FR2464495A1 - HIGH SPEED DIRECTIONAL SPATIAL FILTER FOR FAULT INSPECTION APPARATUS - Google Patents

Abstract

THE INVENTION RELATES TO A DIRECTIONAL HIGH-CUT SPACE FREQUENCY FILTER. THE FILTER ACCORDING TO THE INVENTION IS OBTAINED BY COMBINATION OF A SPATIAL FREQUENCY FILTER 23 FORMED BY PHOTOGRAPHIC RECORDING OF THE DISTRIBUTION OF THE INTENSITIES OF THE OPTICAL SPECTRUM OBTAINED BY TRANSFORMATION OF FOURIER OF A REGULAR DRAWING, WITH A DIRECTIONAL FILTER WITH HIGH CUT 24 D THE HIGH-PASS CHARACTERISTIC IS INCREASED ACCORDING TO THE PROPORTION OF DIRECTIONAL COMPONENTS RECORDED ON THE SPACE FILTER. APPLICATION TO FAULT INSPECTION EQUIPMENT OF INDUSTRIAL OBJECTS.

Description

The present invention relates to a system of

  spatial frequency filtering of a diffraction pattern

  A laser beam (hereinafter referred to as a spatial filtering system) for detecting defects of a regular pattern by using optical diffraction phenomena, and more specifically a spatial frequency filter. (hereinafter referred to as a spatial filter as used herein) prepared by photographic recording

  of the intensity distribution of the optical spectrum of the trans-

  Fourier formation of a normal regular pattern, in

the system.

  Recently, spatial filtering systems have been used in practice using laser beams for detecting pattern defects arranged regularly, by

  example in various metal filters and

  integration of integrated circuits. There is also known a system in which a two-dimensional image is scanned by a thin luminous brush, with formation of output signals that are processed in a computer, the assembly forming a two-dimensional image processing system. Although this classic system has drawbacks because

  the treatment is complicated because the treatment time

  is relatively large and because the colt is

  the above-mentioned system of spatial filtering is

  happy because the spatial parallel processing of an image

  two-dimensional can be easily achieved with a

  colt optic head of relatively low manufacturing and high speed. As a result, a system was used

  spatial filtering in an inspection device of

  industrial products with two-dimensional designs

  such as metal filters, integrated circuit covers, fabrics, etc. We now consider, with reference to FIG. 1 which is a diagram representing the fundamental arrangement CA

  a defect inspection device using a system

  spatial filtering, how is achieved essential-

such a system.

  A laser beam 2 coming from a laser oscillator 1 reaches a collimator 3 and transforms into an enlarged parallel beam 4 which is transmitted to an object to be inspected. This object is placed in the front focal plane of a Fourier lens while the spectrum of the Fourier transformation object is formed in the rear focal plane of this lens by a

  light beam 6 which is obtained by diffraction of the

  parallel beam 4 caused by its passage through the object 5. A spatial filter 8 is placed in the rear focal plane of the Fourier lens. The spatial filter 8 is a negative film on which the re-partition of the Fourier spectrum intensities, for the normal drawing of the object to be inspected, is recorded (the inspection being provided by the lens 7). Thus, only a spectrum corresponding to the normal pattern

  absorbed by the spatial filter 8 and a corresponding spectrum

  laying down a defect drawing may be transmitted by the

spatial filter 8.

  In this respect, it should be noted that the position of the

  spatial filter 8 corresponds to the focal plane before a

  gate 10 providing inverse Fourier transformation.

  Consequently, a light beam 9 transmitted by the spatial filter 8 is subjected to an inverse Fourier transformation and therefore appears in the rear focal plane of the lens 10 in the form of an image having undergone an inverse Fourier transformation obtained by spatial filtering, that is to say in the form of an image indicating

  defective parts of the object to be inspected.

  The image is detected by an optical detector 11. A screen

  used in place of the optical detector 11 allows the pre-

  faulty parts in the form of clear dots. So, in the latter case, the defects of the object

can be inspected by eye.

  Referring now to FIGS. 2a to 2c, FIG. 2a is a diagram showing two-dimensional crossed rows, FIG.

  a plan view of a spatial filter obtained by recording

  photographic representation of the distribution of Fourier optical spectrum intensities of the two-dimensional cross-rows of FIG. 2a and FIG. 2c being a graphical representation indicating the relationship between the geometric dimensions of the crossed rows and the distribution of Fourier spectrum intensities, an example of a conventional spatial filter 8. FIG.

  two-dimensional cross-rows. Figure 2b shows

  a spatial filter 8 obtained by

  graph of the distribution of the intensities of the Fourier optical spectrum of two-dimensional cross-rows, the spatial filter 8 having an opaque portion 12 and a transparent portion 13 which are in black form and in non-black form depending on the spectral intensities. Figure 2c shows the relationship between the geometric dimension

  two-dimensional cross-rows and the distribution of

  the intensities of the Fourier optical spectrum in the x direction. In FIG. 2c, the spatial frequencies X = x / (X.f) (where A is the wavelength and f is the focal length) are plotted on the horizontal axis and the distribution

  of intensities I (X), is plotted on the ordinate and is

  given by the formula: 1 (X) = (s AfrX) 2 sin C X) 2 (1) A1TX sin BurX The distribution of intensities has a shape such that the

  interference fringes (sin CBXx) 2 corresponding to several

  Slots are modified by the diffraction image (sin AirX 2 A7rX) (a dot in the figure) of a slit. In

  the equation (1) above, A, B and C represent the

  first side of a square network, the distance between

  adjacent networks and the total length respectively

  as shown in Figure 2a.

  In general, during the inspection of products or industrial objects having a regular design,

  for the determination of defects by the

  space map, we use a spatial filter obtained by

  photographic recording on a 35 mm film or

  logue, the distribution of optical spectrum intensities

  formed by Fourier transform of a normal drawing.

  This spatial filter is advantageous because it can be easily

  manufactured so that its cost is low; however, it has drawbacks on the following points. The optical system sets criteria extremely demanding on mechanical accuracy so that the spatial filter can not be offset from the optical axis or can not be rotated around the optical axis. In particular, in the

  the case of the use of an ordinary convex lens

  Fourier lens, the focal plane is curved, so that the high order diffraction light beam is no longer focused. In addition, as the light beam of

  high order diffraction at an excessive spectral intensity

  sively low, recording in the dynamic range of 102 to 103 lux seconds of the photographic film is not possible. So, the spatial filter can not filter the

  light beam of high order diffraction, and

  it is not necessary to obtain a sufficient S / N signal-to-noise ratio

high level.-

  We have already proposed a method of using a directional filter or a special Fourier lens giving a Fourier transformation plane which is flat,

  to remedy these disadvantages.

  An example of a spatial filter, called a "directional filter", is shown in FIG. 3 of the accompanying drawings. The spatial filter carries a drawing in which, as in the case of two-dimensional cross-rows, the main components of its Fourier spectrum are distributed in the O degree and 90 degree directions. The spatial filter has an opaque portion 14 (hatched) and the remaining transparent portion 15 allows the optical system to have reduced mechanical precision, i.e.

  the use of this spatial filter is more convenient in

  tick. When using this spatial filter, the com-

  posantes in the directions 0 degrees and 90 degrees, that is,

  say the linear components of the drawing, are considered normal components and are removed by filtering while the remaining components which are considered as fault components, are transmitted by the spatial filter. Thus, it is not possible for the spatial filter to filter out a non-directional spectrum of a false defect such as a radius formed at a corner (such as

  indicated by the reference R in Figure 6) and consequently

  the S / N signal-to-noise ratio of the spatial filter is

not high enough.

  On the other hand, the second method indicated is not convenient in practice since the special Fourier high resolution resolution lens giving a flat Fourier transformation plane is very expensive, several tens of thousands of francs. Even when using such a lens, the problem posed by

  the dynamic range of the photographic film is not resolved.

  Consequently, it is not possible to provide a filtering effect of the high-order diffraction light beam of the

desired manner.

  The invention overcomes the aforementioned drawbacks

  presented by a classical spatial filter.

  More specifically, the invention relates to a directional-type and high-cutoff spatial frequency filter, allowing the inspection of defects of industrial products having regular two-dimensional patterns, this filter giving high signal-to-noise ratios, high output signals. representative of defects and excellent image quality with relatively low mechanical accuracy, without the drawbacks of conventional spatial filtering systems, the parallel spatial processing of two-dimensional images can be easily achieved with a manufacturing cost optical system

relatively weak.

  Other features and advantages of the

  will become apparent from the following description,

  Referring to the accompanying drawings, in which FIGS. 1, 2a, 2b, 2c and 3 have already been described: FIG. 4a is a plan view of a spatial filter obtained by photographic recording of a

  industrial product with two-dimensional crossed

  tional; FIG. 4b is an enlargement of the essential part of the spatial filter of FIG. 4a; FIG. 5a is a plan view of an example of FIG.

  high-directional directional spatial filter according to the invention

  obtained by combining the spatial filter of FIG. 4a with a high-cut directional filter; Figure 5b is an enlargement of the essential portion of the filter of Figure 5a;

  FIG. 6 is a diagram given for explanatory

  tif, representing holes or holes, and a defect, in the case of an industrial product having two-dimensional cross-rows;

  Fig. 7a is a diagram showing originals

  fices or holes as well as a defect formed in a metal filter; Figure 7b is an enlargement of the most important part of the metal filter; Fig. 8 is an enlargement of a directional high cutoff spatial filter used for the metal filter of Fig. 7a;

  FIG. 9 is a graph showing the range of

  S / N signal-to-noise ratio ratio, as a function of diameter, a directional high-cut spatial filter and a circular low-pass spatial filter; FIG. 10 is a graph showing the variation of the optical intensities relating to a fault,

  ordinate, according to the diameter carried in abs-

  in the case of a directional high-cut spatial filter and a circular low-pass spatial filter; FIG. 1 is a graph representing the variation of the signal / noise ratio S / B, carried on the ordinate, as a function of the displacement of the optical axis, taken as abscissa, in the case of the directional high-cut spatial filter and the spatial filter. classic; and FIG. 12 is a graph showing the variation of the signal / noise ratio S / B carried on the ordinate as a function of the rotation around the optical axis.

  in abscissa in degrees, when using the filter spa-

  directional tial with high cutoff and spatial filter

classic.

  We first consider a classical spatial filter,

  with reference to FIGS. 4a and 4b, so that understanding

  of the invention is facilitated.

  The classic spatial filter has parts 16 that

  are blackened by recording the distribution of

  of the Fourier optical spectrum of an industrial product.

  triel, for example a metal filter having two-dimensional cross-rows, the filter thus comprising opaque portions 16 and transparent portions 17-1 and 17-2 which are not blackened because the intensity of the spectrum is low. As shown in Figure 4a, the

  main components of the cross-row spectrum

  are distributed in the 0 degree and 90

  and the components are much smaller in a direction O between 0 degrees and

  degrees. Thus, a distribution of dark areas, reflected

  as the distribution of the intensities of the spectrum as it is, is formed on a photographic film. In this way, in the 0 degree and 90 degree directions, the blackening is ensured for the region corresponding to the high order diffracted light whereas in the direction e, only the region corresponding to the lower order diffracted light is blackened. The diameter of the black point

  is different from the diameter corresponding to the limit of

  fraction, depending on the light source (eg

  several microns). Thus, in a different light region

  Fragment of high intensity, having an order close to the fundamental order, the dark point has a large diameter of the order

  from 100 to 150 microns, given the overexposure and ef-

  halo fet during the registration process; this-

  during, when the order of diffraction of light increases

  the point diameter decreases. When an object 5 to be inspected, having a defect 18 as shown in FIG. 6, is subjected to spatial filtering in the system described with reference to FIG. 1, using the spatial filter-shown in FIGS.

  4b, among the spectral components of the defect 18, the

  high order posers and lower order components

  pass in the transparent parts 17-1 and 17-2 res-

  respectively and appear in the form of clear dots

  in the inverse Fourier transformation plane. Some

  industrial products, such as metal filters having apertures or windows, generally have R-radii in the corners, as indicated by reference 19

  Figure 6. As a result, during spatial filtering,

  above mentioned, the spectral components of high order,

  in the direction O, for the radius of the corners, are trans-

  put by transparent parties 17-1 and may also

  appear as light dots in the transmission plan.

  inverse formation of Fourier so that signals corresponding to faults appear and the ratio

signal / noise S / B is reduced.

  The noise components of a conventional spatial filtering system can be classified into

  noise from the optical system (due to optical dispersion

  evoked by the irregular surface of the lens and the dust carried by this surface), and in noise components corresponding to the high order diffracted light beams transmitted in the region of the order spectrum

  in which the photographic film is not suf-

  obscured by a spectral intensity indi-

  sufficient and poor focus. A big

  part of each of these two different types of components

  noise rents can be transmitted by the trans-

  parents 17-1. These occupy most of the spatial filter. Therefore, although the noise components of the optical system and the high order diffracted light have a very low intensity, they can not be neglected when the transmitted light is integrated on the surface. The invention is based on these phenomena. High-order diffracted light beams

  crossing the spatial filter include those that

  rectilinear peripheral parts of the orifices forming square gratings. This phenomenon can be explained from the theory of marginal waves that high-order diffracted light beams are those originating from the periphery of an orifice (as described in Hiroshi Kubota's "Wave Optics" p. 257 Iwanami Shoten, 1971). These high order diffracted light beams pass through parts of the spatial filter that are not darkened, in the 0 degree and 90 directions

  degrees. As is clear from the description which pre-

  yields, high-order diffracted light beams,

  constituting part of the spectral components of a

  are important, but they are nothing more than the main components of noise when considered together. Thus, an improvement in the signal-to-noise ratio S / B of the spatial filter requires the formation of a spatial filter which transmits the components of the diffracted light by a defect, as far as possible, and which removes the high order diffracted light beams. which form noise components, to the extent that

possible.

  Figures 5a and 5b show an example of spatial filter according to the invention. This filter can be obtained by adding a high-cut directional filter 20 having portions of the spatial filter of Fig. 4 obtained by photographic recording (shown as hatched in Figs. 5a and 5b). The configuration of the hatched portions shows that the high cutoff filter hides the lower order diffracted light region whose spectrum is recorded, in the direction O in which the high order diffracted light region is not recorded; in the directions of the main spectral components, that is, in the directions

  0 and 90 degrees, the spectrum is recorded up to the

  diffracted light of higher order, so that filtering is ensured. Thus, the high cut filter 20 tends to have a high pass type opening and

  has a configuration corresponding to the proportion of

  Directional Positives of the Fourier Optical Spectral Distribution Distribution for Two-Dimensional Cross-Rows Almost all higher order diffracted light beams around the O direction can be removed by adding a high-cut directional filter

  having such a configuration to a conventional spatial filter

  obtained by photographic recording. As a result

  almost all the effects of the noise components of the optical system, the radii formed at the corners and

  marginal diffracted light can be eliminated.

  On the other hand, among the spectral components of the defect

  18, only the low-order diffracted light beams

  are transmitted by the filter in the direction O and, in the directions 0 and 90 degrees, the light beams

  diffracted, to higher order diffracted beams

  are transmitted by the filter to the extent possible.

  depending on the filtering ability of the given spatial filter by photographic recording, so that the reduction of the output optical signal corresponding to the

  defect 18 is avoided to the extent possible and the report

  S / B signal / noise port is remarkably improved.

  The effects of the high-cut directional filter reside not only in the increase of the optical output signal of the defect but also in the improvement of the signal-to-noise ratio S / B. This characteristic is described

  in the form of improving the quality of the image.

  In the combination of a circular low-pass filter and

  of a spatial filter obtained by photographic recording.

  The S / B signal-to-noise ratio is improved, but the resulting image has a very low quality due to diffraction by circular holes. On the other hand, the high-cut directional filter, having a high-pass tendency in the direction of the main spectral components, is only slightly affected by the diffraction so that the image obtained by Fourier inverse transformation is excellent. In addition, in the region of light

  low order diffraction which is used

  the diameter of the black spots in the spectrum

  is 10 to 20 times larger than the corresponding points

  corresponding to the diffraction limit which is determined by the size of the light source. As a result, the possibilities of rotation around the optical axis and

  displacement relative to the optical axis are relative-

  the mechanical precision of the filter

  space can be significantly reduced.

  The high-cut directional filter can be formed of any material that can absorb light. In

  Consequently, it can be prepared by darkening, followed by

  a predetermined configuration, of the spatial filter obtained by photographic recording, or by gluing a black paper cut to the desired configuration on the spatial filter. Thus, the high-cut directional filter can be manufactured easily and at a low price. The black paper cut to the predetermined pattern can be placed several millimeters from the spatial filter in

the direction of the optical axis.

  As the above description indicates, the use of

  the directional high-cutoff spatial filter obtained according to the invention by combining the filter

  space obtained by photographic recording of the

  partition of the Fourier spectrum intensities of a regular pattern with the high-cut directional filter having a high-pass trend in the directions in which there are more components as a function of the proportion of the directional components of the spectrum, makes it possible

  the formation of a mechanical precision device

  significantly reduced but allowing the inspection of industrial products with regular two-dimensional designs so that defects are determined with precision

  high, without loss of the characteristic of the filament system.

  spatial mapping that allows easy spatial parallel processing of the two-dimensional image with an optical system

relatively low cost.

  The effects obtained according to the invention are now considered. Figures 7a and 7b are explanatory diagrams showing a metal filter and an enlargement of

  the most important part of this metal filter is

  tively. The metal filter has a thickness of 0.2 mm and a regular two-dimensional pattern. The filter is used

  for determining the characteristics of the invention.

  In FIGS. 7a and 7b, the reference-21 denotes a hole

  or orifice of the metal filter and the reference 22 a de-

  need. FIG. 8 represents in enlarged form an example of directional high-cutoff spatial filter

  according to the invention obtained by combining a direct filter

  high-cut 24 and a spatial filter 23 obtained

  himself by photographic recording of the distribution

  Fourier optical spectrum intensities of the metal filter shown in Fig. 7, the light source being a He-Ne laser and the de-Fourier lens being a convex lens having a focal length f of 250 mm and a diameter of 100 mm. . The main components of the Fourier spectrum are in the 0 and 90 directions

  degrees. However; like most of the spectral components

  are in the 0 degree direction, the

  The high-pass filter of the high-cut directional filter 24 is improved in the 0 degree direction so that the F / E ratio is equal to 2 (F and E respectively denoting the lengths in the 0 degree and 90 degree directions).

  as shown in Figure 8).

  Figure 9 shows graphically the variation

  the ratio of the optical intensity of the defect 22 to the

  optical noise intensity due to the wedge measured in the plane

  reverse Fourier transform, ie the ratio of

  signal / noise port S / B, depending on the diameter D $ indicated

  in Figure 8, in the case of the spatial filter direction-

  FIG. 8 (curve A) and a low-pass circular spatial filter formed by combining a circular low-pass filter (of DO dimension) and a spatial filter 23 obtained by photographic recording. (curve B). The S / B ratio of the spatial filter

classical is equal to 5 (S / B = 5).

  Figure 10 graphically represents the variation

  the optical intensity corresponding to the defect 22, measured as indicated in FIG. 9 (ordinates) in

  function of the diameter Do, for the spatial filter direction-

  high cutoff (curve A) and for the circular low pass spatial filter (curve B). The optical intensity of the

  conventional spatial filter is 2.3.

  As shown in Figures 9 and 10, the report

  S / B port of the directional high-cut spatial filter according to the invention is about 5 times larger than that

  of the classic spatial filter. The S / N ratio, when

  This directional spatial filter is higher than that obtained by using the low-pass circular spatial filter and the optical intensity of the defect 22 is greater. The optical intensity of the defect 22 for a value equal to 3, corresponding to the peak value of the S / N ratio in the case of the directional spatial filter, is only 20% lower than the optical intensity.

  in the case of the classical spatial filter; the intensity

  tick in the case of the low-pass circular spatial filter

is 80% lower.

  Figure 11 shows the effects of displacement

  of the optical axis, plotted on the abscissa, on the S / N ratio.

  It is assumed that the S / N ratio limit for a satisfactory inspection is 4. Therefore,

  the classical spatial filter (curve B) obtained by recording

  very photographic can not be used when the optical axis is displaced by more than 10 microns while the high-cut directional spatial filter (curve A) can be used even when the optical axis is moved

35 microns.

  Figure 12 shows the effects of rotation around the optical axis, plotted on the abscissa, on the S / N ratio, plotted on the ordinate. If we assume that the limit

  effective ratio is S / B = 4, we note that the spe-

  tial (curve B) can not be used when

  that it rotates around the optical axis by more than 0.2 degrees while the possible rotation around the optical axis in the case of directional filter with high cutoff

  according to the invention (curve A) is 1.0 degree.

  As the description clearly indicates

  precedes, the spatial filtering system comprising the filter

  directional high-cut space according to the invention

  puts inspection of industrial products with regular two-dimensional drawings and determination of defects with high S / B signal-to-noise ratios, with high signals representing defects and with high quality

  image, for a relatively low mechanical accuracy.

  Thus, the invention has very appreciated practical effects.

Claims (5)

  1. directional high-cutoff spatial frequency filter, characterized in that it comprises: a spatial frequency filter (23) prepared by photographic recording of the distribution of the intensities of the optical spectrum formed by Fourier transform of a drawing and a high-cut directional filter (24) whose high-pass characteristic is increased as a function of the proportion of directional components of said   distribution of the intensities of the trans-   Fourier formation, in directions in which spectral components are greater than those other directions.   2. Filter according to claim 1, characterized in that the directional cut-off filter (24) is   formed of a light absorbing material.   3. Filter according to claim 2, characterized in that said material is a black sheet cut to a predetermined configuration.   4. Filter according to claim 2, characterized in that the high-cut directional filter (24) is prepared by blackening the spatial frequency filter.   (23) according to a predetermined configuration.   Filter according to Claim 2, characterized in that the high-cut directional filter (24)   formed of a light-absorbing material is placed at   spatial frequency filter (23).