GB1589872A

GB1589872A - Comparator

Info

Publication number: GB1589872A
Application number: GB3061176A
Authority: GB
Original assignee: SIRA INSTITUTE
Current assignee: SIRA INSTITUTE
Priority date: 1977-07-21
Filing date: 1977-07-21
Publication date: 1981-05-20

Description

(54) COMPARATOR (71) We, SIRA INSTITUTE LIMITED, a British Company of South Hill, Chislehurst, Kent BR7 5EH do hereby declare the invention, for which we pray that a patent may be granted to us, and the method by which it is to be performed, to be particularly described in and by the following statement: The present invention relates to comparators and to a method of comparison. Comparators are widely used, for example, in production processes where a number of identical articles are produced, the articles being compared with a standard article or with each other, to determine whether or not faults have occurred in the manufacturing process.

For example, minute defects in a printed circuit board can lead to circuit failure after expensive components have been mounted on the board; small printing errors on a sheet of tabulated printed figures can result in illegibility; other examples can be cited where similar small defects have no functional effect but are aesthetically dissatisfying.

It is possible by a number of electrooptical means to interrogate the light reflected from neighbouring small areas of a two dimensional article which is simply treated as a pattern; such means include flying spot scanners, flying image scanners, TV camera systems and "static" image scanners such as solid state detector arrays.

In order to retain adequate sensitivity to the smallest defects of interest in the article and hence in the pattern it is necessary to divide the total area of inspection into a mesh of much smaller areas, the reflectivity of each area being individually measured and compared with the value obtained from the corresponding area on a 'perfect' specimen.

Typically, the mesh might need to contain 500 elements or more along each side in order to resolve the pattern into a sufficiently large number of bits.

To decide whether a test pattern derived from a test article is acceptable or not, one has to check whether every element on the test pattern appears similar to the corresponding element of a pattern from the perfect article. A convenient way of achieving this comparison which avoids the need for a vast information-storage capability, is to use a pair of synchronised scanners to examine respectively the nominally identical test and reference patterns. Scanner outputs are subtracted, and any resulting difference exceeding a preset threshold causes pattern rejection. However, the performance of any two channel scanner is ultimately limited by residual mutual distortion or misregister, either inherent in the electro-optical devices used or inherent in the patterns under inspection, or both.In order to get the best out of such a system it is necessary to be able to correct for residual distortions.

The present invention provides, according to one aspect, a method for comparing two nominally identical patterns comprising using means to compare a first array of elements from one of the patterns with each of a plurality of arrays of elements from the other pattern generally similar to each other and to the first array, the plurality of arrays being slightly misaligned with respect to one another whereby, if the two patterns are identical one of the plurality of arrays will correspond within predetermined limits to the first array. Such a method, and apparatus for carrying out the comparison is useful where the two patterns are likely to be slightly misaligned.The effect is that the array from the first pattern is compared with a number of arrays which are slightly misaligned with respect to one another from the second pattern, the misalignment of the second arrays being chosen so that it is likely that the misalignment of one of the second arrays will correspond to the misalignment of the first array.

The invention also provides according to a further aspect apparatus for comparing two nominally identical patterns comprising means for providing a first array of signals corresponding to a first array of elements from one of the patterns, means for providing a plurality of arrays of signals corresponding to a plurality of arrays of elements from the other pattern, the plurality of arrays being generally similar to each other and to the first array of elements, the plurality of arrays of elements being slightly misaligned with respect to one another, means for comparing the first signal array with each of the other signal arrays, whereby if the two patterns are identical, one oof the plurality of signal arrays from the other pattern will correspond within predetermined limits with the first signal array.

In one arrangement the two patterns may be virtually superimposed on one another and the plurality of arrays from the other pattern may be provided by relative move mcnt of the two virtually superimposed patterns. In this case, however, it will take time for each of the plurality of arrays to be produced and until the correct array is found.

In a preferred arrangement, however, the comparison takes place in real time so that all of the plurality of arrays from the other pattern are continuously compared with the array from the one pattern. In this case, if the two patterns are identical then at least one of the plurality of arrays from the second pattern will always correspond with the array from the first mentioned pattern.

Preferably the plurality of arrays is provided by a larger array of elements and the plurality of arrays each comprise a selection of elements from within that larger array.

Although it is conccivable to compare two patterns corresponding to, for example, a complete sheet of printed matter or a complete printed circuit board, this would in practice require a very large number of elements to be inspected simultaneously.

Preferably, therefore, the patterns are derived from corresponding small interrogation areas of larger patterns, which interrogation areas are scanned across the larger patterns. As difficulty of relative alignment between the scanned areas of the two patterns has been a particular problem in the past, the present invention is particularly applicable.

Although by pattern we generally mean an arrangement in two dimensions, and the method and apparatus to be described will be limited to two dimensional patterns, they may be derived from three dimensional objects. This may be arranged by comparing holograms of three dimensional objects or by utilising sensing probes to produce the array.

The invention is particularly applicable to optical type scanners.

The arrays may comprise a square matrix of elements or a rectangular matrix of elements or, for certain purposes, a circular, hexagonal or other matrix array of elements might be used.

A preferred embodiment of the invention will now be described by way of example only and with reference to the accompanying drawings in which: Figure 1 shows in diagrammatic form an optical scanner with which the present invention might be used; Figure 2 illustrates a test pattern and a reference pattern with the relative size and disposition of the arrays marked thereon, and Figure 3 shows two arrays in the form of square matrices.

Referring firstly to Figure 1, there is illustrated an optical scanner incorporating a reference pattern 10, which may be, for example, a perfect specimen of a printed circuit board and the nominally identical test pattern 11 may comprise a circuit board which it is desired to compare with the reference pattern 10. The two patterns 10, 11 are illuminated by suitable lamps 12, 13 respectively. The two patterns 10, 11 are arranged parallel and face to face with each other and mounted between the two patterns is a double sided plane mirror 14 rotatable about an axis 15 by means of a motor not shown. The axis 15 is generally parallel to the planes of the patterns 10, 11.

Two similar lens systems 16, 17 are arranged on opposite sides of the mirror 14, the axes of the two lens systems 16, 17 being coincident and generally at right angles to a line from the axis 15 to each of the patterns 10, 11. The lens systems 16, 17 are such as to focus at the points 18, 19 images of the patterns 10, 11 respectively, reflected from opposite faces of the mirror 14. Optical detectors are placed at the points 18, 19 as will be described in detail.

As will be well understood as the mirror 14 rotates, the image passed to the points 18, 19 will comprise a small inspection area which is scanned across the patterns 10, 11 by means of the rotating mirror so that the complete area of the patterns 10, 11 can be interrogated. The advantage of the scanning arrangement is, of course, that a small area and consequently great detail can be examined.

Such a scanning arrangement is generally well known. Conventional photoelectric detectors are placed at the points 18, 19 and the signals from these two detectors are compared. Providing the signals from the two detectors remain generally similar during the scanning process, then the test pattern may be considered to be acceptable.

However, there have arisen problems of alignment with such scanners. It will be understood that it is difficult to mount all the optical components and the patterns 10, 11 sufficiently accurately to ensure that the detectors at points 18, 19 are interrogating exactly corresponding areas of the two patterns 10,11. Any slight inaccuracy in the positioning of the various components will cause the test pattern 11 to be rejected when it is within acceptable limits. Furthermore, some slight misalignment of different parts of a pattern might be acceptable.

In apparatus according to the present invention, the detectors placed at 18, 19 are not simple photodetectors but comprise solid state detector matrices comprising n x n and m x m elements respectively. In the preferred embodiment described n = 5 and m = 9 but other numbers might be used.

The effect of this is illustrated in Figure 2 in which, as can be seen, the test pattern is scanned by an array comprising a matrix of 5 x 5 elements and the reference pattern is scanned by an array comprising a matrix of 9 x 9 elements. It will be noted that the elements in the 9 x 9 matrix are the same size as in the 5 x 5 matrix so that the area covered by the matrix scanning the reference pattern is larger than that covered by the matrix scanning the test pattern.

Each element in the matrices produces a separate electric output and these outputs are fed to a suitable computer and analysed in the manner to be described. For simplicity, referring to Figure 3, it will be seen that the elements in the two matrices are num bered all, a12, a13 a55 and bull, bl2, b13 b99 respectively.

As has been mentioned above, the areas on the two patterns being interrogated sometimes will not coincide where single detectors are being used. In the present arrangement, it is intended that the scanner should be arranged so that the area of the test pattern interrogated should coincide with the middle of the area of the reference pattern interrogated (that is correspond to square 10 of Figure 3). However it will be clear from Figure 2 that even if the area of the test pattern interrogated does not exactly coincide with the centre of the area of the reference pattern interrogated owing to faults, it is still likely that the interrogated area of the test pattern will still be within the interrogated area of the reference pattern.

It is desirable, therefore, to compare the matrix of signals from the test pattern with all possible arrangements of that size of matrix within the matrix from the reference pattern. In other words, to virtually move the interrogated area of the test pattern around within the interrogated area of the reference pattern until coincidence is found.

This may be done by virtually superimposing the matrix of signals from the test pattern on the matrix of signals from the reference pattern and causing this relative movement but this would mean that it would take some time before the coincidence would be found. In practice, we prefer to provide at all times a set of signal matrices from the matrix from the reference pattern which set of signal matrices will correspond to every possible position of the matrix from the test pattern within the matrix from the reference pattern. Some of this set of signal matrices is number 11 - 14 on Figure 3. Then the matrix of signals from the test pattern is continuously compared with this large number of signal matrices from the reference pattern so that continual coincidence can be observed.

The mathematical way of considering this is as follows.

The basic algorithm is outlined as follows.

If the test pattern matrix A as illustrated in the upper half of Figure 3, and the matrix B from the reference pattern is illustrated in the lower half of Figure 3 and a decision is required as to whether all the elements of the A matrix can be matched within a prescribed tolerance e to corresponding elements of any sub-matrix of 5 x 5 neighbouring elements taken from the 9 x 9 B matrix I a,, bl,l < e and I a12 - b12 I < e and I a13 b,3 l < e and | I a14 b l < e and a15 - b15 l < e and ( a21 - b21 l < e and . . and a55 - b55 l < e (i.e. in this case the A matrix corresponds to the upper left hand elements of the B matrix within box 11.) or l all - b12 l < e and a12 - b1 l < e and | a13 - bl4 l < e and I al4 - b,5 l < e and l a15 b16 l < e and l a21 - b22 l < e and . . and a55 - b56 or (or a further 23 possibilities) At least one of these total of 25 possibilities must produce a "Yes" result if the A matrix corresponds with one of the possible matrices from the B matrix, that is if the test pattern corresponds to the reference pattern.

Which one of the twenty five possibilities provides the "Yes" result may vary as the test pattern is scanned.

All the above comparisons can readily be performed at very high speeds and in real time using standard electronic signal processing techniques without recourse to a digital computer.

The invention is not restricted to the details of the foregoing example. For example, in place of a plane mirror there may be provided a rotating polygon with an even number of plane facets each pair of diametrically opposite facets being mutually parallel. The detector matrices may not be square; if a prior knowledge of polarisation of distortion is available then reactangular matrices might be more useful to obtain adequate distortion allowance with a minimum number of parallel channels. The size of the matrices have been chosen to be 5 x 5 and 9 x 9 and these are probably about the optimum arrangement.If the matrices are too small then it becomes increasingly more likely that an erroneous chance match be tween the two patterns could be obtained while if the matrices are too large the problem of dealing with vast numbers of parallel signal channels becomes an uneconomic engincering proposition.

Although the invention has been described with reference to test patterns and a reference pattern, in a production process in which several identical items are produced simultaneously, it may be sufficient to comparc two test patterns for mismatch.

With some production items it might be desirable to view the surface always in normal incidence in which case, additional optical components, such as a cylindrical lens could be incorporated in both channels.

The optimum spectral response to the measuring system will probably simulate the Overall eye response. However, any desired combination of colour and heat filters could be installed, for example, immediately after or between the lenses of the optical system in order to correct the overall system re sponsc.

It will be understood also that it is not necessary to place the photocells themselves at the positions 18, 19. In some circumstances, it may be preferable to arrange at these two positions, an array of optical fibres which may lead the light from that point to photodetcctors at a separate point.

Furthermore, in place of an array of 9 x 9 detectors from which are produced in this casc, 25 arrangements of 5 x 5 arrays, the light received from the reference pattern could be split, for example, into 25 different components by suitable beam splitters, each of the separate 25 beams being passed to a separate 5 x 5 array of photodetectors. It will, of coursc, be arranged that the optical beam will be misaligned with respect to each of the arrays. The advantage of such an arrangement would be that, whereas in the arrangement so far described the minimum relative misalignment between two reference arrays, (considering them. for the momcnt, as superimposed) is by 1 detector width whereas the use of beam splitters allows a reduction of this width. In this way the resolution of the apparatus can be increased or decreased as desired to a particular degree.

In cases where it is desirable to compare all test items on a production line with an independent reference item (as opposed to testing each item against its neighbour) the complexity of providing a second, synchronised line carrying reference items could be avoided by employing two-dimensional optical scanning of a single reference item, synchronised with the one dimensional (optical)/one dimensional (mechanical) scanning of the production items, i.e. it may be more convenient to simulate motion in the transport direction by appropriate optical scanning. One great benefit of this approach would be the elimination of damage to the reference items by continual retransportation.

The inverted-reference configuration shown in Figure 1 is by no means the only way of achieving suitably synchronised scanning. Arrangements allowing both reference and test patterns to be of the same general orientation are possible using, for example, a rotating polygon extended in length to allow simultaneous use of different areas of the same mirror facet.

The invention is not restricted to the details of the foregoing examples.

WHAT WE CLAIM IS: 1. A method for comparing two nominally identical patterns comprising using means to compare a first array of elements from one of the patterns with each of a plurality of arrays of elements from the other pattern generally similar to each other and to the first array, the plurality of arrays being slightly misaligned with respect to one another whereby, if the two patterns are identical one of the plurality of arrays will correspond within predetermined limits to the first array.

2. A method as claimed in claim 1 in which the comparison takes place in real time so that all of the plurality of arrays from the other pattern are simultaneously compared with the first array from said one pattern.

3. A method as claimed in claim 1 or 2 in which the plurality of arrays is provided by a larger array of elements and the plurality of arrays each comprise a selection of elements from within that larger array.

4. A method as claimed in any of claims 1 to 3 in which all of the arrays are simultaneously scanned across their respective patterns in synchronism.

5. A method as claimed in claim 1 substantially as hereinbefore described.

6. Apparatus for comparing two nominally identical patterns comprising means for

**WARNING** end of DESC field may overlap start of CLMS **.

Claims

**WARNING** start of CLMS field may overlap end of DESC **. computer. The invention is not restricted to the details of the foregoing example. For example, in place of a plane mirror there may be provided a rotating polygon with an even number of plane facets each pair of diametrically opposite facets being mutually parallel. The detector matrices may not be square; if a prior knowledge of polarisation of distortion is available then reactangular matrices might be more useful to obtain adequate distortion allowance with a minimum number of parallel channels. The size of the matrices have been chosen to be 5 x 5 and 9 x 9 and these are probably about the optimum arrangement.If the matrices are too small then it becomes increasingly more likely that an erroneous chance match be tween the two patterns could be obtained while if the matrices are too large the problem of dealing with vast numbers of parallel signal channels becomes an uneconomic engincering proposition. Although the invention has been described with reference to test patterns and a reference pattern, in a production process in which several identical items are produced simultaneously, it may be sufficient to comparc two test patterns for mismatch. With some production items it might be desirable to view the surface always in normal incidence in which case, additional optical components, such as a cylindrical lens could be incorporated in both channels. The optimum spectral response to the measuring system will probably simulate the Overall eye response. However, any desired combination of colour and heat filters could be installed, for example, immediately after or between the lenses of the optical system in order to correct the overall system re sponsc. It will be understood also that it is not necessary to place the photocells themselves at the positions 18, 19. In some circumstances, it may be preferable to arrange at these two positions, an array of optical fibres which may lead the light from that point to photodetcctors at a separate point. Furthermore, in place of an array of 9 x 9 detectors from which are produced in this casc, 25 arrangements of 5 x 5 arrays, the light received from the reference pattern could be split, for example, into 25 different components by suitable beam splitters, each of the separate 25 beams being passed to a separate 5 x 5 array of photodetectors. It will, of coursc, be arranged that the optical beam will be misaligned with respect to each of the arrays. The advantage of such an arrangement would be that, whereas in the arrangement so far described the minimum relative misalignment between two reference arrays, (considering them. for the momcnt, as superimposed) is by 1 detector width whereas the use of beam splitters allows a reduction of this width. In this way the resolution of the apparatus can be increased or decreased as desired to a particular degree. In cases where it is desirable to compare all test items on a production line with an independent reference item (as opposed to testing each item against its neighbour) the complexity of providing a second, synchronised line carrying reference items could be avoided by employing two-dimensional optical scanning of a single reference item, synchronised with the one dimensional (optical)/one dimensional (mechanical) scanning of the production items, i.e. it may be more convenient to simulate motion in the transport direction by appropriate optical scanning. One great benefit of this approach would be the elimination of damage to the reference items by continual retransportation. The inverted-reference configuration shown in Figure 1 is by no means the only way of achieving suitably synchronised scanning. Arrangements allowing both reference and test patterns to be of the same general orientation are possible using, for example, a rotating polygon extended in length to allow simultaneous use of different areas of the same mirror facet. The invention is not restricted to the details of the foregoing examples. WHAT WE CLAIM IS:

1. A method for comparing two nominally identical patterns comprising using means to compare a first array of elements from one of the patterns with each of a plurality of arrays of elements from the other pattern generally similar to each other and to the first array, the plurality of arrays being slightly misaligned with respect to one another whereby, if the two patterns are identical one of the plurality of arrays will correspond within predetermined limits to the first array.

5. A method as claimed in claim 1 substantially as hereinbefore described.

providing a first array of signals corresponding to a first array of elements from one of the patterns, means for providing a plurality of arrays of signals corresponding to a plurality of arrays of elements from the other pattern, the plurality of arrays of elements being generally similar to each other and to the first array of elements, the plurality of arrays of elements being slightly misaligned with respect to.one another, means for comparing the first signal array with each of the other signal arrays, whereby if the two patterns are identical, one of the plurality of signal arrays from the other pattern will correspond within predetermined limits with the first signal 'array.

7. Apparatus as claimed in claim 6 in which the means for comparing the signals is adapted to simultaneously compare the first signal array and each of the other signal arrays.

8. Apparatus as claimed in claim 6 or 7 in which the plurality of arrays of elements is provided by a larger array of elements and the plurality of array of elements each comprise a unique selection of elements within that larger array.

9. Apparatus as claimed in any of claims 6 to 8 in which scanning apparatus is provided whereby all the arrays of elements are simultaneously scanned across their respective patterns in synchronism.

10. Apparatus as claimed in any of claims 6 to 9 in which the arrays are square.

11. Apparatus as claimed in any of claims 6 to 9 in which the arrays are rectangular.

12. Apparatus as claimed in any of claims 6 to 11 in which the signals in each signal array are provided by a photosensitive cell.

13. Apparatus as claimed in claim 6 substantially as hereinbefore described with reference to the accompanying drawings.