight patterns P _{1 to} P ₈ (the prediction accuracy improves as the number of patterns increases) as shown by broken lines in FIG. 9 are stored in the memory 9 as reference data. Hereinafter, this embodiment will be described with reference to the flowchart shown in FIG. 8 and the explanatory diagram of FIG. In FIG. 9, the reference data is exaggerated.

When the power is turned on at t = 0, the timer 7 for measuring the elapsed time after the power is turned on is set to the timer signal q for the time T2.
Is output, and the stop determination unit 3 determines whether or not the piezoelectric vibrating gyroscope 2 is in a stationary state. When it is determined that the piezoelectric vibrating gyroscope 2 is in a stationary state, the prediction calculation unit 6 outputs a sensor output signal output from the A / D conversion unit 5 at times t ₁ , t ₂ , and t ₃ , that is, a startup drift. Measured value δ of
(T ₁ ), δ (t ₂ ), δ (t ₃ ) are sampled (S 31). The sampling time does not need to be at a fixed time interval, and may be any time as long as the time T2 has not elapsed. Also, the number of samplings is not limited to three as described herein.

Next, the same time t ₁ , t ₂ ,
extracting the values of the reference data P ₁ to P ₈ in the t ₃ (S
32). Reference data P at each time t ₁ , t ₂ , t _3
j (j = 1 to 8) are functions P _j (t ₁ ), P _j (t ₂ ), P
_j (t ₃ ), the actual measurement value δ of the starting drift at each time for each reference data P _j
(T ₁ ), δ (t ₂ ), difference ΔP from δ (t ₃ )
_j (t ₁ ), ΔP _j (t ₂ ), and ΔP _j (t ₃ ) are obtained (S3
3). ΔP _j (t ₁ ) = | P _j (t ₁ ) −δ (t ₁ ) | ΔP _j (t ₂ ) = | P _j (t ₂ ) −δ (t ₂ ) | ΔP _j (t ₃ ) = | P _j (t ₃ ) −δ (t ₃ ) | (6)

For each of the reference patterns P _{1 to} P ₈ , the average value of these differences ΔP _j ^MEAN = [ΔP _j (t ₁ ) + ΔP _j (t ₂ ) + ΔP _j (t
₃ )] / 3 is calculated (S34), and the average value Δ of these deviations is calculated.
^{_{P 1 MEAN, ΔP 2 MEAN,}} ..., to find the minimum value by comparing the magnitude of [Delta] P ₈ ^M _E ^AN extracted (S35).

For example, in the case of FIG. 9, ΔP ₂ ^MEAN
Is the smallest, it is determined that the function represented by the reference data P ₂ best approximates the starting drift, and the starting drift of the starting drift at an arbitrary time after the time t ₃ is determined based on the reference data P ₂ . The value is predicted (S36). That is, the predicted value of the starting drift is set to d (t) = P ₂ (t).

When the reference data approximating the starting drift δ (t) of the piezoelectric vibrating gyroscope 2 is determined in this way, the prediction operation unit 6 predicts the starting drift d based on the reference data, for example, P ₂ (t). (T) = P ₂ (t) is output to the synthesizer 12. As a result, the synthesizer 12 samples the sensor output signal S (t) output from the A / D conversion unit 5 (S37), and starts up drift calculated by the prediction calculation unit 6 from the sensor output signal S (t). Predicted value d (t)
Is subtracted (S38), and a correction signal A (t) = S (t) -d (t) as shown by a solid line in FIG. 8 is sent to the integration processing unit 10 (S39). Since the start-up drift has been removed from the correction signal A (t), the correction signal A (t) substantially represents only the angular velocity signal ω (t).

(Third Embodiment) As a result of creating reference data for many inertial sensors such as piezoelectric vibrating gyroscopes,
It has been found that the reference data can be approximately approximated by a logarithmic function. Therefore, in this embodiment, a method of approximating the starting drift of the piezoelectric vibrating gyroscope 2 by the following logarithmic function (natural logarithm or common logarithm) δ (t) = C ₁ · log (C ₂ · t + C ₃ ) Using. Here, C ₁ , C ₂ , and C ₃ are constants, and actual measured values δ (t ₁ ), δ (t ₂ ) of the starting drift at arbitrary times t ₁ , t ₂ , and t ₃ (≦ T2). δ
(T ₃ ). That is, three simultaneous equations δ (t ₁ ) = C ₁ · log (C ₂ · t ₁ + C ₃ ) δ (t ₂ ) = C ₁ · log (C ₂ · t ₂ + C ₃ ) δ (t ₃ ) = C _1, C _2, C ₃ are determined by solving a _{C 1 · log (C 2 ·} t 3 + C 3). Thus, C ₁ , C ₂ , C _{3 are obtained} from the output signal of the piezoelectric vibration drift.
Is determined, the prediction calculation unit 6 predicts the value of the starting drift by d (t) = C ₁ · log (C ₂ · t + C ₃ ) (7), and calculates the predicted value d.
(T) is subtracted from the sensor output signal S (t) to output a correction signal A (t), that is, an angular velocity signal ω (t).

(Fourth Embodiment) FIG. 10 is a block diagram showing a configuration of a signal processing device 21 of a piezoelectric vibrating gyroscope 2 according to still another embodiment of the present invention. 2 shows a signal processing device 21 having a function of correcting drift.

In this embodiment, the sensor output signal S (t) output from the piezoelectric vibrating gyroscope 2 is converted into a digital signal by the A / D converter 5 via the analog circuit 4, and then converted to a digital signal. The light is branched in three directions at a branching portion 11. Sensor output signal S (t) branched to one of them
Is led to the synthesizer 12, the sensor output signal S (t) branched to the other is guided to the stop determination unit 23, and the sensor output signal S (t) branched to the other is passed through the low-pass filter (LPF) 22. After passing, it is guided to the prediction calculation unit 6.

Based on the sensor output signal S (t) output from the A / D converter 5 and the predicted value d (t) of the temperature drift output from the prediction calculator 6, the stop determination unit 23 calculates Thus, it is determined whether or not the piezoelectric vibrating gyroscope 2 is stopped. Temperature drift δ (t) [The same symbol is used for the temperature drift as the startup drift]
Is generally considered to be a frequency component sufficiently lower than the angular velocity signal ω (t) to be detected.
After passing through the low-pass filter 22, the sensor output signal S
By separating low frequency components from (t), only the temperature drift δ (t) can be separated. Accordingly, only the temperature drift δ (t) of the sensor output signal S (t) passes through the low-pass filter 22 and enters the prediction calculation unit 6, and the prediction calculation unit 6 outputs a predicted value d of the temperature drift as described later.
(T) is output.

The sensor output signal S from the A / D converter 5
When (t) and the predicted value of the temperature drift δ (t) from the prediction calculation unit 6 are transmitted to the stop determination unit 23, the stop determination unit 23
Evaluates the magnitude of the sensor output signal S (t) on the basis of the predicted value d (t) of the temperature drift output from the prediction calculation unit 6. In other words, the stop determination unit 23 subtracts the predicted value d (t) of the temperature drift, which is the output of the prediction calculation unit 6, from the sensor output signal S (t) to obtain an approximately angular velocity signal ω (t).
≒ S (t) -d (t) is obtained, and as shown in FIG. 11, the magnitude of the angular velocity signal ω (t) is within a certain range, that is, −Vth <ω (t) <Vth. When it is detected that there is, the piezoelectric vibrating gyroscope 2 is determined to be stopped, and the stop determination time measuring timer 24 is started.
Then, the predetermined time T3 is measured by the timer 24, and when it is detected that the temperature drift δ (t) remains within the predetermined range [−Vth <ω (t) <Vth] during the predetermined time T3. Then, it confirms that the piezoelectric vibrating gyroscope 2 is in a stationary state, and outputs a stop determination signal u urging the measurement of the temperature drift δ (t) to the prediction calculation unit 6.

The stop determination unit 23 may be configured to determine whether the piezoelectric vibrating gyroscope 2 is stopped by directly monitoring the angular velocity signal ω (t) output from the synthesizer 12. No problem (not shown).

Upon receiving the stop determination signal u from the stop determination unit 23, the prediction calculation unit 6 extracts the temperature drift δ () extracted through the low-pass filter 22 at a fixed time interval ΔT measured by the sampling time measurement timer 8. t) is sampled. As shown in FIG. 11, this sampling time is defined as t ₁ , t ₂ = t ₁ + ΔT, t ₃
= T ₂ + ΔT, the prediction calculation unit 6 calculates three temperature drift values δ ₁ = δ (t ₁ ), δ ₂ = δ (t ₂ ), and δ ₃ = δ.
(T ₃ ) is sampled, and a temperature drift prediction value d (t) at time t is calculated based on the temperature drift sampling data δ ₁ , δ ₂ , δ ₃ .

Also in this embodiment, the prediction operation unit 6
Outputs the sampling data δ ₁ , δ ₂ , δ ₃ to the synthesizer 12 as they are while sampling the _three temperature drift values δ ₁ , δ ₂ , δ ₃ after the elapse of the stop determination time T _3. When the sampling of the data δ ₁ , δ ₂ , δ ₃ for the prediction calculation is completed, the predicted temperature drift predicted value d (t) is output to the synthesizer 12.

FIG. 12 is a diagram showing an algorithm for calculating the predicted value d (t) of the temperature drift in the prediction calculation section 6 as described above. That is, the stop determination unit 23 determines the temperature drift δ output from the prediction calculation unit 6.
(T) and the sensor output signal S (t) from the A / D converter 5
From this, it is monitored whether or not the piezoelectric vibrating gyroscope 2 is in a stationary state (S41). When it is determined that the piezoelectric vibrating gyroscope 2 is in a stationary state (time t ₀ ), the stop determination time measuring timer 24 is started. It is determined whether the piezoelectric vibrating gyroscope 2 is in a stationary state during the stop determination time T3 (S
41, S42). When it is confirmed that the piezoelectric signal gyro 2 is in a stationary state continuously during the stop determination time T3,
The counter is reset to j = 1 (S43), and the value of the temperature drift δ (t) output from the low-pass filter 22 is sampled (S44). Next, the sampling value δ (t) of the temperature drift is stored in the memory 9 (S
45). Next, the timer 8 for measuring the sampling time is started and waits until the sampling time ΔT has elapsed (S46). When the sampling time ΔT elapses and the sampling time measuring timer 8 times out, the count value j of the counter is increased by 1 (S47), and it is determined whether or not the count value j of the counter is 3 or less (S47).
48). If the count value j of the counter is 3 or less, the process returns to step S44 to repeat the processing. Thereby, the sampling data δ ₁ , δ ₂ , δ ₃ of the temperature drift are sampled.

The starting drift sampling data δ ₁ ,
When δ ₂ and δ ₃ are stored in the memory 9, the count value of the counter becomes j = 4 (S 49), and thereafter, at every predetermined time ΔT (S 47), the predicted value δ of the starting drift
(T) is predicted (S49).

The method of obtaining the predicted value d (t) of the temperature drift in the prediction operation unit 6 is the same as that of the case of the start-up drift. In the case of gradual, the above (5)
A method of approximation by a power series of an equation, a method of approximation by reference data previously obtained based on actually measured values as shown in FIG. 9, a method of approximation by a logarithmic function as in the above equation (7), and the like can be used. .

The other sensor output signal S (t) output from the A / D converter 5 and split by the signal splitter 11 is used in the synthesizer 12 to obtain sampling data of the temperature drift output from the prediction calculator 6. The predicted value d (t) of δ ₁ , δ ₂ , δ ₃ or the temperature drift is subtracted. Thus, the predicted value d of the temperature drift is obtained from the sensor output signal S (t).
Correction signal A (t) from which (t) is subtracted = S (t) -d
(T) ≒ ω (t) is input to the integration processing unit 10 and is converted into an angle signal by the integration processing unit 10.

Therefore, in this embodiment, it is possible to reduce offset fluctuations due to environmental changes such as temperature drift and to improve the accuracy of detecting angular velocity and angle.

In any of the above embodiments, the integration processing section 10 may be omitted, and the angular velocity signal may be directly output.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an example of a sensor output signal output from a piezoelectric vibrating gyroscope after power is turned on.

FIG. 2 is a diagram showing a startup drift output from a piezoelectric vibrating gyroscope after power is turned on.

FIG. 3 is a diagram showing a temperature drift included in a sensor output signal of a piezoelectric vibrating gyroscope.

FIG. 4 is a block diagram illustrating a configuration of a signal processing device for a piezoelectric vibrating gyroscope according to an embodiment of the present invention.

FIG. 5 is an explanatory diagram illustrating a principle for obtaining a predicted value of a startup drift and outputting only an angular velocity signal to an integration processing unit in the above signal processing device.

FIG. 6 is a flowchart showing an algorithm for a startup drift prediction process in the prediction calculation unit of the above.

FIG. 7 is a flowchart showing an algorithm for a correction value calculation process in the combiner of the above.

FIG. 8 is a flowchart illustrating an algorithm for a correction value calculation process in a signal processing circuit according to another embodiment of the present invention.

FIG. 9 is an explanatory diagram of the above.

FIG. 10 is a block diagram showing a configuration of a signal processing device of a piezoelectric vibrating gyroscope according to still another embodiment of the present invention.

FIG. 11 is an explanatory diagram illustrating a principle for obtaining a predicted value of a temperature drift and outputting only an angular velocity signal to an integration processing unit in the signal processing device of the above.

FIG. 12 is a flowchart showing an algorithm for a temperature drift prediction process in the prediction calculation unit according to the third embodiment.

[Explanation of symbols]

 2 Piezoelectric vibrating gyroscope 3 Stop judging unit 6 Prediction calculating unit 10 Integral processing unit 22 Low-pass filter 23 Stop judging unit S (t) Sensor output signal ω (t) Angular velocity signal δ (t) Starting drift or temperature drift d (t) Starting Estimated drift or temperature drift

Claims (5)

[Claims] 1. An inertial sensor signal processing device for performing signal processing by reducing unnecessary signal components from a sensor output signal containing unnecessary signal components, wherein the signal processing device is kept stationary after power supply of the inertial sensor is turned on. A sensor output signal output from the inertial sensor while measuring the sensor output signal output from the inertial sensor, and predicting a value of an unnecessary signal component thereafter based on the obtained measurement data. A signal synthesizing unit for subtracting a predicted value of the unnecessary signal component predicted by the prediction calculation unit. 2. An inertial sensor signal processing apparatus for performing signal processing by reducing offset fluctuations from a sensor output signal including offset fluctuations due to an environmental change, wherein the signal processing apparatus outputs a low frequency signal from a sensor output signal output from the inertial sensor. Means for extracting a component and comparing the sensor output signal with its low frequency component to determine whether or not the inertial sensor is in a stationary state.If it is determined that the inertial sensor is in a stationary state, Prediction calculation means for predicting a value of a subsequent offset fluctuation based on a low frequency component of the sensor output signal; and a predicted value of an unnecessary signal component predicted by the prediction calculation means from a sensor output signal output from an inertial sensor. Signal processing means for an inertial sensor, comprising: 3. The predicting operation unit measures a low-frequency component of the sensor output signal or a sensor output signal output from an inertial sensor determined to be in a stationary state, and approximates the obtained measurement data by a power series. 3. The signal processing device for an inertial sensor according to claim 1, wherein a value of an unnecessary signal component or an offset fluctuation thereafter is predicted. 4. An inertial sensor which has data holding means for holding data relating to unnecessary signal components or offset fluctuations of an inertial sensor, wherein said predictive calculating means determines that the sensor output signal has a low frequency component or is in a stationary state. 3. A sensor output signal output from the sensor is measured, and obtained measurement data is compared with the data to predict a value of an unnecessary signal component or an offset fluctuation thereafter. 3. A signal processing device for an inertial sensor according to claim 1. 5. The predicting operation means measures a low-frequency component of the sensor output signal or a sensor output signal output from an inertial sensor determined to be in a stationary state, and approximates the obtained measurement data by a logarithmic function. 3. The signal processing device for an inertial sensor according to claim 1, wherein a value of an unnecessary signal component or an offset fluctuation thereafter is predicted.