JPH09179895A - Circuit analysis device - Google Patents

Abstract

(57) [Abstract] [PROBLEMS] To obtain a circuit analysis device that automatically obtains a solution without requiring the user to reset initial values and convergence conditions. A data file input processing unit (2) receives connection information (1) of a circuit to be simulated and obtains a circuit equation based on the connection information (1) of the circuit. The circuit equation solver 3 obtains the solution of the circuit equation. Output part 4
Outputs the solution. In the circuit equation solving section 3,
After obtaining the solution in the flow 100, the solution is the flow 10
If the convergence condition is not satisfied in 1, the flow 1
At 02, a different solver is selected and a solution is obtained again based on the selected solver. When the solution does not satisfy the convergence condition in this way, the user does not need to reset the initial value and the convergence condition by re-solving the solution with another solver.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a circuit analysis device for performing circuit simulation used in, for example, LSI circuit design, and more particularly to a circuit analysis device for efficiently performing circuit simulation.

[0002]

2. Description of the Related Art FIG. 3 is a conceptual diagram showing a conventional circuit analysis device. First, the operation will be described. The data file input processing unit 2 reads an input file in which the connection information 1 of the circuit to be simulated is described, and generates a circuit equation such as a simultaneous nonlinear equation that describes the circuit operation based on the connection information 1. The circuit equation is expressed as in Equation 1 below. In addition, X is an unknown variable including the node voltage and the current flowing through the voltage definition branch, and n represents the number thereof.

[0003]

[Equation 1]

The circuit equation solving section 3 receives the equation 1. Further, conventionally, the circuit equation solving section 3 includes only one solver for finding the solution of the circuit equation. The solver is software for a solution method for finding the solution of the circuit equation. The operation of the circuit equation solving section 3 mainly includes a flow 100 and a flow 101.

In the flow 100, the value of the unknown variable X, which is the solution of the equation 1, is obtained based on only that one solver. Usually, solvers often use Newton's method as a solution method.

Next, the Newton method will be briefly described. The Newton's method is an iterative calculation, and Equation 2 shown below is repeated to obtain a solution. Equation 2 represents the solution at the (k + 1) th iteration. The initial value of the solution when k = 1 is set in advance by the user.

[0007]

[Equation 2]

The flow 101 determines whether or not the solution satisfies the convergence condition preset by the user. If it is satisfied, the solution is output. If not satisfied, a report indicating that the solution has not converged is output.

The above convergence condition will be described in detail.
There are first and second convergence conditions. First, the first condition is that in the iterative calculation, the difference between the solution at the k-th iteration and the solution at the (k-1) -th iteration is converged within an allowable error range preset by the user. . Next, the second condition is that the number of repetitions k is less than the number of repetitions preset by the user. When the k-th solution satisfies the first and second conditions, the solution is output as described above. When the k-th solution does not satisfy the first and second conditions, the fact that the solution has not converged is output as a report as described above.

The output unit 4 receives the output of the circuit equation solving unit 3 and outputs a solution or a report, for example, as a display.

[0011]

Since the conventional circuit analysis device is constructed as described above, it has the following problems. That is, when the solution that satisfies the convergence condition is not obtained, the user needs to reset the initial value and the convergence condition and perform the simulation again. In addition, Newton's method has a remarkable initial value dependence, and since the initial value does not exist in the vicinity of the true solution, there is a characteristic that the convergence decreases.
The work of resetting initial values and convergence conditions becomes very complicated. In particular, the larger the circuit, the more complicated it becomes.

The present invention has been made in order to solve the above problems, and it is an object of the present invention to obtain a circuit analysis device for automatically obtaining a solution without the need for the user to reset initial values and convergence conditions. To aim.

[0013]

According to a first aspect of the present invention, there is provided a data file input processing unit for receiving connection information of a circuit to be simulated and obtaining a circuit equation based on the connection information. And a plurality of solvers for obtaining a solution of the circuit equation, receiving the circuit equation, until the solution has a predetermined convergence condition,
And a circuit equation solver for continuously activating the plurality of solvers in a predetermined order to obtain the solution.

In the problem solving means according to claim 2 of the present invention, the plurality of solvers are Newton's method, Damped-New
At least 2 of ton method, Gmin-Stepping method and homotopy method
Including one.

In the problem solving means according to the third aspect of the present invention, the homotopy method makes the equation based on the homotopy method closer to the circuit equation in finer steps.

[0016]

BEST MODE FOR CARRYING OUT THE INVENTION

Embodiment 1 FIG. 1 is a conceptual diagram showing a circuit analysis device according to a first embodiment of the present invention. In FIG. 1, 102
Is a flow for selecting a solver, and other symbols correspond to the symbols in FIG.

Next, the configuration will be described. The main configuration is the same as the configuration of the circuit analysis device shown in FIG. The different part is that the circuit equation solving section 3 includes a plurality of different solvers for finding solutions of preset circuit equations. If the convergence condition is not satisfied in the flow 101, the process moves to the flow 102 for selecting another solver and then to the flow 100.

Next, the operation will be described. The main operation is the same as the operation of the circuit analysis device shown in FIG. The difference is that the k-th solution obtained in flow 101 is the first solution.
If the second condition is not satisfied, the process moves to the flow 102 and one of the solvers is selected. Next, the process moves to flow 100, and the solution of Expression 1 is obtained based on the selected solver. Even if the selected solver does not satisfy the convergence condition in the flow 101, the flow 10
Go to 2 and select another solver.

Therefore, when the solution does not satisfy the convergence condition, the solution is obtained by another solver, so that the user does not need to reset the initial value and the convergence condition, and the solution can be automatically obtained.

FIG. 2 is a diagram showing details of the operation of the circuit equation solving section 3 of FIG. In FIG. 2, m is a flag for selecting a solver. First, before entering the flow 103,
As an initial condition, 0 is assigned to each of the flag m and the repeat count k. In addition, the number of flags m is 0, 1,
Cases 2 and 3 correspond to a solver using the Newton method, a solver using the Damped-Newton method, a solver using the Gmin-Stepping method, and a solver using the homotopy method, respectively.

Next, referring to the flow 103, the solution of the k-th circuit equation is obtained, and then 1 is added to k. Before entering the flow 103, a solver corresponding to the number of m is selected in advance. In particular, when the flow 103 is first entered (the flag m is 0), the Newton method is selected in advance.

Next, referring to the flow 104, the first solution is obtained.
It is determined whether or not the condition of is satisfied. The first condition is the same as the conventional technique. When the k-th solution satisfies the first condition, the iterative calculation is ended and the converged solution is output to the output unit 4. If the k-th solution does not satisfy the first condition, the process moves to flow 105.

Next, referring to the flow 105, the second condition is judged. The second convergence condition is the same as in the conventional technique. When the k-th solution satisfies the second condition, the process moves to flow 103. When the second convergence condition is not satisfied, 1 is added to the flag m, and the process proceeds to flow 106.

A flow 106 is a flow in which a solver corresponding to the number of m is selected in advance before entering the flow 103.

In this way, the flows 103, 104, 10
5. Next, the solution is obtained by performing repeated calculation by the loop returning to the flow 103. If the solution does not converge and the number of iterations k is equal to or greater than a preset number of iterations, different solvers perform iterative calculation by the above loop. Thus, when a converged solution cannot be obtained,
By choosing different solvers, the converged solution is found based on successively different solvers until a converged solution is obtained.

In the flow 106, if the flag m is 4, that is, the solution is flow 1 by all the solvers.
If it does not converge in 04, the iterative calculation is terminated and, for example, the fact that the solution has not converged is output as a report.

The solver uses the Newton method as a solution method,
Damped-Newton method, Gmin-Stepping method, homotopy method 4
Use the type.

The four types of solvers will be described. First, the Newton method is similar to the description in the conventional art.

Next, the Damped-Newton method will be briefly described. The Damped-Newton method is an iterative calculation. Using the damping parameter t that continuously changes in the interval (0, 1), the solution at the (k + 1) th iteration is obtained by the following Expression 3.

[0030]

(Equation 3)

The damping parameter t is determined so as to satisfy the following equation (4).

[0032]

(Equation 4)

Next, the Gmin-Stepping method will be briefly described. The Gmin-Stepping method is also an iterative calculation. The exponential function that describes the diode characteristics is a factor that causes the circuit equation to have strong nonlinearity. A semiconductor element such as a MOSFET that can be handled by a simulator in a circuit analysis device includes a plurality of diodes in its equivalent circuit expression. For example, pn junctions between the source and the substrate and between the drain and the substrate of the MOSFET correspond to them.
If a resistor is inserted in parallel with this diode, the nonlinearity will be relaxed and the convergence will be improved, but in the Gmin-Stepping method, the parallel resistance will start from a small value and then gradually increase (~ 1.0E6) To the original open state (original circuit configuration).

Next, the homotopy method will be briefly described. The homotopy method is also an iterative calculation, and is a method of tracing a solution curve while checking the norm (residual error) of a nonlinear equation in order to improve convergence when no initial value is specified. For the nonlinear equation (Equation 1), the interval (0,
The homotopy is defined by the following equation 5 based on the homotopy method using the parameter t that changes in 1). Note that a
Is an initial value, and W is a scalar quantity for expressing a weight.

[0035]

(Equation 5)

Equation 5 becomes Equation 6 shown below when the parameter t = 0. The expression 5 approaches the expression 1 as the parameter t approaches 1, and coincides with the expression 1 when the parameter t = 1.

[0037]

(Equation 6)

The homotopy method uses H with the parameter t = 0.
Based on the solution X = a of (t, X) = 0, the solution curve of H (t, X) = 0 is traced while gradually changing the value of the parameter t, and finally the parameter t = This is a method aiming at reaching a solution for 1 (that is, satisfying F (X) = 0).

In order to efficiently trace the solution curve, the norm SR shown in the following equation 7 is obtained by using the solution of the equation 5.

[0040]

(Equation 7)

When the following equation 8 is obtained for the norm SR (k) at the time of the k-th tracking, the initial value a in the equation 5 is obtained.
Is replaced with a solution of H (t, X) = 0 at the time of k-th tracking.

[0042]

(Equation 8)

After the replacement, the homotopy function is reconstructed to return the parameter t to 0, and then tracking is started. If the value of SR becomes sufficiently small, it is an approximate solution of Equation 1 to be originally solved. By checking the size of the norm SR, it is possible to prevent the local minimum value from being recognized as a solution, and to more efficiently find the solution of the nonlinear equation.

According to this embodiment, there is an effect that a solution can be automatically obtained without designating a solver.

In this embodiment, the Newton method, Da
Although the mped-Newton method, the Gmin-Stepping method, and the homotopy method are selected in this order, they may be selected in another order. That is, the Newton method, Damped-N
ewton method, Gmin-Stepping method, homotopy method
There may be three orders. Further, in the present embodiment, the above four solvers are used, but other solvers other than the above four types may be used. In addition, the circuit equation solver 3
It only needs to have one or more different solvers. Also, the two or more solvers may be in different orders.

Embodiment 2 Embodiment 2 will be described.
The range (0, 1) in which the parameter t of the homotopy method described in the first embodiment can change is divided into m sections, and δ =
1 / m. A new parameter p is introduced in order to determine the parameter t, and p (0) = 0 is set, and t at the time of the (k + 1) th tracking is determined by Expression 9 and Expression 10.

[0047]

[Equation 9]

[0048]

(Equation 10)

When the parameter t is changed by using the equations 9 and 10, when the equation 5 is approximated to the circuit equation, the closer it is, the closer it is to the circuit equation.

In the present embodiment, by using the conversion function as shown in the equations 9 and 10, the initial value (t =
0) is coarse in the vicinity of 0) and can be traced in fine intervals in the vicinity of the true solution (t = 1), and the solution can be efficiently reached.

[0051]

According to claim 1 of the present invention, there is an effect that a solution can be automatically obtained without designating a solver.

According to claim 2 of the present invention, when the homotopy method is included, it is possible to prevent recognizing a local minimum value as a solution and to efficiently obtain a solution of a circuit equation. Play.

According to claim 3 of the present invention, there is an effect that a solution can be efficiently reached.

[Brief description of the drawings]

FIG. 1 is a diagram showing a circuit analysis device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing details of the operation of the circuit equation solving section 3 shown in FIG.

FIG. 3 is a diagram showing a conventional circuit analysis device.

[Explanation of symbols]

1 connection information, 2 data file input processing part, 3 circuit equation solving part, 4 output part.

Claims (3)

[Claims] 1. A data file input processing unit for receiving connection information of a circuit to be simulated and obtaining a circuit equation based on the connection information, and a plurality of solvers for obtaining a solution of the circuit equation, A circuit equation solver for receiving the circuit equation and continuously activating the plurality of solvers in a predetermined order to obtain the solution until the solution has a predetermined convergence condition,
Circuit analyzer equipped with. 2. The circuit analysis device according to claim 1, wherein the plurality of solvers includes at least two of a Newton method, a Damped-Newton method, a Gmin-Stepping method, and a homotopy method. 3. The circuit analysis device according to claim 2, wherein in the homotopy method, when the equation based on the homotopy method is brought closer to the circuit equation, the closer to the circuit equation, the closer to the circuit equation.