JPH0476450B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application] The present invention relates generally to a device for automatically adjusting the focus of an objective lens used for inspecting an object, and more particularly to a device for automatically adjusting the focus of an objective lens used for inspecting an object, and more particularly to a device for automatically adjusting the focus of an objective lens used for inspecting an object. Utilizing a multi-aperture dosing reticle and a pair of multi-aperture return masks;
Utilizing a corresponding detector that generates a control signal that is used to vary the physical distance between the microscope objective and the reticle or object to be inspected so as to move the microscope to a position of focus relative to the object to be inspected. It relates to electro-optical devices.

[Prior art]

2. Description of the Related Art In an automatic inspection apparatus for objects such as photomasks, it is important to include means for automatically focusing a lens used for inspection work. One conventional device for automatically focusing a lens for photomask inspection takes advantage of the fact that the magnification of a reticle seen through a lens system is a function of the distance of the reticle from the image plane of that lens. There is. In that device, as the reticle moves away from the lens and away from the lens' ideal focal point, a point within the periphery of the field of view moves inward toward the optical axis of the lens; , a point within the periphery of the field of view moves outward away from its optical axis. Thus, by focusing a narrow beam of light onto a point within the periphery of the field of view and then detecting the position of that point relative to a known position, the distance between the objective and the object to be examined can be determined. The device works well if the surface of the object being inspected is generally smooth, but it is not acceptable in the case of multi-level surfaces of the type present in silicon wafers on which integrated circuits etc. are formed. Accuracy cannot be obtained.

An automatic lens focusing apparatus configured for inspecting semiconductor wafers and the like is disclosed in U.S. Patent Application No. 582,584, filed February 22, 1984. The invention disclosed by that U.S. patent application is assigned to the applicant of the present application, and the U.S. patent application upon which this application claims priority is a continuation-in-part of the above-mentioned U.S. patent application.

[Summary of the invention]

Another object of the invention is to provide an improved autofocus device for use in wafer inspection equipment and the like.

It is an object of the present invention to provide the above-mentioned method which can focus with high precision on a target such as a patterned silicon wafer which has considerable local surface roughness and whose reflectivity varies. It is to obtain different kinds of equipment.

In summary, a preferred embodiment of the present invention includes a light source, a dual channel optical fiber bundle, a means for intermittent entry of light into the channels, and application of alternating images onto a reticle and an object to be inspected. an associated optical device for examining the return mask and the alternating reticle image, and an associated detector for driving the microscope in response to the output of the detector to focus the microscope of the device. and a control circuit that operates. This preferred embodiment also includes a flipping pupil to receive a specific objective of the microscope.

A major advantage of the present invention is that it provides a device that can focus on a patterned silicon wafer with many level changes and large contrast variations. This advantage is achieved because both channels return the same image when both channels are in focus. When the channels are subtracted from each other, the influence of the wafer pattern is eliminated. A particular objective of the microscope is therefore focused on the local average roughness of the surface area to be examined.

〔Example〕

Hereinafter, the present invention will be explained in detail with reference to the drawings.

First, refer to FIG. A microscope turret, indicated generally at 10, is held by a turret mounting and positioning assembly 12. As shown in FIG. The turret mounting and positioning assembly 12 is of the type disclosed in the above-mentioned pending U.S. patent application and is capable of mounting a selected one of the microscope objectives.
It is for positioning on the wafer W to be inspected. The wafer W is supported by a movable table indicated by the reference numeral 13 in the notched portion. Turret mounting and positioning assembly 12 includes a turret drive and positioning mechanism, indicated generally at 14, and a microscope imaging lens system, indicated generally at 16. The optical axis 18 of this device is connected to one of the objective lenses 20 of the microscope, the imaging lens 16, the field lens 22, 148
mm lens 24, dichroic beam splitter 26, 300mm lens 28, and pentaprism 3
0 and the input lens 32 of the television camera 34. It should be noted that the light is collimated between lenses 24 and 28. TV camera 34
and the operation of pentaprism 30 are also described in the aforementioned US patent application. However, if the present invention is an optical device that requires automatic focusing of a microscope,
It will be seen that it can be applied to any device.

A controller 36, responsive to signals generated by an autofocus preamplifier 38, controls the turret drive and positioning mechanism 14, as will be explained in more detail below.
Give a drive signal to.

The autofocus optical system, indicated generally by the reference numeral 40, includes means for forming a first optical path or axis 42. As shown in FIG. A first optical path 42 begins with a quartz halogen lamp 44 and a spherical mirror 46 and passes through a condenser lens 50 and a hot mirror 52. First optical path 4
2 is split through mirror 52 into a pair of initially separated parallel optical paths or channels by an aperture in chopper ring 54. These channels are fiber optic bundles 56 (channel A) and 58
(Channel B). Light emerging from the distal ends of fiber optic bundles 56, 58 is passed through a flipping pupil 60, a red filter 62, and a red filter 62.
It passes through a 50mm pupil lens 64 and a multiple aperture projection reticle 66. Then, the first optical path 42 is 5
% beam splitter 68 and 50% beam splitter 70
and imaging lens 72 before entering the optical path or optical axis 18 of the system at dichroic beam splitter 26.

A portion of the light passing through the reticle 66 is bent by a beam splitter 68 and sent to the detector imaging lens 74.
and a reflector 76 along a second optical path or optical axis 72 to a forward photodetector 78 .

Returning from wafer W, the light reflected along optical path 42 is bent by beam splitter 70 and
Another 50% beam splitter 80, a first return mask 88, a detector lens 90, and an S-CURVE (S-CURVE).
-curve) along the third optical path or optical axis 79 including the detector 92; The light returning along this third optical axis 79 is bent by a beam splitter 80 and continues along a fourth optical path or optical axis 81 that includes a second return mask 82 and a detector lens 84. , SUM (sum) detector 86 .

Note that a portion of fiber optic bundle 56 is split at 94 to send light to channel A selector 96. Detector 78, 86, 9
All outputs of 2,96 are input to preamplifier 38.

Generally speaking, the autofocus optics 40 alternately deposits a pair of red light images of the axially offset projection reticle 66 onto the wafer W being inspected, as will be explained further below. These axially offset images are passed back through the device along optical axes 79 and 81 and focused onto a photodetector by means of a slotted mask. The position of the slotted image relative to the associated mask will then indicate the state of focus of the microscope.

As mentioned earlier, this autofocus device is 100W
A quartz halogen lamp 44 is used as a light source. The light from this source is collected by spherical mirror 46 and then focused by mirror 52 and chopper ring 54 onto the entrance ends of fiber optic bundles 56, 58 (FIG. 2). Outlet ends 57, 59 of the fiber optic bundle
is placed at the conjugate image (pupil image) of the rear aperture of the microscope objective. This conjugate image is created by placing the exit end of the fiber optic bundle at the focal point of one of the pupil lenses 64 with a focal length of 50 mm so that the light exiting the fiber bundle passes through the pupil lens 64 into a plane. , is imaged. The light then passes through a projection reticle 66 placed at the other focal point of the pupil lens. This reticle 66 is 100
It is also placed at one focal position of the mm autofocus imaging lens 72. The autofocus imaging lens 72 collimates the light passing through the reticle. red filter 6
The collimated light, containing only wavelengths longer than 650 nm as a result of passing through the beam splitter mount 26, enters the microscope apparatus through a dichroic mirror 27 forming part of the beam splitter mount 26. The dichroic mirror 27 reflects light with a wavelength longer than 650 nm and transmits light with a shorter wavelength. Since the light passing through the reticle 66 is parallelized when it enters the optical path inside the microscope, the 148 mm microscope lens 24 and the 300 mm microscope imaging lens 28
In the parallel light region in between, the reticle image is focused on the wafer W when the microscope is in focus.

As best shown in FIG. 3, the exit ends 57, 59 of the optical fiber bundles 56, 58 are covered by relatively large fixed apertures 61.
Its aperture 61 is 5x, 10x, 20x and
It is slightly smaller than the conjugate image of the rear aperture of the 50x objective. The rear aperture of the 100x objective is much smaller than the others, so when the 100x objective is used, the smaller aperture 63 is placed above the larger aperture. The dimensions of this aperture are slightly smaller than the conjugate image of the rear aperture of the 100x objective. The assembly that flips the smaller aperture into and out of the optical path is called a "flupil" (flipping pupil).
pupil).

In this embodiment, a tipper ring 54 is placed directly in front of the fiber optic bundle, and two sets of openings 52
a, 53b successively alternately uncover each fiber optic channel (see also FIG. 2). When the Chiyotsupa wheel 54 is rotated at 1800 RPM,
The configuration is such that the light passing through the aperture is switched from "channel" to "channel" at a rate of 1600 Hz. Furthermore, only one channel is illuminated at any given time. At any one time, only the top or bottom half of the rear aperture of any objective will be illuminated. To that end, as the objective lens of the microscope is moved up or down relative to the wafer W, i.e., as the objective lens is moved in or out of its focus, the dosing reticle 66 associated with each fiber optic channel is The return image will be moved down or up. In other words, when the channel A image moves up, the channel B image moves down, and when the channel A image moves down, the channel B image moves up.

As mentioned above, 50% of the autofocus red light returned from the wafer W is placed in the beam splitter 7 between the imaging lens 72 and the dosing reticle 66.
0 from the main beam. The light is then split by a beam splitter 80 into two equal beams, each beam forming an image of the reticle on a return mask 82, 88 in an alternating manner.

As the microscope objective moves up or down from the focal position, each return image is reflected by a return mask 82,8.
It will be seen that there is a transverse movement (relative to the longitudinal axis of each aperture) on each of the eight apertures. The light passing through each return mask is focused by detector lenses 84 and 90 onto silicon photodetectors 86 and 92, respectively.

Next, refer to FIG. This figure shows how the preferred embodiment of the projection reticle consists of five transparent bars 67 within an opaque field of view. Images of those five bars are focused on an area on the wafer. The height of that area is 34
is equal to the width of the field of view, and the width of the area is three times its length. However, it should be noted that the shape and dimensions of the aperture are not limited thereto.

Each return mask 82, 88 (FIG. 5) consists of a similar array of transparent bars 87 on an opaque background. When the microscope is focused, the reticle images from both channels coincide and are shown at 8 in FIG.
9, aligned to the S-CURVE return mask. As the objective lens moves upward from its focal position, the reticle image associated with channel A moves upward (the reticle image from channel B moves downward) and is therefore incident on the S-CURVE detector 92. Light from channel A (and B) increases. As the microscope objective moves down, the image of channel A moves down (and the image of channel B moves up), reducing the light incident on the detector.

As explained in more detail below, the current output of S-CURVE detector 92 is routed through a current-voltage amplifier. Therefore, the voltage output of the S-CURVE detector/amplifier exhibits a variation with the position of the focus approximately as shown in FIG. When the objective lens is almost in focus, the B channel image moves in the opposite direction to the A channel image, so the S-CURVE is opposite to the channel B signal.

When the detector/amplifier outputs are multiplexed by chopper wheel 54 and one voltage is subtracted from the other, the result is a curve similar to that shown in FIG. Due to background light levels that are constant in one channel but different between channels, the curve does not pass through zero when in focus. This focus bias is later subtracted by an autofocus controller. Note that when the microscope is in focus, the return images of both channels are identical, so the return light is independent of the pattern on the wafer. This is very important because the reticle image returned from the wafer includes not only the reticle pattern but also the wafer pattern.

The second return mask is a SUM mask where the projected bar is inserted into the transparent slot 8 of the SUM return mask 82.
S-7, except that the S-
Same as CURVE mask. Therefore, the light in both channels behaves the same way. The lights are added together to form the signal shown in FIG.

The intensity of the autofocus projected light is S-CURVE
In order not to affect the levels of the and SUM signals, and to eliminate the effects of differences in the efficiency of the optical fibers of channels A and B, the device includes a forward detector 78. As previously mentioned, 4% of the light passing through the projection reticle 66 is split from the main projection beam by the plate beam splitter 68 and transmitted to the detector 78 by the third detector lens 74.
I can focus on. The output of detector 78 is converted to a voltage and split into the S-CURVE and SUM detector/amplifier voltage outputs. The signal from the front detector 78 is also used to synchronize multiplexer signals in the autofocus preamplifier.

A fourth detector 9 to facilitate synchronization of the device.
6 is included. Several optical fibers 94 are separated from channel A fiber bundle 56 and used for direct input to channel selection detector 96. Forward detector 78 records correctly when any channel is on, but channel selection detector 96 records when only channel A is on. The voltage output of the autofocus preamplifier is summarized as follows:

There is a certain background associated with each Vsa, Vsb, Vsuma, and Vsumb. They can be represented by the following definitions.

Vi=Vi′+Vi″, i=Sa, Sb, suma, sumb (1) Here, Vi′ is the part of Vi that depends on the focal position, and Vi″ is the scattered light and the part that is independent of the focal length of the objective lens. The associated background signal is due to curvature from the lens surface.

(Vi′ and Vi″ are of similar magnitude) Also, if Fa and Fb represent the forward detector signals for channels A and B, respectively, then S−
The CURVE signal is given by S−CURVE = Vsa′+Vsa″/Fa−Vsb′+Vsb″/Fb (2), and the SUM signal is given by SUM=Vsuma′+Vsuma″/Fa +Vsumb′+Vsumb″/Fb (3). Given.

As explained in the specification of the aforementioned pending US patent application, when the apparatus is in macro mode, there is no wafer under the microscope and the S-
All that is read from the CURVE and SUM detectors is the background signal. This signal is a function of the objective lens magnification and is otherwise constant.

S−CURVE(MACRO)=Vsa″/Fa−Vsb″/Fb (4) and SUM(MACRO)=Vsuma″/Fa+Vsumb″/Fb (5) are obtained. Substituting equation (4) into equation (2) and equation (5) into equation (3), S−CURVE=Vsa′/Fa−Vsb′/Fb+
(S-CURVE(MACRO)) (6) and SUM=Vsuma'/Fa +Vsumb'/Fb+(SUM(MACRO)) (7).

The autofocus controller controls the S-CURVE for each objective lens when the instrument is in macro mode.
(MACRO) and SUM (MACRO) are recorded, and when the device is in focus, S-CURVE and SUM are recorded.
Subtract them from the output of . As a result, Vsa′/Fa−Vsb′/Fb=(S−CURVE)
−(S−CURVE(MACRO)) (8) and Vsuma′/Fa+Vsumb′/Fb = SUM−SUM(MACRO) (9) are obtained.

The final feedback signal is obtained by dividing equation (8) by equation (9) to compensate the signal for wafer reflectivity variations and background signals. That is, S−CURVE(NORM)=Vsa′/Fa−Vsb′/Fb/V
suma′/Fa−Vsumb′/Fb = (S−CURVE)−(S−CURVE(
MACRO))/SUM−SUM(MACRO)(10) Next, refer to FIG. This figure is a functional block diagram of autofocus preamplifier 38. This autofocus preamplifier is a focus (S-CURVE) detector 92
and a SUM detector 86, and generates a difference (S-CURVE) output and a SUM (reflectance) output from these signals to output terminals 100 and 102, respectively, and sends these outputs to an automatic focus controller 36. (Figure 1). Processing of the two input signals to the autofocus controller is performed in substantially the same manner. The output of the focus detector 92 consists of alternating current pulses corresponding to each of the two interrupted pupils separated by a gap such that the two interrupted pupils are "off", and the current-to-voltage converter is input to a chopper stabilizing amplifier 104 configured as a. As a result, a voltage waveform W1 is generated at the point shown. (See also Figure 10).

The basic function of an autofocus preamplifier is to determine the difference between two time-multiplexed values of its signal. However, since the two and the pupil are not perfectly balanced in intensity, the focus signal must first be normalized with respect to the intensity of the forward traveling light. This is done using the output of the front sensor 78. The forward sensor measures a small fraction of the light traveling forward. As shown in FIG.
The output of 8 is input to a current-voltage converter 106 to generate a waveform W3 (FIG. 10). Focus signal W1 is then divided by signal W3 to normalize the focus return signal to the forward pupil intensity.

Dividing signal W1 by signal W3 is performed by analog divider 108. To prevent the device from saturating when both pupils are off (divider input equal to zero), a hold function 110 is added to the divisor input to hold the previous channel value during the off period. . This generates a normalized focus signal W5 (FIG. 10). In addition to parallelizing the two channels, this normalization also makes the device independent of the brightness of the focal illuminator.

A difference signal is calculated for each pair of sensor inputs.
A fourth sensor 96 (A-selection) is used to indicate which pupil is active at any given time.
To generate waveform W4, that sensor is buffered like the other sensors. The waveform W4 is
Used in conjunction with forward signal W3 to control the timing logic for the sum and difference circuits. The normalized focus signal is provided to a balanced modulator 112. The balanced modulator has a gain equal to minus one for the first channel of the pair of channels. Then the output W7 is the integrator 11
4, the result is integrated over a time determined by timing logic 116 and the result is temporarily held. second of a pair of channels
For the first channel, the modulator gain is switched to +1 and the signal W7 is re-integrated for the same integration time as for the first channel. Therefore, the output of the integrator gradually rises for the first channel and falls gradually for the second channel. The net output of the integrator after this processing represents the difference between the two time multiplexed signals. That value is sampled and held by a sample and hold circuit (switch 118 and capacitor 120) and then amplified by amplifier 122 to produce an S-CURVE (difference) output at terminal 100. Integrator 114 is then reset and the cycle repeats. This integration/sampling/resetting process minimizes the effects of sensor noise and preamplifier noise.

The SUM output produced at terminal 102 indicates that the normalized signal is integrated in the same direction for both channels to produce a signal that in this case represents the sum of the individual time-multiplexed signals. Therefore, S-
Generated in the same way as the CURVE (difference) output.

Reference is now made to FIG. 11, which is a block diagram showing the major components of autofocus controller 36. It can be seen that the autofocus controller 36 processes the S-CURVE and GUM signals generated by the preamplifier 38 and the encoded position signal to drive the objective lens mount to the focus position. Dew. Operation of autofocus controller 36 is initiated by instructions sent via a serial data link from a main computer (not shown). This autofocus controller is
In addition to the functional components shown in FIG.
Contains 0.

Rotation of the shaft of the DC motor used to drive the microscope positioning mechanism 14 is monitored by an incremental angle encoder (not shown). This incremental angle encoder provides a biphasic combined pulse to the encoder logic 122 of the autofocus controller. The logic then stores the 16-bit absolute position count on the data bus 12.
4 to the microcomputer 120. 2000 counts corresponds to one revolution of the motor shaft. Rotation of that axis is translated into a 20 micron movement of the objective lens mount via a micrometer drive mechanism and lever arm included in device 12 (FIG. 1).

The SUM signal is applied to input terminal 126 as a differential signal. The differential signal is converted to single-ended form by amplifier 128 and then converted to an 8-bit digital signal by analog-to-digital converter 130.

The S-CURVE signal at input terminal 132 is also converted from digital to single-ended form by amplifier 134. The signal is then provided to an analog divider 136 in which the S-SURVE offset voltage is subtracted therefrom and the difference is divided by the sum normalized voltage. The S-CURVE signal thus "conditioned" is then converted to a 10-bit digital signal by bipolar analog-to-digital converter 138. As shown, the S-CURVE offset signal and the SUM normalization signal are set by microcomputer 120 and converted to analog form by digital-to-analog converters 140 and 142, respectively. Analog-digital converter 1
generated at 30,138 output terminals, respectively.
The SUM and S-CURVE signals are read by microcomputer 120.

Autofocus controller 36 generates an analog motor drive signal at output terminal 146 via an 8-bit D/A converter 144. This signal generated is applied to another power amplifier and then to a DC motor that drives the objective lens mount 12 via a mechanical linkage as previously described. The motor movement is fed back as changes in the encoder and/or changes in the S-CURVE signal so that the microcomputer can perform closed loop servo control.

Reference is now made to FIGS. 12 and 13. It will be appreciated that the autofocus device of the present invention employs two types of control loops. FIG. 12 shows the position control loop used by the controller to monitor and control the encoder counter to a target value. Optical limit sensors (not shown) monitor the upper and lower limits of lever arm travel within the autofocus device 12 and encoder counts are initialized with reference to the upper limits. In order to stabilize the control loop, the microcomputer configures a second type of digital filter through software.

FIG. 13 shows the S-CURVE control loop. This servo loop can only operate in the linear region of S-CURVE. The specimen curve for each objective shows that the slope of the curve becomes steeper with objective and magnification. The amplitude of the curve is also proportional to the reflectance of the wafer surface. Because of these factors, if no compensation is performed, the open loop gain of the focus servo will be changed. Therefore, the microcomputer sets the gain of the software digital filter. This compensates for changes in the slope of the curve due to the magnification of the objective lens.
The change in reflectance is normalized by dividing the incoming S-CURVE by the SUM normalization voltage. SUM
Since the signal obtained by is also proportional to the reflectance of the wafer, the output from the analog divider is the reflectance normalized S-CURVE signal.

The resulting input to the S-CURVE analog-to-digital converter is (S-CURVE(NORM) = (S-CURVE) - (S-CU
RVE(MACRO))/SUM−SUM(MACRO)). Their "macro" values are first subtracted. This is because these "macro" values are offsets resulting from optical ghosts that are not related to the signal reflected from the wafer.
This autofocus servo loop, in contrast to the position (encoder) loop, includes additional elements and thus:
For stability, the compensating type 2 digital filter has poles and zeros arranged differently. An advantage of software digital filters is that the coefficients can be easily changed to optimize the configuration of the device.

In operation, automatic focus controller 36 responds to commands from the main computer to locate a focus point and move the objective lens mount to a desired (target) position (e.g., if the objective lens does not interfere with the wafer) in order to locate and orient the focus point to that focus point. or perform any other actions required to make this autofocus device work with the rest of the mechanism. Most of these operations are simple, but the "finding focus" routine and the routine for maintaining that focus position must be mentioned.

All functions of the autofocus controller occur within a repeating time cycle of approximately 700 microseconds. Therefore, for each cycle the autofocus controller
The A/D converter, S-CURVE/A/D converter, position encoder, and input command line are sampled.
The automatic focus controller also outputs a drive signal for the DC position control motor and updates the output status line. The timer cycle has separate time sampling periods for the digital filters used in the position and focus loops.

The operation of the focus finding routine is shown in FIG. In FIG. 14, the vertical axis represents the vertical position of the objective lens mount. Also shown is the S-CURVE signal. The horizontal time axis represents the amplitude and direction of the drive voltage. The focus search begins from a retracted position above the start of the S-CURVE. The movement of the objective lens mounts is then plotted relative to their axes.

Once you receive a command that allows you to find your focus,
The autofocus controller applies the highest drive voltage to lower the objective toward the S-CURVE. S-
When the autofocus controller detects a zero crossing of CURVE, the autofocus controller records the corresponding encoder position as the desired focus position. At the same time, the maximum driving voltage is first applied in an increasing direction to stop the objective lens mount, and then the movement of the objective lens mount is reversed. At this point the autofocus controller records the stop position and calculates the target position between the stop position and the focus. The automatic focus controller then applies a full drive voltage to drive the objective lens mount to the target position, and a final downward drive voltage to bring it to a final stop at the focus. At this point the autofocus controller exits open loop control mode and
Close the focus loop to servo control the focus position. In practice, the above operation is a little more complex in order to correct for factors such as elastic elongation of mechanical parts and additional zero-crossings of the S-CURVE.

When the focus loop is closed at the selected filter value, the autofocus device servos at the null position on the S-CURVE. However, there is a definite misalignment between the optimal focus position measured by the autofocus device (infrared) and the focus position monitored by the television camera (in visible light).
This positional shift is called color shift, and differs depending on the objective lens. The color shift is measured for each objective during initial calibration. Once the focus position is reached, the automatic focus controller moves the objective lens mount to its calibrated color shift position. Another offset may be applied to control the height of the objective lens relative to the focus of the S-CURVE. This is useful, for example, when viewing features deeper than the depth of the field of view at 100 magnification.

[Brief explanation of the drawing]

1 is an exploded perspective view of the autofocus device of the present invention; FIG. 2 is a diagram detailing the illuminated end of the optical chopper ring and associated optical fiber bundle; and FIG.
The figures are diagrams showing details of the dual aperture pupil mechanism used in the present invention, Figure 4 is a diagram showing the projection reticle of the invention, and Figure 5 is a diagram showing the image of the projection reticle when the microscope is in focus. Diagrams showing how the S-CURVE masks are aligned; FIGS. 6, 7 and 8 are curves showing the operation of the autofocus device of the present invention; FIG. 9 is a diagram showing how the S-CURVE masks are aligned; 10 is a waveform diagram of a signal at a specific point in the block diagram of FIG. 9. FIG. 11 is a block diagram showing the main functional parts of the automatic focus controller of the present invention. The block diagrams shown in Figs. 12 and 1
FIG. 3 is a functional block diagram showing the control loop for position feedback and focus return, and FIG. 14 is a diagram showing the focus finding operation of the present invention. 14...turret drive and positioning mechanism, 2
6, 68, 70, 80...beam splitter, 34...
...TV camera, 38...autofocus preamplifier,
44... Light source, 54... Chiyotsupa ring, 56, 58
... optical fiber bundle, 60 ... pupil, 66 ... projection reticle, 74 ... detector imaging lens, 76 ... reflector, 78 ... forward light detector, 82, 88 ... return mask, 86 ... ...SUM detector, 92...S-
CURVE detector, 102...Chopper stabilization amplifier, 106...Current-to-power converter, 112...Balanced modulator, 114...Integrator, 120...Microcomputer, 122...Encoder logic, 1
30,138...analog-digital converter, 1
40,142...Digital-to-analog converter.

Claims (1)

[Scope of Claims] 1. A microscope of the type used for inspecting objects such as semiconductor wafers, having an optical inspection axis extending between an inspection instrument and an electrically controllable microscope device. In an autofocusing device of an optical inspection apparatus used, means for forming a first optical axis having a light source at one end and a first beam splitter arranged along the optical inspection axis at the other end; , for transmitting alternately spatially separated bursts of light rays to pupil means arranged at the conjugate image of the rear aperture of the objective lens of the microscope in the first optical axis; means disposed in the optical axis of the pupil means; and said pupil means having said aperture means such that each burst of light passing through said pupil means projects its image through an aperture means onto an object to be examined under a microscope. a reticle means disposed between the first beam splitter and the pupil means in the optical axis of the reticle; a second beam splitter disposed between the reticle means and the first beam splitter; , a third beam splitter, and an electronic controller, the first beam splitter sending light from the first optical axis to a microscope and reflecting light returned from the microscope. and the second beam splitter reflects light containing an image of the aperture means returning from the object to a second optical axis between itself and the S-curve detection means; In the second optical axis, half of the image of the aperture means is passed to the S-curve detection means when the object is in focus, and more than half of the image of the aperture means is passed when the object is out of focus. disposing a first return mask that passes a portion or a smaller portion to the S-curve detection means; said third beam splitter is arranged between said second beam splitter and said first return mask from said object; It is arranged to guide a part of the returned light to a third optical axis having a SUM detection means at its tip, and when an object is focused on the third optical axis, the aperture means A second return mask is arranged that allows only a part of the image to pass through the SUM detection means when the image is out of focus, while leaving the entire image intact, and the electronic controller is configured to control the output signal from the S-curve detection means. and the output signal from the SUM detection means, and outputs a drive signal for controlling the positioning of the microscope with respect to the object being inspected so as to focus the microscope on the surface of the object being inspected. An automatic focusing device for an optical inspection device, characterized in that: