Surface inspection scanning system and method

(42) Date of commencement of the patent: 15.10.1991 (45) Date of issue: 25. 5.1992

<td>(30) Priority:</td><td>(73) Patent owner:</td>

<td>12. 5.1981 US 262866 claims.</td><td>EASTMAN KODAK COMPANY</td>

<td>(56) Documents: DE-OS 2936689 EP-A1-0008010 US-PS 3922093 US-PS 3971956</td><td>14650 ROCHESTER (US).</td>

(54) METHOD AND DEVICE FOR EXAMINING THE POLISHED SURFACE OF AN OBJECT

QQ

AT 394 632

MR 0078318

AT 394 632 B

The invention relates to a method for examining the polished surface of an object for small surface defects, wherein the article is transported at a predetermined speed along a test surface and its polished surface is repeatedly scanned along a scan line perpendicular to the transport direction with a narrow beam of electromagnetic radiation and wherein the presence and type of surface defects is determined at least from the scattered radiation reflected from the surface . wherein a first channel output signal is generated, which is representative of the scattered radiation reflected from the surface.

Furthermore, the invention relates to a device for carrying out the method, consisting of a test area, a transport device for transporting objects at a predetermined speed over the test surface, a scanning device for repeatedly scanning a scan line extending transversely to the transport direction over the test surface with a tightly focused beam of electromagnetic radiation, a first channel device with a detector, which is designed to detect scattered electromagnetic radiation which is present on the polished surface of the object and to emit a first channel output signal corresponding to the scattered radiation.

In the manufacture of semiconductors, silicon plates are used as a substrate for solid state electronic devices such as transistors and integrated circuits

The occurrence of various surface defects, such as dust, dirt, crystalline imperfections, scratches, haze and depressions, in the micrometer range, are extremely disadvantageous to the production process of such components and negatively affect the yield of individual components.

Currently, most manufacturers use manual, optical inspection techniques to control silicon plates and integrated circuits. In a manual process, the plate surface is irradiated with an intense light source. An operator sets the silicon plate at a suitable angle to the light beam and observes the plate surface for several seconds to determine the surface finish of the plate. Although this method is sufficient to detect dust and dirt particles as well as scratches of the order of 5 μτη, surface defects of smaller size are not detected

In an article entitled Laser Scanning Method for Evaluating the Surface of D. Electronic Materials. R. Oswald and D. F. Monroe, published in the Journal of Electronic Materials, Vol. 3, no. 1,19474, pages 225 to 241 there is a description of a method using a laser beam for surface scanning of silicon plates. In this case, a beam of electromagnetic radiation of a laser of 3 mW is first spread from its original diameter to a larger diameter and then directed to a vibrating mirror, which directs the beam at 90 ° in a lens. The axis of rotation of the mirror lies in the front focal plane of the lens and intersects its optical axis. The plate surface to be examined is placed on the other side of the lens at the focal point. The focused beam is perpendicular to the disk surface and is deflected in a line across the same, wherein the plate transverse to the scanning direction of the beam or this line moved forward.

If there is no error, all light is reflected perpendicularly from the surface along the direction of incidence. However, when the incident light encounters an error, it is scattered at the defect so that the spatial area immediately surrounding the reflected beam contains the stray beam energy. In the beam path between the laser generator and the oscillating mirror, a mirror aperture is arranged, which allows the passage of the beam from the generator to the oscillating mirror, but reflects the reflected scattered radiation.

The scattered radiation reflected by the mirror aperture is focused by a lens onto a detector. The changes in the stray radiation incident on the detector are used to indicate surface defects of the silicon plate.

To display the situation or to accumulate the occurrence of errors, a storage oscilloscope and a counter are provided.

This known device is satisfactory for detecting and counting errors whose diameter is greater than 8 gm and in some cases as low as 1 gm, provided that optimum conditions exist for scattered radiation.

Despite efforts to develop automatic test methods that have led to prototypes and laboratory facilities with limited success, the industry has not yet accepted such facilities as a replacement for the manual inspection procedures. The basic needs of the semiconductor manufacturers are in fact in a complete test device that detects tiny errors (1 gm and larger) of silicon plates, but also allows a classification and automatic discrimination of such errors and the examined plates as in order or depending on the Manufacturer or customer required tolerances as committee is located

The types of defects of interest to the industry are defined by the American National Standard Institute, Inc. ASTM F 154-76 and classified as: scratches, pits, embedded abrasive grains, haze, soils, edge chips, cracks, goose feet, or crow's feet , Craters, scars, grooves, caterpillars, orange peel, saw marks and stripes.

The method for producing silicon plates comprises the steps of sawing off a disk from

-2AT 394 632 B drawn block, polishing, cleaning and further processing of the disc. During these operations, one or more of the aforementioned errors may arise. In some cases, for. B. if only dust occurs, it is not necessary to excrete the entire disc, but only necessary to repeat the cleaning. On the other hand, the disc can be put in order if z. B. the defects are dimples, haze or scratches by repeating the polishing and cleaning. Coarser defects, such as saw marks or other pronounced deviations, usually lead to the rejection of the relevant target.

In EP-A1-0 008 010 a method for detecting surface defects is described in which the surface is scanned several times with a narrow beam of electromagnetic radiation and the occurrence of Obeflächenfehlem by evaluation of the reflected radiation and the reflected scattered radiation is determined. This produces a first channel output indicative of the surface aberration-induced reflected scattered radiation and a second channel output signal indicative of the surface-aberrations of the mirrored radiation. The narrow beam is passed several times over a predetermined portion of a tube while simultaneously advancing and rotating the tube so that the entire tube surface is scanned along a helical line. The output signals of the two channels are amplified and converted from their analog form into digital form, each sample signal corresponding to longitudinal scanning by the beam being divided into a large number of discrete values, each representing a predetermined location of the sample. For each sample point, a dynamic average is calculated by averaging the discrete values ​​associated with the same location from a predetermined number of previously derived samples. Each time a new sample line is sampled and divided into discrete values, each of these values ​​is compared to the corresponding dynamic average Signal formed, which is characteristic for the surface texture.

Although this method is satisfactorily applicable in the field of nuclear engineering of certain steam and fuel pipes, it is not sufficiently accurate for the examination of silicon plates for minute surface defects of the order of 1 pm. One reason for this lack of accuracy is that the stored and used to calculate the dynamic average scanning signals are not correlated with the respective locations on the pipe surface in a special way.

The aim of the invention is the elimination of said defect and the provision of an examination method and a test device for detecting the smallest surface defects in the order mentioned and also for the classification of the detected errors according to their nature.

Another object of the invention is to provide an automatic inspection device for detecting various surface defects of an object and evaluating the surface depending on the number and type of defects detected.

Another object of the invention is to provide an inspection method and apparatus for detecting surface defects of the order of 1 pm and for determining whether the object being inspected is in order, needs to be cleaned or polished again, or to be precipitated

These goals are achieved on the one hand by a method of the type specified, characterized that according to the invention a second channel output signal is generated, which corresponds to a reduction of the reflected radiation caused by surface defects, in that the second channel output signal is compared with a first predetermined threshold value and a signal having a predetermined level is then generated in a corresponding first threshold category channel in each case, when the second channel output signal crosses the first predetermined threshold, that portions of the second channel output signal having frequency components within a predetermined frequency range are compared to a bandpass threshold and a predetermined level signal is generated in a corresponding second threshold category channel at each pass of the portions of the second channel output signal through the bandpass threshold; in that the first channel output signal is compared with a plurality of predetermined threshold values ​​and a corresponding signal of a predetermined level in a corresponding threshold category channel is respectively generated on each passage of the first channel output signal through the corresponding threshold value, that the threshold category channels are scanned several times during each scan and during several consecutive scans, and therefrom several scanned area units are derived on the surface of the examined object, that the output signal levels occurring in the sampled threshold category channels in the respective threshold category channels, in which the output signal levels occur, converted into assigned digital values, in that a running digital value is compared with the digital value stored at the same time within a corresponding scanning area unit and the larger of the compared digital values ​​is stored as the value indicative of the scanning area unit, in that the occurring identically designated scanning area units are summed and the surface quality of the object is determined according to preset summation values.

On the other hand, these objects are achieved with a device of the type described above according to the invention by a second channel device with a detector which detects the reflected from the polished surface of the article scattered electromagnetic radiation and to deliver one of surface defects caused by the polished surface reduction of the mirrored radiation second

-3AT 394 632 B

Channel output signals is formed, a first comparator, formed on the second channel means and for comparing the second channel output signal with a first predetermined threshold value and for outputting a predetermined level output signal in a corresponding first threshold category channel at each passage of the second channel output signal through the first predetermined threshold value; a second comparator, connected to the second channel device and for comparing those transmitted parts of the second channel output signal, have the frequency components lying within a predetermined frequency range, is formed with a bandpass threshold and for outputting a predetermined level output signal in a corresponding second threshold category channel at each pass of the transmitted portions of the second channel output signal through the bandpass threshold; a third comparator, adapted to be connected to the first channel means and adapted to compare the first channel output signal with a plurality of predetermined threshold values ​​and to output a corresponding output signal of predetermined level in a corresponding threshold category channel at each passage of the first channel output signal through a respective threshold value; a timed scanner, which is designed for multiple sampling of the threshold category channels during each scan and during several consecutive scans and for the determination of a plurality of scan area units, a converter for converting the output of the sampled threshold category channels into digital values ​​corresponding to the respective threshold category channels, in which the output signals occur, a memory for storing these digital values, a fourth comparator for comparing a current digital value with the digital value stored in memory at a same time within a corresponding sample area unit and for outputting the larger of the compared digital values ​​and for storing the same as the corresponding sample area unit value; and summing and determining means for summing the occurrence of identically marked scanning area units and for determining the surface quality of the object in accordance with preset summation values.

As explained below, the device can be installed in a system comprising an automatic transport system, an optical inspection station and a signal processing unit with electronic circuits for analyzing the examination results and for controlling the sorting of the individual silicon plates.

The transport system leading through the system delivers the individual silicon plates from storage cassettes to the examination site and, depending on the examination result, to individual quality cassettes. The transport system has feed conveyor modules for automatically loading the plates, a conveyor belt for transporting the plates through the inspection station, and discharge conveyor modules for automatically sorting and placing the plates in the respective quality cassettes, which are labeled OK, clean, polish, and reject.

The optical inspection station has a low power helium-neon laser focused on a plane that matches the surface of the transported silicon plates. The laser beam is repeatedly deflected along a line transverse to the transport direction of the plates to ensure 100% inspection. The reflections from the plate are detected by a fiber optic dual receiver, wherein the receivers are spaced to each reflect the reflected mirror radiation or To detect scattered radiation. The receiver for the mirror radiation has a narrow slit, so that any deviation in the parallelism of the surface causes a deflection of the mirrored beam with respect to the slit and a decrease in the amount of energy reaching the receiver. The scattered radiation receiver is disposed approximately perpendicular to the disk surface and receives a portion of the light scattered from particles on or off the surface.

The light received by the fiber optic receivers is passed through dichroic filters to turn off ambient light other than the laser wavelength. The light is then fed to a pair of photomultiplier tubes whose electrical output signals are supplied to the electrical signal processing unit.

The signal processing unit has analog circuits for receiving the electrical output signals of the photomultiplier tubes and for providing seven channels of real-time information in the form of seven-bit addresses. The information in the seven channels is derived by comparing the electrical output signals with different thresholds for stray radiation; In the case of the mirror radiation, this is done by Ausfiltem different frequency components and comparison of these components with predetermined thresholds.

Three of these seven channels contain information obtained from the output of the scattered photomultiplier tube by comparison with three different thresholds. The remaining four channels contain information obtained from the output of the mirror-photomultiplier tube by filtering out individual frequency components and comparing them with different threshold values ​​in the case of three channels or Comparison of the output signal with the threshold value can be obtained in the case of the fourth channel.

The seven-bit address is fed to a logic network configured to create a digital surface map of the disk under test. The scanned plate is divided into a large number of small area units, each of which is assigned a binary code characterizing the most serious error detected on the area unit. Those area units on which no error is detected, z. B. characterized by a coding with all zeros.

The location and error coding for each area unit is temporarily stored in a memory in which it is accessible as information essential to the quality and sorting of the disk.

Depending on the user's instructions, an algorithm for assessing the disk under test is used to compare the accumulated number of stored error codes according to a priority program. The number of codes of the most serious errors are compared with the number of proper disks and added to the total of the codes of the lighter errors, Then, if this number is within the number of proper disks, the number of codes of less serious errors is compared with the number of proper disks for this error code and added to the total of the number of smaller errors. if this number is within the number of proper plates this procedure will continue until one of the correct numbers is exceeded or all error codes are checked thoroughly and none of the correct numbers have been exceeded. In the first case, the plate is evaluated as a scrap or queued for re-polishing or cleaning, and in the second case, the plate is classified as being proper

In addition to allocating the plates to the corresponding quality cassettes, the means for optically indicating the occurrence of defects on the particular plate being inspected and for printing a chart of each plate is adapted to elaborate diagnostic information with which the causes of the defects during manufacture or can be eliminated before the examination.

The invention will be explained in more detail below with reference to a preferred embodiment, which is shown schematically in the drawing; Fig. 1 a perspective view of the device according to the invention, Fig. 2 a block diagram of the entire device, Fig. 3A is a block diagram of the dark channel analogue circuit, FIG. 3B is a block diagram of the HellkanaT 'analog circuit, FIG. 4 a view of the scanning matrix of the device, Fig. 5 a block diagram of a Randfolger algorithm, which is provided in the inventive method, Fig. 6 a block diagram of an algorithm for error identification, which is applied in the inventive method, and Fig. 7A and 7B together form a block diagram of the classification algorithm used in accordance with the invention.

The embodiment shown in Fig. 1 of the examination device (10) according to the invention comprises a table (12) on which a transport system (14) and an optical examination place (16) are mounted.

The silicon plates to be examined (W_s) are stored in a conventional supply cassette (18) which can hold up to 25 individual plates (W_s) contains. The cassette (18) is placed on a vertical conveyor (20), which is commercially available and represents an industrial object. The vertical conveyor (20) advances the cassette (18) in the vertical direction stepwise, the plates (W_G) successively on a transport path (21). On the transport path (21), the plates (21) are transported at a constant speed of about 3.6 cm / s to the optical examination place (16). The plates (21) enter the examination place (16) through an opening (34) in a housing (32) surrounding the place (16). The plate examined therein is designated by reference numeral (Wj).

Within the housing (32) is arranged an optical scanning system which emits a beam (B) which focuses to a point of about 50 to 75 gm diameter on a test surface corresponding to the top of the plate (Wj). The beam (B) is repeatedly deflected along a line on the test surface whose direction is transverse to the transport direction of the disk (Wj).

The beam (B) is emitted by a low-power helium-neon laser generator (70), deflected by two mirrors (68), (66) and focused by means of lenses (64), (62). Then the beam (B) is illuminated a rotatable prism mirror (60) with 18 faces rotating at about 3600 rpm.

Before the beam (B) reaches the test area, it falls on a scan start detector (72), which outputs a scan start signal used in the evaluation of the information occurring during the examination of the disk (Wj). For receiving the light reflected from the top of the plate (Wj) of the scanning beam two separate fiber optic receiver (52) and (56) are provided, of which one (52) is arranged above the surface being examined such that it on the beam path ( R) receives reflected light. The receiver (52) contains a narrow slit mask so that any change in the parallelism of the slab (Wj) which causes a deflection of the reflected beam (R) across the slit is detected. That in the

Receiver (52) incident light is passed through a dichroic filter (55) to turn off extraneous light other than the laser wavelength (632 nm), and then detected by a photomultiplier tube (54). Since light of varying intensity is normally always present in the mirrored beam, the receiver (52) and the photomultiplier (54) are referred to below as the bright channel detector.

A scattered radiation receiver (56) serves to receive a part of the light on the beam path (S),

-5AT 394 632 B which is diffusely controlled due to particles or surface defects. The receiver (56) contains a light collector (59) and a dichroic filter (57) which is equal to the filter (55). The scattered light passing through the dichroic filter (57) is detected by a photomultiplier tube (58). The receiver (56) and the photomultiplier tube (58) are hereinafter referred to as the dark channel detector.

The electrical outputs of the photomultiplier tubes (54) and (58) are thresholded and analyzed for amplitude and frequency to detect surface defects. The information analysis and analysis circuit will be described in detail below.

The analysis results of the analysis represent sorting instructions for the transport system (14), with which the examined plate (W ·) is guided into one of several cassettes (22), (24), (26) or (28).

In the cassette (22), the plates (Wp) are stored, which require a Nachpolierens; she is on a vertical conveyor (23) and is gradually moved up to record each next classified as similar plate (Wp).

In the cassette (24) there are plates (W_c), which have been classified as requiring cleaning. This cassette (24) is also arranged on a vertical conveyor (25) and is moved up step by step to the next as similarly classified plate (W_c).

The cassette (26) takes plates (W_r), which have been classified as a committee. The cassette (26) is also mounted on a vertical conveyor (27) and is gradually moved up to the next board (W_r).

The cassette (28) finally serves to accommodate proper plates (W_fl); she sits on a vertical conveyor (29) and is gradually lifted to the next plate (W_fl), which has been found to be in order

The electrical part of the device is shown in the form of a block diagram in Fig. 2 and consists of transport, Oberflächenabtast-, electronics and control devices.

The Oberflächenabtasteinrichtung agrees with the in Fig. 1 shown in the nature with which the plate to be examined (Wj) in the transport direction (T) forward transported and reflected light along the beam path (R) to HellkanaT 'photomultiplier tube (54) is passed. Each surface defect present on the plate (W ·) causing scattering of the scanning beam onto the beam path (S) is detected by the DunkelkanaT 'photomultiplier tube (58). The scan start detector (72) outputs a scan start pulse before the scan of the disk (Wj) begins. The Hellkanal- and DunkelkanaT'-output signals via the line (L) or (D) are supplied to an analog circuit (100) of the electronic device.

In the electronic device 7 channels for real-time information in the form of seven-bit addresses over threshold category channel lines (T j to T₇) intended. The surface error information in the seven-bit address channels is derived in the case of dark channel signals by comparing the electrical output signals of the light and dark channels with different threshold values ​​and in the case of light channel signals by filtering different frequency components and by comparing these components with predetermined threshold values.

The seven-bit address is passed to a digital fault identification and processing network (200) in which it is used to form a digital matrix or a digital graph of the disk surface is processed. In the elaboration of this diagram, the scanned plate is divided into a large number of minute area units, each of which is assigned a binary code representing the most severe type of surface defects occurring on the area unit concerned. A video monitor (400) and a chart printer (402) may be provided for optically indicating the occurrence of surface defects on the inspected disk and providing the display information. The digital network (200) also serves to classify and count the surface error types. Further, quality grading of each examined disk is accomplished according to predetermined rules or norms input through a keyboard (408) connected via a peripheral interface (302), an auxiliary computer (300) and an interface (301). The quality information at the output of the network (200) is sent via the interface (301) to the auxiliary computer (300), which issues appropriate instructions to the transport device to feed the disk of the respective cassette and by means of the associated vertical conveyor by means of the peripheral interface (300). 302) to take one step to its next position.

The position of the Plaue is tracked by Lagefühlem, which are arranged along the transport path.

Further, a printer (404) and a character display (406) are provided to list the inspected plates according to the sorting, quality and number of types of surface defects.

A dark channel analog circuit (100θ) is designed to process the light changes that occur when surface irregularities cause an increase in scattered radiation. Its block diagram is shown in FIG. 3A.

-6AT 394 632 B

For processing in this channel are signal conditioning amplifier, u. a preamplifier (102), an impedance matching amplifier (104) and amplifiers (113) and (114) are used. Threshold comparison circuits (116), (118) and (120) compare the amplified output signal of the dark channel detector with thresholds each having a signal-to-noise ratio of approximately 1: 1.4: 1 and 20: 1, respectively, and digital-to-analog converters (115). , (117) and (119).

The reference voltages for the thresholds are provided by a microprocessor in which the individual digital clamp values ​​(DTj to DT-j) are stored and sent to the analog comparators via the above mentioned digital to analogue converters (115), (117) and (119). The output signals of the threshold category channels are denoted by (T j to T j).

To ensure long-term stability of the signal levels by compensating for the reduction of the intensity of the laser beam, the photomultiplier tubes and / or circuits, a gain control loop is provided for the photomultipliers. The output of the photomultiplier tube (58) is scanned when the laser beam is not directed at the disc surface and corresponds to the reference noise at the comparator input. The microprocessor controls an analog switch (105) which samples a noise level, which is then amplified and rectified by a filter (106) to provide a DC voltage corresponding to the noise amplitude. With an analog-to-digital converter four samples are formed and delivered to the microprocessor, in which an average value is formed. This average is then averaged even with four previous averages to form a new digitally coded value representing the gain. This value is converted via a digital-to-analog converter (110) to an analog voltage which is applied to the feed (112) of the photomultiplier tubes, with which the high voltage (and the gain of the tube (58)) to maintain a constant amplitude of the Background noise is set.

In Fig. 3B is a HellkanaV analog circuit (IOOjO) designed to process the light variations caused by inaccuracies of the surface of the inspected disk (Wj), causing a fall (change) of the reflected light. The processing in the analog circuit (100p) of FIG. 3B for the bright channel is similar to that in the analog circuit (100p) of FIG. 3A, wherein the photomultiplier tube (54) outputs an output signal to a preamplifier (142). The output of this preamplifier (142) is fed to an impedance matching amplifier (144), the output of which is fed to an amplifier (154).

The automatic gain control of the magnitude of the output of the photomultiplier tube is performed, except as described in the context of the DunkelkanaT 'analog circuit (100p), with one exception; the signal strength control is determined by sensing the intensity of the light as the laser beam impinges on the disk, in contrast to the noise level control of the dark channel. The automatic gain control loop present in the analog circuit (100p) consists of an analogue switch (146) which receives the output of the impedance matching amplifier (144) and a scan control signal of the

Microprocessors are supplied. The output of the analog switch (146) is fed to an analog-to-digital converter (148). whose output signal is fed to the microprocessor, in which it is subjected to a comparison and the averaging, as explained in the description of the dark channel analog circuit (100p) The digitally coded value representing the gain is converted into an analog voltage by means of a digital-to-analog converter (150). which is applied to the feed (152) of the photomultiplier tube, whereby the high voltage is adjusted to maintain a normally constant signal amplitude.

The output of the impedance matching amplifier (144) is applied to a bandpass amplifier (162) whose passband is in the middle range of 170 to 1400 kHz and further to a comparison circuit (164). The comparison circuit (164) is further supplied on the input side with a bandpass threshold with a signal to noise ratio of approximately 1: 1 of a digital-to-analog converter (163) receives a digital Sch well value (DTg) from the microprocessor. The output state of the comparator (164) changes each time the output of the bandpass amplifier (162) exceeds the 1: 1 bandpass threshold, the state change being due to an output signal on the line (T₅) is displayed accordingly.

The output of the impedance matching amplifier (144) is also fed to a bandpass amplifier (166) having a passband in the high range of 800 to 5000 kHz, the output of which is fed to a comparison circuit (168) in which it is connected to a bandpass clamp (16). DT ^) with a

Signal to noise ratio of approximately 8: 1 is compared. Whenever the output of the bandpass amplifier 55 (166) exceeds the 8: 1 bandpass threshold, the comparator (168) changes state and issues a signal to the line (T₆).

The output of the impedance matching amplifier (144) is also provided to an amplifier (154), the output of which is fed to a bandpass amplifier (158) having a low pass band

-7AT 394 632 B

Range of 50 to 200 kHz is supplied. The output of the bandpass amplifier (158) is provided to a comparison circuit (160) in which it outputs a bandpass threshold (DT₄) is compared with a signal-to-noise ratio of approximately 4: 1. Whenever the output of the bandpass amplifier (158) exceeds the 4: 1 bandpass threshold, the comparator (160) changes state and outputs a signal of changed level to the line (T ^).

In addition, the output of the amplifier (154) is fed to a compare circuit (156) and compared therewith to a margin indication threshold (DTy) having a signal to noise ratio of about 4: 1. Each time the output of the amplifier (154) indicates a serious deviation in the amplitude of the received light, the comparator (156) changes state, thereby sending to the line (Τγ) a signal indicating the edge of the disk.

The following table shows a list of detectable surface defects in order of priority. Each type of error is briefly described and characterized by an output address on the lines (Tj to Τγ) as well as a correspondingly designated hexadecimal code

table

Priority of surface defect types

<td>hexadecimal</td><td>priority</td><td>threshold</td><td>kind</td>

<td>G</td><td>1</td><td>^T7 (Tj to Tg disregarded)</td><td>Edge and local errors &gt; 50 um Π 000 0001</td>

<td>F</td><td>2</td><td>Tg, T₄ and Tß (Tj and T₂ disregarded)</td><td>indefinite deformation / abrasion (101100)</td>

<td>e</td><td>3</td><td>T₆ and T.₄(Tj and T₂ disregarded)</td><td>indefinite deformations (1010QQ)</td>

<td>D</td><td>4</td><td>Tg, T₅ and T.₃(Tj and T₂ disregarded)</td><td>Dust 50 gm (HOlfiQ)</td>

<td>C</td><td>5</td><td>T5, T₄ and Tß (Tj and T₂ disregarded)</td><td>Heavy deformation with abrasion 10111001</td>

<td>B</td><td>6</td><td>Tj and T₄(Tj, T₂, Tg unconsidered)</td><td>Sawmills, polished (0110QQ)</td>

<td>A</td><td>7</td><td>T5 and Tß (Tj, T₂ disregarded)</td><td>local deformation errors with abrasion (0101001</td>

<td>9</td><td>8th</td><td>^T5 (T_p T2, Tg excluded)</td><td>Deformation errors, high frequency, such as craters or dimples (Q10QQQ)</td>

<td>8th</td><td>9</td><td>T₄ and Tß (T j, T₂ disregarded)</td><td>Deformation with abrasion (OOllßQ)</td>

<td>7</td><td>10</td><td>^T4 (Tj, T₂ disregarded)</td><td>Deformation low level, low frequency like waves or caterpillars (0010QQ)</td>

-8AT394 632B

Table (continued)

<td>hexadecimal</td><td>priority</td><td>threshold</td><td>kind</td>

<td>6</td><td>11</td><td>^T3 (T_p T2, Tg excluded)</td><td>Pits, abrasions&gt; 2 gm (Q001QQ)</td>

<td>5</td><td>12</td><td></td><td>not used</td>

<td>4</td><td>13</td><td>^T2 (Tj disregarded)</td><td>Pits, scratches&gt; 20 gm (00001Q)</td>

<td>3</td><td>14</td><td>Tj and Tg (T₂ disregarded)</td><td>dust (1000Q1)</td>

<td>2</td><td>15</td><td>^T1</td><td>Haze, scores, scratches (000001)</td>

<td>1</td><td>16</td><td>^T6</td><td>Noise (100000)</td>

Referring now to Fig. 5, there is shown the block diagram of the digital error detecting circuit 200 ^ which is for making a map-like chart of the surface of the plate being inspected.

The matrix of the surface diagram has shown in FIG. 4 a field of investigation of about 14 x 14 cm, which allows the examination of plates up to a diameter of 125 mm. As the plate rests on a dark background, significant changes in light reflection occur as the scanning beam passes the plate edge. In this way, the plate edge can be detected, which is indicated by signals on the line (Ty). The edge indication signals are only for sampling control of the threshold signals (Tj to Ty) between the detected disk edges. The area matrix according to FIG. 4 consists of 172 lines of 172 area units each (i.e. H. a matrix of 29,584 area units).

According to FIG. 5, in the digital error detection circuit (200_A) analyzes the scans of each information line to provide the individual unit areas of 0.813 x 0.813 mm size, each of which is labeled with a binary code associated with the most severe surface defect in that area unit. An area unit without errors is marked with a code of all zeros. The formation of the surface graph occurs in the course of a predetermined time base, and the digital error detection circuit (200_A) outputs a single line of 172 area units to an associated microprocessor memory every 24 headers of the scan beam.

The three DunkelkanaT 'lines (Tj, T₂ and Tj) of the analog circuit (100p) are connected to a holding register (220). The four threshold channel lines (T₄ to Ty) of the bright channel analog circuit (100p) are also connected to the holding register (220). The line (Ty) of the HellkanaT analog circuit (100p) is further connected to a edge follower circuit (200β) which effects control of the data on the lines (Tj to Ty) between detected edges of the disk (W x)

The register (220) is a temporary latching register which holds the seven-bit information of the lines (Tj to Ty) of the analog circuits and outputs a seven-bit address code for each of the 172 area units during each beam scan. The contents of the register (220) are used as a seven bit address for a 128 x 5 read-only memory (224) containing pre-developed 5-bit codes in each of its 128 memory locations. Special cases are memory location 0, which contains a code of all zeros to indicate no errors, and the top 64 memory locations, each containing the one on the disk. The memory locations 1 to 63 contain codes which are assigned to the respective combination of active threshold values ​​with respect to each test point. The result is a five-bit code read from the read-only memory (224), which indicates the respective type of error which is represented by the correspondingly addressed memory location of the read-only memory (224). The five-bit code is determined so that the numerical value of the code is larger,

-9AT 394 632 B the more serious the error is The order of priority is given in the table.

The five-bit code read from read-only memory (224) will be referred to hereinafter as the new error code, which is compared to a corresponding previously acquired five-bit code taken from a random access memory (238). The random access memory (238) is synchronously addressed to dump the old error code stored at the memory location corresponding to the same area unit of the currently scanned and processed scan beam. A comparison circuit (228) is for display when the new error code is greater than the old "error code, and accordingly outputs an enable signal via a line (230) or (232) to an AND gate (234) or (236). The enabled AND gate (234) passes the new five-bit code to the random access memory (238) and optionally turns on the old error code, thereby identifying the most severe error on that area unit. Otherwise, the AND gate (236) is turned on to reinsert the old error code into the random access memory (238).

Since the disk transport system advances the disks by 0.813 mm during the time required for 24 scans, the scan / compare operation is performed 24 times for each area unit, or 4.128 times for each of the 172 lines containing the array of area units form.

A clock and control network (254) is provided for controlling and switching through a 6.48 MHz sample start, a disk ready and a clock signal to the respective logic circuits.

The random access memory (238) is addressed 172 times per sample, and this address is provided by a area unit position counter (250) which receives clock signals sampled by the clock and control network (254). From the clock and control network (254) are nested unloading or Load commands on lines (246) or (244). When the 172nd Address in each scan line, a decoder monitoring the eight-bit output of the counter (250) outputs a signal enabling an AND gate (258). The enabled AND gate (258) passes a clock pulse to a trace counter (262) which monitors the number of traces of the scan beam. The output of the decoder (256) also outputs a reset signal to the position counter (250). At the beginning of the appearance of the 24. Scanning by the beam, the trace counter (262) feeds a decoder (264) which provides an output to enable an AND gate (266). The enabled AND gate (266) passes a clock pulse to a line counter (268) which monitors the number of detected lines delivered to the temporary holding memory of the microprocessor. The output of the decoder (264) also outputs a reset signal to the trace counter (262).

At the beginning of the 24th scan of the 127th line, a decoder (267) issues a signal to the clock and control network (254) which in turn issues a reset signal to the line counter (268).

The decoder (264) provides an output to the clock and control network (254) which causes not only the reset of the trace counter (262), but also the delivery of control commands to three lines (270), causing the temporarily holding memory of the associated microprocessor reads the next line of 172 data and outputs five address lines (240) and eight position address lines (260) word addresses.

At the beginning of every twenty-four sample, one row of 172 area unit codes is transferred over the data lines (240) to the microprocessor memory for subsequent analysis.

Upon transmission of each area unit code, the respective memory space of the temporary holding random access memory is set to a zero all error code in preparation for the next line processing. After transferring 172 such lines, the circuit (200) stops active processing and waits for the following disk.

The disk edge coordinates are used in the microprocessor to determine the boundary of the disk within the entire 172 x 172 matrix. These coordinates are obtained from the edge follower logic circuit (200g) shown in FIG. 6 and used to trigger a bistable flip-flop, enables or disables the AND gate for the signals (Tj to Ty) of the analog circuit (100) to the digital error detection and processing circuit (200). In this way, no extraneous signals can be processed from outside the disk surface as errors, on the other hand appeared in the accumulated total error number.

The in Fig. 6 shown circuit (200g) is used to record the first on the disk recording and the last of the disk recording and to the fact that all other signals are disregarded, which could occur as Τγ recording. A holding device prevents the recording of all but the first on the disk '' - recording during the scanning cycle. That of the disk recording records the location of the scan where the hold interrupt occurs, this coordinate being used during the next scan to not erroneously evaluate the occurrence of dirt or ash on the disk as the edge location of the disk. Plug housing switches are provided as parts of the network for the purpose of positionally adjusting the detected edges of the recovered to the disk signal. The setting determines an edge zone on the disk that is disregarded for purposes of error detection

-10AT 394 632 B remains. The outer zone of the plate is not normally used for the manufacture of semiconductor integrated elements because the edges very often have many detectable defects. Otherwise, these unwanted signals would result in false error information outside or along the edges of the plate, and thus in incorrect judgment of the plate.

The time interval between the set edges is hereinafter referred to as the time portion of the true errors and is used to control the output from the analog circuit information.

The edge follower circuit (200g), according to FIG. 6 an eight-step down counter (202) controlled by clock pulses (8 per unit area) and preset by the plug housing switches (204) connected to the various counter stages. Upon receipt of the load instruction corresponding to the Τγ transition representing the transition of the scanning beam from outside the disk to the disk, the counter (202) starts to count the clock pulses. When the counter (202) counts down from a total number corresponding to 2 or 3 area units by default by means of the switches (204), it generates a pull-out output signal as a command for switching a true-error gate to latch the signals on the lines (Tj to Τγ).

To generate a true error disable signal, the circuit (200g) uses the information from the previous line scan at the last one of the disk picks to determine the end of the true error signal for the current scan.

Upon the occurrence of the scan start signal output from the detector 72, a twelve-stage counter 206 is reset to zero and begins counting the clock pulses (8 per unit area). The contents of the counter (206) are stored in a register (208) each time one of the disk recording signals is received. This signal represents a transition on the Ty threshold line, which, as previously explained, can produce false signals due to dust or ash.

An arithmetic logic unit (212) receives the information stored in register (208) and subtracts a digital value preset by plug housing switch (210) corresponding to the edge zone of the disk for which the error detection is undesirable.

At each occurrence of the disk recording signal, the register (208) is loaded with a new number from the counter (206) and outputs that number to the logic unit (212). This value is pre-stored in a twelve-step down counter (214) upon receipt of the next scan line scan start signal output from the detector (72). When the counter (214) counts down from the prestored value to zero, a true fault disable signal is asserted to disable reception of information from the analog circuit over the lines (Tj to Τγ).

The processing unit of the digital error detection and processing circuit (200) consists of FIG. 2 from a microprocessor programmed to deliver many of the above-mentioned output signals and also to analyze the 172 lines of information output to its temporary holding memory. The in Fig. 7A and 7B show the quality grading algorithm that is programmed into the microprocessor to issue sorting commands to the conveyor and rules for further processing of the particular plate being inspected.

According to FIG. 7A, after issuing a start command, it is examined how many type E errors have been detected and stored in the temporary memory. The maximum number of acceptable errors of type Έ is preset by the user and is denoted by Εθ. If the number E of the errors of the type E is greater than Εθ, the command for grading the plate is given out as being in need of polishing. If the number E is not greater than Εθ, the number E is added to the number of errors of the type 7, which have a subordinate priority than the errors of the type E. Then the number of errors of type 7 including the number E is examined. The number 7 of errors of the type "7 is compared with a preset maximum number 7θ. If the number 2 is larger than 7θ, the command for grading the plate is given as being in need of polishing. If the number 2 is less than or equal to 7θ, then any error of the type 7 is renamed as such type existing in the immediately preceding place

It will be recalled that a type 7 error according to the table represents a low frequency distortion which sometimes occurs even when previous positional errors are detected. A low-frequency oscillation occurs, which sometimes leads to an error of the type "7 being considered to have been detected after detection of a different kind. Therefore, if a small number of such Type 7 errors are detected, they will be reclassified as errors of the type detected at the next preceding digit.

Following the rearrangement of the type 7 errors, the number of type G errors is examined. If the number G is larger than the preset maximum number Gq, a command to classify the disk as a scrap is issued.

The following instructions in the flow chart of Figures 7A and 7B follow the general priority list according to the table until the low priority errors have been tested. In the case of type 3 error, if the number 2 of these errors is greater than the preset maximum number 3θ, a command for

-11AT 394 632 B

Classification of the plate given as requiring cleaning, the number 2 is not greater than 3θ, then a command to display the number 2. on the character display (406), which means dust. At the same time, there is an investigation into the number of Art 2 errors. The number 2 of these errors is compared with a preset maximum number 2θ to determine if the occurrence of type 2 errors indicates haze on the disk surface. If the number 2 is larger than 2θ, a command for grading the disk is given as being in need of polishing. If the number 2 is not larger than 2θ, it is compared with another preset number 2j representing the maximum number of dimples possibly appearing on the disk surface. If the number 2 is greater than 2j, it is displayed on the character display (406), and at the same time, a command for grading the disk is given as being in need of polishing. On the other hand, if the number 2 is not larger than 2 j, a command for grading the disk is given as OK and the number 2 is displayed.

Of course, after submitting the respective commands, such as order, cleaning, polishing or scrap, the temporary memory is cleared and the device is returned to the ready state until the next disk enters the test site.