JP6460458B2 - Digital temperature compensated oscillator - Google Patents

Description

  The present invention relates to a digital temperature compensated oscillator incorporating a temperature voltage sensor composed of a ring oscillator.

In recent years, as mobile environments such as high-performance portable terminals such as smartphones become small cells such as femtocells, picocells, and microcells, requirements for frequency stability of base station side clocks have become strict. . The femtocell requires a 50 ppb stability clock, and a temperature compensated oscillator equipped with a temperature sensor capable of detecting 0.05 ° C. is required to satisfy this frequency stability.
Patent Document 1 uses a ring oscillator that is driven by a current source that varies with temperature and a ring oscillator that is driven by a current source that is stable with respect to temperature, and uses a ring oscillator that is stable with respect to temperature as a count clock. This is a method of counting the pulse width of the ring oscillator clock that changes in temperature, and the temperature is first-order approximated by the count value. A digital temperature compensation oscillator incorporating a temperature sensor is described in Patent Document 2, for example.

  The applicant previously filed a “digital temperature compensated oscillator” (Japanese Patent Application No. 2014-026219) as a temperature compensated oscillator. This uses one ring oscillator as a temperature sensor (temperature-sensitive oscillator), and uses a clock of a voltage-controlled oscillator as a reference clock for frequency counting. The temperature compensation circuit of the temperature compensated oscillator is composed of a digital circuit, the circuit unit for converting the temperature compensated digital data to D / A and applying a voltage to the voltage controlled oscillator is composed of a ΔΣ modulator and a passive LPF (low pass filter). To do. Furthermore, digital temperature data input to the temperature compensation circuit is obtained from a temperature sensor (temperature-sensitive oscillator) composed of a ring oscillator and a subsequent temperature conversion circuit. With such a configuration, this temperature compensated oscillator can obtain a digital temperature compensated oscillator that has high resolution and that does not cause fluctuations in output frequency due to aging or temperature drift over a long period of time.

JP 2007-187659 A

JP-A 64-73823

In the ring oscillator used in Patent Document 1, since the current source also fluctuates when the power supply voltage fluctuates, the output frequency has voltage dependency, and therefore the accurate temperature cannot be measured when the power supply voltage fluctuates.
Further, the previous digital temperature compensated oscillator proposed by the applicant can improve the frequency fluctuation due to the aging and temperature drift of the D / A converter. In addition, by using a ring oscillator for the temperature sensor, it can be realized by a circuit in a semiconductor IC chip other than a crystal resonator. However, the ring oscillator of the temperature sensor used in this digital temperature compensated oscillator also has voltage dependency on the oscillation frequency, and it has been difficult to measure temperature with high accuracy.
The present invention has been made under such circumstances, and a digital temperature compensated oscillator has a temperature sensor capable of measuring a temperature with high accuracy, and a compensation voltage based on the measured temperature information (with respect to the oscillator). A temperature compensated oscillator capable of realizing high frequency stability by a control voltage signal is provided. In addition, by using this temperature sensor, it is possible to measure the power supply voltage that fluctuates sequentially at the same time as the temperature measurement, so that more stable oscillation can be achieved by generating the compensation voltage in consideration of the information on the measured power supply voltage. A temperature compensated oscillator that maintains frequency is provided.

The present invention provides a high-accuracy temperature voltage sensor for a high-accuracy temperature-compensated oscillator (TCXO) for femtocells of mobile phone base stations (clock accuracy of about 50 ppb) and the like without using external temperature sensor components externally. A digital temperature-compensated oscillator that can be realized by a circuit on an IC (chip) and is highly accurate and miniaturized is obtained. The crystal oscillator (crystal oscillator) is originally externally attached to the semiconductor IC.
The digital temperature-compensated oscillator of the present invention is characterized in that temperature compensation data is created based on temperature digital information obtained by using the following temperature sensor, and temperature compensation of the oscillation frequency is performed. The temperature sensor uses two ring oscillators with different temperature characteristics that are sensitive to the temperature formed on the semiconductor IC. The oscillation frequency (f1, f2) is stable in the same temperature (T) and voltage (V) environment. Measured at a good reference frequency f0 (crystal oscillation clock, etc.), and calculated in advance from the data obtained by measuring the oscillation frequency (f1, f2) of each ring oscillator by changing the temperature T and the voltage V. Original high-order polynomial approximation: Using the coefficients of f1 (T, V) and f2 (T, V), T and V are calculated from the two simultaneous equations of two ring oscillators with T and V as unknowns. It is characterized by solving with the above arithmetic circuit.

According to one aspect of the digital temperature compensated oscillator of the present invention, an oscillation signal that generates a first oscillation frequency that changes in frequency with respect to a change in environmental temperature and that can control the first oscillation frequency with a control voltage is provided. A voltage-controlled oscillator to be generated; a first ring oscillator that generates a second oscillation frequency having a rate of change greater than the first oscillation frequency with respect to a change in the environmental temperature; and a change in the environmental temperature. A second ring oscillator that generates a third oscillation frequency having a rate of change larger than the first oscillation frequency and having a temperature characteristic different from that of the second oscillation frequency, and based on the first oscillation frequency. By measuring the second and third oscillation frequencies at the set count time, a quadratic or higher-order polynomial approximation expression of temperature and voltage is set for the obtained frequency count data. In addition, the first coefficient set and the first coefficient set in the polynomial approximation formula of the first ring oscillator and the second ring oscillator calculated based on the frequency count data measured by changing the temperature and the power supply voltage in advance A coefficient storage unit that stores two coefficient sets, and the second and third oscillation frequencies of the first ring oscillator and the second ring oscillator that operate in an environment of the same temperature and power supply voltage. By measuring with time, a frequency counter that generates first frequency count data and second frequency count data corresponding to the oscillation frequency at that time, the first frequency count data, and the first coefficient set The first equation applied to the polynomial approximation formula, the second frequency count data, and the second coefficient set are the polynomial. From simultaneous binary equations with a second equation applied to the approximate expression, a temperature voltage converter to obtain the temperature and power supply voltage when generating the first and second frequency count data by calculation, the temperature-voltage converter A temperature compensation circuit for generating temperature compensation digital data for compensating for the change of the first oscillation frequency with respect to the environmental temperature based on the temperature obtained by the above, and a passive multistage low-pass filter comprising a ΔΣ modulator and a passive element. And a D / A converter for converting the temperature compensation digital data into the control voltage represented by an analog voltage.

The temperature compensation circuit outputs digital data related to temperature compensation and digital data related to voltage compensation, respectively. The digital data related to temperature compensation is input to the D / A converter, and digital data related to voltage compensation is output. May be input to a power supply voltage control circuit that controls a stabilized power supply that supplies the power supply voltage of the voltage controlled oscillator.
The temperature compensation circuit may output compensation data based on both the temperature and the power supply voltage obtained by the temperature voltage conversion unit as the temperature compensation digital data, and input the compensation data to the D / A converter. .

  The digital temperature compensated oscillator of the present invention has a temperature sensor capable of measuring temperature with high accuracy, and realizes high frequency stability by a compensation voltage (control voltage signal for the oscillator) based on the temperature data. Can do. In addition, since the power supply voltage that sequentially changes can be measured simultaneously with the temperature measurement, a more stable oscillation frequency can be maintained by generating the compensation voltage in consideration of the voltage data. Also, the output voltage of the stabilized power supply can be compensated based on the voltage data, and a stable oscillation frequency can be maintained by this.

1 is a block diagram illustrating a temperature sensor according to Embodiment 1. FIG. FIG. 2 is a block diagram of a digital temperature compensation oscillator incorporating the temperature sensor shown in FIG. 1. The characteristic view explaining the frequency characteristic of the temperature of the ring oscillator used for the temperature sensor of FIG. 1, and a voltage. FIG. 4 is a waveform diagram for explaining a relationship between an output of the ring oscillator shown in FIG. 3 and a reference clock. The block diagram explaining the ring oscillator shown in FIG. FIG. 2 is a circuit diagram illustrating a circuit configuration of the ring oscillator shown in FIG. 1. The circuit diagram explaining the ring oscillator used in an Example. The figure which shows the polynomial approximation formula (Formula 1) showing the oscillation frequency of the ring oscillator shown in FIG. The figure which shows the determinants (Formula 12) and (Formula 13) which show the variation | change_quantity (DELTA) T and (DELTA) V of the oscillation frequency of the ring oscillator shown in FIG. The flowchart which calculates temperature (T) and voltage (V) from frequency count value (F1, F2). FIG. 2 is a circuit block diagram of a digital temperature compensation oscillator incorporating the temperature sensor shown in FIG. 1. FIG. 6 is a circuit block diagram of a digital temperature compensated oscillator according to a second embodiment.

Hereinafter, embodiments of the invention will be described with reference to examples.

First, Embodiment 1 will be described with reference to FIGS. 1 to 6 and FIGS. 8 to 11. The digital temperature compensated oscillator in this embodiment uses the temperature sensor shown in FIG. 1, and its schematic configuration is as shown in the block diagram of FIG. Further, a more detailed configuration corresponding to the block diagram of FIG. 2 is shown in the block diagram of FIG. As shown in FIG. 2, this temperature-compensated oscillator is composed of a semiconductor IC (chip) with an externally attached crystal oscillator (crystal oscillator) and a built-in oscillation circuit.
The temperature sensor used in this embodiment uses the oscillation frequencies (f1, f2) of two ring oscillators 2a, 2b that are sensitive to temperature and have different temperature characteristics in the same temperature (T) and voltage (V) environment. Measurement was made at a reference frequency f0 (reference clock signal) with good stability, and the oscillation frequencies (f1, f2) of the ring oscillators 2a, 2b were calculated in advance from data measured by changing the temperature T and voltage V. Binary elements of two ring oscillators 2a and 2b with T and V as unknowns using the coefficients of T and V binary high-order polynomial approximations (f1 (T, V) and f2 (T, V)) It is characterized in that T and V are solved from simultaneous equations by an arithmetic circuit formed in a semiconductor IC (chip).

  As described in “Means for Solving the Problems” (see FIG. 1), the temperature sensor has different temperature characteristics and voltage characteristics of two oscillation frequencies that are sensitive to temperature and have different temperature characteristics. The count time (Tosc) for measuring the oscillation frequencies f1 and f2 of the first ring oscillator 2a and the second ring oscillator 2b and the first ring oscillator 2a and the second ring oscillator 2b is set. A reference frequency supply unit (reference clock source) 1 for supplying a reference clock signal f0, and a temperature T and a power supply voltage with respect to count data of frequencies obtained by measuring the oscillation frequencies f1 and f2 at the count time Tosc. The frequency measured by changing the temperature T and the power supply voltage V in advance after setting a binary approximation formula of binary second order or higher of V A coefficient storage unit (ROM) 34 for storing the first coefficient set and the second coefficient set in the polynomial approximation formula of the first ring oscillator 2a and the second ring oscillator 2b calculated based on the count data; By measuring the oscillation frequencies f1, f2 of the first ring oscillator 2a and the second ring oscillator 2b operating in the environment of the same temperature and power supply voltage with the count time Tosc, oscillation at that time A frequency counter (Counter) 32 that generates first frequency count data F1 and second frequency count data F2 corresponding to the frequency, the first frequency count data F1, and the first coefficient set are represented by the polynomial approximation formula And the second frequency count data F2 and the second coefficient set applied to the polynomial. From simultaneous binary equations with a second equation applied to the approximate expression, and a temperature voltage converter 33 for determining the temperature and power supply voltage when generating the first and second frequency count data by computation.

A reference frequency supply unit (reference clock source) 1 sets a reference clock for setting a count time (Tosc) for measuring the oscillation frequencies f1 and f2 of the first ring oscillator 2a and the second ring oscillator 2b. In this embodiment (FIG. 11), since the temperature sensor is included in the temperature compensated oscillator, the reference frequency supply unit 1 includes the voltage controlled oscillator 1 constituting the temperature compensated oscillator. And The voltage controlled oscillator 1 generates a first oscillation frequency whose frequency changes with a change in environmental temperature, and generates an oscillation signal capable of controlling the first oscillation frequency with a control voltage.

  Here, the oscillation frequency generated by the voltage controlled oscillator 1 is the first oscillation frequency, the oscillation frequency f1 of the first ring oscillator 2a is the second oscillation frequency, and the oscillation frequency f2 of the second ring oscillator 2b is Is the third oscillation frequency.

Further, as described in “Means for Solving the Problems” (see FIG. 11), the digital temperature compensated oscillator has a first frequency that changes in response to a change in environmental temperature in addition to the temperature sensor. And the same frequency when the first and second frequency count data are generated, and the voltage controlled oscillator 1 that generates an oscillation signal that can control the first oscillation frequency by a control voltage. The temperature compensation circuit 4 generates temperature compensation digital data for compensating for the change of the first oscillation frequency with respect to the environmental temperature based on the temperature, and a passive multistage low-pass filter 6 including a ΔΣ modulator 5 and passive elements. And a D / A converter 7 for converting the temperature compensation digital data into the control voltage represented by an analog voltage.

3, FIG. 3 (a) shows the temperature characteristics and voltage characteristics of the oscillation frequencies f1, f2 of the first and second ring oscillators 2a, 2b in this embodiment, and FIG. 3 is another example of temperature and voltage characteristics. FIG. 3C shows another example of the temperature characteristics of the oscillation frequencies f1 and f2. The oscillation frequency f1 has a positive characteristic and the other f2 has a negative characteristic.
FIG. 4 shows the outputs f1 and f2 of the first and second ring oscillators and the reference clock f0. Each frequency is counted by the count time Tosc set by the reference clock f0, and the frequency count data F1. , F2 is obtained. To measure the frequency, count the rising edge of the waveform.

5 and 6 show the ring oscillator 2 constituting the temperature voltage sensor. The ring oscillator includes two ring oscillators 2a and 2b to which a power source 2c is supplied from the outside. These are denoted as RO1 and RO2, respectively. Each of the two ring oscillators includes a plurality (N) of inverters (FIG. 6) connected in a ring shape. The first ring oscillator (RO1) 2a is connected to a current source for current driving and outputs an oscillation frequency f1 (FIG. 6 (a)). The oscillation frequency f1 is a function of the temperature T and the voltage V and is expressed as f1 (T, V). The second ring oscillator (RO2) 2b is connected to the power source 2c for voltage driving and outputs an oscillation frequency f2 (FIG. 6 (b)). The oscillation frequency f2 is a function of the temperature T and the voltage V, and is expressed by f2 (T, V). The oscillation frequency (f) of the ring oscillator is determined by the number of inverters (N) and the delay time (td), and the delay time (td) depends on the temperature (T). That is, the oscillation frequency f is represented by 1 / (2N · td (T)).
FIG. 5 illustrates a current-driven ring oscillator and a voltage-driven ring oscillator using a common power supply 2c. The power supply line of the ring oscillator is connected to a current source (current source) formed separately from the circuit of the ring oscillator to be current driven (RO1). An example of an actual circuit is shown in FIGS. 7 (c) and 7 (d).

Hereinafter, a temperature detection method using the temperature sensor by the two ring oscillators of this embodiment will be described.
The oscillation frequency of the ring oscillator formed in the semiconductor IC has voltage dependence in addition to temperature dependence. Clock accuracy required by the femtocell (about 50 ppb) The temperature accuracy required for the temperature sensor required for the temperature compensation circuit of the temperature compensation oscillator is calculated from the characteristics of the temperature sensor used in the temperature compensation oscillator. 0.05 ° C is required. Although it is conceivable to stabilize the power supply voltage of this ring oscillator with a regulator, the power supply voltage within a wide range of −50 to 125 ° C. is within a voltage fluctuation corresponding to a temperature accuracy of 0.05 ° C. (much below 1 mV). It is not easy to stabilize. The invention described by this embodiment solves such a problem.

The oscillation frequency f of one ring oscillator can be expressed by a polynomial approximation formula f (T, V) (Equation 1) of temperature (T), voltage (V) alone, and T · V terms. Equation 1 is set forth in FIG. In this approximate expression, m, n, p, q, M, N, P, and Q are all natural numbers, and a _m , b _n , c _pq , and c ₀ are coefficients. Here, the first ring oscillator is expressed as RO1, and the second ring oscillator is expressed as RO2. In this embodiment, it is assumed that the approximate expression is composed of a linear combination of a fourth order related to T, a second order related to V, and a first order related to T · V.

The values obtained by measuring the oscillation frequencies f1 and f2 of the two ring oscillators with the counter 32 shown in FIG. 1 are F1 (T, V) and F2 (T, V).
The frequency count value F1 of RO1 is expressed by Equation 2.
F1 (T, V) = a ₄₁ T ⁴ + a ₃₁ T ³ + a ₂₁ T ² + a ₁₁ T + b ₂₁ V ² + b ₁₁ V + c ₁₁ TV + c ₀₁ (Formula 2)
a41, a31, a21, a11, b21, b11, c11, c01 are coefficients of each term.
Further, the frequency count value F2 of RO2 is expressed by Expression 3.
F2 (T, V) = a 42 T 4 + a 32 T 3 + a 22 T 2 + a 12 T + b 22 V 2 + b 12 V + c 12 TV + c 02 ( Equation 3)
a42, a32, a22, a12, b22, b12, c12, c02 are coefficients of each term.

  Next, the frequency count values F1 and F2 data of the RO1 and RO2 are respectively collected while changing the power supply voltage while changing the temperature of the RO1 and RO2 in the same temperature environment such as a thermostat. The number of measurement points is set to a point number equal to or greater than the minimum number at which a coefficient that is an unknown number can be calculated by the least square method. Then, the data obtained from the measurement of RO1 is [F1 (Ti, Vj), Ti, Vj] (i = 1 to m, j = 1 to n), and the data obtained from the measurement of RO2 is [F2 ( Ti, Vj), Ti, Vj] (i = 1 to m, j = 1 to n). However, i, j, m, and n are all natural numbers. M and n used here are irrelevant to m and n in Equation 1 shown in FIG.

From the data relating to RO1 obtained by these measurements, unknowns [a41, a31, a21, a11, b21, b11, c11, c01] relating to RO1 are calculated by the method of least squares, and similarly RO2 obtained by the measurements. The unknowns [a42, a32, a22, a12, b22, b12, c12, c02] related to RO2 are calculated from the data related to (2) by the least square method. The calculation by the least square method is performed at the time of shipping test of the semiconductor IC or at the time of product incorporation by the user. Data can be collected from the semiconductor IC and can be calculated by an arithmetic device such as an external personal computer.
The coefficient obtained by such processing is written in the coefficient memory of the semiconductor IC in which the temperature voltage sensor is formed. The coefficient memory corresponds to the memory (ROM) 34 in FIG. As the coefficient memory, a nonvolatile memory such as an EPROM, an EEPROM, or a flash memory is appropriate.

Next, RO1 and RO2 are placed in the same temperature environment with the respective coefficients written in the memory in the semiconductor IC, and the same power supply voltage is applied. The frequency count values of RO1 and RO2 are F1 (T, V) and F2 (T, V), and unknowns T and V are calculated from these two values. That is, since the coefficients relating to F1 and F2 of RO1 and RO2 are known, the unknowns T and V are calculated by solving the binary (T, V) simultaneous quaternary equations of the following equations (Equations 4 and 5). Is done.
F1 = a ₄₁ T ⁴ + a ₃₁ T ³ + a ₂₁ T ² + a ₁₁ T + b ₂₁ V ² + b ₁₁ V + c ₁₁ TV + c ₀₁ (Formula 4)
F2 = a ₄₂ T ⁴ + a ₃₂ T ³ + a ₂₂ T ² + a ₁₂ T + b ₂₂ V ² + b ₁₂ V + c ₁₂ TV + c ₀₂ (Formula 5)
However, since the solution of the binary (T, V) quaternary simultaneous equation is not formulated like the quadratic equation, the solution is obtained by iterative calculation using a recurrence formula using the Newton-Raphson method. .

First, a value obtained by subtracting the left side from the right side of Equation 4 is Y1 (T, V), and similarly, a value obtained by subtracting the left side from the right side of Equation 5 is Y2 (T, V). Y1 and Y2 are expressed as Expression 6 and Expression 7.
Y1 (T, V) = a 41 T 4 + a 31 T 3 + a 21 T 2 + a 11 T + b 21 V 2 + b 11 V + c 11 TV + c 01 -F1 ( Equation 6)
Y2 (T, V) = a 42 T 4 + a 32 T 3 + a 22 T 2 + a 12 T + b 22 V 2 + b 12 V + c 12 TV + c 02 -F2 ( Equation 7)

Further, in Expression 6 and Expression 7, the differential values are expressed in Expression 8 to Expression 11.
∂Y1 / ∂T = 4a ₄₁ T ³ + 3a ₃₁ T ² + 2a ₂₁ T + a ₁₁ + c ₁₁ V (Formula 8)
∂Y2 / ∂T = 4a ₄₂ T ³ + 3a ₃₂ T ² + 2a ₂₂ T + a ₁₂ + c ₁₂ V (Formula 9)
∂Y1 / ∂V = 2b ₂₁ V + b ₁₁ + c ₁₁ T (Formula 10)
∂Y2 / ∂V = 2b ₂₂ V + b ₁₂ + c ₁₂ T (Formula 11)
From these equations, the changes ΔT and ΔV of T and V are expressed by Equations 12 and 13 shown in FIG.

Next, from Expression 12, ΔT and ΔV are expressed by the following expressions.
ΔT = − (Y1 * ∂Y2 / ∂V−Y2 * ∂Y1 / ∂V) / det
(Formula 14)
ΔV = (Y1 * ∂Y2 / ∂T−Y2 * ∂Y1 / ∂T) / det
(Formula 15)
The recurrence formula by the Newton-Raphson method using the calculation formulas of ΔT and ΔV shown here is expressed by the following formula.
Tn + 1 = Tn + ΔT (Formula 16)
Vn + 1 = Vn + ΔV (Formula 17)
At the beginning of the calculation, it is necessary to give T0 and V0 corresponding to Tn of Expression 16 and Expression 17 and n = 0 of Vn as initial values. As T0 and V0, it is also possible to give the approximate center value of the assumed fluctuation range of T and V.

Since the Newton-Raphson method repeats the calculation based on the recurrence formula, if a result within a desired error is obtained, the calculation is stopped and Tn and Vn at that stage are output as the calculation results.
Specifically, as shown in the following equations 18 and 19, it is determined that the convergence has occurred when ΔT and ΔV are less than preset error ranges Et and Ev.
ABS (ΔT) <Et (Formula 18)
ABS (ΔV) <Ev (Equation 19)
ABS (X) represents the absolute value of X.
Further, as shown in the following equation 20, when the number of calculations (Nctv) exceeds the limit number (Mtv), the calculation is stopped as time is over, and the final result at that time is output. Or, it is possible to output a default value without any problem.
Nctv> Mtv (Formula 20)

Further, division is necessary for the calculation of (1 / det) described in Expressions 14 and 15, but the calculation itself can also use the Newton-Raphson method.
The above calculation by the Newton-Raphson method can be realized by a digital arithmetic circuit and an arithmetic sequence (program) mounted on the semiconductor IC. This program is stored in a memory formed in the semiconductor IC.

Next, a method for calculating the temperature (T) and the voltage (V) from the frequency count values (F1, F2) will be described with reference to FIG.
The coefficients for RO1 (a41, a31, a21, a11, b21, b11, c11, c01, 4a41, 3a31, 2a21, 2b21), the coefficients for RO2 (a42, a32, a22, a12, b22, b12, c12, c02, 4a42, 3a32, 2a22, 2b22), det reciprocal coefficient (2), temperature (T) error limit value (Et), voltage (V) error limit value (Ev), det reciprocal error limit value (Edet), T and V calculation count limit values (Mtv), and det reciprocal calculation count limit values (Mdet) are written in the memory (ROM) (coefficient storage section) 34 shown in FIG.
The calculation method is divided into two flows. The left flow (FIG. 10A) is the main flow, the right flow (FIG. 10B) is the sub flow, and Idet (= 1 / det) is calculated. To do.

First, the main flow will be described.
[Start Calculation] When the calculation is started, the frequency count data F1 and F2 of RO1 and RO2 are acquired. The count data is normally acquired from a register because data acquired from the counter (32 in FIG. 1) is stored in a register or the like.
[Initial Value Setting] Next, T0 and V0 are given as initial values of temperature data T and voltage data V in the case of calculation by the Newton-Raphson method. The initial value can be stored in advance in a memory (34 in FIG. 1) and read out from this memory.
[Counter Reset] Next, the counter of the number of repetitions of the Newton-Raphson method is reset (that is, the calculation number (Nctv) is set to 0). Then, Y1 (T, V) and Y2 (T, V) are calculated by substituting the already given initial values T0, V0 into T, V in Expressions 6 and 7.

Next, dY1T, dY2T, dY1V, and dY2V (abbreviated symbols) of first-order derivatives with respect to T and V of Y1 (T, V) and Y2 (T, V) are similarly calculated by Expressions 8 to 11. These abbreviations are expressed by Formula 21 to Formula 24 as shown below.
dY1T = ∂Y1 / ∂T (Formula 21)
dY2T = ∂Y2 / ∂T (Formula 22)
dY1V = ∂Y1 / ∂V (Formula 23)
dY2V = ∂Y2 / ∂V (Formula 24)
Next, det is calculated. Calculation is performed by substituting Equations 8 to 11 into Equation 13 representing det.

Next, ΔT and ΔV derived from Expression 12 are expressed in Expression 14 and Expression 15, and values obtained by adding det to these are expressed in Expression 25 and Expression 26. Then, the right side of the two equations corresponding to ΔT * det and ΔV * det is calculated.
ΔT * det = − (Y1 * ∂Y2 / ∂V−Y2 * ∂Y1 / ∂V)
(Formula 25)
ΔV * det = (Y1 * ∂Y2 / ∂T−Y2 * ∂Y1 / ∂T)
(Formula 26)
The calculation results of Y1, Y2, dY1T, dY2T, dY1V, dY2V, det, ΔT * det, and ΔV * det calculated above are stored in the temporary storage register after calculation.
Next, calculation of Idet (= 1 / det) will be described, but since the details are a sub-flow (FIG. 10B), the description will be given here assuming that Idet has been calculated.

Next, ΔT * det and ΔV * det that have already been calculated in Expressions 25 and 26 are multiplied by Idet to calculate ΔT and ΔV.
ΔT = Idet * (ΔT * det) (Formula 27)
ΔV = Idet * (ΔV * det) (Equation 28)
Next, T = T1 and V = V1 are calculated by Expression 16 and Expression 17. Specifically, it is expressed by the following formulas 29 and 30 where n = 0.
T1 = T0 + ΔT (Formula 29)
V1 = V0 + ΔV (Formula 30)
[Convergence Determination] Next, in order to determine whether or not the calculated result corresponds to the convergence referred to in the Newton-Raphson method, a magnitude comparison between ABS (ΔT) <Et and ABS (ΔV) <Vt is calculated. If the result is Yes, T1 and V1 calculated above are output as calculation results, and the calculation ends.

On the other hand, if the result is No (unconvergence), the calculation count Nctv is incremented by one to repeat the calculation again. When Nctv is initially reset, Nctv = 1 is counted up.
Next, it is calculated whether the calculation count Nctv exceeds the calculation count limit value Mtv. If it exceeds (Yes), the T and V calculation results are output, and the calculation ends. If not (No), the calculation is repeated from the calculation of Y1 and Y2.

Next, the subflow will be described. Here, the reciprocal calculation flow of Idet will be described.
[Calculation Start] First, a method for calculating the inverse of det using the Newton-Raphson method will be described.
For example, when the reciprocal of the numerical value a is x, x = 1 / a can be obtained by solving f (x) = 0 as shown in Equation 31. The differential value of f (x) is shown in Equation 32.
f (x) = (1 / x) −a (Equation 31)
f ′ (x) = − 1 / x ² (Formula 32)
Using these equations, the recurrence equation shown in Equation 33 is obtained. The recurrence formula is expressed as Expression 33.
xn + 1 = xn-f (xn) / f '(xn) = xn (2-axn)
(Formula 33)
First, det data stored in a register is acquired during the main flow.

[Initial Value Setting] The initial value Idet0 is read from the memory (ROM) shown in FIG. 1 and set as the Idet initial value.
[Counter Reset] Next, the calculation number counter Ncdet is reset (Ncdet = 0). Then, Idet is calculated according to Equation 33. At that time, the det reciprocal calculation coefficient (2) of ROM is used.
[Convergence Determination] Next, in order to determine whether or not the calculated result corresponds to the convergence referred to in the Newton-Raphson method, it is calculated whether or not ABS (ΔIdet) <Edet. ΔIdet is “xn + 1−xn” in Expression 33. If ABS (ΔIdet) <Edet is Yes (convergence), the above calculated Idet is output as the calculation result, and the calculation ends. On the other hand, if the result is No (unconverged), the calculation count Ncdet is incremented by one in order to repeat the calculation again. If Ncdet is initially reset, Ncdet = 1 is counted up.
Next, it is calculated whether the calculation number count Ncdet exceeds the calculation number limit value Mtv. If it exceeds (Yes), the Idet calculation result is output and the calculation is completed. When it does not exceed (No), the calculation is repeated from the calculation of Idet.

Next, a method for temperature compensation and voltage compensation of an oscillator will be described using a digital temperature compensated oscillator (see paragraph 0017) in which the temperature sensor described above is incorporated in the same semiconductor IC. (See Figure 11)
In this embodiment, the temperature compensation of the voltage controlled oscillator (VCO) 1 and the voltage compensation of the power supply voltage of the PWM output circuit 57 are performed by a D / A converter 7 constituted by a ΔΣ modulator 5 and a passive multistage low-pass filter 6. How to do it. The PWM output at the final stage of the digital output of the D / A converter 7 depends directly on the power supply voltage. Therefore, it is necessary to fix the power supply voltage, but it is difficult to completely stabilize it. Therefore, in this embodiment, the voltage compensation amount is calculated in addition to the temperature compensation amount using the voltage data of the temperature voltage sensor output, and both the compensation amounts are added and input to the D / A converter 7. is there.

Next, in FIG. 11, 1 is a voltage controlled oscillator (VCO), 2a and 2b are ring oscillators 1 and 2 having different temperature characteristics, 3 is a temperature voltage digital data generation unit, 4 is a temperature compensation circuit, 5 is A ΔΣ modulator, 6 is a passive LPF (low-pass filter), 7 is a D / A conversion circuit, 8 is a buffer, 9 is a frequency dividing circuit, and 10 is an output buffer.
The voltage-controlled oscillator 1 includes a crystal resonator (crystal oscillator) 11 and a voltage-controlled oscillation circuit 12 that are externally attached to a semiconductor IC, and an oscillation frequency (first frequency) that changes with a change in environmental temperature. And an oscillation signal capable of controlling the first oscillation frequency by the control voltage output from the passive four-stage LPF 6. The generated oscillation signal having the first oscillation frequency is output to the buffer 8 and input to the control circuit 31 of the temperature / voltage digital data generation unit 3 through the frequency divider 9, while the buffer 8 outputs the output buffer 10. Is output to the outside.

The temperature compensation circuit 4 includes a temperature voltage compensation calculation circuit 41 and a memory 42.
The ΔΣ modulator 5 includes a multiplier 51, adders 52a, 52b, and 52c, delay circuits 53a, 53b, and 53c, multipliers 54a and 54b, a quantizer 55, a PWM modulator 56, and a PWM output circuit 57. It consists of. A dither signal is input to the adder 52b. The dither signal is composed of lower bit data of the output of the counter that has received the output of the ring oscillator. For example, if the output of the counter that has received the output of the ring oscillator is 18-bit data, it is preferable that the lower 4 bits of data be the dither signal.
The quantizer 55 is a multi-value quantizer with three or more values. For example, a quantized PDM signal quantized with four levels of “00”, “01”, “10”, and “11” is used. The signal is output to the PWM modulator 56 and output to the adder 52c. The PWM modulator 56 outputs a binary PWM signal with a multi-level pulse width of 3 levels or more. For example, if the quantizer 55 is 4 values (4 levels), “0” is set in the same manner. Of the four pulse widths “1”, “2”, and “3”, the pulse width is converted into a binary PWM signal having a pulse width of a level corresponding to the level of the input PDM signal. Buffered and output to the passive 4-stage LPF 6. On the other hand, the adder 52c receives a signal input to the adder 52b as well as the output of the quantizer 55, and the adder 52c outputs a quantization error by the quantizer 55.

The delay circuits 53a, 53b, and 53c delay the quantization error by one cycle, two cycles, or three cycles. The output of the delay circuit 53a is multiplied by a predetermined coefficient by the multiplier 54a and then input to the adder 52a. The output of the delay circuit 53b is multiplied by a predetermined coefficient by the multiplier 54b and input to the adder 52a. On the other hand, the output of the delay circuit 53c is directly input to the adder 52a. In addition to these inputs, the output of the multiplier 51 is input to the adder 52a, and the inputs are added and output to the adders 52b and 52c.
The passive four-stage LPF 6 is configured by four stages of LPFs composed of a resistance element and a capacitance element. For example, the total resistance value is 1 GΩ, the first-stage resistance element is 700 MΩ, and the other three resistance elements are each set to 100 MΩ. The capacitance value of each capacitive element is set, for example, such that the sum of the capacitance values is 100 pF, the capacitive element of the fourth stage as the final stage is 70 pF, and the other three capacitive elements are each set to 10 pF. Yes. In this way, if the first-stage resistance value and the final-stage capacitance value are set larger than other resistance values or capacitance values, the amount of attenuation in the low frequency region can be increased. The PWM signal is converted into an analog signal by passing through the passive four-stage LPF 6. This analog signal is input to the voltage controlled oscillation circuit 12 of the voltage controlled oscillator 1 as a control voltage signal.

The voltage VA from the stabilized power supply 100 that stabilizes the external power supply is applied to the crystal oscillator (XO), the PWM output circuit 57, and the ring oscillator 1 (2a) and the ring oscillator 2 (2b) that are temperature sensors. Is done. A circuit block or the like not described with voltage application is a digital circuit block whose voltage value is not necessarily VA.
Next, the temperature voltage digital data generation unit 3 will be described.
Outputs having frequency information of the ring oscillator 1 (2a) and the ring oscillator 2 (2b) become frequency count data F1 and F2 in the corresponding counters 32a and 32b, respectively. The count times of these counters 32a and 32b are generated in the control circuit 31 using the output clock of the frequency dividing circuit 9 as a reference clock.

In the control circuit 31, synchronization and control signals of three asynchronous signals having different outputs from the frequency dividing circuit 9 and outputs from the ring oscillators 1 and 2 are generated. The frequency count data F1 and F2 corresponding to each of the ring oscillators 1 and 2 outputted through the control circuit 31 are converted into temperature voltage data (T, V) by the temperature voltage conversion circuit 33 by calculation. The memory 34 is a circuit that stores coefficients and the like necessary for the calculation.
The temperature / voltage compensation calculation circuit 41 constituting the temperature compensation circuit 4 described above receives the temperature data T and voltage data V from the temperature / voltage conversion circuit 33, and the temperature of the voltage controlled oscillator 1 stored in the memory 42 in advance. The digital data TVC for temperature voltage compensation is calculated using a coefficient corresponding to the compensation characteristic of the characteristic, the power supply voltage characteristic of the PWM output circuit 57, the power supply voltage characteristic of the voltage controlled oscillator 1, or both. And output to the D / A converter 7.

  As described above, in this embodiment, the digital temperature-compensated oscillator has a temperature sensor that can measure temperature with high accuracy, and is a compensation voltage (control voltage signal for the oscillator) based on the temperature data. Stability is realized. In addition, the power supply voltage that sequentially varies at the same time as the temperature measurement is also measured. Since the compensation voltage is obtained by taking the voltage data into account, a more stable oscillation frequency can be maintained.

  In the first embodiment, the example in which the compensation voltage (control voltage signal for the oscillator) is generated based on both the temperature data T and the voltage data V obtained by the temperature sensor measurement has been described. It is not something that can be done. That is, the temperature sensor according to the present invention can perform temperature measurement with higher accuracy than before, so that the compensation voltage may be generated using only the temperature data T. In that case, the temperature compensation circuit 4 calculates the temperature compensation digital data using a coefficient corresponding to the compensation of the temperature characteristic of the voltage controlled oscillator 1 stored in advance in the memory 42, and outputs it to the D / A converter. .

Next, Example 2 will be described with reference to FIG.
This embodiment is characterized in that voltage correction of the stabilized power supply 100 that supplies the power supply voltage VA of the PWM output circuit or the voltage controlled oscillator 1 is input to the power supply voltage control circuit 110.
In this embodiment, the output of the temperature voltage compensation calculation circuit 41 constituting the temperature compensation circuit 4 that generates temperature compensation digital data is divided into data TC related to temperature compensation and data VC related to voltage compensation. The data TC is input to the D / A converter 7 as in the first embodiment (see FIG. 11), and temperature compensation of the voltage controlled oscillator 1 is performed. On the other hand, the data VC is output from the temperature voltage compensation arithmetic circuit 41 and input to the power supply voltage control circuit 110, and the output performs voltage compensation of the stabilized power supply 100. The power supply voltage control circuit 110 converts the data VC, which is digital data, into an analog voltage value and outputs the analog voltage value to the stabilized power supply 100.

The stabilized power supply output VA thus compensated for voltage is applied to the PWM output circuit 57 of the ΔΣ modulator 5, the voltage controlled oscillator 1, the ring oscillator 1 (2a) that performs temperature voltage sensing, and the ring oscillator 2 (2b). The
As described above, in this embodiment, the digital temperature compensated oscillator maintains a stable oscillation frequency by incorporating a temperature voltage sensor that can measure temperature with high accuracy and can also measure power supply voltage fluctuation at the same time. can do.

Next, Embodiment 3 will be described with reference to FIG.
In this embodiment, a ring oscillator used in this temperature / voltage sensor will be described.
The ring oscillator is a circuit that is usually used as a monitor for evaluating the performance of a semiconductor process, and is generally used in a CMOS process that is widely spread. Further, in a CMOS circuit, it is used for temperature measurement and voltage drop measurement in a circuit block. However, the temperature measurement by the conventional ring oscillator is at most a first-order (linear) approximation, and there is nothing aiming at the high accuracy as intended by the present invention.
In addition, papers that use ring oscillators to measure the power supply voltage drop due to current consumption are known, but those that sense both temperature and voltage using a ring oscillator are not known.

    FIG. 7A shows a ring oscillator composed of CMOS circuits (C1, C2,... Cn) composed of N-channel MOSFETs and P-channel MOSFETs. The ring oscillator is formed by connecting an odd number of inverting amplifiers in a loop, and three stages are the minimum number of stages. VDD (plus power supply) is connected to the PMOSFET, and VSS (ground potential) is connected to the NMOSFET.

FIG. 7B shows a ring oscillator in which inverting amplifiers composed of NPN transistors (B1, B2, B3) and resistors (R1, R2, R3) as bipolar transistors are connected in a loop in three stages. Each inverting amplifier has a resistor connected to VDD (plus power supply) and VSS (ground potential) connected to an emitter.
When comparing aging of elements, a threshold (threshold) voltage, which is generally an important characteristic of a MOS transistor, tends to drift. Therefore, a bipolar ring oscillator is better than a CMOS ring oscillator when the oscillation frequency of the ring oscillator changes over time and the temperature sensor characteristics change over time.

As described in the first embodiment, in order to greatly change the temperature characteristics of the two ring oscillators (RO1, RO2), one is driven by voltage and the other is driven by current. Specifically, the ring oscillator shown in FIGS. 7A and 7B is directly connected to VDD (plus power supply) to drive the voltage, and the other is shown in FIGS. 7C and 7D. Like the ring oscillator, the power line of the ring oscillator is connected to a current source (current source) formed separately from the circuit of the ring oscillator for current driving (see FIG. 5). The current source includes a bias circuit and a current mirror circuit described at the left end of the circuits illustrated in FIGS. The bias circuit and the current mirror circuit include a PMOS transistor P1, a resistor R4 and a bipolar transistor B4 provided between power supplies (VDD, VSS), a PMOS transistor P2 and a bipolar transistor B5 provided between power supplies (VDD, VSS), The PMOS transistor P2 is composed of a PMOS transistor P3 having a gate connected in common.

DESCRIPTION OF SYMBOLS 1 ... Reference frequency supply part (reference clock source), voltage control type oscillator 2 ... Ring oscillator 2a ... 1st ring oscillator (RO1)
2b. Second ring oscillator (RO2)
2c ... Power supply 3 ... Temperature voltage digital data generation circuit 4 ... Temperature compensation circuit 5 ... ΔΣ modulator 6 ... Passive low-pass filter (LPF)
7 ... D / A conversion circuit 8 ... buffer 9 ... frequency divider circuit 10 ... output buffer 11 ... crystal oscillator (crystal oscillator)
DESCRIPTION OF SYMBOLS 12 ... Voltage controlled oscillation circuit 31 ... Control circuit 32a, 32b ... Frequency counter 33 ... Temperature voltage conversion part (temperature voltage conversion circuit)
34 ... Coefficient storage unit (memory)
41 ... Temperature voltage compensation arithmetic circuit 42 ... Memory 100 ... Stabilized power supply 110 ... Power supply voltage control circuit

Claims (3)

A voltage-controlled oscillator that generates a first oscillation frequency that changes in frequency with respect to a change in environmental temperature and generates an oscillation signal that can be controlled by the control voltage; and a change in the environmental temperature A first ring oscillator that generates a second oscillation frequency having a rate of change larger than the first oscillation frequency, and a rate of change greater than the first oscillation frequency with respect to a change in the environmental temperature. And a second ring oscillator for generating a third oscillation frequency having a temperature characteristic different from that of the second oscillation frequency, and the second and third at a count time set based on the first oscillation frequency. The frequency and frequency data obtained by measuring the oscillation frequency is set with a temperature and voltage binary quadratic approximation formula, and the temperature and power supply voltage are changed in advance. A coefficient storage unit that stores the first coefficient set and the second coefficient set in the polynomial approximation formula of the first ring oscillator and the second ring oscillator calculated based on the frequency count data Corresponding to the oscillation frequency at that time by measuring the second and third oscillation frequencies of the first ring oscillator and the second ring oscillator operating in the environment of temperature and power supply voltage with the count time A frequency counter that generates first frequency count data and second frequency count data, a first equation that applies the first frequency count data and the first coefficient set to the polynomial approximation, Simultaneous binary by a second equation applying the second frequency count data and the second coefficient set to the polynomial approximation From equation, the temperature voltage converter to the temperature and supply voltage when generating the first and second frequency count data obtained by calculation, the first oscillation on the basis of the obtained temperature by the temperature-voltage converter A temperature compensation circuit for generating temperature compensation digital data for compensating a change in frequency with respect to environmental temperature, and a passive multistage low-pass filter including a ΔΣ modulator and a passive element, and the temperature compensation digital data is represented by an analog voltage. A digital temperature-compensated oscillator comprising a D / A converter for converting to the control voltage. The temperature compensation circuit outputs the temperature compensation digital data and digital data related to voltage compensation, the temperature compensation digital data is input to the D / A converter, and the digital data related to voltage compensation is the 2. The digital temperature compensated oscillator according to claim 1, wherein the digital temperature compensated oscillator is input to a power supply voltage control circuit that controls a stabilized power supply that supplies a power supply voltage of the voltage controlled oscillator. The temperature compensation circuit outputs compensation data based on both the temperature and the power supply voltage obtained by the temperature voltage converter as the temperature compensation digital data, and inputs the compensation data to the D / A converter. The digital temperature-compensated oscillator according to claim 1.

Images