JPS5855806A - Pattern position detecting device - Google Patents

Abstract

PURPOSE:To detect the pattern position highly accurately, by obtaining the pattern data of the pick up signals of images in detecting windows from the preset sizes of the detecting windows and the distances from chip corners, and selecting the pattern whose difference from the reference projection value is smallest among the patterns which are shifted by a plurality of bits. CONSTITUTION:A pad pattern 90 is processed 10 by a pick up system such as a TV camera. The image signal is converted 11 into digital data having three values at every picture element, and the result is stored in a frame memory 12. The corners of a semiconductor chip is detected 13 from said data, and the pattern detecting windows 91 and 92 are determined by the corners. Then the data in the detecting windows are summed for every column unit, and a numerical data A0 for one line is obtained. By using this, the shift processing of + or -n bits in the row direction is performed. Then the difference between the data A1-An and A-1-A-n and the preset reference projection value 17 for every bit is average by a reference projection value subtracting part 16 and a bit difference total sum averaging part 18. This is judged as the optimum address. The position detection is made possible by a minimum value detecting part 19, an optimum matching address output part 20, and a pad center address computing part 21.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a recognition device having an imaging system, and more particularly to a pattern position detection device for detecting the position of a pattern of a transistor, integrated circuit, etc.

Conventionally, in order to detect the position of a pattern in a transistor integrated circuit, etc., the necessary position information has to be obtained by calculation from the contact address.

FIG. 1 shows a two-contact address for a transistor chip butter/. The nine-picture image obtained by the imaging system is sequentially scanned horizontally from the top, for example, from 1-1° to 1-3, and the point 3 that first touches the chip 5 is set as the "first contact address," Conversely, scan from the bottom sequentially by scanning amount -1 to 2-4, and use point 4, which first touched chip 5, as the $20 contact address, and use these two addresses and distances a and b from the contact, which are calculated in advance. Find the required coordinates 6.7 No.

In the above method, the y-axis direction can be determined using the scanning accuracy of a television camera, but the x-axis direction can be determined by, for example, one pad
High accuracy is impossible because it changes depending on the car etc.

FIG. 2 shows the erroneous attachment in the X-axis direction.

As is clear from the comparison with the first ml, the contact points sl and al change in the X-axis direction depending on the scanning position.

As a method to improve the above method, eight pre-charged detection windows are used to cut out the video signal obtained from the imaging system, and as a result, the change points of the 11 data are used as the pattern section. There is a way to find it from two directions.

Although the above method improves accuracy, it may cause false detection of sections due to noise, such as P-de flaws.

The present invention is intended to solve the above-mentioned problems, and its purpose-
It is an object of the present invention to provide a pattern position detection device that has few errors and is resistant to noise.

A feature of the present invention is that since it has an imaging system, the recognition device has a digitization circuit that binarizes or ternaryizes the video signal, and the output signal of the digitization circuit is processed through a specific detection window. A comparison is made to determine the difference between the projection result and the predetermined projection data. Pattern position detection characterized by determining the pattern position from a plurality of directions using at least two detection windows. It's in the device.

Hereinafter, the present invention will be explained in detail using the drawings.

FIG. 3 is a flowchart showing the operation of the pattern position detection device of the present invention.

The nine image signals obtained by the imaging system processing 1110 are converted into three-value digital data by the ternarization circuit 11 and stored in the frame memory 12.

The corner of the semiconductor chip is detected by the chip corner detection 1s using the stored digital data, and the pattern/detection window is determined using the data.

Next, using the detection window, a detection window projection value number of stitches 15 is performed.

In the projection value count ls, data within the detection window is summed column by column and converted into one row of numerical data, that is, projection data A.

Using the projection datum O, a bit shift in the row direction (J611) is performed.

The projection data Am with a +1 bit shift has a corresponding number of bits of M-1 bits, with a +n bit shift it has a corresponding number of M-grains, and with a -n bit shift it has a corresponding number of bits of [1].

Corresponding bits mean valid bits within the nine valid ranges, taking into account deviations from the valid range caused by bit shifting, and the bits are summed in ^ units to show nine values.

Next, datums 1 to k, A obtained by shift processing
-1=A --m and the reference projection value 17 obtained in advance using projection data A.subtraction 16 is performed, and then the total average calculation of the bit difference -18, that is, the subtraction result corresponding to each bit. The absolute values are summed and divided by the corresponding number of bits.

Next, the minimum value of the result of the calculation data 18 is detected 19.

The result is output 2 to the pad center address calculation 31 as the optimal matching address.

In pad center address calculation 21, a pad center address is calculated and output using the result.

In other words, the data obtained by the minimum value detection 19 indicates the position of the projection data with the closest value to the reference projection value 17.1 This position data makes it difficult to detect the position in the pattern η, so the center of the bat is In the address calculation, the center of the bat is determined using the relationship between the pattern and the center of the bat determined in advance and the pattern position data. .

FIG. 4 shows an embodiment of the invention.

Reference projection value register 230 on Pasco wing of Combi, -Yu etc.
Input 24. Input 26 of projection value data register 2S, clock input 21, clock input 29 of shift number counter 2s, output 31 of digital position register 30,! Inputs 48 of four Lunaplexers are connected.

The output 32 of the reference projection value register 23 is the data register s
Enters input s4 of s.

The data register 33 (the parallel output 3s is the subtraction circuit S
Enter the first input 37 of @.

Projection value data register ago output 38 is connected to input 4 of data shift register 39.

Data shift register 39 output 41 company subtraction - road sso
Enter the third input 42.

The subtraction circuit 3/O output 43 is connected to the first input 45 of the i-multiplexer 44, and the output 4- of the multiplexer 4γ is connected to the shift register 500A output 61, respectively.

The output 52 of the shift register 50 enters the third input s3 of the multiplexer 44.

The output 54 of the multiplexer 44 is the input S of the adder circuit II.
Connected to @.

The output 4@ of the shift circuit counter 2$ is
eta first human power 68. Subtraction II 5 years' worth of first-time skills io,
Enter the first input 61 of the shift register 30, number of registers N memory 30, first output 63, second input 64 of the comparator s7, second input 66 of the second otb power 151d subtract @is, respectively. I can't connect.

Output 4 of comparator Is1 enters a second input 86 of multiplexer 47.

The output 61 of the adder circuit IS is the first manual input of the division @SS.
-, respectively enter the 20th input 11 of the output cuff 0F1 division -68 of the subtractor sI.

Divider 6 $ output cuff 3 is division value register 730 person cuff 4
connected to.

The first output cuff I of the division value register 73 is connected to the input cuff 7 of the minimum value register 7@, and the second output cuff 1 is connected to the comparator 19.
followed by the first person cam respectively.

fi Output 80 of small value register 7 is co/valley -79
The second input IK enters.

The output sl of the comparator 79 goes into the load input 12 of the minimum value register 76 and the load input 12 of the shift position register 30.

The binary or ternary digital signal, ie, digital data, is stored in the projection value data register 3SK.

The nine reference projection values are stored in the data register 39 via the reference projection value register 2s.

The digital data transferred to the projection value data register 2s is sequentially stored in the data shift register 31 by a tarot that is outputted from the IP path 33. At the same time, the tarok is entered in the number of shifts cuff/re 2-, and the same number of clocks are shifted sequentially.

In the comparator 57, the data stored in advance in the register number N memory 6 and the shift a number counter yf
1 until the contents of the shift circuit counter 38 become larger than the contents of the register number N memory.
The multiplexer sequentially outputs clocks representing 1 to the shift register 60 and stores them therein.

That is, the shift register sOK stores a corresponding bit indicating permission.

The subtracting circuit s6 shifts the reference measurement value to the right or left and continuously subtracts it from the shadow value data.
Output to 4.

The multiplexer 44 outputs only the subtraction result corresponding to the corresponding pinpoint, that is, 1, stored in the shift register sOK to the adder circuit.

In the addition circuit, 4, which is the result of subtracting 9, is added corresponding to the significant bit, and 6siK of division is output.

The subtracter 6 uses the contents of the number of shifts and the contents of the register number N register 63 to calculate the corresponding number of bits, and outputs the result to the divider 6Ik. N is the number of registers, N is the content of memory 63,
n is the content of the number of shifts.

Division 6Sg performs an operation of dividing the addition result by the corresponding number of bits, and stores the result in the division value register 73.

The contents of the stored minimum value register 76 are compared with the comparator 7 times, and the divided value register ? If the contents stored in the register 3 are small, the contents are stored in the minimum value register 76 and the contents of the shift position register 110 are stored in the shift position register.

The above-mentioned operations are sequentially performed on the crosshairs outputted from the path 22, and the result, that is, the pattern O optimum position data is stored in the shift position register.

In FIG. 4, the detection window 1') was explained, but the accurate pattern position can be detected by performing the foreboard motion and obtaining the detection window from multiple directions multiple times.

Figures S (a) and Cb) show the operation of the Wi401 [Example O].

The digital data of the pad butter 790 is extracted by the detection window 91 and is projected in each direction. The projection value datum O indicates the detection window 910 projection value.

The sum Ate of the absolute value of the difference between each bit of the reference projection output and the projection value data AO is determined. Furthermore, when the shift occurs by 1 bit to the left, find the difference between 10 and 10 bits.

In the same way, Muta is determined from the reference projection value and the projection value data.

As the shift amount increases, the corresponding number of bits decreases, so the total absolute value Ate decreases dramatically.

Therefore, find the difference in bit units.

When A-1, the data (sum of absolute values ttAt 1
) 10 (corresponding bit number mu - 1) is the minimum, and the position of sh' is determined to be a match.

As mentioned above, the present invention calculates a plurality of pre-input detection windows and distances from the chip corner, and calculates a reference projection value in which butter/data within a plurality of detection windows is recorded in advance. This provides a pattern position detection device that compares and subtracts the mutual comparison position with the one shifted by a plurality of bits, and then divides by the corresponding number of bits and detects the minimum value as the pattern position. According to -, it is possible to detect a pattern position with high precision that is not affected by P-P scratches.

[Brief explanation of drawings]

Fig. 112 is a pad pattern diagram showing a conventional pattern position detection method, Fig. 3 is a flowchart showing the processing process of the present invention, Fig. 4 is a circuit configuration diagram showing an embodiment of the present invention, Fig. S (a) , 佽) are operation principle diagrams showing the operation of the present invention in detail. 2♂・・・・・・−・−・−・・・−・−・・・Shift I! Number of cows/ri33・−・−・・−・・・・・・・・・・
・・・・・・・・・・・・・・・Data register 30.3
9.50・・・・・・−・Shift register 36・・・・
・・・・・・－・－・・・・・・・・・・・・・・・・
・・Subtraction circuit −′44・・・・・・・・・・−・・−
・・・・・・・・・・・・Multiplexer 55-
−−・−・−・・・・・・・・・・・・・・−・Addition circuit 6 t・−m−・−・・・・・−・=・・・・Division [1 Road 41 Applicant Fujitsu Limited Figure 1 Figure 2

Claims (1)

[Claims] l) Digitization circuit that converts the video signal into 3-value or 3-value data in a IIK device having an imaging system and brightens the output signal of the digitization circuit by a specific detection process. Then, a circuit for projecting the good signal output by the sand removal circuit, a comparison circuit for calculating the difference between the projection result and the projection data obtained in advance, and a comparison circuit for determining the minimum value of the comparison circuit 0III. At least two detection circuits detect the position as the pattern position.
A pattern detection device characterized in that the position of an O pattern is determined from a plurality of O patterns using .