JP2734469B2 - Circuit Constant Optimization Method for Automatic Circuit Constant Design System - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for optimizing circuit constants in an automatic circuit constant design system for automatically determining constants of circuit components of an arbitrary circuit in a state where the topology of the circuit is fixed. It is about.

[0002]

2. Description of the Related Art In general, analog electronic circuits have a wide variety of evaluation items for defining their characteristic specifications, such as AC (alternating current), DC (direct current), and transient characteristics. Therefore, it is necessary to consider these characteristics when determining circuit constant values such as a resistance value and a capacitance value. However, in the current manual-centered analog circuit design, considerable skill is required to optimally determine each circuit constant value in consideration of the total characteristics of AC, DC, and transient. It takes a lot of time to determine the circuit constant value. There are problems such as.

[0003] In order to solve these problems,
Conventionally, a system that automatically designs optimal circuit constant values for required specifications (hereinafter abbreviated as circuit constant optimization system)
As a conventional system, for example, "Electronic Institute of Japan / Electronic Circuit Research Society ECT-88-24" Optimal Design of CMOS Operational Amplifier Circuit Applying Simplex Method "" and "Nikkei Electronics 1990.10.15
No. P195 to P207 “Analog CAD System for Automatically Changing Circuit Constants According to Target Characteristics”. Hereinafter, this system will be described with reference to the drawings.

FIG. 6 is a block diagram showing a conventional circuit constant optimizing system based on the system shown in the former document. This conventional system 11 performs analysis of circuit characteristics such as AC, DC, and transient characteristics. Circuit characteristic analysis program 1 to be executed
2, a design program 13 for determining an initial value of a circuit constant in response to a required specification from a designer, and evaluating whether the characteristics analyzed by the circuit characteristic analysis program 12 satisfy the required specification. An optimization search program 14 is provided for searching for which circuit constant value will be closer to the optimum value if the circuit constant value is changed to larger or smaller using the characteristic analysis program 12.

Next, the operation will be described with reference to the flowchart of FIG. First, in step S5, the designer inputs design specification values such as target circuit data, variable designation data, measurement data, and characteristic specification data to the design program 13 shown in FIG. Next, in step S6, an initial value is set for each circuit constant to be optimized. Then
In step S7, characteristic analysis is performed with the given circuit constant value using the circuit characteristic analysis program 12 such as a circuit simulator. Next, in step S8, an objective function O for optimization is performed based on the result of the characteristic analysis.
Calculate BJ. I in the specification input step S5
The lower limit specification value of the second circuit characteristic is Sib, and the upper limit specification value is S
where it and Sio are the target values and Yi is the circuit characteristic value calculated in the characteristic analysis step S7.

(1) When there is a lower limit

[0007]

(Equation 1)

(2) When there is an upper limit

[0009]

(Equation 2)

(3) When there are upper and lower limits

[0011]

(Equation 3)

Using Yio standardized by
The objective function OBJ is, for example,

[0013]

(Equation 4)

However, (1) when there is one of an upper limit and a lower limit and Yio <−1

[0015]

(Equation 5)

(2) There is one of an upper limit and a lower limit, and Yi
When o> -1

[0017]

(Equation 6)

(3) When there are both upper and lower limits

[0019]

(Equation 7)

Is represented by As can be seen from these equations, the objective function is configured to have a smaller value as the circuit characteristics are closer to the target value entered by the designer.

Next, in step S9, the optimization search program 14 determines which of the current circuit constant values is to be larger or smaller.
An optimization search is performed to determine whether the value changes to a smaller one or approaches an optimum value. That is, a search is made for a direction in which the objective function becomes smaller. For this search, various methods such as a steepest descent method using a derivative of an objective function and a revised simplex method using no derivative are proposed. In any case, the optimal value search requires the calculation of the objective function value.

Therefore, in this step S9, a plurality of characteristic analyzes by a circuit simulator or the like are necessary.
After all, in step S9, after each circuit constant value is changed by a small amount to perform characteristic analysis, the objective function value is calculated, and a set of circuit constants that minimizes the objective function value among the changed ones is calculated. Is to look for. The value of the objective function at this time is OBJ ^N.

Next, in step S10, the objective function OBJ calculated in step S8 and the objective function value OBJ ^N obtained as a result of the optimization search are used.

[0024]

(Equation 8)

Here, TOL is a constant value representing an allowable range of convergence.

If the expression is true, the process proceeds to step S12.
If false, the process proceeds to step S11.

In step S12, since the last set constant value is the optimum value, the value is output. On the other hand, in step S11, the constant value is updated to the circuit constant value found in the optimization search step S9.

As described above, the loop of steps S11 → S7 → S8 → S9 → S10 is repeated until the objective function converges, and the objective function value decreases. Therefore, each time this loop is repeated, the circuit characteristics approach each specification value input by the designer, and when the convergence condition (Equation (8)) is satisfied, the circuit constant value has been optimized.

FIG. 8 is a block diagram showing a conventional circuit constant optimization system based on the system shown in the latter document. This conventional system 31 calculates circuit characteristics such as AC, DC, and transient characteristics. A circuit characteristic calculation program 32 such as a circuit simulator to be executed, an initial value of a circuit constant is determined based on a target specification from a designer, and it is evaluated whether the characteristic calculated by the circuit characteristic calculation program 32 satisfies the target specification. And a constant value changing program 34 for sequentially changing all constant values which are constants to be changed for optimization.

Next, the operation will be described with reference to the flowchart of FIG. First, in step N5, the designer inputs design specification values such as target circuit data, optimization target constants, and characteristic specification data to the design program 33 shown in FIG. Next, in step N6, an initial value is set for each circuit constant to be optimized. This is defined as a reference circuit constant value (a ₁ , a ₂ ,..., A _n ). Next, in step N7, the reference circuit characteristic _A0 is calculated by the circuit characteristic calculation program 32 such as a circuit simulator at the given circuit constant value. Next, in step N8, it is determined whether the circuit constant currently being optimized is an optimum value. An optimum value is determined when the following condition is satisfied.

<Conditions> Even if all the circuit constants to be optimized are slightly changed, either large or small, the circuit characteristics do not approach the target specification more than the current reference circuit characteristics. Therefore, in the first determination step N8, NO
And the process proceeds to step N9. At step N9, the constants (a ₁ , a ₂ ,..., A _n ) which are the circuit constants to be optimized are taken out one by one by the constant value changing program 34 and minutely changed into large or small. . That is, sensitivity analysis is performed. For example, let the circuit constant a ₁ be a ₁ + △
a ₁ (however, △ a ₁ is a positive minute amount). Thereafter, the flow returns to step N7 again, and the circuit constant a ₁
, That is, the circuit characteristic A in the state of a ₁ + △ a ₁
To calculate the ₁ ^+. Next, it is determined again whether the value is the optimum value in step N8. In this case, the previously calculated reference circuit characteristic A
Compare ₀ with A ₁ ⁺ above

(1) When A ₀ is closer to the target specification than A ₁ ^{+ It} is determined that changing a ₁ to a ₁ + △ a ₁ does not contribute to the optimization process, and the process proceeds to step N9. The following constant value change is performed. For example, a ₁ is changed to a ₁ − △ a ₁ , and the circuit characteristic A ₁ ⁺ is calculated in step N7 in the same manner as described above, and the process returns to step N8. Things to find close to the target specification from the reference circuit characteristic A ₀ In this way, continuing to vary constant value.

(2) When A ₁ ⁺ is closer to the target specification than A _{0 It} is determined that changing a ₁ to a ₁ + △ a ₁ contributes to the optimization process, and the above A ₁ ⁺ is used as a reference. As circuit characteristics, the circuit constant value at this time is replaced with the previous reference circuit constant value, that is, a
_A value obtained by changing ₁ to a ₁ + △ a ₁ is set as a new reference circuit constant value. Thereafter, the process again proceeds to step N9, and the circuit constants to be optimized are extracted one by one, and a small change is made to either large or small, and the above processing is performed in steps N9 → N7 → N
Repeat along the loop of 8.

[0034] For all the circuit parameters are subject to optimization in step N8 as described above, a large, if it is determined that not approach target specifications than the reference circuit characteristic A ₀ be changed to a small time shift Only in step N10, since the reference circuit constant value at that time is the optimum value, these are output.

[0035]

Since the conventional circuit constant optimizing system shown in FIG. 6 is configured as described above, in the process of optimizing the circuit constants, the system is optimized by a characteristic analyzing means such as a circuit characteristic analyzing program. Since each circuit characteristic is calculated with high accuracy, there is a problem that the calculation is performed with an accuracy which is not originally required in an initial stage in which the circuit characteristic fluctuates greatly, and a useless calculation time is required.

The optimization loop step N11 → N
Since the constant value is changed among 7 → N8 → N9 → N10, it is necessary to perform an optimization search (step N9) each time the constant value is changed. As described above, in this step N9, since various constant values are changed and the characteristic analysis is performed each time, a considerable amount of calculation time is required even in step N9 alone. Therefore, there is a problem that a large amount of calculation time is required for each change of the constant value.

On the other hand, since the conventional circuit constant optimizing system shown in FIG. 8 is configured as described above, in the process of optimizing the circuit constants, the circuit constants to be optimized are set to the target specifications and There is a problem in that a circuit constant having no relation, that is, not contributing to the optimization is minutely changed every time, either large or small, and the circuit characteristics at that time are calculated, so that a wasteful calculation time is required.

The present invention has been made in order to solve the above-mentioned problems. In order to find the optimum value of the circuit constant, the operation time can be shortened, and the obtained optimum constant value has a high precision as in the prior art. It is an object of the present invention to provide a method for optimizing circuit constants in an automatic circuit constant design system capable of obtaining the above.

[0039]

According to a first aspect of the present invention, there is provided a method for optimizing circuit constants, comprising: a circuit characteristic analyzing means for analyzing circuit characteristics with low accuracy and high speed when analyzing circuit characteristics; Analysis program 13) and circuit characteristic analysis means for analyzing circuit characteristics with higher accuracy than the general circuit characteristic analysis means (circuit characteristic high accuracy analysis program 2)
And, for the objective function defined using each value to represent the degree of coincidence between the design specification value of each circuit characteristic in optimization and each characteristic value obtained from each of the above circuit characteristic analysis means, Optimization search means (optimization search program 4) for determining which one of the circuit constants due to the circuit characteristics should be increased or decreased in order to reduce this function; By setting necessary values for the search means and changing the values set in the circuit characteristic analysis means and the optimization search means based on the characteristic values obtained by the analysis and the results of the optimization search means, An automatic circuit constant design system including a control means (design program 3) for searching for a constant value is provided. With road characteristic analyzing means, in the final stage is obtained as to use a said circuit characterization means switches the analyzer circuit characteristics depending on the stage of the search.

In the method for optimizing circuit constants according to the present invention, when the automatic design system brings the circuit constants closer to the optimum circuit constant values, the optimization searching means uses
If it is clarified that the objective function can be reduced by changing any of the circuit constants, either large or small, the change in the value of the arbitrary circuit constant can be performed without using the optimization search means as long as the objective function value keeps decreasing. Is to be repeated.

According to a third aspect of the present invention, there is provided a circuit constant optimizing method, comprising: a circuit characteristic calculating means (circuit characteristic calculating program 24) for calculating circuit characteristics such as DC characteristics, AC characteristics, and transient characteristics of a circuit; Effective contribution learning means (effective contribution learning program 21) for storing which circuit constant of the circuit constant based on the circuit characteristics calculated by the characteristic calculation means has effectively contributed to the optimization processing; Constant value changing means (constant value changing program 2) for changing a circuit constant value based on information obtained by the learning means;
2) and setting necessary values for the circuit characteristic calculating means and the constant value changing means, and controlling the circuit constants to approach the optimum values by changing the set values based on the circuit characteristic calculation results. A control means (design program 23) for changing the circuit constant value is determined by the effective contribution learning means.

In the circuit constant optimizing method according to the fourth aspect of the present invention, in order to search for a change in a circuit constant value effective for optimization, the value of all circuit constants to be optimized is slightly changed. Without doing this, in the process of optimizing the circuit constants, the target circuit constants to be minutely changed according to the effective contribution obtained for each circuit constant by the effective contribution learning means are limited or minutely changed. The processing order of the circuit constants is changed.

[0043]

According to the first aspect of the present invention, the circuit characteristic analysis means reduces the accuracy of the circuit characteristic analysis to a certain extent at the initial stage of optimization when the circuit characteristic analysis accuracy is not so required, thereby shortening the operation time. Analyze, and after reaching the optimal solution with this accuracy, switch the analysis means,
This time, the circuit characteristics are analyzed with high accuracy by the circuit characteristics analysis means, and the optimum solution with high accuracy is obtained again.

According to the second aspect of the present invention, once the change of the circuit constant is selected in the optimization search means for searching for the minimum value of the objective function, as long as the objective function keeps decreasing due to the change of the constant value, The constant change is repeated unconditionally, and when the reduction of the objective function has stopped, which circuit constant is to be changed to a larger or smaller one is again searched by the optimization search.

According to the third aspect of the present invention, the effective contribution learning means stores which of the circuit constants based on the circuit characteristics calculated by the circuit characteristic calculation means has effectively contributed to the optimization processing, and stores the constant value. The changing means changes the circuit constant value based on the information obtained by the effective contribution learning means. The control means sets necessary values for the circuit characteristic calculating means and the constant value changing means and controls the circuit constants to approach the optimum values by changing the set values based on the circuit characteristic calculation results. In particular, the target constant for changing the circuit constant value is determined by the effective contribution learning means.

According to the fourth aspect of the present invention, the circuit constants that contribute little to the optimization by the effective contribution learning means are excluded from the elements subject to the minute change, and the number of loops for performing the circuit characteristic calculation is reduced.

[0047]

【Example】

Embodiment 1 FIG. FIG. 1 is a block diagram showing an automatic circuit constant design system according to an embodiment of the present invention. In FIG. 1, the system 10 includes a circuit characteristic high accuracy analysis program 2 as a circuit characteristic analysis means for analyzing circuit characteristics with high accuracy,
A circuit characteristic analysis program 13 as an outline circuit characteristic analysis means for analyzing at high speed, and an optimization search means as an optimization search means for examining which of the current circuit constant values should be changed to larger or smaller to approach an optimum value. Search program 4
And a design program 3 as control means. The design program 3 receives the required specifications from the designer, determines the initial values of the circuit constants, and analyzes the circuit characteristics analyzed by the circuit characteristic outline analysis program 13 or the circuit characteristic high accuracy analysis program 2 and the optimization search program 4. After trying each of the constants, it is determined based on the circuit characteristics when the constant selected to reduce the objective function is changed, whether each current circuit constant value is an optimum value.

Next, the operation will be described with reference to the flowchart of FIG. First, in step S5, the designer instructs the design program 3 of DC,
Input target specification values such as AC and transient. Next, in step S6, an initial value of each circuit constant is set. Next, in step S14A, the rough characteristic analysis step S1 has already been performed.
It is determined whether or not the objective function calculated based on each characteristic of the circuit obtained by 5A and S15B has converged. The first time of this determination step S14A is always NO,
Proceed to step S15A. In step S15A, each characteristic of the circuit considered in the optimization is calculated at high speed by the circuit characteristic outline analysis program 13.

Therefore, the characteristic calculation here does not need to be performed with high accuracy. In an analog circuit, when a semiconductor element such as a transistor or a diode is used as the element, the element characteristic expression is usually expressed by a nonlinear expression, so that the calculation of the circuit characteristic becomes non-linear and convergence calculation is required. In addition, the convergence calculation is usually given its allowable value.
Therefore, by increasing the allowable value used for the convergence calculation, the number of calculations can be reduced, and the characteristic calculation can be performed at high speed. However, the greater the tolerance, the lower the accuracy of each characteristic. As described above, in step S15A, a general characteristic analysis is performed in which the calculation accuracy is low but the calculation is performed at high speed, for example, by increasing the allowable value of the convergence calculation.

After step S15A, step S8
Proceeding to A, the objective function value for optimizing the circuit constant is calculated based on each circuit characteristic value obtained in the above-described general characteristic analysis step S15A. As the objective function, for example, Expressions (1) to (7) of the above-described conventional example can be considered.

Thereafter, only the first time, step S9
And search for the optimal value. In this step S9, it is checked which of the current circuit constant values is changed to a larger value or a smaller value to approach an optimum value. That is, in this step S9, after analyzing the characteristics by changing each circuit constant value by a very small amount, the objective function value is calculated, and a set of circuit constants that minimizes the objective function value among the changed ones is calculated. Is to look for. At this time, when performing the characteristic analysis, the characteristic analysis method used for calculating the objective function value in step S8A, that is, the same one of the general characteristic analysis and the high-precision characteristic analysis is used.

Next, in step S18, when the respective characteristics for calculating the objective function value in steps S8A and S9 are obtained by the approximate characteristic analysis, the objective function value OBJ obtained in step S8A and the objective function value OBJ obtained in step S9 are obtained. Using the objective function value OBJ ^N

[0053]

(Equation 9)

Here, TOL is a constant value indicating an allowable range of convergence.

If the expression has become true at least once, the process proceeds to step S19. If the expression has never become true, the process proceeds to step S11A.

In step S11A, the circuit constant value is changed in accordance with the change in the constant value selected in the optimum value search step S9, that is, the change in the constant value that minimizes the objective function value. Thereafter, the flow returns to step S14A again, and whether the convergence has been achieved even once in the rough characteristic analysis, that is, step S18.
It is determined whether or not Yes has been determined. Here, if the convergence has been made even once by the rough characteristic analysis, the process proceeds to the high-precision characteristic analysis by the circuit characteristic high-precision analysis program 2 in step S7A. The high-precision characteristic analysis refers to, for example, calculating the circuit characteristics required for optimization with high accuracy by reducing the allowable value of the convergence calculation described in step S15A.

Next, the process proceeds to step S8A in which an objective function is calculated based on each circuit characteristic obtained in step S7A.

Next, when the process passes through step S8A for the second time or later, the process always proceeds to step S16. In step S16, when the process of step S11A is performed last, which constant value has been changed to large or small is stored, and the same constant value is changed to the same direction, that is, large or small. That is, the same constant change as the previous time is repeated. The change in the constant value in this step is only temporary, and when the calculation of the objective function is completed in step S8B by this temporary change, it is necessary to return to the previous constant value.

Next, in step S14B, the same determination as in step S14A is used to determine whether or not convergence has been performed at least once, that is, whether the process has proceeded to Yes in step S18. Therefore, the determination in step S14A and the determination in step S14B are always the same. That is, in the characteristic analysis step, once the general characteristic analysis step S15A is selected, the next step S15B is always selected.

In the next step S15B, step S7B and step S8B, respectively,
A, the same processing as in steps S7A and S8A is performed. Therefore, in step S8B, the objective function value OBJ ^C when the same constant value change as in step S11A or step S11B is repeated is calculated. When step S8B is completed, the constant value temporarily changed in step 16 needs to be returned to the previous value.

Next, in step S17, the objective function value OBJ calculated in step S8A and the objective function value OBJ ^C calculated in step S8B are used.

[0062]

(Equation 10)

If the expression (10) is true, step S11B
If not, the process proceeds to step S9. Step S11
When the process proceeds to B, it is determined that the temporary change of the constant value in step S16 approaches the optimum value, and the constant value is changed. Thereafter, the process proceeds to step S16, where the same constant is again changed to the same direction, that is, larger or smaller, and the value is temporarily changed again.

Therefore, the same circuit constant value is set in the same direction, that is,
If the change to large or small is repeated, step S16 → S
14B → S15B (or S7B) → S8B → S17 → S
The loop of 11B → S16 is repeated, and the circuit constant approaches the optimum value only by going around this loop. Since there is no optimum value search step in this loop, the number of calculations of the circuit characteristics is considerably reduced, and the calculation time is shortened.

As described above, the optimization of the circuit constant value is advanced. When the equation (9) becomes true at least once in step S18, the process always passes step S18 and proceeds to step S19. . In step S19, whether or not the objective function calculated based on each circuit characteristic calculated using the high-precision characteristic analysis steps S7A and S7B has converged using the above equation (9) in the same manner as in step S18. Make a judgment. If not converged, the process proceeds to step S11A to change the constant value and continue the optimization.
If the convergence is reached, the process proceeds to step S12, and the last constant value of each circuit constant is the respective optimum value.
Output the value.

Note that steps S5 and S in this embodiment
6, S7A and S7B, S8A and S8B, S9, S11A
, S11B, and S12 correspond to steps S5, S6, S7, S8, S9, S11, and S12 of the conventional example, respectively, and indicate the same contents.

Embodiment 2 FIG. In the first embodiment, the circuit characteristics that are the basis of the optimization calculation are calculated by the rough characteristic analysis in the initial stage of the optimization, and once the circuit characteristics converge by the rough characteristic analysis, the calculation of each circuit characteristic is made high. Although the accuracy characteristic analysis has been performed, the analysis of each circuit characteristic is roughly analyzed as shown in the flowchart of FIG. 3 (steps S15A and S15).
B), medium-precision analysis (steps S21A and S21B), and high-precision analysis (steps S7A and S7B) can be optimized at high speed even when divided into three stages. In these three characteristic analysis methods, the operation accuracy is low and the operation time is short in the order of general, medium accuracy, and high accuracy characteristic analysis. In steps S20A and S20B, the process branches to the above three types of characteristic analysis in accordance with the convergence states in steps S18, S22, and S19, respectively. Which one to select is in accordance with the first embodiment, and the first time starts with a rough characteristic analysis, and then shifts to medium precision and high precision analysis.

Embodiment 3 FIG. In the first embodiment, the difference between the rough characteristic analysis and the high-precision characteristic analysis is expressed by the allowable value at the time of the convergence calculation. However, the present invention is not limited to this. If it is included in the characteristics to be considered in the conversion, the general characteristics can be obtained by reducing the number of sample points for calculating the function value, and the calculation time can be shortened. Conversely, if the number of sample points is increased, high-accuracy characteristics are obtained, and the same effect is obtained.

Embodiment 4 FIG. FIG. 4 is a block diagram showing an automatic circuit constant design system according to an embodiment of the present invention. In FIG. 4, the system 20
A circuit characteristic calculation program (circuit characteristic calculation means) 24 for calculating circuit characteristics such as C, DC, and transient characteristics; an initial value of a circuit constant is determined based on a target specification from a designer; A design program (control means) 23 for evaluating whether the circuit characteristics calculated in 24 satisfies the target specification and controlling the processing, and whether the constant value change effectively contributed for optimization in the course of optimization Effective learning program (effective contribution learning means) 21 that learns for each circuit constant, and the contribution information of each circuit constant that the effective contribution learning program (effective contribution learning means) 21 has. A constant value changing program (constant value changing means) 2 which determines circuit constants for minutely changing constant values and takes into account the degree of contribution of sequentially changing the constant values for them.
2 is provided.

Next, the operation will be described with reference to the flowchart of FIG. First, in step N5, the designer sends the design program 23
Input target specification values such as C, AC, and transient. Next, at step N6, initial values of the respective circuit constants are set. This is defined as a reference circuit constant value (a ₁ , a ₂ ,..., A _n ). Next, at step N7, the reference circuit characteristic _A0 is calculated by the above-described circuit characteristic calculation program 24 such as a circuit simulator at the given circuit constant value. Next, in step N15, it is determined whether the circuit constant currently being optimized is an optimum value. An optimal value is determined when the following condition is satisfied. <Conditions> The circuit characteristics do not approach the target specification more than the current reference circuit characteristics even if all the circuit constants to be optimized are slightly changed, either large or small.

Therefore, the first determination step N15 always results in NO, and the process proceeds to step N16. Step N16 is a process that operates when the change of each constant effectively contributes to the optimization. No operation is performed in the first time, and the process proceeds to the next step N17. In step N17, all the circuit constants (a ₁ , a _{1) to be} optimized are in the initial stage of the optimization, that is, while it is unknown which change in circuit constants contributes to the optimization. ₂ ,..., A _n ) are taken out one by one and minutely changed into large or small. For example _{a 1 a 1 + △ a 1} ( where, △ a ₁ is a positive very small value) and was varied. After this, return to step N7 again,
The circuit characteristic A ₁ ⁺ in a state where the circuit constant a ₁ is changed, that is, in the state of a ₁ + △ a ₁ is calculated. Next, it is determined again whether the value is the optimum value at step N15. At this time, the previously calculated reference circuit characteristics A ₀ and A ₁ ⁺ are compared.

(1) When A ₀ is closer to the target specification than A ₁ ^{+ It} is determined that changing a ₁ to a ₁ + △ a ₁ does not contribute to the optimization process, and the process proceeds to steps N16 and N17. And the next constant change is performed. In step N16, no operation is performed because the constant change does not contribute to the optimization. At step N17, the following constants values change for example, the a ₁ a ₁ -
Δa ₁ , and the circuit characteristic A ₁ ⁺ is calculated in step N7 in the same manner as described above, and the process returns to step N15. In this way, closer to target specifications from the reference circuit characteristic A ₀ is continuing to vary constant value to find.

(2) A₁ ⁺Is A₀Closer to target specification
Case a₁A₁+ △ a₁Change contributes to the optimization process
Judging that, A₁ ⁺Is the reference circuit characteristic.
Replace the circuit constant value with the reference circuit constant value,
a₁A₁+ △ a₁With the new reference circuit constant
Value. Thereafter, the process proceeds to step N16. Step N
At 16, the effective contribution learning program 21 executes a₁
Has effectively contributed to the optimization process,
A among constants₁Is the degree of contribution, for example, one contribution
Remember. Thereafter, the process proceeds to step N17, and the optimization target
One for each of the circuit constants, large and small
A slight change is made to the process, and the above processing is performed in step N17.
It repeats along the loop of → N7 → N15 → N16.

In this way, steps N17 → N7 → N
As the loop from 15 to N16 is repeated, the effective contribution learning program 21 clarifies the effective contribution to the optimization of each circuit constant, that is, the number of contributions. Therefore, after the circuit constants that repeatedly contribute to the above-mentioned loop become almost clear, in the conventional step N17, all the circuit constants to be optimized are taken out one by one and minutely changed to either a large or small one. The above processing is performed by excluding circuit constants having a small effective contribution to each circuit constant from the circuit constants to be optimized. That is, in step N17, constant values are changed for all circuit constants in the initial stage of optimization by the constant value changing program 22 in consideration of the degree of contribution. From this point, the constant value is changed only with respect to the circuit constants that have contributed effectively up to that time, that is, the circuit constants with a high effective contribution.

However, only the circuit constants that have contributed effectively reach the optimum value while repeating the minute change of the constants.
Determining an optimum value at step N15, i.e. before for their circuit constants large, if it is determined that not approach target specifications than the reference circuit characteristic A ₀ be changed In any small, excludes the circuit constant is not again effectively contribute It is checked in a loop of steps N16 → N17 → N7 → N15 whether or not all of the first circuit constants to be optimized have optimum values. At this time, if the reference circuit characteristic A ₀
If there is something more similar to the target specification A, the above loop is repeated to change the circuit constant value. In this way, even if all the circuit constants to be optimized in step N15 are changed, either large or small, the reference circuit characteristic A ₀
Only when it is determined that the target specification does not approach the target specification again, the process proceeds to step N10, and since the reference circuit constant value at that time is the optimum value, these are output.

Embodiment 5 FIG. In the fourth embodiment, the effective contribution to the optimization of each circuit constant by the effective contribution learning program 21 shown in FIG. 4 is counted as the number of contributions. However, how close the circuit characteristics have to the target specification is quantified. Thus, every time an effective contribution is generated, the sum of them for each circuit constant can be used as the effective contribution. For example,
The target specification is A, the current reference circuit characteristic is A ₀ , and the circuit constant a ₁
The when the circuit characteristics when the a ₁ + △ a ₁ and the minute amount greater change A ₁ ⁺ and, towards the A ₁ ⁺ is close to the target specifications A than A _0. At this time, changing a ₁ to a ₁ + △ a ₁ is an effective contribution. Also, the effective contribution E ₁ of a ₁ at this time is

[0077]

[Equation 11]

The effective contribution of the change of each circuit constant can be quantified. By calculating this value every time an effective contribution occurs for each circuit constant and adding them, the contribution of each circuit constant to the optimization can be calculated as the effect of the constant change. A more detailed contribution is obtained than the contribution count of the fourth embodiment.

Embodiment 6 FIG. In the fourth embodiment, after learning the validity of optimization of each circuit constant by the effective contribution learning program 21 of FIG. 4, circuit constants having a small effective contribution are excluded from the circuit constants to be optimized thereafter and optimized. However, this effective contribution is used as the order of taking out circuit constants for changing the constant value in step N17 of the flowchart shown in FIG. In other words, the optimization process is performed by preferentially taking out circuit constants having a high effective contribution and making small changes in large or small.
The loop of steps N16 → N17 → N7 → N15 in FIG. 5 becomes more effective. In this way, the effective contribution learning program 21 in FIG. 4 controls the order in which the circuit constants are prioritized and the circuit constants are changed based on the effective contribution of each circuit constant.

[0080]

As described above, according to the first and second aspects of the present invention, when calculating the characteristics, there are an initial stage that does not require much high precision and a final stage of optimization that requires high accuracy. By exchanging the characteristic analysis method, there is an effect that the calculation time in the initial stage can be shortened while maintaining high accuracy of the final result. Also, since the same constant value as the previous time was repeatedly changed, and if it reduced the objective function and contributed to the optimization, the constant change was adopted, so the same constant value had to be repeatedly changed many times. In some cases, there is an effect that the operation time can be greatly reduced. In order to improve the accuracy of optimization, it is necessary to reduce the width of one change of the constant value. However, the smaller the value, the longer the calculation time usually required to reach the optimal solution. By repeating the same constant value change as before, the width of one change is small,
In other words, there is an effect that the operation time is not so long even if the accuracy is improved.

As described above, according to the third and fourth aspects of the present invention, for each circuit constant to be optimized, the effective contribution of the optimization to the target specification at that time is learned. The optimization process is configured so that circuit constants that are not effective for the optimization are excluded from the target of change or priorities are set as targets for change from more effective circuit constants. is there.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an automatic circuit constant design system according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a circuit constant optimizing method by the system of FIG. 1;

FIG. 3 is a flowchart showing another circuit constant optimizing method by the system of FIG. 1;

FIG. 4 is a block diagram showing an automatic circuit constant design system according to an embodiment of the present invention.

FIG. 5 is a flowchart showing a circuit constant optimizing method by the system of FIG. 4;

FIG. 6 is a block diagram showing a conventional automatic system for designing circuit constants.

FIG. 7 is a flowchart showing a circuit constant optimizing method by the system of FIG. 6;

FIG. 8 is a block diagram showing a conventional automatic circuit constant design system.

FIG. 9 is a flowchart showing a circuit constant optimizing method by the system of FIG. 8;

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Automatic circuit-constant design system 2 Circuit characteristic high-precision analysis program (circuit characteristic analysis means) 3 Design program (control means) 4 Optimization search program (optimization search means) 13 Circuit characteristic schematic analysis program (schematic circuit characteristic analysis means) 20 Automatic circuit constant design system 21 Effective contribution learning program (Effective contribution learning means) 22 Constant value change program considering constant (Constant value change means) 23 Design program (Control means) 24 Circuit characteristic calculation program ( Circuit characteristic calculation means)

Claims (4)

(57) [Claims] When analyzing circuit characteristics, the circuit characteristics analysis means for analyzing the circuit characteristics with low accuracy and high speed, and the circuit characteristic analysis means for analyzing the circuit characteristics with higher accuracy than the general circuit characteristic analysis means. For an objective function defined using each value to represent the degree of coincidence between the design specification value of each circuit characteristic in the optimization and each characteristic value obtained from each of the circuit characteristic analysis means, Optimization searching means for checking which one of the circuit constants according to the circuit characteristics should be increased or decreased in order to reduce the function, and values required for the circuit characteristic analyzing means and the optimization searching means are set. In addition, based on the characteristic values obtained by the analysis and the results of the optimization search means, the optimum circuit is changed by changing the values set in the circuit characteristic analysis means and the optimization search means. An automatic circuit constant design system comprising control means for searching for a circuit constant value, wherein the circuit characteristic analysis means is used in an initial stage of a process of searching for an optimum circuit constant value, and the circuit characteristic analysis is performed in a final stage. Circuit optimization method in an automatic circuit constant design system, characterized in that means is used and the circuit characteristic analysis means is switched in accordance with a search stage. 2. A circuit characteristic analyzing means for analyzing circuit characteristics with low accuracy and high speed when analyzing circuit characteristics, and a circuit characteristic analyzing means for analyzing circuit characteristics with higher accuracy than the general circuit characteristic analyzing means. For an objective function defined using each value to represent the degree of coincidence between the design specification value of each circuit characteristic in the optimization and each characteristic value obtained from each of the circuit characteristic analysis means, Optimization searching means for checking which one of the circuit constants according to the circuit characteristics should be increased or decreased in order to reduce the function, and values required for the circuit characteristic analyzing means and the optimization searching means are set. In addition, based on the characteristic values obtained by the analysis and the results of the optimization search means, the optimum circuit is changed by changing the values set in the circuit characteristic analysis means and the optimization search means. An automatic circuit constant design system comprising control means for searching for a circuit constant value, and when the circuit constant approaches the optimum circuit constant value, the optimization search means sets any circuit constant to large or small. If it is clarified that changing the objective function can be reduced by changing to, the value of the above-mentioned arbitrary circuit constant is repeatedly changed without using the optimization search means as long as the objective function value keeps decreasing. Method for optimizing circuit constants in an automatic design system for circuit constants. 3. A circuit characteristic calculating means for calculating circuit characteristics such as a DC characteristic, an AC characteristic, and a transient characteristic of a circuit, and a circuit constant based on the circuit characteristics calculated by the circuit characteristic calculating means is optimized. Effective contribution learning means for storing whether or not it has effectively contributed to the processing; constant value changing means for changing a circuit constant value based on information obtained by the effective contribution learning means; And control means for setting necessary values for the constant value changing means and controlling each circuit constant to approach an optimum value by changing the set value based on the circuit characteristic calculation result. A circuit constant optimization method in an automatic circuit constant design system, characterized in that a constant to be changed is determined by the effective contribution learning means. 4. A circuit characteristic calculating means for calculating circuit characteristics such as a DC characteristic, an AC characteristic, and a transient characteristic of a circuit, and which circuit constant among circuit constants based on the circuit characteristics calculated by the circuit characteristic calculating means is optimized. Effective contribution learning means for storing whether or not it has effectively contributed to the processing; constant value changing means for changing a circuit constant value based on information obtained by the effective contribution learning means; And control means for setting necessary values for the constant value changing means and controlling the respective circuit constants to approach the optimum value by changing the set values based on the calculation results of the circuit characteristics. In order to search for a change in the circuit constant value that is effective in order to optimize the circuit constants without making small changes in the values of all the circuit constants to be optimized In the process, the target circuit constant to be minutely changed is limited according to the effective contribution obtained for each circuit constant by the effective contribution learning means, and the processing order of the circuit constant to be minutely changed is changed. A circuit constant optimizing method in an automatic circuit constant design system, characterized in that: