CN107040209A - Circuit arrangement, oscillator, electronic equipment and moving body - Google Patents

Abstract

Circuit devices, oscillators, electronic equipment, and moving objects. The circuit device includes: a processing unit that performs signal processing on input frequency control data to output the frequency control data; and an oscillation signal generation circuit that uses a vibrator to generate an oscillation signal at an oscillation frequency set by the frequency control data. In the update process of the prior estimate value of the Kalman filtering process, the processing unit performs addition processing of the posterior estimate value x ^{^} (k-1) and the correction value D(k-1) at time step k-1 to obtain The process of obtaining the prior estimate value x ^{^‑} (k) of time step k, and according to the result of Kalman filter processing, performs aging correction of frequency control data.

Description

Circuit devices, oscillators, electronic equipment, and moving objects

technical field

The present invention relates to a circuit device, an oscillator, electronic equipment, and a mobile body.

Background technique

Conventionally, oscillators such as an OCXO (oven controlled crystal oscillator) and a TCXO (temperature compensated crystal oscillator) have been known. For example, OCXOs are used as reference signal sources in base stations, network routers, measurement equipment, and the like.

In oscillators such as such OCXOs and TCXOs, high frequency stability is desired. However, there is a problem that the oscillation frequency of the oscillator changes with time called aging, and the oscillation frequency fluctuates over time. For example, there is a technique disclosed in Japanese Patent Application Laid-Open No. 2015-82815 as a conventional technique for suppressing fluctuations in oscillation frequency when a reference signal such as a GPS signal cannot be received and a so-called hold-over state is reached. . In this prior art, a storage unit for storing correspondence relation information (aging characteristic data) between a correction value of a control voltage of an oscillation frequency and an elapsed time is provided, and an elapsed time measuring unit is provided. Also, when the hold mode is detected, burn-in correction is performed based on the correspondence relation information between the correction value and the elapsed time stored in the storage unit and the elapsed time measured by the elapsed time measuring unit.

However, in this prior art, there is a problem that a storage unit having a large storage capacity is required as a storage unit for storing the correspondence relation information between the correction value and the elapsed time, leading to an increase in the size of the circuit device. For example, in order to achieve higher-precision burn-in correction, it is necessary to store a larger amount of correspondence information in a storage unit, resulting in an increase in the size of the circuit device.

Contents of the invention

According to some aspects of the present invention, it is possible to provide a circuit device, an oscillator, an electronic device, a mobile body, and the like capable of achieving higher-precision burn-in correction with a smaller-scale circuit.

One aspect of the present invention relates to a circuit device including: a processing unit that performs signal processing on frequency control data; In an oscillation signal of an oscillation frequency set by the frequency control data, the processing unit adds an a posteriori estimated value and a corrected value at a previous time in update processing of the a priori estimated value in the Kalman filter process. Processing is performed to obtain a priori estimated value at this time, and aging correction of the frequency control data is performed based on the result of the Kalman filter processing.

According to one aspect of the present invention, the frequency control data is signal-processed by the processing unit, and an oscillation signal having an oscillation frequency set by the frequency control data is generated using the vibrator and the frequency control data from the processing unit. Furthermore, in one aspect of the present invention, in the updating process of the priori estimated value in the Kalman filtering process, the priori estimated value at the previous time is added to the corrected value to obtain the priori estimated value at the current time. test estimate. Then, based on the result of the Kalman filter processing, aging correction of the frequency control data is performed. In this way, for example, compared with the case of using extended Kalman filter processing, the processing load on the processing unit is reduced, and an increase in the circuit scale of the circuit device is suppressed. Therefore, burn-in correction with higher accuracy can be realized with a smaller-scale circuit.

Furthermore, in one aspect of the present invention, the processing unit may obtain the correction value from an observation residual in the Kalman filtering process.

In this way, burn-in correction can be realized using a correction value updated to reflect the observation residual in the Kalman filter process, and more accurate burn-in correction can be realized.

In addition, in one aspect of the present invention, the processing unit performs the a posteriori estimated value x^(k-1) and the correction value D(k- 1) Addition process, by x^ ^- (k)=x^(k-1)+D(k-1) to obtain the said a priori estimated value x^ of the time step k at this time ^- (k).

In this way, a priori estimated value x^ ^{- (} k ), realizing miniaturization of circuit devices, etc.

In addition, in one aspect of the present invention, the processing unit may obtain the correction value D(k-1) at the time step k-1 and the observation residual error in the Kalman filter processing. Get the correction value D(k) of the time step k.

In this way, burn-in correction can be realized using the correction value D(k) that reflects the observation residual in the Kalman filter process and is updated at each time step.

In addition, in one aspect of the present invention, when the observation residual is ek and the constant is E, the processing unit may use D(k)=D(k-1)+E·ek The correction value D(k) is obtained.

In this way, the correction value D(k) can be obtained by simple arithmetic processing such as D(k)=D(k-1)+E·ek, and the processing load on the processing unit can be reduced.

Furthermore, in one aspect of the present invention, the circuit device may include a storage unit that stores the constant E.

In this way, for example, update processing of the correction value D(k)=D(k-1)+E·ek can be realized using an appropriate constant E corresponding to each product, and more accurate burn-in correction can be realized.

Furthermore, in one aspect of the present invention, the processing unit may perform the signal processing on the frequency control data based on a phase comparison result of an input signal based on the oscillation signal, the processing unit performs the process of estimating the frequency control data based on the phase comparison result by Kalman filter processing until the hold mode is detected due to the disappearance or abnormality of the reference signal. The true value of the observed value of , when the hold mode is detected, the true value at the time corresponding to the time when the hold mode is detected is stored, and by performing arithmetic processing based on the true value, an aging The corrected frequency control data.

In this way, burn-in correction can be realized based on the true value estimated by the Kalman filter process and stored at the time corresponding to the detection time of the hold mode. Therefore, highly accurate burn-in correction that has not been possible conventionally can be realized.

Furthermore, in one aspect of the present invention, the processing unit may generate the aging-corrected frequency control data by performing the arithmetic processing of adding the correction value to the true value.

In this way, an arithmetic process of adding a correction value for compensating for, for example, a frequency change due to an aging rate to the true value estimated by the Kalman filter processing is performed, whereby burn-in correction is realized. Therefore, highly accurate burn-in correction can be realized with simple processing.

Furthermore, in one aspect of the present invention, the processing unit may perform the arithmetic processing of adding the corrected value obtained by filtering to the true value.

In this way, it is possible to effectively suppress a decrease in the accuracy of burn-in correction due to adding the correction value with fluctuating fluctuations to the true value.

In addition, in one aspect of the present invention, the circuit device may further include a storage unit that stores a system noise constant for setting the system noise of the Kalman filtering process and an observation value of the Kalman filtering process. Observed noise constant for noise setting.

In this way, burn-in correction that reduces the influence of element variations of system noise and observation noise can be realized.

Furthermore, in one aspect of the present invention, the circuit device may include a digital interface unit for monitoring the a priori estimated value and the observed value.

In this way, an external device such as an inspection device can monitor a priori estimated values and observed values via the digital interface unit. Thereby, for example, setting processing for making the observation error corresponding to the difference value between the observed value and the a priori estimated value smaller, etc. can be realized.

Furthermore, another aspect of the present invention relates to an oscillator including: the circuit device described in any one of the above aspects; and the vibrator.

Furthermore, another aspect of the present invention relates to electronic equipment including the circuit device described in any one of the above-mentioned aspects.

Furthermore, another aspect of the present invention relates to a moving body including the circuit device described in any one of the above-mentioned aspects.

Description of drawings

FIG. 1 is an explanatory diagram of element variation with respect to aging characteristics.

FIG. 2 is a basic configuration example of a circuit device according to this embodiment.

FIG. 3 is an explanatory diagram of the method of this embodiment.

FIG. 4 is an explanatory diagram of the method of this embodiment.

FIG. 5 is an explanatory diagram for burn-in correction in hold mode.

FIG. 6 is an explanatory diagram for a hold mode.

FIG. 7 is an explanatory diagram for a hold mode.

FIG. 8 is an explanatory diagram for a hold mode.

FIG. 9 is a detailed configuration example of the circuit device of this embodiment.

FIG. 10 is an explanatory diagram of burn-in correction using Kalman filter processing.

FIG. 11 is an explanatory diagram of burn-in correction using Kalman filter processing.

Fig. 12 is a detailed configuration example of a processing unit.

FIG. 13 is an explanatory diagram of temperature compensation processing.

FIG. 14 is an explanatory diagram of temperature compensation processing.

FIG. 15 is an explanatory diagram of temperature compensation processing.

FIG. 16 is an explanatory diagram of the operation of the processing unit.

FIG. 17 is an explanatory diagram of the operation of the processing unit.

Fig. 18 is a configuration example of a burn-in correction unit.

Fig. 19 is a model example of Kalman filtering.

FIG. 20 is a configuration example of a Kalman filter unit.

FIG. 21 is a diagram showing an example of a predicted frequency deviation and a measured frequency deviation according to the present embodiment.

Fig. 22 is a structural example of a temperature sensor.

FIG. 23 is a configuration example of an oscillation circuit.

FIG. 24 is an explanatory diagram of a modified example of the present embodiment.

FIG. 25 is an explanatory diagram of a modified example of the present embodiment.

Fig. 26 is a structural example of an oscillator.

Fig. 27 is a configuration example of an electronic device.

Fig. 28 is a structural example of a mobile body.

Fig. 29 is a detailed configuration example of an oscillator.

Fig. 30 is a configuration example of a base station as one of electronic devices.

detailed description

Hereinafter, preferred embodiments of the present invention will be described in detail. In addition, the present embodiment described below does not unduly limit the content of the present invention described in the claims, and not all configurations described in the present embodiment are necessarily solutions of the present invention.

1. Oscillation frequency changes due to aging

In oscillators such as OCXO and TCXO, the oscillation frequency fluctuates due to time-dependent changes called aging. In addition, in the characteristics of the aging fluctuation of the oscillation frequency among individual oscillators, there are individual variations due to the performance of the components constituting the oscillator, the mounting state of the components and the oscillator, or the environment in which the oscillator is used (hereinafter referred to as is the difference caused by component deviation).

A1 to A5 in FIG. 1 are examples of measurement results of aging characteristics of a plurality of oscillators having the same or different shipment lot numbers. As shown in A1 to A5 of FIG. 1 , there is a difference in the form of aging variation due to element variation.

The cause of the change in oscillation frequency due to aging is considered to be the shedding and adhesion of dust generated in the hermetically sealed container to the vibrator, environmental changes based on some outgassing, or the adhesive used in the vibrator. changes over time.

As a countermeasure for suppressing such a change in oscillation frequency due to aging, there is a method of performing initial burn-in of operating the oscillator for a certain period of time before shipment, and then shipping the oscillator after initially changing the oscillation frequency. However, for applications that require high frequency stability, it is not enough to take such an initial aging countermeasure, and aging correction that compensates for fluctuations in the oscillation frequency due to aging is desired.

In addition, when an oscillator is used as a reference signal source of a base station, there is a problem of so-called hold mode. For example, in a base station, frequency fluctuations are suppressed by using a PLL circuit to synchronize the oscillation signal (output signal) of an oscillator with a reference signal from GPS or a network. However, when the reference signal from the GPS or the network (Internet) disappears or becomes abnormal in a hold mode, a reference signal for synchronization cannot be obtained. Taking GPS as an example, when the positioning signal cannot be received due to the installation position or direction of the GPS antenna, the positioning signal cannot be received accurately due to interference waves, or the positioning signal is not transmitted from the positioning satellite , a hold mode occurs, and synchronization processing using the reference signal cannot be performed.

When such a hold mode occurs, the oscillation signal generated by the self-oscillation of the oscillator becomes a reference signal source of the base station. Therefore, hold mode performance is required that suppresses fluctuations in oscillation frequency due to self-oscillation of the oscillator during the hold mode period from when the hold mode is generated to when the hold mode returns (release time).

However, as described above, since the oscillation frequency of the oscillator fluctuates to a non-negligible degree due to aging, there is a problem that high hold mode performance cannot be realized due to this. For example, during a hold mode period such as 24 hours, when an allowable frequency deviation (Δf/f) is specified, if there is a large fluctuation in the oscillation frequency due to aging, the allowable frequency deviation cannot be satisfied.

For example, various methods such as FDD (Frequency Division Duplex) and TDD (Time Division Duplex) have been proposed as communication methods between a base station and a communication terminal. In addition, in the TDD method, uplink and downlink use the same frequency to transmit and receive data in a time-division manner, and a guard time is set between time slots allocated to each device. Therefore, in order to realize appropriate communication, it is necessary to perform time synchronization in each device, and accurate absolute time measurement is required. That is, in order to provide a wireless communication system for communication over a wide area, such as mobile phones and terrestrial digital broadcasting, it is necessary to install a plurality of base stations, and if the counting time varies between these base stations, appropriate communication cannot be realized. However, when the reference signal from GPS or the network disappears or an abnormal hold mode occurs, the oscillator needs to time the absolute time in the absence of the reference signal. If the counted time deviates, communication will fail. . Therefore, for an oscillator used in a base station or the like, very high frequency stability is required also during the holdover mode. Therefore, high-precision correction is also required for aging correction that compensates for frequency fluctuations due to aging.

2. The structure of the circuit device

FIG. 2 shows the basic circuit configuration of the circuit device of this embodiment. As shown in FIG. 2 , the circuit device of the present embodiment includes a processing unit 50 and an oscillation signal generating circuit 140 . In addition, the configuration of the circuit device according to the present embodiment is not limited to the configuration of FIG. 2 , and various modifications such as omitting some components or adding other components are possible.

The processing unit 50 performs various signal processing. For example, signal processing is performed on frequency control data DFCI (frequency control code). Specifically, the processing unit 50 (digital signal processing unit) performs, for example, burn-in correction processing, Kalman filter processing, and, if necessary, signal processing (digital signal processing) such as temperature compensation processing. And, the signal-processed frequency control data DFCQ is output. The processing unit 50 can include a Kalman filter unit 54 (circuit or program module for Kalman filter processing) and a burn-in correction unit 56 (circuit or program module for burn-in correction processing). The processing unit 50 may be realized by an ASIC circuit such as a gate array, or may be realized by a processor (DSP, CPU) and a program (program module) operating on the processor.

The vibrator XTAL is, for example, a quartz vibrator of a thickness-shear vibration type such as an AT cut type or an SC cut type, or a piezoelectric vibrator of a bending vibration type or the like. As an example, the vibrator XTAL is of a type installed in a constant temperature bath of an oven-controlled oscillator (OCXO), but it is not limited thereto, and may be a vibrator for a TCXO of a type that does not have a constant temperature bath. The oscillator XTAL can also be a resonator (electromechanical resonator or electrical resonant circuit). In addition, as the vibrator XTAL, a SAW (Surface Acoustic Wave: Surface Acoustic Wave) resonator, a MEMS (Micro Electro Mechanical Systems: Micro Electro Mechanical Systems) vibrator which is a silicon vibrator, or the like can be used as a piezoelectric vibrator. As the substrate material of the vibrator XTAL, piezoelectric single crystals such as quartz, lithium tantalate, and lithium niobate, piezoelectric materials such as piezoelectric ceramics such as lead zirconate titanate, or silicon semiconductor materials can be used. As the excitation means of the vibrator XTAL, either means based on piezoelectric effect or electrostatic drive based on Coulomb force can be used.

The oscillation signal generation circuit 140 generates an oscillation signal OSCK. For example, oscillation signal generation circuit 140 generates oscillation signal OSCK having an oscillation frequency set by frequency control data DFCQ using frequency control data DFCQ (signal-processed frequency control data) from processing unit 50 and oscillator XTAL. As an example, the oscillation signal generating circuit 140 oscillates the vibrator XTAL at an oscillation frequency set by the frequency control data DFCQ to generate an oscillation signal OSCK.

In addition, the oscillating signal generation circuit 140 may be a circuit for generating the oscillating signal OSCK by means of a direct digital synthesizer. For example, the oscillation signal OSCK of the oscillation frequency set by the frequency control data DFCQ may be digitally generated by using the oscillation signal of the oscillator XTAL (an oscillation source with a fixed oscillation frequency) as a reference signal.

The oscillation signal generation circuit 140 may include a D/A conversion unit 80 and an oscillation circuit 150 . However, the oscillation signal generating circuit 140 is not limited to such a configuration, and various modifications such as omitting some of the constituent elements or adding other constituent elements can be implemented.

The D/A conversion unit 80 performs D/A conversion of the frequency control data DFCQ (output data of the processing unit) from the processing unit 50 . The frequency control data DFCQ input to the D/A converter 80 is frequency control data (frequency control code) after signal processing by the processing unit 50 (for example, aging correction, temperature compensation, or Kalman filter processing). As the D/A conversion method of the D/A conversion unit 80, for example, a resistor series type (resistor division type) can be employed. However, the D/A conversion method is not limited to this, and various methods such as a resistance ladder type (R-2R ladder type, etc.), a capacitor array type, or a pulse width modulation type may be employed. In addition, the D/A converter 80 may include its control circuit, modulation circuit (dither modulation, PWM modulation, etc.), filter circuit, etc. in addition to the D/A converter.

The oscillation circuit 150 generates an oscillation signal OSCK using the output voltage VQ of the D/A converter 80 and the oscillator XTAL. The oscillation circuit 150 is connected to the transducer XTAL via the first and second terminals for the transducer (pads for the transducer). For example, the oscillation circuit 150 generates an oscillation signal OSCK by oscillating a vibrator XTAL (piezoelectric vibrator, resonator, etc.). Specifically, the oscillation circuit 150 oscillates the vibrator XTAL at an oscillation frequency using the output voltage VQ of the D/A converter 80 as a frequency control voltage (oscillation control voltage). For example, when the oscillation circuit 150 is a circuit (VCO) that controls the oscillation of the vibrator XTAL by voltage control, the oscillation circuit 150 may include a variable capacitance capacitor (varicap diode, etc.) whose capacitance value changes according to the frequency control voltage. ).

In addition, as described above, the oscillation circuit 150 can be realized by a direct digital synthesizer method. In this case, the oscillation frequency of the oscillator XTAL is a reference frequency, which is different from the oscillation frequency of the oscillation signal OSCK.

In this way, the circuit device of this embodiment includes: a processing unit 50 that performs signal processing on the frequency control data DFCI; The oscillation signal OSCK of the oscillation frequency set by the data DFCQ.

In addition, the processing unit 50 performs the process of adding the a priori estimated value and the correction value at the previous time in the update process of the a priori estimated value in the Kalman filtering process to obtain the a priori estimated value at the current time. processing. Then, aging correction of the frequency control data is performed based on the result of the Kalman filter processing. In addition, the addition process of this embodiment includes the process of adding a negative value, ie, the subtraction process.

That is, in the Kalman filter process, observation update and time update are repeated to estimate the state. In the observation update, Kalman gain, posterior estimated value, posterior covariance, etc. are updated according to the observation value and time update results. In the time update, prediction processing is performed in which a priori estimated value and a priori covariance at this time (time step k) are obtained from the results of the observation update.

Furthermore, in the present embodiment, as shown in FIG. 3 , the processing unit 50 passes through the time after the previous time (time step k-1) in the time update of the update process of the prior estimate value of the Kalman filter process. The prior estimation value at this time (time step k) is obtained by adding the prior estimation value and the correction value. Furthermore, the aging correction of the frequency control data is performed according to the result of the Kalman filtering process (true value, or a true value and a corrected value, etc.).

In this case, the processing unit 50 obtains the correction value from the observation residual in the Kalman filtering process. The observation residuals correspond to the difference between the observed value and the prior estimate. For example, a prediction process is performed in which the correction value obtained in the update process based on the observation residual in the Kalman filter process is added to the posterior estimated value at the previous time to obtain the current A priori estimate of time. In this way, burn-in correction can be realized using an updated correction value that reflects the observation residual in the Kalman filtering process, and higher-precision burn-in correction can be realized.

More specifically, as shown in FIG. 4 , the processing unit 50 performs addition processing of the posterior estimated value x^(k-1) and the correction value D(k-1) at the previous time step k-1. , through x^ ^- (k)=x^(k-1)+D(k-1) to obtain the prior estimate x^ ^- (k) of time step k at this moment. In addition, here, a hat-shaped symbol " ^{^} " indicating an estimated value is appropriately arranged and described in two characters.

That is, as will be described later, in normal Kalman filter processing ^, a priori estimated value x^ ^- (k). Here, A relates the state of the system at time step k to the state of the system at time step k+1 in the absence of system noise, and is called the system matrix. In this case, in order to obtain the value of A accurately, processing called extended Kalman filter processing is required.

Extended Kalman filter processing sometimes has a very heavy processing load. In this case, the circuit area of the processing unit 50 tends to be very large when the extended Kalman filter process is to be realized, which is not suitable in a situation where miniaturization of the circuit device built in the oscillator is strongly demanded. On the other hand, when a fixed value is used as the value of A, the difficulty in realizing appropriate burn-in correction increases.

Therefore, as a solution when it is necessary to avoid such a situation, in this embodiment, as will be described in detail below, instead of x^ ^- (k) = Ax^(k-1), x^ ^- (k )=x^(k-1)+D(k-1), realizing the time update process of obtaining the prior estimate value x^ ^- (k) of time step k. The process using this x^ ^- (k)=x^(k-1)+D(k-1) can be done by adding the correction value D(k-1) to the posterior estimated value x^(k-1) Such light-load processing is achieved. Therefore, since there is no need to perform heavy processing such as extended Kalman filter processing, the processing load on the processing unit 50 is reduced, and an increase in the circuit scale of the circuit device is suppressed. That is, more accurate burn-in correction can be realized with a smaller-scale circuit.

Furthermore, in this embodiment, as shown in FIG. 4 , the processing unit 50 obtains the correction value D(k-1) at time step k-1 and the observation residual error in the Kalman filtering process, and obtains the correction value D(k-1) at time step k. Value D(k). Specifically, when the observation residual is ek and the constant is E, the correction value D(k) is obtained by D(k)=D(k-1)+E·ek. In addition, when the Kalman gain is G(k), the correction value D(k) can also be obtained by D(k)=D(k-1)+G(k)·ek. Here, when the observed value y(k) is assumed, the observation residual can be expressed as ek=y(k)-x^ ^- (k).

For example, as described later, the correction value D(k) is a correction value for canceling and compensating for a frequency change due to an aging rate corresponding to the slope of C3 in FIG. 10 . Furthermore, it can be seen from A1 to A5 in FIG. 1 and FIG. 10 that the aging rate fluctuates with time. Therefore, when a fixed value is used as a correction value for burn-in correction, proper burn-in correction cannot be achieved.

In contrast, in the present embodiment, a process of updating the correction value of burn-in correction using the observation residual in the Kalman filtering process is performed as D(k)=D(k-1)+E·ek. Therefore, even when the aging rate fluctuates with the passage of time as shown in A1 to A5 in FIG. 1 and FIG. 10 , it is possible to obtain A correction value that compensates for frequency changes based on a varying aging rate. Therefore, burn-in correction with higher accuracy can be realized with a smaller-scale circuit.

In addition, the constant E in D(k)=D(k-1)+E·ek is desirably stored in advance in the storage unit 34 in FIG. 9 described later. For example, when a product (oscillator, etc.) is manufactured or shipped, an appropriate constant E corresponding to the product is written in the storage unit 34 realized by, for example, a nonvolatile memory, and stored. Thereby, updating processing of the correction value D(k)=D(k-1)+E·ek can be realized using an appropriate constant E corresponding to each product, and more accurate burn-in correction can be realized.

3. Hold mode

Next, the hold mode will be described in detail. FIG. 5 is a diagram illustrating burn-in correction in a hold mode. The frequency control data generating unit 40 performs phase comparison (comparison operation) between an input signal (input clock signal) based on an oscillation signal and a reference signal (reference clock signal) from GPS or a network to generate frequency control data. During normal operation, the selector 48 outputs the frequency control data from the frequency control data generation unit 40 to the oscillation signal generation circuit 140 . The D/A converter 80 of the oscillation signal generation circuit 140 converts the frequency control data into a frequency control voltage, and outputs it to the oscillation circuit 150 . The oscillation circuit 150 oscillates the oscillator XTAL at an oscillation frequency corresponding to the frequency control voltage to generate an oscillation signal. A loop of the PLL circuit is formed by the frequency control data generation unit 40 and the oscillation signal generation circuit 140, and an input signal based on the oscillation signal and a reference signal can be synchronized.

The detection circuit 47 performs a detection operation of a reference signal, and detects a hold pattern in which the reference signal disappears or is abnormal. When the hold mode is detected, the burn-in correcting unit 56 performs burn-in correction for compensating frequency fluctuations due to burn-in to the frequency control data stored in the register 49 . Then, the oscillation signal generation circuit 140 oscillates the vibrator XTAL at the oscillation frequency corresponding to the aging-corrected frequency control data to generate an oscillation signal. Thus, an oscillation signal in self-oscillation can be supplied as a reference signal source for electronic equipment such as a base station.

B1 in FIG. 6 shows the aging characteristics of the ideal oscillation frequency when the hold mode occurs. On the other hand, B2 (dotted line) shows the characteristics of fluctuations in the oscillation frequency due to aging. B3 is the fluctuation width of the oscillation frequency due to aging. In addition, B4 in FIG. 7 shows transition of the frequency control voltage for approaching the characteristic of B1 when the hold mode occurs. On the other hand, B5 (dotted line) indicates a state in which the frequency control voltage is constant from the time when the reference signal disappears or an abnormality occurs.

Burn-in correction is performed in order to correct the characteristic indicated by B2 in FIG. 6 to the ideal characteristic indicated by B1. For example, if the frequency control voltage is changed as shown in B4 in FIG. 7 by aging correction, then it is possible to correct the characteristic shown in B2 in FIG. 6 close to the ideal characteristic shown in B1. For example, if the correction Accuracy, the characteristic shown in B2 can be corrected to the ideal characteristic shown in B1. On the other hand, when aging correction is not performed as shown in B5 of FIG. 7, as shown in B2 of FIG. B1 shown in Figure 6 cannot meet this requirement.

For example, the hold mode time θ _tot representing the amount of time shift (total amount) due to fluctuations in the oscillation frequency during the hold mode period can be represented by the following equation (1).

Here, T ₁ represents the elapsed time of aging caused by the hold mode. f ₀ is the nominal oscillation frequency, and Δf/f ₀ is the frequency deviation. In the above formula (1), T ₁ ×f ₀ represents the total number of clocks, and (Δf/f ₀ )×(1/f ₀ ) represents the amount of time shift within one clock. Furthermore, the frequency deviation Δf/f ₀ can be expressed as in the above formula (2) using the hold mode time θ _tot and the elapsed time T ₁ .

As shown in B6 of FIG. 8 , it is assumed that the frequency deviation Δf/f ₀ changes with a constant slope as a linear function with respect to the elapsed time. In this case, as shown in B7 of FIG. ₈ , as the elapsed time T1 becomes longer, the hold mode time θ _tot becomes longer according to a quadratic function.

For example, in the case of the TDD scheme, in order to prevent overlapping of slots for which a guard time is set, the hold mode time is required to be, for example, θ _tot <1.5 μs. Therefore, it can be seen from the above formula (2) that a very small value is required as the allowable frequency deviation Δf/f ₀ of the oscillator. _In particular, the longer the elapsed time T1 is, the smaller the value of the allowable frequency deviation is required. For example, when the time assumed as the time from the generation of the hold mode to the time of recovery from the hold mode by maintenance work is, for example, T ₁ =24 hours, a very small value is required as the allowable frequency deviation . Furthermore, since the frequency deviation Δf/f ₀ includes, for example, a temperature-dependent frequency deviation and a frequency deviation caused by aging, very high-accuracy aging correction is required in order to satisfy the above-mentioned requirements.

4. Detailed structure example of circuit device

FIG. 9 shows a detailed configuration example of the circuit device of this embodiment. In FIG. 9, the structure of FIG. 2 is further provided with a temperature sensor 10, an A/D conversion unit 20, an I/F unit 30, a register unit 32, a storage unit 34, and a frequency control data generation unit 40 (in a broad sense, a phase comparison section). In addition, the configuration of the circuit device is not limited to the configuration of FIG. 9 , and various modifications such as omitting some of its constituent elements (for example, a frequency control data generation unit) or adding other constituent elements can be implemented. For example, a temperature sensor provided outside the circuit device may be used as the temperature sensor 10 .

The temperature sensor 10 outputs a temperature detection voltage VTD. Specifically, a temperature-dependent voltage that changes according to the temperature of the environment (circuit device) is output as temperature detection voltage VTD. A specific configuration example of the temperature sensor 10 will be described later.

The A/D conversion unit 20 performs A/D conversion of the temperature detection voltage VTD from the temperature sensor 10, and outputs temperature detection data DTD. For example, digital temperature detection data DTD (A/D result data) corresponding to the A/D conversion result of the temperature detection voltage VTD is output. As the A/D conversion method of the A/D conversion unit 20, for example, a successive comparison method or a method similar to the successive comparison method can be employed. Also, the A/D conversion method is not limited to this method, and various methods (counting type, parallel comparison type, series-parallel type, etc.) can be adopted.

The digital I/F unit (interface unit) 30 is an interface for inputting and outputting digital data between the circuit device and an external device (microcomputer, controller, etc.). The digital I/F unit 30 can be realized by, for example, a synchronous serial communication method using a serial clock line and a serial data line. Specifically, it can be realized by an I2C (Inter-Integrated Circuit) method, a 3-wire or 4-wire SPI (Serial Peripheral Interface: Serial Peripheral Interface) method, or the like. The I2C method is a synchronous serial communication method that communicates through two signal lines, a serial clock line SCL and a bidirectional serial data line SDA. Multiple slave devices can be connected to the I2C bus, and the master device communicates with the slave device after specifying the address of the independently determined slave device and selecting the slave device. The SPI method is a synchronous serial communication method that communicates through a serial clock line SCK and two unidirectional serial data lines SDI and SDO. Multiple slave devices can be connected on the SPI bus, and in order to determine these slave devices, the master device needs to use the slave device selection line to select the slave device. The digital I/F unit 30 is composed of an input/output buffer circuit, a control circuit, and the like that realize these communication methods.

The register unit 32 is a circuit composed of a plurality of registers such as a status register, a command register, and a data register. An external device of the circuit device accesses each register of the register unit 32 via the digital I/F unit 30 . Furthermore, the external device can use the registers of the register unit 32 to confirm the status of the circuit device, issue a command to the circuit device, transmit data to the circuit device, read data from the circuit device, and the like.

The storage unit 34 stores various information necessary for various processes and operations of the circuit device. This storage unit 34 can be realized by, for example, a nonvolatile memory. As the nonvolatile memory, for example, EEPROM or the like can be used. As the EEPROM, for example, a MONOS (Metal-Oxide-Nitride-Oxide-Silicon: Metal-Oxide-Nitride-Oxide-Silicon) type memory or the like can be used. For example, a flash memory using a MONOS type memory can be used. Alternatively, as the EEPROM, other types of memory such as a floating gate type can be used. In addition, the memory|storage part 34 should just be a memory which can store and memorize information even if a power supply is not supplied, and it can implement|achieve also by a fuse circuit etc., for example.

In this case, in addition to the Kalman filter unit 54 and the burn-in correction unit 56, the processing unit 50 further includes a hold mode processing unit 52 (circuit or program module for hold mode processing), and a temperature compensation unit 58 (circuit or program module for temperature compensation processing). program module). The hold mode processing unit 52 performs various processes related to the hold mode. The temperature compensation unit 58 (processing unit 50 ) performs temperature compensation processing of the oscillation frequency based on the temperature detection data DTD from the A/D conversion unit 20 . Specifically, the temperature compensator 58 performs calculation for the presence of a temperature change based on temperature detection data DTD (temperature dependent data) that changes in accordance with the temperature, coefficient data (coefficient data of an approximation function) for temperature compensation processing, and the like. In this case, the temperature compensation process reduces the fluctuation of the oscillation frequency.

The reference signal RFCK is input to the circuit device via a terminal TRFCK (pad) which is an external connection terminal of the circuit device. A signal PLOCK notifying whether or not the external PLL circuit is in the locked state is input to the circuit device via a terminal TPLOCK (pad) which is an external connection terminal of the circuit device.

Furthermore, the storage unit 34 stores a system noise constant (V) for setting the system noise of the Kalman filter process and an observation noise constant (W) for setting the observation noise of the Kalman filter process. For example, measurement (inspection) for monitoring various information such as oscillation frequency is performed during manufacture and shipment of products (oscillators, etc.). Then, a system noise constant and an observation noise constant are determined based on the measurement results, and written into the storage unit 34 realized by, for example, a nonvolatile memory. In this way, it is possible to set the system noise constant and the observation noise constant in which adverse effects due to element variations are reduced.

In this case, the circuit device of the present embodiment has a digital I/F unit 30 for monitoring a priori estimated values and observed values. That is, the external device can monitor the a priori estimated value x ^̂− (k) and the observed value y(k) in the Kalman filtering process via the digital I/F section 30 .

Specifically, in the inspection process at the time of product manufacture and shipment, the inspection device as an external device monitors the a priori estimated value x̂ ⁻ (k) and the observed value y(k) via the digital I/F unit 30 . For example, in the Kalman filter unit 54 shown in FIG. 20 described later, the inspection device can monitor the signal value and the observed value y(k) of the node of the prior estimated value x^ ^- (k) via the digital I/F unit 30. The signal value of the node. For example, monitoring registers for monitoring the signal value of the node of the a priori estimated value x̂−(k) and the signal value of the node of the observed value ^y (k) are provided in the register unit 32 . Furthermore, by accessing the monitoring register via the digital I/F unit 30, the inspection device can acquire a priori estimated value x^ ^- (k) and observed value y(k).

Furthermore, the inspection device can obtain, for example, the observation residual ek=y(k)-x^ ^- (k) from the acquired a priori estimated value x^ ^- (k) and the observed value y(k). Furthermore, the system noise constant V and the observation noise constant W that make the observation residual ek smaller are obtained. Alternatively, the optimum constant E in the arithmetic processing of correction value D(k)=D(k-1)+E·ek is obtained. In addition, the system noise constant V, observed noise constant W, or constant E obtained is written in the storage unit 34 in the inspection process at the time of product manufacture and shipment. Thereby, in each product, the optimum constants V, W, and E can be written in the storage unit 34 . Further, during normal operation, by executing the Kalman filter processing of the present embodiment using these constants V, W, and E, it is possible to realize more accurate burn-in correction.

The frequency control data generator 40 generates frequency control data DFCI. For example, an input signal based on the oscillation signal OSCK is compared with the reference signal RFCK to generate frequency control data DFCI. The generated frequency control data DFCI is input to the processing unit 50 . Here, the input signal based on the oscillating signal OSCK may be the oscillating signal OSCK itself, or a signal generated from the oscillating signal OSCK (such as a frequency-divided signal). Hereinafter, the case where the input signal is the oscillating signal OSCK itself is mainly used as an example for description.

The frequency control data generation unit 40 includes a phase comparison unit 41 and a digital filter unit 44 . The phase comparison unit 41 (comparison operation unit) is a circuit that performs phase comparison (comparison operation) between the oscillation signal OSCK as an input signal and the reference signal RFCK, and includes a counter 42 and a TDC 43 (time-to-digital converter).

The counter 42 generates digital data corresponding to the integer part of the result obtained by dividing the reference frequency (for example, 1 Hz) of the reference signal RFCK by the oscillation frequency of the oscillation signal OSCK. The TDC 43 generates digital data corresponding to the fractional part of the division result. The TDC 43 includes, for example: a plurality of delay elements; a plurality of latch circuits which latch a plurality of delayed clock signals outputted from the plurality of delay elements at the edge (high) timing of the reference signal RFCK; The encoding of the output signal of the latch circuit generates digital data corresponding to the fractional part of the division result. Then, the phase comparator 41 adds the digital data corresponding to the integer part from the counter 42 and the digital data corresponding to the fractional part from the TDC 43 to detect a phase error with the set frequency. Furthermore, the digital filter unit 44 generates frequency control data DFCI by performing smoothing processing of the phase error. For example, when the frequency of the oscillation signal OSCK is FOS, the frequency of the reference signal RFCK is FRF, and the frequency division number (division ratio) corresponding to the set frequency is FCW, the relationship of FOS=FCW×FRF is established. The frequency control data DFCI is generated in a manner. Alternatively, the counter 42 may count the number of clocks of the oscillation signal OSCK. That is, the counter 42 performs a counting operation based on an input signal of the oscillation signal OSCK. In addition, the phase comparison unit 41 may compare the count value of the counter 42 in n cycles (n is an integer that can be set to 2 or more) of the reference signal RFCK with an expected value (n×FCW) of the count value using an integer. . For example, the difference between the expected value and the count value of the counter 42 is output from the phase comparison unit 41 as phase error data.

In addition, the structure of the frequency control data generation part 40 is not limited to the structure shown in FIG. 9, Various modifications are possible. For example, the phase comparison unit 41 may be constituted by a phase comparator of an analog circuit, or the digital filter unit 44 may be constituted by a filter unit (loop filter) of an analog circuit. In addition, the processing unit 50 may perform the processing of the digital filter unit 44 (smoothing processing of phase error data). For example, the processing unit 50 performs the processing of the digital filter unit 44 time-divisionally with other processing (hold mode processing, Kalman filter processing, etc.). For example, the processing unit 50 performs filtering processing (smoothing processing) on the phase comparison result (phase error data) of the phase comparison unit 41 .

In addition, in FIG. 9 , the circuit device has a built-in frequency control data generation unit 40 , but the frequency control data generation unit may be a circuit provided outside the circuit device. In this case, the frequency control data DFCI may be input to the processing unit 50 via the digital I/F unit 30 from an external frequency control data generating unit.

Thus, in the present embodiment, the processing unit 50 (processor) performs signal processing on the frequency control data DFCI based on the phase comparison result of the input signal based on the oscillation signal OSCK and the reference signal RFCK. That is, the processing unit 50 performs signal processing on the frequency control data DFCI based on the phase comparison result in the phase comparison unit 41 . For example, frequency control data DFCI from frequency control data generator 40 that compares an input signal based on oscillation signal OSCK with reference signal RFCK to generate frequency control data DFCI is input to processing unit 50 . The processing unit 50 may input the phase comparison result from the phase comparison unit 41 and perform filtering processing (processing by the digital filter unit 44 ) on the phase comparison result. In addition, the processing unit 50 (processor) performs a process of estimating an observation of the frequency control data DFCI based on a phase comparison result by Kalman filter processing until the hold mode due to disappearance or abnormality of the reference signal is detected. The truth value of the value. This true value is a true value estimated by Kalman filter processing, and is not limited to a true true value. The Kalman filter processing is executed by the Kalman filter unit 54 . In addition, the control processing based on the hold mode detection is executed by the hold mode processing unit 52 .

Furthermore, the processing unit 50 (processor) stores the truth value at the time corresponding to the detection time of the hold mode when the hold mode is detected. The timing at which this true value is stored may be the detection timing of the hold mode itself, or may be a timing before that timing, or the like. Furthermore, the processing unit 50 generates aging-corrected frequency control data DFCQ by performing arithmetic processing based on the stored true value. The generated frequency control data DFCQ is output to the oscillation signal generation circuit 140 . The aging corrected frequency control data DFCQ generation process is executed by the aging correction unit 56 .

For example, during normal operation, the processing unit 50 performs signal processing such as temperature compensation processing on the frequency control data DFCI based on the phase comparison result, and outputs the signal-processed frequency control data DFCQ to the oscillation signal generation circuit 140 . The oscillation signal generation circuit 140 generates an oscillation signal OSCK using the frequency control data DFCQ and the oscillator XTAL from the processing unit 50, and outputs it to the frequency control data generation unit 40 (phase comparison unit 41). Thus, a loop is formed by the PLL circuit such as the frequency control data generation unit 40 (phase comparison unit 41 ), the oscillation signal generation circuit 140 , and an accurate oscillation signal OSCK synchronized with the phase of the reference signal RFCK can be generated.

Furthermore, in the present embodiment, even during the normal operation period before the hold mode is detected, the Kalman filter unit 54 of the processing unit 50 operates to perform Kalman filter processing on the frequency control data DFCI. That is, a process of estimating the true value of the observed value for the frequency control data DFCI by Kalman filter processing is performed.

When the hold mode is detected, the true value at the time corresponding to the detection time of the hold mode is stored in the processing unit 50 . Specifically, the burn-in correcting unit 56 holds the true value. Then, the burn-in correcting unit 56 generates burn-in-corrected frequency control data DFCQ by performing arithmetic processing based on the stored true value.

In this way, since burn-in correction is performed based on the true value at the time corresponding to the detection time of the hold mode, the accuracy of burn-in correction can be greatly improved. That is, it is possible to implement burn-in correction that takes into account the effects of observation noise and system noise.

Also, when returning from the hold mode, the oscillation signal generation circuit 140 generates the oscillation signal OSCK based on the frequency control data DFCQ based on the phase comparison result. For example, the oscillation signal OSCK is generated based on the frequency control data DFCQ input from the frequency control data generation unit 40 (phase comparison unit 41 ) via the processing unit 50 . For example, when the disappearance state or abnormal state of the reference signal RFCK is eliminated, the state of the hold mode is canceled and the state of the hold mode is restored. In this case, the operation of the circuit device returns to normal operation. Moreover, the oscillation signal generating circuit 140 is not based on the frequency control data DFCQ generated by the processing unit 50 by performing aging correction, but based on the frequency control data DFCQ input from the frequency control data generating unit 40 via the processing unit 50 (after signal processing such as temperature compensation processing). frequency control data) to generate the oscillation signal OSCK.

Furthermore, the processing unit 50 generates age-corrected frequency control data DFCQ by performing arithmetic processing of adding a correction value to the stored true value (calculation processing of compensating for a frequency change due to aging). For example, at each predetermined time, the correction value corresponding to the aging rate (gradient of aging, aging coefficient) (the correction value for eliminating the frequency change caused by the aging rate) and the time corresponding to the detection time of the hold mode are sequentially combined. The true values are added to generate aging-corrected frequency control data DFCQ.

For example, let the correction value of time step k be D(k), and the aging-corrected frequency control data of time step k be AC(k). In this case, the processing unit 50 obtains aging-corrected frequency control data AC(k+1) at time step k+1 by AC(k+1)=AC(k)+D(k). The processing unit 50 performs such addition processing of the correction value D(k) at each time step until the time of return from the hold mode (release time).

In addition, the processing unit 50 performs arithmetic processing of adding a corrected value after filtering processing to the true value. For example, filter processing such as low-pass filter processing is performed on the correction value D(k), and arithmetic processing is performed in which the filter-processed correction value D'(k) is sequentially added to the true value. Specifically, calculation processing of AC(k+1)=AC(k)+D'(k) is performed.

Furthermore, the processing unit 50 obtains a correction value based on the observation residual in the Kalman filtering process. For example, the processing unit 50 performs a process of estimating the correction value of the burn-in correction from the observation residual before the hold mode is detected. For example, when the observation residual is ek, the correction value D(k) is estimated by performing a process of D(k)=D(k-1)+E·ek. Here E is, for example, a constant, but it is also possible to use a Kalman gain instead of the constant E. Then, the correction value at the time corresponding to the detection time of the hold mode is stored, and arithmetic processing of adding the stored correction value to the true value is performed to generate aging-corrected frequency control data DFCQ.

Furthermore, the processing unit 50 determines whether or not the hold mode is in the hold mode based on the voltage of the input terminal to which the hold mode detection signal is input or the hold mode detection information input via the digital I/F unit 30 . These determination processes are performed by the hold mode processing unit 52 . For example, the hold mode processing unit 52 has a circuit of a state machine whose state transition is performed based on various signals and information. And, when it is judged to be in the state of the hold mode based on the voltage of the input terminal to which the detection signal of the hold mode is input and the detection information of the hold mode input via the digital I/F section 30, etc., the state of the state machine changes to Holds the state of the schema. Various processing (burn-in correction, etc.) in the hold mode is then performed.

For example, a reference signal RFCK and a signal PLOCK can be assumed as detection signals of the hold mode. In this case, the processing unit 50 determines whether or not the hold mode is in the state based on the voltage of the terminal TRFCK to which the reference signal RFCK is input and the voltage of the terminal TPLOCK to which the signal PLOCK is input.

For example, when a PLL circuit is formed by the frequency control data generator 40 provided inside the circuit device, it can be determined whether or not it is in the hold mode based on the voltage of the terminal TRFCK to which the reference signal RFCK is input. For example, when the processing unit 50 detects that the reference signal RFCK has disappeared or is in an abnormal state based on the voltage of the terminal TRFCK, it determines that the state is in the hold mode.

On the other hand, when the PLL circuit is formed by a frequency control data generator provided outside the circuit device, it can be determined whether or not the hold mode is in the state based on the voltage of the terminal TPLOCK to which the signal PLOCK is input. For example, an external device (a device that controls an external PLL circuit) outputs a signal PLOCK that notifies whether or not the external PLL circuit is locked, to the circuit device. Furthermore, for example, when it is determined by the signal PLOCK that the external PLL circuit is not in the locked state, the processing unit 50 determines that it is in the hold mode. In addition, in addition to the signal PLOCK, it is also possible to use the reference signal RFCK to determine whether or not the state is in the hold mode. In addition, the external PLL circuit is a PLL circuit constituted by, for example, a frequency control data generating unit provided outside the circuit device, an oscillation signal generating circuit 140 of the circuit device, and the like.

In addition, when the PLL circuit is formed by a frequency control data generating unit provided outside the circuit device, it is possible to determine whether or not the hold mode is in the hold mode based on the hold mode detection information input via the digital I/F unit 30 . For example, when an external device (for example, a microcomputer) controlling an external PLL circuit determines that it has entered the state of the hold mode based on the disappearance or abnormality of the reference signal, the detection information of the hold mode is set via the digital I/F section 30. Registers in the register section 32 (notification registers). The processing unit 50 judges whether or not it is in the hold mode by reading the hold mode detection information set in the register. In this way, there is no need to newly provide a terminal for detecting the hold mode, and reduction in the number of terminals of the circuit device can be realized.

5. Aging correction using Kalman filter processing

In this embodiment, a burn-in correction method using Kalman filter processing is employed. Specifically, in the present embodiment, the true value of the observed value of the frequency control data (oscillating frequency) is estimated by Kalman filter processing until the hold mode is detected. Then, when the hold mode is detected, a true value at a time (time point) corresponding to the detection time of the hold mode is stored, and arithmetic processing based on the stored true value is performed, whereby burn-in correction is realized.

FIG. 10 is a graph showing an example of measurement results of fluctuations in oscillation frequency due to aging. The horizontal axis represents the elapsed time (aging time), and the vertical axis represents the frequency deviation (Δf/f ₀ ) of the oscillation frequency. As shown by C1 in FIG. 10 , there are large variations in the measured values as observed values due to system noise and observation noise. The deviation due to the ambient temperature is also included in this deviation.

In such a situation where there is a large variation in the measured value, in order to accurately obtain the true value, in this embodiment, state estimation by Kalman filter processing (for example, linear Kalman filter processing) is performed.

FIG. 11 shows a state-space model of time series, and the discrete-time state equation of the model is given by the state equation and observation equation of the following equations (3) and (4).

x(k+1)=A·x(k)+v(k)···(3)

y(k)=x(k)+w(k)···(4)

x(k) is the state at time k, and y(k) is the observed value. v(k) is the system noise, w(k) is the observation noise, and A is the system matrix. When x(k) is an oscillation frequency (frequency control data), A corresponds to, for example, an aging rate (aging coefficient). The aging rate indicates the rate of change of the oscillation frequency with respect to the elapsed period.

For example, it is assumed that the hold mode occurs at the timing indicated by C2 in FIG. 10 . In this case, burn-in correction is performed based on the actual state x(k) at the time C2 when the reference signal RFCK is interrupted, and the burn-in rate (A) corresponding to the slope shown by C3 in FIG. 10 . Specifically, as compensation (correction) for reducing the frequency change caused by the aging rate shown by C3, for example, the oscillation frequency at the time of C2 is adjusted (frequency control) with a correction value that cancels (cancels) the frequency change. Aging correction for sequential changes in the true value x(k) of the data). That is, the correction value that eliminates the frequency change at the aging rate shown by B2 in FIG. 6 so that the ideal characteristic shown by B1 changes the true value x(k). In this way, for example, when the hold mode period is 24 hours, the FDV in FIG. 10 , which is a change in the oscillation frequency after 24 hours has elapsed, can be compensated by aging correction.

Here, the fluctuations in the oscillation frequency (frequency deviation) indicated by C1 in FIG. 10 include fluctuations due to temperature fluctuations and fluctuations due to aging. Therefore, in this embodiment, for example, by employing an oscillator (OCXO) having a constant temperature bath structure with a constant temperature bath, fluctuations in the oscillation frequency due to temperature fluctuations are suppressed to a minimum. In addition, temperature compensation processing for reducing fluctuations in the oscillation frequency due to temperature fluctuations is performed using the temperature sensor 10 in FIG. 9 or the like.

Furthermore, during the period in which the PLL circuit (internal PLL circuit, external PLL circuit) is synchronized with the reference signal RFCK (normal operation period), the frequency control data (frequency control code) is monitored, and the error (system noise, observation noise) is calculated The true value of and stored in the register. Also, when the lock of the PLL circuit is released due to disappearance or abnormality of the reference signal RFCK, burn-in correction is performed based on the true value (true value of the observed value for the frequency control data) held at the time of the lock release. For example, as compensation for reducing the frequency change caused by the slope of C3 in FIG. Here, the frequency control data DFCQ at the time of self-oscillation during the hold mode is generated, and the vibrator XTAL is oscillated. In this way, since the truth value at the time of entering the hold mode can be obtained with the minimum error and aging correction can be performed, it is possible to realize hold mode performance in which adverse effects due to aging fluctuations are minimized.

6. Structure of processing department

FIG. 12 shows a detailed configuration example of the processing unit 50 . In addition, the configuration of the processing unit 50 is not limited to the configuration of FIG. 12 , and various modifications such as omitting some of the constituent elements or adding other constituent elements can be implemented.

As shown in FIG. 12 , the processing unit 50 includes a Kalman filter unit 54 , an aging correction unit 56 , a temperature compensation unit 58 , selectors 62 and 63 , and an adder 65 .

The Kalman filter unit 54 is input with frequency control data DFCI (frequency control data from which environmental fluctuation components have been removed), and executes Kalman filter processing. Furthermore, an a posteriori estimated value x^(k) corresponding to the true value estimated by the Kalman filtering process is output. In addition, in this specification, the hat-shaped symbol " ^{^} " indicating an estimated value is appropriately arranged and described in two characters.

The Kalman filtering process refers to a process of estimating the optimal state of the system using observed values obtained from the past to the present, assuming that noise (error) is included in observed values and variables representing the state of the system. Specifically, observation update (observation process) and time update (prediction process) are repeated to estimate the state. Observation update is the process of updating Kalman gains, estimates, and error covariances using the results of observations and time updates. Time updating is the process of using the results of observation updates to predict the estimated value and error covariance of the next moment. In addition, in this embodiment, a method using linear Kalman filter processing is mainly described, but extended Kalman filter processing can also be used. The details of the Kalman filter processing in this embodiment will be described later.

The burn-in correcting unit 56 receives the a posteriori estimated value x^(k) and the corrected value D'(k) from the Kalman filter unit 54. Then, AC(k), which is aging-corrected frequency control data, is generated by performing an arithmetic process of adding a correction value D'(k) to an a posteriori estimated value x^(k) corresponding to the true value of the frequency control data. Here, D'(k) is a correction value D(k) after filtering (after low-pass filtering). That is, when the correction value (correction value after filtering) at time step k (time k) is D'(k), and the aging-corrected frequency control data at time step k is AC(k), the aging The correction unit 56 obtains aging-corrected frequency control data AC(k+1) at time step k+1 (time k+1) by AC(k+1)=AC(k)+D′(k).

The temperature compensation unit 58 receives temperature detection data DTD as input, performs temperature compensation processing, and generates temperature compensation data TCODE (temperature compensation code) for keeping the oscillation frequency constant with respect to temperature fluctuations. The temperature detection data DTD is data obtained by A/D converting the temperature detection voltage VTD from the temperature sensor 10 by the A/D conversion unit 20 of FIG. 9 .

For example, examples of initial oscillation frequency temperature characteristics are shown in FIG. 13 , FIG. 14 , and FIG. 15 . In these figures, the horizontal axis is the ambient temperature, and the vertical axis is the frequency deviation of the oscillation frequency. As shown in FIGS. 13 to 15 , the temperature characteristics of the oscillation frequency vary greatly depending on the samples of each product. Therefore, the temperature characteristics of the oscillation frequency and the change characteristics of the temperature detection data corresponding to the ambient temperature are measured in the inspection process at the time of manufacture and shipment of the product (oscillator). Furthermore, the coefficients A ₀ to A ₅ of the polynomial (approximate function) of the following formula (5) are obtained based on the measurement results, and the information of the obtained coefficients A ₀ to A ₅ is written into the storage unit 34 of FIG. 9 (non-volatile stored in volatile memory).

TCODE＝A ₅ X ⁵ +A ₄ X ⁴ +A ₃ X ³ +A ₂ X ² +A ₁ X+A ₀ (5)

In the above formula (5), X corresponds to the temperature detection data DTD (A/D conversion value) obtained by the A/D conversion unit 20 . Since the change in the temperature detection data DTD with respect to the change in the ambient temperature is also measured, the ambient temperature can be associated with the oscillation frequency by the approximate function represented by the polynomial of the above equation (5). The temperature compensating unit 58 reads the information of the coefficients A ₀ to A ₅ from the storage unit 34 , performs the calculation processing of the above formula (5) based on the coefficients A ₀ to A ₅ and the temperature detection data DTD (=X), and generates temperature compensation data. TCODE (Temperature Compensation Code). Thereby, temperature compensation processing for keeping the oscillation frequency constant with respect to changes in ambient temperature can be realized.

The selectors 62 and 63 select the input signal of the terminal on the "1" side when the logic level of the input signal of the selection terminal S is "1" (active), and output it as an output signal. In addition, when the logic level of the input signal of the selection terminal S is "0" (inactive), the input signal of the terminal on the "0" side is selected and output as an output signal.

Signal KFEN is an enable signal for Kalman filter processing. Kalman filter unit 54 executes Kalman filter processing when signal KFEN is at logic level “1” (hereinafter, abbreviated as “1”). The signal PLLLOCK is a signal that becomes "1" when the PLL circuit is in the locked state. The signal HOLDOVER is a signal that becomes "1" during the hold mode in which the hold mode is detected. These signals PLLLOCK and HOLDOVER are generated by the circuit of the state machine of the hold mode processing unit 52 in FIG. 9 .

Signal TCEN is an enable signal for temperature compensation processing. Hereinafter, the case where the signal TCEN is “1” and the selector 63 selects an input signal on the “1” side will be described as an example. In addition, the signal KFEN is also "1".

During normal operation, since the signal HOLDOVER is logic level "0" (hereinafter, abbreviated as "0"), the selector 62 selects the frequency control data DFCI on the side of the "0" terminal. Furthermore, the adder 65 The temperature compensation data TCODE is added to the frequency control data DFCI, and the temperature compensation-processed frequency control data DFCQ is output to the subsequent oscillation signal generation circuit 140 .

On the other hand, during the hold mode, the signal HOLDOVER is "1", and the selector 62 selects AC(k) on the "1" terminal side. AC(k) is the age-corrected frequency control data.

FIG. 16 is a truth table illustrating the operation of the Kalman filter unit 54 . When the signals PLLLOCK and KFEN are both "1", the Kalman filter unit 54 executes a truth value estimation process (Kalman filter process). That is, when the PLL circuit (internal or external PLL circuit) is in the locked state during the normal operation period, the process of estimating the true value of the frequency control data DFCI as the observed value continues.

Furthermore, when the lock of the PLL circuit is released and the signal PLLLOCK is "0" in the state of the hold mode, the Kalman filter unit 54 holds the previous output state. For example, in FIG. 12, the value at the detection time of the hold mode (the time when the lock of the PLL circuit is released) is stored and continuously output as the a posteriori estimated value x^(k) and the true value of the frequency control data DFCI. Correction value D'(k) for aging correction.

The burn-in correction unit 56 performs burn-in correction using the a posteriori estimated value x^(k) and the correction value D'(k) from the Kalman filter unit 54 during the hold mode period. Specifically, the a posteriori estimated value x^(k) and the corrected value D'(k) at the time of detection of the hold mode are stored, and burn-in correction is performed.

In addition, in FIG. 12 , the frequency control data DFCI from which temperature variation components (environmental variation components in a broad sense) and aging variation components are removed are input to the Kalman filter unit 54 . The Kalman filter unit 54 performs Kalman filter processing on the frequency control data DFCI from which the temperature fluctuation component (environmental fluctuation component) has been removed, and estimates the true value of the frequency control data DFCI. That is, the posterior estimated value x^(k) is obtained. Furthermore, the burn-in correcting unit 56 performs burn-in correction based on the estimated true value, that is, the a posteriori estimated value x^(k). More specifically, aging-corrected frequency control data AC(k) is obtained from the a posteriori estimated value x^(k) and the corrected value D'(k) from the Kalman filter unit 54. Then, AC(k), which is the frequency control data after aging correction, is input to the adder 65 via the selector 62, and the adder 65 adds temperature compensation data TCODE (data for compensating environmental fluctuation components) to AC(k). .

For example, as shown in the schematic diagram of FIG. 17 , when the temperature fluctuates, the frequency control data also fluctuates correspondingly as shown in E1 . Therefore, when Kalman filter processing is performed using frequency control data that fluctuates with temperature fluctuations like E1 , the true value at the time of holding the mode detection also fluctuates.

Therefore, in the present embodiment, the frequency control data from which the temperature variation component has been removed is acquired and input to the Kalman filter unit 54 . That is, the frequency control data from which temperature fluctuation components (environmental fluctuation components) and aging fluctuation components have been removed are input to the Kalman filter unit 54 . That is, the frequency control data indicated by E2 in FIG. 17 is input. The frequency control data of E2 is frequency control data in which the temperature fluctuation component is removed and the aging fluctuation component remains.

The Kalman filter unit 54 performs Kalman filter processing on the frequency control data DFCI with the temperature variation component removed and the aging variation component remaining in this way, to obtain the posterior estimated value x^(k) of the estimated true value and the aging-corrected Correction value D'(k). Furthermore, the a posteriori estimated value x^(k) and the corrected value D'(k), which are true values estimated at the detection time of the hold mode, are stored in the burn-in correcting unit 56, and are used for burn-in correction.

For example, the temperature compensation data TCODE is added by the adder 65, and the frequency control data DFCQ becomes temperature-compensated frequency control data. Therefore, the oscillation signal generating circuit 140 to which the frequency control data DFCQ is input outputs the oscillation signal OSCK of the temperature-compensated oscillation frequency. Therefore, the frequency control data generation unit 40 of FIG. 9 constituting a PLL circuit together with the oscillation signal generation circuit 140 supplies the frequency control data DFCI from which the temperature fluctuation component has been removed as shown in E2 of FIG. 17 to the processing unit 50 . Then, as shown in E2 of FIG. 17 , the frequency control data DFCI from which this temperature fluctuation component has been removed remains an aging fluctuation component that changes with the elapse of time. Therefore, the Kalman filter unit 54 of the processing unit 50 performs Kalman filter processing on the frequency control data DFCI in which the aging fluctuation component remains, and if the aging correction unit 56 performs aging correction based on the result of the Kalman filter processing, high accuracy can be achieved. aging correction.

12, instead of adding the temperature compensation data TCODE in the adder 65, calculation processing for removing the temperature fluctuation component (environmental fluctuation component) of the frequency control data DFCI may be performed, and the The calculated frequency control data DFCI is input to the Kalman filter unit 54 . For example, the structure of the adder 65 and the selector 63 in FIG. 12 is omitted, and a subtracter for subtracting the temperature compensation data TCODE from the frequency control data DFCI is provided in the preceding stage of the Kalman filter unit 54, and the output of the subtracter is input to the Kalman filter unit 54. Mann filter unit 54 . Also, an adder for adding the output of burn-in corrector 56 to temperature compensation data TCODE is provided between burn-in corrector 56 and selector 62 , and the output of the adder is input to the “1” side terminal of selector 62 . With such a configuration, it is also possible to input the frequency control data DFCI from which the temperature variation component has been removed and only the aging variation component remains, to the Kalman filter unit 54 .

FIG. 18 shows a detailed configuration example of the burn-in correction unit 56 . Since the signal HOLDOVER is "0" during the normal operation period, the selectors 360 and 361 select the "0" terminal side. Thus, during the normal operation period, the a posteriori estimated value x^(k) calculated by the Kalman filter unit 54 and the corrected value D′(k) (corrected value after filter processing) are stored in the register 350, respectively. 351.

When the hold mode is detected and the signal HOLDOVER is "1", the selectors 360, 361 select the "1" terminal side. Thus, the selector 361 continuously outputs the correction value D'(k) stored in the register 351 at the time of detection of the hold mode during the hold mode period.

Furthermore, the adder 340 performs a process of sequentially adding the a posteriori estimated value x^(k) stored in the register 351 and output from the selector 361 to the a posteriori estimated value x^(k) stored in the register 350 at the time of detection of the hold mode for each time step. Correction value D'(k) (correction value). Thereby, burn-in correction represented by the following formula (6) is realized.

AC(k+1)=AC(k)+D'(k)···(6)

That is, aging correction is realized by performing the following process: adding a correction value D'(k) sequentially to the true value stored at the time point C2 in FIG. ) is used to cancel (compensate) the frequency change caused by the aging rate corresponding to the slope of C3.

7. Kalman filter processing

Next, details of the Kalman filter processing in this embodiment will be described. FIG. 19 shows a model example of Kalman filtering. The state equation and observation equation of the model in FIG. 19 are represented by the following equations (7) and (8).

x(k+1)=A·x(k)+v(k)···(7)

y(k)=C ^T x(k)+w(k)···(8)

K represents a time step as discrete time. x(k) is the state of the system at time step k (time k), and is, for example, an n-dimensional vector. A is called the system matrix. Specifically, A is an n×n matrix that relates the state of the system at time step k to the state of the system at time step k+1 in the absence of system noise. v(k) is the system noise. y(k) is the observed value and w(k) is the observation noise. C is the observation coefficient vector (n dimension), T represents the transpose matrix.

In the Kalman filter processing of the models of the above equations (7) and (8), the processing of the following equations (9) to (13) is performed to estimate the true value.

P ^- (k) = A·P(k-1)· ^AT +v(k)···(10)

P(k)＝(1-G(k)·C ^T )·P ^- (k)···(13)

x^(k): Posterior estimate

x^ ^- (k): prior estimate

P(k): posterior covariance

P ^- (k): prior covariance

G(k): Kalman gain

The above formulas (9) and (10) are expressions of time update (prediction process), and the above formulas (11) to (13) are formulas of observation update (observation process). Each time the time step k that is a discrete time advances by one, time update (Equations (9), (10)) and observation update (Equations (11) to (13)) of the Kalman filter process are performed once.

x^(k), x^(k-1) are the posterior estimates of the Kalman filtering process at time steps k and k-1. x^ ^- (k) is an a priori estimate of the forecast before the observations are made. P(k) is the posterior covariance of the Kalman filtering process, and P-(k) is the prior covariance predicted before the observations are obtained. G(k) is the Kalman gain.

In the Kalman filtering process, the Kalman gain G(k) is obtained by the above equation (11) in the observation update. In addition, according to the observed value y(k), the posterior estimated value x^(k) is updated through the above formula (12). In addition, the posteriori covariance P(k) of the error is updated through the above formula (13).

In addition, in the Kalman filtering process, in the time update, as shown in the above formula (9), the next time step is predicted according to the posterior estimated value x^(k-1) of time step k-1 and the system matrix A The prior estimate of k x^ ^- (k). In addition, as shown in the above formula (10), according to the posterior covariance P(k-1) of time step k-1, the system matrix A, and the system noise v(k), the prior covariance of the next time step k is predicted P-(k).

In addition, when the Kalman filter processing of the above expressions (9) to (13) is to be performed, the processing load of the processing unit 50 may be too large, resulting in an increase in the size of the circuit device. For example, in order to obtain A of x^ ^- (k)=Ax^(k-1) in the above formula (9), extended Kalman filter processing is required. Furthermore, the processing load of the extended Kalman filter processing is very heavy, and when the processing unit 50 is realized by hardware capable of performing the extended Kalman filter processing, the circuit area of the processing unit 50 tends to become very large. Therefore, it is not suitable in a situation where miniaturization of a circuit device built in an oscillator is strongly required. On the other hand, when a scalar value of a fixed value is used as the system matrix A, the degree of difficulty in realizing appropriate burn-in correction increases.

Therefore, as a solution when it is necessary to avoid such a situation, in this embodiment, Kalman filtering is realized by processing based on the following expressions (14) to (19) instead of the above expressions (9) to (13). deal with. That is, the processing unit 50 (Kalman filter unit 54 ) executes Kalman filter processing based on the following expressions (14) to (19).

P- ⁽ k)=P(k-1)+v(k)···(15)

P(k)＝(1-G(k))·P- ⁽ k)···(18)

In addition, in the present embodiment, x(k) which is the target of the estimation process of the true value is frequency control data, and observed value y(k) is also frequency control data, therefore, C=1. In addition, since the scalar value of A is infinitely close to 1, the above equation (15) can be used instead of the above equation (10).

As described above, compared with the case where the extended Kalman filter is used as the Kalman filter, in the Kalman filter of this embodiment, as shown in the above formula (14), the posteriori of time step k-1 The estimated value x^(k-1) and the correction value D(k-1) are added to obtain the prior estimate value x^ ^- (k) of time k. Therefore, there is no need to use extended Kalman filter processing, which is excellent in reducing the processing load on the processing unit 50 and suppressing an increase in circuit scale.

In the present embodiment, the above formula (14) is derived by modifying the following formula.

For example, the above formula (20) can be modified like the above formula (21). Here, since (A-1) in the above formula (21) is a very small number, as shown in the above formulas (22) and (23), it is possible to use (A-1)·x^(k-1) The substitution is an approximation of (A-1)·F ₀ . Then, this (A-1)·F ₀ is replaced with the correction value D(k-1).

And as shown in the above formula (19), when updating from time step k-1 to time step k, the correction value D(k)=D(k-1)+E·(y(k)-x^ ^- (k))=D(k-1)+E·ek update process. Here, ek=y(k)-x^ ^- (k) is called an observation residual in the Kalman filtering process. Also, E is a constant. In addition, instead of the constant E, a modification using the Kalman gain G(k) can also be implemented. That is, it may be D(k)=D(k-1)+G(k)·ek.

In this way, in Equation (19), when the observation residual is ek and the constant is E, the correction value D(k) is obtained by D(k)=D(k-1)+E·ek. In this way, it is possible to perform update processing of the correction value D(k) reflecting the observation residual ek in the Kalman filter processing.

FIG. 20 shows a configuration example of the Kalman filter unit 54 . Kalman filter unit 54 includes adders 300 , 301 , 302 , 303 , 304 , multiplier 305 , registers 310 , 311 , 312 , 313 , selectors 320 , 321 , filters 330 , 331 , and arithmetic units 332 , 333 . In addition, the configuration of the Kalman filter unit 54 is not limited to the configuration shown in FIG. 20 , and various modifications such as omitting some of the constituent elements or adding other constituent elements can be implemented. For example, the processing of the adders 300 to 304 and the like can be realized by time-division processing of one arithmetic unit.

With the adder 304 and the register 312, the arithmetic processing of the above formula (14) is executed. In addition, information on the system noise constant V for setting the system noise and the observed noise constant W for setting the observed noise is read from the storage unit 34 in FIG. 9 and input to the Kalman filter unit 54 (processing unit 50 ). And, the arithmetic processing of the above formula (15) is executed by the adder 300 and the register 310 . In addition, the arithmetic unit 332 executes the arithmetic processing of the above formula (16) to obtain the Kalman gain G(k). Then, based on the calculated Kalman gain G(k), the adder 301 , the multiplier 305 , and the adder 302 execute the arithmetic processing of the above formula (17). In addition, the arithmetic unit 333 executes the arithmetic processing of the above formula (18) to obtain the posterior covariance P(k).

In addition, the arithmetic processing of the above formula (19) is executed by the adder 303 , the register 311 and the filter 330 . Information on the constant E input to the filter 330 is read from the storage unit 34 in FIG. 9 . The constant E corresponds to a correction coefficient (filter constant) for the aging rate. For example, the filter 330 can realize E·( ^y (k)-x̂-(k)) in the above formula (19) by performing gain adjustment and the like according to the constant E.

When the signals PLLLOCK and KFEN are respectively "1", the selectors 320 and 321 select the input signal of the terminal on the "1" side. The output signal of the selector 320 is saved in the register 313 . Therefore, after the state of the hold mode is entered and the signal PLLLOCK changes from “1” to “0”, x̂(k), which is a true value at the time of detection of the hold mode, is stored in the register 313 .

The filter 331 performs filter processing on the correction value D(k). Specifically, digital low-pass filtering is performed on the correction value D(k), and the filtered correction value D'(k) is input to the burn-in correction unit 56 in FIG. 18 . The constant J is a filter constant of the filter 331 . According to the constant J, the optimum cut-off frequency of the filter 331 is set.

For example, as can be seen from FIG. 10 , there are fine fluctuations in the correction value D(k) for compensating for the frequency change due to the aging rate. Therefore, when the correction value D(k) fluctuates in this way is added to the true value, the accuracy of burn-in correction decreases.

In this regard, in the present embodiment, the corrected value D'(k) after the filtering process is added to the true value, so more accurate burn-in correction can be realized.

As described above, in the present embodiment, as shown in the above formula (14), the processing unit 50 performs the following process in the update process (time update) of the a priori estimated value of the Kalman filter process: The process of adding the a posteriori estimated value x^(k-1) and the correction value D(k-1) obtains the a priori estimated value x^ ^- (k) at this time. Furthermore, based on the result of the Kalman filter processing, aging correction of the frequency control data is performed. That is, the addition process of the posterior estimated value x^(k-1) and the correction value D(k-1) at the last time, that is, time step k-1, is performed by x^ ^- (k)=x^( k-1)+D(k-1) Calculate the prior estimate x^ ^- (k) of time step k at this moment.

Then, the processing unit 50 (burn-in correcting unit 56 ) performs burn-in correction based on the result (true value, corrected value) of the Kalman filter processing. That is, when the correction value of time step k is D(k) (or D'(k)), and the aging-corrected frequency control data of time step k is AC(k), by AC(k+1 )=AC(k)+D(k) (or AC(k)+D'(k)) Calculate aging-corrected frequency control data AC(k+1) at time step k+1.

Furthermore, the processing unit 50 obtains the correction value D( k). For example, the correction at this time is obtained by adding E·ek (or G(k)·ek), which is a value based on the observation residual, to the correction value D(k-1) at the previous time. Value D(k). Specifically, based on the correction value D(k-1) at time step k-1 at the previous time and the observation residual ek in the Kalman filtering process, the correction value D at time step k at this time is obtained (k). For example, when the observation residual is ek and the constant is E, the correction value D(k) is obtained by D(k)=D(k-1)+E·ek.

For example, in this embodiment, as described in FIG. 17 , environmental change component information such as temperature change component information is obtained, and the environmental change component from which the environmental change component and the aging change component have been removed is obtained by using the acquired environmental change component information. frequency control data. Here, the environmental variation component information may be a power supply voltage variation component, an air pressure variation component, a gravity variation component, or the like. Then, burn-in correction is performed based on the frequency control data from which the environmental fluctuation component has been removed. Specifically, the environmental variation component is assumed to be temperature. The temperature fluctuation component information as the environmental fluctuation component information is acquired based on the temperature detection data DTD obtained by the temperature detection from the temperature sensor 10 as the environmental fluctuation information acquisition unit of FIG. It is obtained from the voltage VTD. Then, using the acquired temperature variation component information, frequency control data from which the temperature variation component has been removed is acquired. For example, the temperature compensation unit 58 in FIG. 12 acquires the temperature compensation data TCODE, and the temperature compensation data TCODE is added by the adder 65, whereby the frequency control data DFCI from which the temperature fluctuation component has been removed is input from the frequency control data generation unit 40, and acquired by the processing unit 50 . That is, as shown in E2 of FIG. 17 , the frequency control data DFCI in which the temperature variation component is removed and the aging variation component remains is acquired and input to the Kalman filter unit 54 .

In addition, the frequency control data from which environmental fluctuation components have been removed includes not only frequency control data in an appropriate state in which environmental fluctuation components have been completely removed, but also frequency control data in which there are negligible environmental fluctuation components in the frequency control data. data.

For example, environmental variation component information such as temperature variation component information and power supply voltage variation component information can be acquired by a temperature sensor, a voltage detection circuit, etc. as an environmental variation information acquisition unit that detects the environmental variation component information. On the other hand, the aging variation component is a variation component of the oscillation frequency that changes with time, and it is difficult to directly detect information of the aging variation component with a sensor or the like.

Therefore, in the present embodiment, environmental change component information such as temperature change component information that can be detected by a sensor or the like is obtained, and the environmental change component information is used to obtain a frequency with the environmental change component removed from the environmental change component and the aging change component. control data. That is, by performing processing (for example, adding processing by the adder 65 ) to remove the environmental variation component from the variation component of the frequency control data, frequency control data in which only the aging variation component remains can be obtained as shown in E2 of FIG. 17 . Then, by performing Kalman filter processing or the like on the frequency control data in which the aging fluctuation component remains, the true value of the aging fluctuation component with respect to the frequency control data can be estimated. Furthermore, if burn-in correction is performed based on the true value estimated in this way, high-accuracy burn-in correction that cannot be realized in conventional examples can be realized.

Thus, in the present embodiment, the frequency control data DFCI from which the temperature variation component (environmental variation component) has been removed and the aging variation component remains is input to the Kalman filter unit 54 . Furthermore, as shown in FIG. 1 and FIG. 10 , if a period is limited, it can be assumed that the oscillation frequency changes at a constant aging rate within the period. It is possible to assume, for example, a constant slope change as indicated by C3 in FIG. 10 .

In the present embodiment, the frequency for compensating (eliminating) the constant aging rate due to such an aging fluctuation component is obtained by the formula of D(k)=D(k-1)+E·ek Changed correction value. That is, the correction value D(k) for compensating the frequency change due to the aging rate corresponding to the slope of C3 in FIG. 10 was obtained. Here, the aging rate is not constant, but changes with time as shown in Fig. 1 and Fig. 10 .

On the other hand, in this embodiment, as D(k)=D(k-1)+E·ek, the observation residual ek=y(k)-x^ ^- (k) from the Kalman filtering process, Update processing of the correction value D(k) corresponding to the aging rate is performed. Therefore, it is possible to realize update processing of the correction value D(k) that also reflects the change in the aging rate corresponding to the elapsed time. Therefore, burn-in correction with higher accuracy can be realized.

For example, in FIG. 21 , the measured frequency deviation and the predicted frequency deviation are compared and shown. D1 is the frequency deviation of the actually measured oscillation frequency, and D2 is the frequency deviation of the oscillation frequency predicted by the estimation process of the Kalman filter in this embodiment. The predicted frequency deviation indicated by D2 falls within the allowable error range with respect to the measured frequency deviation indicated by D1 , indicating that high-accuracy burn-in correction is realized by this embodiment.

8. Temperature sensor, oscillation circuit

FIG. 22 shows a configuration example of the temperature sensor 10 . The temperature sensor 10 of FIG. 22 has a current source IST, and a bipolar transistor TRT whose collector is supplied with a current from the current source IST. The bipolar transistor TRT is diode-connected with its collector and base connected, and outputs a temperature detection voltage VTDI having a temperature characteristic to the node of the collector of the bipolar transistor TRT. The temperature characteristic of the temperature detection voltage VTDI is caused by the temperature dependence of the base-emitter voltage of the bipolar transistor TRT. The temperature detection voltage VTDI of the temperature sensor 10 has, for example, a negative temperature characteristic (primary temperature characteristic with a negative gradient).

FIG. 23 shows a configuration example of the oscillation circuit 150 . This oscillation circuit 150 has a current source IBX, a bipolar transistor TRX, a resistor RX, a variable capacitance capacitor CX1, and capacitors CX2 and CX3.

The current source IBX supplies a bias current to the collector of the bipolar transistor TRX. The resistor RX is disposed between the collector and the base of the bipolar transistor TRX.

One end of the variable capacitance capacitor CX1 with variable capacitance is connected to one end of the vibrator XTAL. Specifically, one end of the variable capacitance capacitor CX1 is connected to one end of the oscillator XTAL via the first terminal for the oscillator (pad for the oscillator) of the circuit device. One end of the capacitor CX2 is connected to the other end of the vibrator XTAL. Specifically, one end of the capacitor CX2 is connected to the other end of the oscillator XTAL via the second terminal for the oscillator (pad for the oscillator) of the circuit device. One end of the capacitor CX3 is connected to one end of the vibrator XTAL, and the other end is connected to the collector of the bipolar transistor TRX.

A base-emitter current generated by oscillation of the oscillator XTAL flows through the bipolar transistor TRX. And, when the base-emitter current increases, the collector-emitter current of the bipolar transistor TRX increases, and the bias current branched from the current source IBX to the resistor RX decreases. Therefore, the collector voltage VCX reduce. On the other hand, when the base-emitter current of the bipolar transistor TRX decreases, the collector-emitter current decreases, and the bias current branched from the current source IBX to the resistor RX increases, so the collector voltage VCX rises. This collector voltage VCX is fed back to the oscillator XTAL via the capacitor CX3.

The oscillation frequency of the vibrator XTAL has a temperature characteristic, and this temperature characteristic is compensated by the output voltage VQ (frequency control voltage) of the D/A converter 80 . That is, the output voltage VQ is input to the variable capacitance capacitor CX1, and the capacitance value of the variable capacitance capacitor CX1 is controlled by the output voltage VQ. When the capacitance value of the variable capacitance capacitor CX1 changes, the resonant frequency of the oscillation loop changes, so the fluctuation in the oscillation frequency caused by the temperature characteristics of the oscillator XTAL is compensated. The variable capacitance capacitor CX1 can be realized by, for example, a variable capacitance diode (varactor: variable capacitance diode) or the like.

In addition, the oscillation circuit 150 of the present embodiment is not limited to the structure shown in FIG. 23 , and various modifications can be made. For example, in FIG. 23 , the case where CX1 is a variable capacitance capacitor has been described as an example. However, CX2 or CX3 may be a variable capacitance capacitor controlled by the output voltage VQ. In addition, a plurality of CX1 to CX3 may be variable capacitance capacitors controlled by VQ.

In addition, the oscillation circuit 150 does not need to include all circuit elements for oscillating the vibrator XTAL. For example, a configuration may be employed in which a part of the circuit element is constituted by a discrete component provided outside the circuit device 500 and connected to the oscillator circuit 150 via an external connection terminal.

9. Modification

Next, various modified examples of the present embodiment will be described. FIG. 24 shows a configuration example of a circuit device according to a modified example of the present embodiment.

In FIG. 24 , unlike FIGS. 2 and 9 , the D/A converter 80 is not provided in the oscillation signal generation circuit 140 . Furthermore, the oscillation frequency of the oscillation signal OSCK generated by the oscillation signal generation circuit 140 is directly controlled according to the frequency control data DFCQ from the processing unit 50 . That is, the oscillation frequency of the oscillation signal OSCK is controlled without going through the D/A conversion unit.

For example, in FIG. 24 , an oscillation signal generation circuit 140 has a variable capacitance circuit 142 and an oscillation circuit 150 . The D/A converter 80 shown in FIGS. 2 and 9 is not provided in this oscillation signal generation circuit 140 . In addition, this variable capacitance circuit 142 is provided instead of the variable capacitance capacitor CX1 of FIG. 23 , and one end of the variable capacitance circuit 142 is connected to one end of the vibrator XTAL.

The capacitance value of the variable capacitance circuit 142 is controlled based on the frequency control data DFCQ from the processing unit 50 . For example, the variable capacitance circuit 142 has a plurality of capacitors (capacitor array), and a plurality of switching elements (switch array) for controlling on and off of each switching element based on frequency control data DFCQ. Each of the plurality of switching elements is electrically connected to each of the plurality of capacitors. And, by turning on or off the plurality of switching elements, the number of capacitors whose one end is connected to one end of the vibrator XTAL among the plurality of capacitors changes. As a result, the capacitance value of the variable capacitance circuit 142 is controlled, and the capacitance value at one end of the vibrator XTAL changes. Therefore, the capacitance value of the variable capacitance circuit 142 can be directly controlled by using the frequency control data DFCQ, and the oscillation frequency of the oscillation signal OSCK can be controlled.

In addition, when a PLL circuit is configured using the circuit device of this embodiment, it can also be a PLL circuit of a direct digital synthesizer system. FIG. 25 shows an example of a circuit configuration in the case of a direct digital synthesizer method.

The phase comparison unit 380 (comparison operation unit) performs phase comparison (comparison operation) of the reference signal RFCK and the oscillation signal OSCK (input signal based on the oscillation signal). The digital filter unit 382 performs smoothing processing of the phase error. The configuration and operation of the phase comparison unit 380 are the same as those of the phase comparison unit 41 in FIG. 9 , and can include a counter and a TDC (time-to-digital converter). The digital filter unit 382 corresponds to the digital filter unit 44 in FIG. 9 . The numerically controlled oscillator 384 is a circuit for digitally synthesizing an arbitrary frequency and waveform using a reference oscillation signal from a reference oscillator 386 having a vibrator XTAL. That is, instead of controlling the oscillation frequency according to the control voltage from the D/A converter like VCO, digital frequency control data and the reference oscillator 386 (oscillator XTAL) are used to generate an oscillation of an arbitrary oscillation frequency through digital operation processing. Signal OSCK. With the configuration of FIG. 25, an ADPLL circuit of a direct digital synthesizer method can be realized.

In addition, in the method of the present embodiment described above, Kalman filter processing is performed on the aging fluctuation component obtained by removing the environmental fluctuation component from the fluctuation component of the frequency control data, and the true value after removing the bias is estimated. Further, by calculating a correction value (oscillation characteristic variation coefficient) for burn-in correction, and performing burn-in correction using the true value and the corrected value, control to keep the oscillation frequency constant is realized. However, the method of this embodiment is not limited to such a method, and various modifications are possible.

For example, in the method of the first modified example of this embodiment, the value of the aging fluctuation component obtained by removing the environmental fluctuation component from the measured value is stored in the storage unit (memory). Furthermore, the oscillation frequency can be controlled based on the stored values of a plurality of aging fluctuation components and an approximate expression of a first-order linear expression or a multi-degree polynomial prepared in advance.

In addition, in the method of the second modification of this embodiment, for example, when the value of the aging fluctuation component obtained by removing the environmental fluctuation component from the measured value is greater than or equal to a predetermined value, switching to Another oscillator is set to suppress the aging effect within a constant range.

In addition, the circuit device of this embodiment can be used not only as a circuit device constituting a loop of a PLL circuit but also as a circuit device for a self-excited oscillator.

10. Oscillators, electronic equipment, moving objects

FIG. 26 shows a configuration example of an oscillator 400 including the circuit device 500 of this embodiment. As shown in FIG. 26 , the oscillator 400 includes a vibrator 420 and a circuit device 500 . The vibrator 420 and the circuit device 500 are installed in the package 410 of the oscillator 400 . Furthermore, the terminals of the vibrator 420 and the terminals (pads) of the circuit device 500 (IC) are electrically connected by internal wiring of the package 410 .

FIG. 27 shows a configuration example of electronic equipment including the circuit device 500 of this embodiment. This electronic device includes the circuit device 500 of this embodiment, a vibrator 420 such as a quartz resonator, an antenna ANT, a communication unit 510 , and a processing unit 520 . In addition, an operation unit 530 , a display unit 540 , and a storage unit 550 may be further included. The oscillator 400 is constituted by the vibrator 420 and the circuit arrangement 500 . In addition, the electronic device is not limited to the configuration of FIG. 27 , and various modifications such as omitting some of the constituent elements or adding other constituent elements can be implemented.

As the electronic equipment shown in FIG. 27 , for example, network-related equipment such as base stations or routers, high-precision measurement equipment, GPS built-in clocks, vital information measurement equipment (pulse meter, pedometer, etc.), or head-mounted display devices, etc. can be assumed Various devices such as wearable devices, smartphones, mobile phones, portable game devices, portable information terminals (mobile terminals) such as notebook PCs and tablet PCs, content providing terminals for distributing content, and video equipment such as digital cameras and video cameras.

The communication unit 510 (wireless circuit) performs processing of receiving data from the outside or transmitting data to the outside via the antenna ANT. The processing unit 520 performs control processing of electronic devices, various digital processing of data transmitted and received via the communication unit 510 , and the like. The function of the processing unit 520 can be realized by a processor such as a microcomputer, for example.

The operation unit 530 is used for the user to perform an input operation, and can be realized by operating buttons, a touch panel display, and the like. The display unit 540 is used to display various information, and can be realized by a display such as liquid crystal or organic EL. In addition, when a touch-panel display is used as operation unit 530 , the touch-panel display functions as both operation unit 530 and display unit 540 . The storage unit 550 is used to store data, and its function can be realized by a semiconductor memory such as RAM and ROM, or an HDD (Hard Disk Drive).

FIG. 28 shows an example of a mobile object including the circuit device of this embodiment. The circuit device (oscillator) of this embodiment can be incorporated into various mobile bodies such as vehicles, airplanes, motorcycles, bicycles, and ships, for example. A mobile body is, for example, a device or device that has a driving mechanism such as an engine or a motor, a steering mechanism such as a steering wheel or a rudder, and various electronic devices (vehicle devices), and moves on land, in the air, or at sea. FIG. 28 schematically shows a car 206 as a specific example of a mobile object. An oscillator (not shown) including a circuit device and a vibrator according to the present embodiment is incorporated in the automobile 206 . The control device 208 operates based on the clock signal generated by this oscillator. The control device 208 controls the stiffness of the suspension or controls the braking of the individual wheels 209 according to, for example, the attitude of the vehicle body 207 . For example, the automatic operation of vehicle 206 can be realized by means of control device 208 . In addition, the device incorporating the circuit device or the oscillator according to the present embodiment is not limited to such a control device 208 , and may be incorporated in various devices (vehicle devices) installed in mobile bodies such as automobile 206 .

FIG. 29 is a detailed configuration example of the oscillator 400 . The oscillator 400 in FIG. 29 is an oscillator with a double constant temperature bath structure (in a broad sense, a constant temperature bath structure).

The package 410 is composed of a substrate 411 and a case 412 . Various electronic components not shown are mounted on the substrate 411 . The second container 414 is provided inside the casing 412 , and the first container 413 is provided inside the second container 414 . Furthermore, a vibrator 420 is attached to the inner surface (lower surface) of the upper surface of the first container 413 . In addition, the circuit device 500 of this embodiment, the heater 450 and the temperature sensor 460 are mounted on the outer surface (upper surface) of the upper surface of the first container 413 . For example, the temperature inside the second container 414 can be adjusted by the heater 450 (heating element). Furthermore, for example, the temperature inside the second container 414 can be detected by the temperature sensor 460 .

The second container 414 is provided on the substrate 416 . The substrate 416 is a circuit substrate on which various electronic components can be mounted. A heater 452 and a temperature sensor 462 are mounted on the surface of the substrate 416 opposite to the surface on which the second container 414 is provided. The temperature of the space between the casing 412 and the second container 414 can be adjusted by, for example, the heater 452 (heating element). Furthermore, the temperature of the space between the casing 412 and the second container 414 can be detected by the temperature sensor 462 .

As heating elements of the heaters 450 and 452 , for example, heating power bipolar transistors, heating heater MOS transistors, heating resistors, Peltier elements, and the like can be used. Control of heat generation by these heaters 450 and 452 can be realized by, for example, a thermostat control circuit of the circuit device 500 . As the temperature sensors 460 and 462, for example, a thermistor, a diode, or the like can be used.

In FIG. 29 , since the temperature adjustment of the vibrator 420 and the like can be realized by the constant temperature bath of the double constant temperature bath structure, the stabilization of the oscillation frequency of the vibrator 420 and the like are achieved.

FIG. 30 shows a configuration example of a base station (base station device) as one of electronic devices. The physical layer circuit 600 performs physical layer processing in communication processing via a network. The network processor 602 performs processing of layers higher than the physical layer (link layer, etc.). The switch unit 604 performs various switching processes of communication processing. The DSP 606 performs various digital signal processing necessary for communication processing. The RF circuit 608 includes: a receiving circuit constituted by a low noise amplifier (LNA); a transmitting circuit constituted by a power amplifier; a D/A converter, an A/D converter, and the like.

The selector 612 outputs any one of the reference signal RFCK1 from the GPS 610 and the reference signal RFCK2 (clock signal from the network) from the physical layer circuit 600 to the circuit device 500 of this embodiment as the reference signal RFCK. The circuit device 500 performs a process of synchronizing the oscillation signal (an input signal based on the oscillation signal) with the reference signal RFCK. Furthermore, various clock signals CK1 , CK2 , CK3 , CK4 , and CK5 with different frequencies are generated and supplied to the physical layer circuit 600 , the network processor 602 , the switching unit 604 , the DSP 606 , and the RF circuit 608 .

According to the circuit device 500 of this embodiment, in the base station shown in FIG. 30 , the oscillation signal can be synchronized with the reference signal RFCK, and the clock signals CK1 to CK5 with high frequency stability generated based on the oscillation signal can be supplied to the base station of the base station. each circuit.

In addition, although this embodiment was described in detail as mentioned above, it should be easy for those skilled in the art to understand various modifications which do not actually deviate from the novel matter and effect of this invention. Therefore, all such modified examples are included in the scope of the present invention. For example, in the specification or drawings, at least once a term (temperature variation component, etc.) described together with a different term (environmental variation component, etc.) in a broader or synonymous sense can be replaced by the different term in any part of the specification or drawings. term. In addition, all combinations of the present embodiment and modifications are also included in the scope of the present invention. In addition, the circuit device, oscillator, electronic equipment, structure or operation of the mobile body, burn-in correction processing, Kalman filter processing, hold mode processing, temperature compensation processing, etc. are not limited to those described in this embodiment, and various out of shape.

Claims (14)

1. A circuit arrangement, wherein the circuit arrangement comprises: a processing unit, which performs signal processing on the input frequency control data, and outputs the frequency control data; and an oscillating signal generating circuit that generates an oscillating signal at an oscillating frequency set by the frequency control data using a vibrator, The processing unit performs an addition of a posterior estimated value x^(k-1) and a correction value D(k-1) at time step k-1 in update processing of the prior estimated value in the Kalman filtering process. The process of obtaining the prior estimation value x ^̂- (k) of time step k, and performing the aging correction of the frequency control data according to the result of the Kalman filtering process. 2. The circuit arrangement according to claim 1, wherein, The processing unit obtains the correction value D(k-1) from the observation residual in the Kalman filtering process. 3. The circuit arrangement according to claim 1 , wherein, The processing section adds the posterior estimated value x^(k-1) of the time step k-1 to the correction value D(k-1), by x ^{^-} (k)=x ^(k-1)+D(k-1), to obtain the prior estimate x ^{^-} (k) of the time step k. 4. The circuit arrangement according to claim 3, wherein, The processing unit calculates the correction value D(k-1) at the time step k based on the correction value D(k-1) at the time step k-1 and the observation residual ek in the Kalman filtering process. ). 5. The circuit arrangement according to claim 4, wherein, When the constant is E, the processing unit obtains the correction value D(k) by D(k)=D(k-1)+E·ek. 6. The circuit arrangement according to claim 5, wherein, The circuit device further includes a storage unit storing the constant E. 7. The circuit arrangement according to claim 1, wherein, The processing section performs the signal processing on the input frequency control data based on a phase comparison result of an input signal and a reference signal, wherein the input signal is based on the oscillation signal, During a period until the hold mode caused by the disappearance or abnormality of the reference signal is detected, processing is performed to estimate a true value of an observed value of the input frequency control data by the Kalman filter processing, When the hold mode is detected, the true value at the time corresponding to the time at which the hold mode is detected is stored, and arithmetic processing based on the true value is performed, thereby generating the burn-in-corrected Frequency control data. 8. The circuit arrangement according to claim 7, wherein, The processing unit generates the aging-corrected frequency control data by performing the arithmetic processing of adding the corrected value to the true value. 9. The circuit arrangement according to claim 8, wherein, The processing unit performs the arithmetic processing of adding the filtered correction value to the true value. 10. The circuit arrangement according to claim 1, wherein, The circuit device further includes a storage unit storing a system noise constant for setting the system noise of the Kalman filtering process and an observed noise constant for setting the observed noise of the Kalman filtering process. 11. The circuit arrangement of claim 1 , wherein, The circuit arrangement also includes a digital interface for monitoring said a priori estimates and observations. 12. An oscillator, wherein the oscillator comprises: A circuit arrangement as claimed in claim 1 ; and the vibrator. 13. An electronic device, wherein the electronic device comprises the circuit device according to claim 1. 14. A moving body comprising the circuit device according to claim 1.