KR19990070870A - Prober autofocus method - Google Patents

Abstract

The present invention relates to an autofocus method of a wafer prober. In particular, the intensity of an image obtained from a camera is measured by a strength measurement function through step movement repeatedly up and down in the Z-axis, and the intensity table is updated to an average value with previous data And a second step of smoothing the table value and moving the Z axis to a position having a maximum energy value to obtain an optimum focal length. This process is repeated until a given condition is satisfied, In many cases, as the table value is updated more frequently, local maxima (maxima, minima) disappear and the focus of the camera for image processing of the wafer prober is automatically adjusted, the same focal length is always guaranteed, and the non-error is minimized.

Description

Prober autofocus method

The present invention relates to a prober, and more particularly to a prober autofocus method that automatically focusses a camera for image processing of a wafer prober and ensures the same focal distance at all times.

Prober is a device that makes contact between a wafer and a probe at a designated position by moving the wafer so that the test can be connected to the tester. BACKGROUND ART [0002] In a wafer test process, which is a final process as a wafer state, which is one of semiconductor device production processes, electrical characteristics and functional characteristics of a semiconductor device formed on a wafer are measured to determine good products and defects, For a memory chip, the wafer or probe card is moved step by step in chip size so that the connected tester can judge whether the defective cell can be saved or not. The wafer prober is a device that repeats the above operation so that the tester can inspect all the chips on one wafer.

1, a probe card 17 having a probe 11 for testing the wafer 11, a probe card 17 having a probe 18 for testing the wafer 11, X, Y, Z, and θ stages 12, 13, 14 and 15 for precisely transferring the wafer 11 to the position of the probe 18 for contact with the stage 18, A loader 20-1 for raising the wafer 11 to the stages 12 to 15, a camera 10 for detecting the position and orientation of the probe 18, And a pin ring 19 and a test head 20 for the probe card 17. [

First, when the wafer 11 is transferred to the XYZ? Stage 12, 13, 14, 15 by the loader, the XYZ? Stages 12 to 15 transfer the wafer 11 to the probe position 18 of the probe card 17 .

Such a wafer prober uses the computer vision to detect the pattern position of the wafer and the position of the pin 18 of the probe card 17. [ At this time, in order to perform image processing, the camera 10 must acquire an image at a position where an image of an object is accurately formed, and a position at which an image is obtained needs a repeat accuracy that always exists within an allowable error. This is why auto focus is required.

Fig. 2 is a block diagram of the configuration of the prober vision system. The output of the camera module 10 attached to the frame of the prober is applied to the frame grabber 24. The frame grabber 24 is connected to a bus of a personal computer (PC) as a main controller 25 and stores an image obtained from the camera 10 in a memory (not shown) in the controller 25. At this time, the Z-axis 14 of the prober operates as the motion controller 26 and the motion controller 26 is connected to the bus of the PC as the main controller 25. [

The intensity value of the image read from the frame grabber 24 indicates the maximum at the focus, and becomes smaller as the distance from the focus is increased. A function that calculates the intensity value of such an image and obtains the energy value is called a standard function. A graph showing the energy distribution according to the focal distance can be obtained by using this function. It can be assumed that the energy graph of this intensity has a normal distribution curve (Gaussian distribution curve) symmetric about the maximum point as shown in FIG. However, it is difficult to obtain the same image and the corresponding image due to the vibration, the perpendicularity between the optical axis and the measured object, and the characteristics of the grabber. Because of this noise, the energy graph using the standard function is asymmetric, or has a very small, maximum value.

In order to focus on online, we use gradient search method. We can not find the maximum point by focusing only on the inclination search using only the output of the standard function. We need to remove the local maximum point (maximum, minimum) .

Conventionally, the gradient search algorithm for on-line focus search is as follows, and FIG.

k = 100

get_degree_of _focus

move_z (? z)

k = k + 1

repeat

get_degree_of_focus

? Mean = T (k) -T (k-1)

if △ mean> 0 then

Δz = Δz

else

Δz = - Δz

move_z (? z)

until finish

That is, energy values of two points are compared in the first search interval, and then the Z axis is moved by a unit length by one step to a position having a larger energy value. However, the value obtained by the intensity measurement function is repeated due to noise or the like, so that the direction of Z moves repeatedly in the same section for the first several times, or repeatedly for a given number of iterations without the optimal focus.

As described above, in the conventional autofocus method described above, it is difficult to find the maximum or minimum by the slope of the function or to apply it to the vision of the wafer prober. That is, since the intensity measurement function of the image has a local maximum or minimum point due to noise or the like, the maximum focus can not be found and converges at a partial maximum / minimum point.

Therefore, the distortion of the measurement data due to the noise makes it impossible to search the inclination, and thus there is a problem that it is not suitable for finding the optimum on-line focus for measuring the distance from the pin or the wafer in the probervision.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method and apparatus for automatically adjusting a focus of a camera for image processing of a wafer prober and ensuring that the same focal distance is always maintained, And to provide an auto-focusing method of the prober.

According to another aspect of the present invention, there is provided an automatic focusing method of a prober, comprising: measuring intensity of an image obtained from a camera through step movement repeated vertically in the Z- And a second step of smoothing the table value and moving the Z axis to a position having a maximum energy value to obtain an optimum focal length. This process is repeated until a given condition is satisfied It is characterized by iteration.

1 is a schematic block diagram of a general wafer prober;

2 is a block diagram of the configuration of the prober vision system

3 is a graph showing the energy distribution according to the focal length

4 is a flowchart showing an auto focus method using a conventional gradient search.

FIG. 5 is a flowchart illustrating an autofocus method using an inclination search according to the present invention.

DESCRIPTION OF THE REFERENCE NUMERALS

10: camera 11: wafer

12 to 15: XYZΘ stage 16: stage base

17: Probe card 18: Probe pin

19: pin ring 20: test head

24: frame grabber 25: main controller

26: Motion controller

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The hardware for carrying out the autofocus method of the wafer prober according to the present invention is the same as that of FIG. 2 described above, and a flow chart for the improved gradient search algorithm is shown in FIG.

That is, the image acquired by the grabber is updated to the previous data and average in the intensity table T (k) through the intensity measurement function, and determines the direction of movement of the Z axis through the previous data and the smoothing function.

The function F _Df (Degree of Focus) for measuring the intensity is expressed by Equation 1 below.

Here, I is the gray level at the pixel (x, y).

At this time, the intensity measurement function is obtained by measuring the intensity of a window having N x N pixel regions equal to or smaller than the entire image in the image acquired by the grabber 24. The intensity difference between one pixel and its neighboring pixels in a window is small in a blurred image and becomes large as it is focused. If the size of the window is too small, the effect of noise becomes serious. If it is too large, the effect of noise becomes small, but the execution speed becomes too slow. When the object to be measured is three-dimensionally, when the window size is too large, the effect of averaging the different height portions of the object to be measured occurs, so that the influence of the window size is seriously affected. However, since the object to be measured in the prober is not a three-dimensional wafer but a wafer plane, the size of the window is not so large. Therefore, in the present invention, the window size is determined to be a 100x100 pixel area by several experiments.

In addition, since the pattern of the object to be measured (wafer surface) may also affect the intensity function, F _Dfx and F _Dfy must be separately _determined considering the continuous pattern in the longitudinal direction or the transverse direction.

Then, the measured value F _Df is stored in the table T (k) as an average value through the following equation (2) through the stepwise movement in the Z axis and the table value is updated frequently in the noisy part, The learning effect of eliminating the points can be obtained.

In this case, the optimal focal length is obtained by finding energy data in a predetermined sampling interval and then performing curve fitting to find an optimal position. However, in this method, the data must be obtained for the whole section and data must be collected repeatedly in order to eliminate noise Therefore, it is not effective in prober. In order to find the optimal focus position on-line, it is necessary to collect a large amount of data in a place where noise is high, and to use the collected data as much as possible to find the Z value to move to the next. Accordingly, in the gradient search algorithm of the present invention, the Z-axis is moved, the intensity table is corrected, the table value is smoothed, and the Z-axis is moved to a position having the maximum energy value. This process is repeated until a given condition is satisfied.

The following is an improved gradient search algorithm, and FIG. 5 is a flowchart of the algorithm.

k = 100

get_degree_of _focus

update_table (k)

move_z (? z)

k = k + 1

repeat

get_degree_of_focus

update_table (k)

? Mean = T (k) -T (k-1)

if △ mean> 0 then

Δz = Δz

else

Δz = - Δz

move_z (? z)

until finish

That is, energy values of two points are compared in the first search interval, and the Z axis is moved by one unit length (6 um) to a position having a larger energy value. When one step is moved, the strength is measured and the strength table T (k) with the step index k is updated. However, the value obtained by the intensity measurement function can be repeatedly moved in the same section for the first several times because the direction of Z is different from the optimal focal direction due to noise, etc. However, this process averages the values of the table to remove noise. The movement direction of the Z axis is determined by smoothing the table value in the previous step and the table value in the current step by smoothing using the fuzzy rule and then comparing the two values to determine the Z axis movement direction.

As described above, according to the autofocus method for the wafer prober according to the present invention, the intensity of the image acquired by the grabber in the process of moving the Z axis is obtained through the intensity measurement function, and the intensity table is updated And the table value is smoothed to move the Z axis to a position having the maximum energy value. As a result, the update frequency of the table value is increased in the portion where the noise is high, and the local maximum point (maximum and minimum) Automatically focus the camera for processing, ensure the same focal distance at all times, and minimize non-error.

Claims (7)

A frame grabber for storing an image obtained from a camera attached to a frame of the prober in a memory of a main controller, and a motion controller for operating a Z-axis of the prober, A first step of measuring intensity of an image obtained from a camera through step movement repeated vertically in the Z axis and updating the intensity table, And a second step of smoothing the table value and moving the Z axis to a position having a maximum energy value to obtain an optimum focal length. This process is repeated until a given condition is satisfied. Focus method. 2. The method of claim 1, wherein the first step Calculating energy values of two points in a first search interval and moving the Z axis by a unit length by one step to a position having a larger energy value, And calculating a strength value of an image obtained from the frame grabber using a strength measurement function every time one step is performed, and updating the strength table as an average of the previous data. 3. The method according to claim 2, wherein the intensity measurement value is obtained by an intensity measurement function expressed by the following equation. Where I is the gray level at the pixel (x, y). 4. The method according to claim 3, wherein the intensity measurement function measures the intensity of the window obtained by the grabber with respect to a window having NxN pixel regions equal to or smaller than the entire image. 4. The method according to claim 3, wherein the intensity measurement function is _obtained by separately _obtaining F _Dfx and F _Dfy considering continuous patterns in the longitudinal direction or the transverse direction. 3. The method according to claim 2, wherein the intensity measurement value of the image obtained by the intensity measurement function is stored as an average value in the intensity table by the following equation. 2. The method of claim 1, And smoothing the table value in the previous step and the table value in the current step by smoothing using the fuzzy rule, and comparing the two values to determine the Z-axis moving direction.