JPH0897271A - Autofocus method - Google Patents

Abstract

PURPOSE: To make it possible to shorten remarkedly the working time to take for an autofocus of a dicer by a method wherein information on the thickness of a wafer is previously registered in a CPU, the focal point of a microscope is moved at a stretch from its initial position to a position to reach the error range of an irregularity in thickness and after that, the autofocus is accurately performed. CONSTITUTION: Previously registered information on the thickness (t1 ) of a wafer (In the case where the wafer is adhered on an adhesive tape, the thickness (t1 ) includes the thickness of the tape.) is outputted from a CPU to an alignment means 7, the focal point P of a microscope 7a is moved at a stretch from its initial position to the vicinity of the surface of the wafer W and after that, is step-fed at a resolution of about 2μm, for example, and each time the focal point is step-fed, a differential processing for finding the maximum value of the contrast of an image and the like are conducted on the basis of an image imaged by a CCD camera, whereby the time to take for an autofocus of a dicer is remarkedly shortened and the improvement of the productivity of the dicer can be contrived.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an autofocus method performed by a dicer or the like.

[0002]

2. Description of the Related Art A semiconductor wafer W on which a plurality of circuits such as ICs are formed as shown in FIG. 3 is divided into individual circuits R (IC chips) by a dicer. Prior to this division, the cutting line S (street) of the wafer is detected by the optical means, and precise alignment with the cutting blade is performed. The alignment by the optical means is performed by imaging a circuit on the wafer by a CCD camera forming the optical means and performing image processing. However, in order to improve the alignment accuracy, the alignment of the wafer of the microscope of the optical means is performed automatically. Focus is essential. For example, the auto focus is the focus of the microscope 2
The image is subjected to stepwise feeding with a resolution of .mu.m, and image processing such as differential processing is performed to obtain the maximum value of the contrast from the image (pixel density information) taken by the CCD camera each time (Japanese Patent Laid-Open No. 61-198204). .

[0003]

According to the conventional autofocus method as described above, if the distance to be autofocused is long, the number of times of imaging and image processing increases and a long time (tens of seconds) is spent, resulting in productivity. However, there is a problem in that The present invention was made in order to solve such conventional problems, focusing on the fact that the wafer shape, thickness, street spacing, etc. are standardized for each user depending on the type of semiconductor wafer, Information to the CPU
If you register it in advance in, the focus of the microscope is moved all the way to the position where the error range of the thickness variation is reached, and then the autofocus is performed precisely, so that the working time can be significantly shortened. The challenge is to provide.

[0004]

As a means for technically solving the above problems, the present invention provides a step of registering standardized semiconductor wafer thickness information in a CPU and a holding means such as a dicer. Positioning the held wafer directly below the optical means, moving the focus of the optical means to a position that reaches an error range of thickness variation based on the thickness information of the wafer, and within this error range The gist of the present invention is an autofocus method including a step of precisely performing autofocus.

[0005]

[Operation] The error range of wafer thickness variation is usually 5
0 μm or less, and even if the maximum is 50 μm, for example, 2 μm
The number of image processings with autofocus at resolution of 2 is 2
It is about 5 times, which is significantly less than the conventional one. Conventionally, a distance of about 2 to 3 mm is auto-focused, so in the case of a resolution of 2 μm, 1000 times to 1500 times.
The image processing has to be performed once, and at least about 20 seconds have been spent, but in the case of the present invention, the time spent for autofocus is only about 1 to 2 seconds, and the productivity can be improved.

[0006]

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings. FIG. 1 shows an example of a dicer for precisely grinding a semiconductor wafer or the like. A cassette 2 is placed in a cassette placing area 1 which moves in the vertical direction.
The wafer W (attached to the frame F via the adhesive tape N) is carried out from the inside to the standby area 4 by the carrying-in / carrying-out means 3 and is carried onto the chuck table 6 by the carrying means 5 having a swing arm. , After chucking the wafer W on the chuck table 6, the chuck table 6
Is moved to position on the alignment means 7, and after alignment, dicing is performed by the cutting means 8 provided with the rotary blade 8a.

The alignment means 7 has an autofocus function, and alignment is performed to align the cutting line S of the wafer W with the blade 8a of the cutting means 8. The autofocus is performed prior to the alignment.

In this auto focus, as shown in FIG. 2, the focus P of the microscope 7a of the alignment means 7, which is an optical means, is adjusted with respect to the wafer W with the frame F sucked and held on the chuck table 6 as described above. For example, 2 μm
The image is subjected to image processing such as differential processing in order to obtain the maximum value of the contrast based on the pixel density information digitally converted from the image obtained by imaging the circuit R etc. by the CCD camera (not shown) each time. Done.

If the distance L from the initial position of the focal point P of the microscope 7a to the surface of the wafer W is long, the number of image processings increases as described above, and it takes a long time. The thickness t of W is registered in the CPU in advance, and the number of times of image processing is reduced based on this thickness information to shorten the processing time.

That is, the thickness t of the wafer W registered in advance is
The CPU outputs information (including the thickness of the tape N when it is attached to the adhesive tape N) from the CPU to the alignment means 7, and moves the focus P of the microscope 7a from the initial position to the vicinity of the surface of the wafer W at once. After that, if the imaging is performed by the step feed as described above, the time spent for autofocus can be significantly shortened. That is, useless image processing for the distance L can be omitted.

However, even though it is a standardized wafer W, there is some thickness error. Therefore, in consideration of the error, the focus P of the microscope 7a is lowered to a position where it reaches the error range of thickness variation, and then To perform precise autofocus.

Incidentally, if the standard thickness t = 600 μm and the thickness error is ± 25 μm, the actual thickness t ₁ is 575 μm.
Since it is within the range of ≦ t ₁ ≦ 625 μm, it is sufficient to move the focus P from the initial position to 625 μm (distance from the upper surface of the chuck table 6) at a stretch, and then step-feed by 2 μm for image processing. Become. As a result, the autofocusing work is completed in 1 to 2 seconds, and the time can be remarkably shortened as compared with the conventional case.

Alignment is performed in a short time by the low-magnification image processing and the high-magnification image processing disclosed in the patent application (Japanese Patent Publication No. 3-27043) previously filed by the applicant. If the technique for performing the above is used together and applied to this, the alignment time can be further shortened.

[0014]

As described above, according to the present invention,
Since the wafer thickness information is registered in advance in the CPU, the focus of the microscope of the alignment means is moved all at once to a position where the thickness variation error range is reached, and then autofocus is performed precisely, so that the dicer The time required for autofocus can be significantly reduced,
This has the effect of improving productivity.

[Brief description of drawings]

FIG. 1 is a perspective view showing an example of a dicer.

FIG. 2 is an explanatory diagram showing an example of an alignment method according to the present invention.

FIG. 3 is a top view showing a wafer with a frame.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Cassette mounting area 2 ... Cassette 3 ... Carrying in / out means 4 ... Standby area 5 ... Conveying means 6 ... Chuck table 7 ... Alignment means 7a ... Microscope 8 ... Cutting means 8a
…blade

Continuation of the front page (51) Int.Cl. ⁶ Identification number Office reference number FI technical display location H01L 21/301

Claims (1)

[Claims] 1. A step of registering standardized semiconductor wafer thickness information in a CPU, a step of positioning a wafer held by a holding means such as a dicer directly under an optical means, and the wafer thickness information. An autofocus method comprising: a step of moving the focus of the optical means to a position that reaches an error range of the thickness variation at once, and a step of precisely performing the autofocus within the error range.