KR100294660B1 - Multi-variable control method - Google Patents

Abstract

PURPOSE: A multi-variable control method is provided to exactly control a temperature in a chamber having a semiconductor test handler. CONSTITUTION: A control part reads temperature data from a sensor which senses an indoor temperature(111), and calculates an error between an input temperature and a set temperature(112). A minimum error is compared with a predetermined positive limited temperature(113). If the minimum error is over the predetermined positive limited temperature, a step jumps (115)steps and a control command is calculated into a minimum value. If the minimum error is below the predetermined positive limited temperature, there is judged whether a maximum error is over a predetermined negative limited temperature(114). If the maximum error is below the predetermined negative limited temperature, a control command is calculated(116). If the maximum error is over the predetermined negative limited temperature, an optimum control command is carried out(117).

Description

Multivariable Control Method and Device

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a semiconductor inspection apparatus, and more particularly, to a multivariate control method and apparatus for performing semiconductor temperature stress inspection in a temperature-related environment.

Semiconductor reliability is a very important factor in determining the lifetime of equipment and other electronic applications. Common considerations for semiconductors are initial failure, constant failure rate, lifetime (long-term reliability) and durability. However, the initial failure, failure rate, and lifespan are the main targets in the semiconductor, and durability does not affect the reliability of the semiconductor in normal operation. In addition, the failure rate of the semiconductor is relatively low, but the electronic application is expensive and has a very important meaning to secure reliability, and is considered as part of the technology for predicting and adjusting the long-term reliability. This is recognized as part of the definable statistical failure distribution.

The lifetime distribution for semiconductors is known to be lognormal and has been used to model long term failure rates. Since this failure rate depends on stress, or accelerated stress, this accelerated stress test is considered as a fundamental technique for predicting and specifying long-term reliability. In this accelerated stress test, the Arrhenius relationship, which determines the effect of temperature, is used. At this time, the failure mechanism is most affected by temperature, and it is a standard for quality that indicates that the lot approval test for long-term reliability and predetermined reliability and objectivity are met.

Early failure failure rates are modeled by lognormal or Weibull statistics, and screening to reduce initial failure failure is by accelerated stress testing. A model for estimating the effects of this screening and burn-in of dynamics is to find fault mechanisms, as is well known in the art. The main factors are temperature, humidity, charge, voltage, current, etc., by using this factor to create one or more complex environments in the semiconductor, semiconductor material, package sealing state, semiconductor surface properties deterioration, semiconductor malfunction And back.

The semiconductor is subjected to quality and lot approval tests, which are also performed in the prototype's internal circuits for the chip's operation in the electrolyte, temperature bias test, temperature humidity bias test and current gain. Find the mechanism and screen normal semiconductors.

In addition, semiconductor burn-in test is an effective method for screening defects that are problematic for long term reliability such as initial failure. This sorting method is performed by combining the electrical stress with temperature and time. Therefore, the accelerated stress is to find the temperature dependent failure mechanism in a relatively short time.

Thus, in order to secure the reliability of the semiconductor, the semiconductor is subjected to a number of inspections and screening for defect factors of the failure mechanism. Nevertheless, the failure mechanism of semiconductors has existed to date, and the selection of semiconductors could not be completed.

At present, the cause of the failure mechanism is mainly influenced by temperature, and it is recognized as the most reliable method of classifying according to the electrical characteristics of the semiconductor at a predetermined temperature.

Accordingly, in terms of improving the reliability of the semiconductor, it is necessary to check the operation state of the semiconductor in the high temperature chamber and select the grade according to the operation state.

For this purpose, the inspection apparatus conventionally used is to place the semiconductor inspection handler 2 in the chamber 1 as shown in FIG. The chamber 1 is in the form of a thermostat for precise temperature control and is precisely temperature controlled. Although not shown, the chamber 1 is provided with a heater and a fan in the upper portion thereof, and the semiconductor inspection handler 2 containing 32 semiconductors is placed at a predetermined position. At this time, the chamber 1 should normally have a constant temperature distribution. In practice, however, the temperature is distributed as shown in the figure due to structural problems in the chamber and temperature loss of the outer wall. As shown on the graph, the temperature in the chamber is distributed from 126 ° C to 132 ° C and the difference reaches 6 ° C. This temperature difference is arranged at a high temperature and a low temperature in relation to the test semiconductor housed in the semiconductor test handler 2, and thus has a temperature difference of 1-2 ° C.

This temperature difference could result in inaccurate or poor grades. In this way, when the grade is misselected or treated poorly due to the temperature difference (1-2 ° C.) at high temperature, the loss cannot be ignored.

This is because, at this stage, the semiconductor has been inspected for all the failure mechanisms, and cannot be treated as defective or unfair in the class unless it is a specific case. For example, if the semiconductor is less than 125 ° C. or above 126 ° C., a defect or a non-low grade semiconductor is failed due to the micro temperature difference, resulting in financial loss.

Therefore, this temperature difference is eliminated, so it is very desirable to accurately classify the semiconductor and to minimize the defect rate.

In view of these points, most controllers use a proportional integral derivative (PID) method to control the temperature in the chamber when the heater is operated and controlled. This method is designed by trial-and-error method, and the appropriate parameter is determined by looking at the frequency response rather than minimizing the given figure of merit. This method did not fundamentally guarantee system stability and performance at design time.

The reason for this is the lack of coping with steady-state error, delay of transient response speed, deterioration of plant performance, and disturbance to the system such as external noise.

Accordingly, an object of the present invention is to provide a multivariate control method and apparatus for precisely controlling the temperature in a chamber containing a semiconductor inspection handler by performing multivariate control to improve the stability and performance of the system.

Another object of the present invention is to install three or more heaters and three sensors for sensing the area temperature inside the chamber heated by the heaters to precisely control the temperature in the chamber, and individually control the heaters according to the sensor input. To provide a multi-variable control method and apparatus.

1 is a view showing a temperature distribution in a chamber in which a handler for semiconductor inspection is provided therein;

2 is a block diagram showing a correlation between a heater, a fan, and a temperature sensor installed in a chamber according to the prior art;

Figure 3 is a block diagram showing the correlation of two or more heaters, fans and temperature sensors installed in the chamber according to the present invention;

4 is a flowchart showing a method for calculating an algorithm of multivariate parameters in accordance with the principles of the present invention;

5A and 5B are flowcharts illustrating multivariable control over temperature distribution in a chamber, in accordance with the principles of this invention;

Fig. 6 is a block diagram showing control means constructed in accordance with the principles of this invention.

<Explanation of main drawing code>

1: chamber 2: inspection site 3: fan 4: heater

5: Sensor 6: Control Unit 7.8.9: State Estimation Computation Unit

It is still another object of the present invention to provide a multivariate control method and apparatus for performing multivariable control to optimize the performance of a system with respect to a steady state error, response speed, and disturbance.

The invention is based on at least three heaters, a sensor that senses the temperature in the region heated by the heater, and a temperature sensing signal from the sensor to maintain the temperature in the chamber containing the semiconductor inspection handler at an absolute temperature. And control the operation of the heater to control the temperature in the chamber precisely.

The control unit applies an input according to the system test and receives the output to determine the temperature model according to the state in the chamber, and accordingly determines the multivariate control parameter according to the state of the chamber. By calculating the optimal control gain, and calculating the optimal state estimator gain by differentiating the state equation, the state differential equation is converted into the digital control parameter so that the optimal control is achieved based on the state estimator gain. Identify, read the data from the temperature sensor every predetermined time, calculate the error with the set temperature, determine whether the error is optimal control, minimum control, and maximum value control to generate a predetermined control signal, In response to this signal, the heater drive unit causes the heater to be driven.

Therefore, after the system modeling is completed, the control unit converts the control signal calculated by the estimated state equation based on the digital control parameter obtained by converting the error signal according to the signal from each sensor into the state differential equation. , It is generated for the heater.

As described above, the present invention provides ARMAX (Auto-Regressive Moving Average With External Input) to allow the control unit to approximate the parameters obtained from the relationship between the irregular input and the output of the plant to the actual plant. System identification is performed by applying the model to improve the stability and performance of the system.

In addition, the present invention includes a control unit which systematically minimizes the performance index according to the temperature difference between the set temperature in the chamber and the actual temperature by applying an output model in which all state vectors are measurable in order to increase the control performance.

This invention will be described in detail with reference to the accompanying drawings.

As shown in Fig. 2, in the prior art, the fan 3 and the heater 4 are sequentially installed in the chamber 1, so that one sensor 5 is separated from the heater 4 and fixed. In addition, the fan 3 and the heater 4 are connected to the output terminal of the control unit 6 for operation control, and the sensor 5 is connected to the input terminal of the control unit 6 and senses the temperature in the chamber 1 to control the control unit 6. ).

On the other hand, according to the principle of the present invention, as shown in FIG. 3, the multi-variable control device according to the present invention has one or more fans 3 and one or more sensors 5 having a predetermined distance with respect to each of the heaters 4 and 4. Are installed in the chamber 1. For example, three heaters 4 may be installed, for example, but the fan 3 may be installed to correspond to the heater 4, or only one may be installed. Of course, the control unit 6 is connected to the heater 4 and the sensors 5 to control the operation or to receive the sensing temperature. This constituted the plant which controls the temperature in the chamber 1 which accommodates a semiconductor test handler so that the control part 6 may have three outputs and three inputs. This means that the entire system of this invention consists of a multivariable input / output system consisting of three inputs and three outputs.

Therefore, in order to precisely control the temperature in the system chamber 1, the control parameters of the system must be obtained as shown in FIG. First, in order to estimate the parameters of the system, in step 101, a system test input is applied, and in step 102, the output thereof is received to measure input / output data. At this time, the input from the control unit is power, and the output is a sensing temperature. In step 103, the system identification is performed using the ARMAX model of a single input-output model well-known in this technique, and the parameter correction is performed using the least square algorithm, that is, the least square method, for parameter identification. This allows the parameter to approximate the characteristics of the actual plant. This ARMAX method allows the identification of a single input / output model that is well simulated with the actual temperature system by the actual experimental results. The whole system is regarded as a multivariate input / output system. The state variables of the system are three temperature points. For these system state variables, the multivariable input / output system was identified using the ARMAX model, which is well known in the art.

In order to perform system I / O control after the system identification is made, the present invention uses LQG / LTR (Linear Quadratic Gaussian / Loop Transfer Recovery) method in the single input / output system model. For the variable I / O model, LQR (Linear Qnadratic Regulator) method can be used. These methods can minimize the performance index of the system and at the same time ensure stability.

As the present invention is regarded as a multivariate input / output system, in step 104, the optimum control gain Kc _i is calculated as follows according to the equation of motion in the chamber.

State equation: x _i (t) = A _i x _i (t) + B _i U _i (t)

Output equation: y _i (t) = C _i x _i (t) + D _i U _i (t)

here

x _i (t): State vector of linear time invariant multivariate

U _i (t): control input vector

A _i , R _i , C _i : System parameters obtained using system identification

J _i : Price function

R _i : Control weighted matrix

State weighted matrix Q _i : C _i ^T C _i- (C _i )

Therefore, by using Equation 1, Equation 2, Equation 3 and Equation 4, an optimal control gain K _ci can be obtained to minimize the price function J _i .

In step 105, the optimal state estimator gain Kf _i : ( _i = 1,2,3) is obtained as follows.

First, the state equation of Equation 1 is differentiated. This refers to the structure of the MBC model of dynamics of state equations.

In Equation (5), the estimated state vector Optimal State Estimation by Minimizing Covariance

The gain K _fi ( _i = 1,2,3) is calculated.

In step 106, these state differential equations are converted into digital control parameters. That is, the plant of the present invention is applied to the invariant discrete time system of linearity, and the optimum control parameter is determined according to the relative spatial model of the system.

Therefore, when the state space model is determined and the control command of the input vector is determined, the state estimation for the temperature in the chamber by each heater and sensor is made and the control signal is generated.

In order to generate such a control signal, the system performs system control as shown in FIG. 5A.

First, the control unit 6 reads the temperature data (T ₁ , T ₂ , T ₃ ) according to the sensor 5 detects the room temperature in step 111. In step 112, the error (er1, er2, er3) between the input temperature and the set temperature is calculated and the process proceeds to step 113.

In step 113, the minimum error is compared with a preset positive limit temperature. If the minimum error is greater than the positive limit temperature, for example, greater than 1 degree, the control command having the minimum value is calculated by jumping to step 115. If the minimum error is less than the positive limit temperature then go to Step 114.

In step 114, if the maximum error is less than the preset negative limit temperature, for example, a negative limit temperature of 1 degree, the control command is calculated by moving to step 116. If the maximum error is not less than the negative limit temperature, step 117 Will execute the optimal control command.

When the control command is executed, the control command for the input vector and the state vector for measuring the state of the state space model is sequentially performed for each heater and sensor. This is done in turn as shown in Figure 5B. This first calculates the state estimation of the plant as shown in FIG. 5B.

The control command is then calculated.

When such a calculation is made, the process proceeds to step 118 to generate a pulse width modulation (PWM) signal according to the control command, and applies it to a heater driver (not shown). The heater driver controls the operation of the heater in step 119.

Meanwhile, as shown in FIG. 5B, the control unit calculates the state estimation by inputting the detected temperature from each sensor and the error temperature for the set temperature. The calculated output is a state command signal and outputs the control signals PWM1, PWM2, and PWM3 shown in Equations 6 and 7. These control signals are applied to a heater driver (not shown) to control the operation of the heater. Therefore, it can be seen that the controller includes state estimation calculators 7, 8 and 9.

In this invention, the parameter estimation and system identification of the temperature model are performed to precisely control the temperature of the chamber according to the design of the system, and the multivariate control is applied to the control unit to maintain stability and robustness while minimizing the performance index. In this regard, optimal control can be achieved considering all the surroundings for controllability, stability, root locus, maximum overshoot, rise time and settling times.

Accordingly, the temperature inside the chamber minimizes the noise and normal error of the disturbance controller due to the internal temperature difference, thereby achieving reliable temperature control.

Claims (8)

In the multivariate control method that configures the temperature chamber as a plant to optimize the performance of the system, In order to precisely distribute the temperature in the chamber, output is made according to the system test, and the load sensing signal is received to determine the temperature model according to the temperature distribution in the chamber. Estimation parameter is estimated according to temperature model, state equation, control input vector and price function are determined and optimal gain is calculated. The optimal state estimator gain is calculated by differentiating the state equation, Based on the state optimal estimator gain, the state differential equation is converted into digital control parameters to identify the system. Read data from temperature sensor every predetermined time, calculate error with set temperature, Multivariate control method to determine whether the error is optimal control, minimum value control and maximum value control to generate a predetermined control signal and to operate the heater according to the control signal The method of claim 1, Multivariable Control Method for Optimal Control Gain Calculation x _i (t) = A _i x _i (t) + B _i U _i (t) U _i (t) =-Kc _i x _i (t) U _i = (K) =-K _ci x _i (k) Calculate Kc1, Kc2, Kc3 to minimize J _i ( _i = 1,2,3). The method of claim 1, Multivariate Control Method for Optimum Estimator Gain _Kfi Calculated as in Calculate K _fi ( _i = 1,2,3) to minimize the covariance of. The method of claim 1, A multivariate control method that transforms the state equation for calculating the gain of the optimal state estimator and converts it into the digital control para meta as follows. Optimal control Constant to be PHI _i GAMMA _i ( _i = 1,2,3) Calculate The method of claim 1, Multivariable control method that makes control command signal become pulse width modulated signal In the multivariate control system that configures the temperature chamber as a plant to optimize the performance of Three or more loads to be a control output to the plant of the system; Three or more sensing means loads serving as control inputs; And a control unit configured to perform optimum control of the load by including a state estimating operation unit for each input in a state in which sensing means are connected to inputs and outputs thereof and system identification modeling is completed. The method of claim 6, Multivariable control that causes the load to become a heater. The method of claim 7, wherein Multivariate control device that allows the relative estimation operation unit to determine the control signal for the error signal according to the input of each sensor as follows