BACKGROUND OF THE INVENTION

This invention relates to opto-electronic scale-reading apparatus for use in measuring relative displacement of two members. In known such apparatus e.g. British Patent No 1,504,691 the scale comprises a diffraction grating co-operating with at least one other grating on a read head to produce interference fringes which move relative to the read head during a said displacement of the members, and said measurement is a count of said fringes. It is clear that such a scale has to be of diffraction quality, i.e. the accuracy and reliability of the measurement depends on such parameters as the regularity of the spacing of the scale marks, the sharp definition of the edges of the marks, and the freedom of the scale from scratches and like imperfec ions. Such a scale can be expensive to produce and protect especially when the scale has to be relatively long.

It is among the objects of this invention to overcome or reduce this difficulty.

It is also known to increase the number of signals obtainable from any two adjacent marks of the scale by phase quadrature interpolation. Known scale-reading apparatus can be subject to phase errors and consequent interpolation errors. It is optionally an object of this invention to overcome or reduce this difficulty.

BRIEF DESCRIPTION OF THE INVENTION

According to this invention there is provided apparatus for measuring displacement between two members, comprising a scale defined by marks provided on one of the members, a read head provided on the other member diffraction means for producing interference fringes having movement relative to said read head responsive to a said displacement, and detecting means for detecting said movement, characterized in that there is provided a spatial filter including said diffraction means and mounted on said read head the filter being tuned to a nominal periodicity lying within a band of periodicities defining the pass band of the filter, the aarks of said scale are defined by light sources positioned to illuminate said diffraction means and having a periodicity lying within said pass band.and the light from said sources interacting with said diffraction means to produce said fringes

It will be seen that in the apparatus according to this invention the diffraction mechanism takes place entirely in the read head. The scale is merely required to provide a pattern of light sources. Thus the scale is not required to be a diffraction grating and the provision of the marks

SUBSTITUTESHEET

on the scale does not have to be of diffraction quality. The scale may have relatively imperfect markings or the markings of the scale may be capable of being produced with greater economy than in known apparatus.

Further, the read head according to this invention is inherently convolutional , i.e. the fringes constitute a convolution of the scale pattern with a substantially sinusoidal pattern. This makes the read head substantially free from phase quadrature errors.

DESCRIPTION OF PREFERRED EMBODIMENTS

Embodiments of apparatus according to this invention, will now be described with reference to the accompanying drawings wherein:-

Fig 1. is a plan view of the apparatus, Fig.2. is a perspective view of Fig.l, Fig.3. is an enlarged detail of Fig.l, Fig.4. is a light ray diagram of a first embodiment, Fig.5. is a light ray diagram of a second embodiment , Fig.6. is a first diagram showing at (a) the response curve of a filter and at (b) a scale periodicity variation, Fig.7. is a second diagram showing at (a) the response curve of a filter, at (b) a band of scale periodicities, and at (c) a different position of the latter band, Fig.8. is a third diagram showing at (a) the response curve of a filter, at (b) a band of scale periodicities, and at (c) a different position of the latter band,

Fig.9. shows at (a) an enlarged representation of a scale showing a modulation of the scale marks, and at (b) demonstrating the ramp characteric of this modulation.

Fig.10. is a diagram of characteristics of the modulation showing different such characteristics at (a) and (b),and

Fig.11. is an enlarged representation of a scale showing amplitude modulation of the scale periodicity and at (b) the binary characteristic of the modulation in this case.

General Description of Apparatus.

Referring to Figs.l to 3, there,is shown a linear scale 3»0 secured to a track 11. A read head 12 is secured to a carriage 13 supported on the track 11 for linear movement in a direction X being the direction of the length of the scale. The scale has marks 14 (Figs.2, 3) defined by lines extending in a direction Y perpendicular to the direction X. The head 12 has an axis 12A extending in a direction Z perpendicular to both the directions X and Y. The head embodies a light source 15 positioned to illuminate the scale over a range 16. The head further comprises, in succession from the scale 10 and along the axis 12A, a first or index grating 17, a second or analyser grating 18, a lens 19, and a sensor assembly 20 preferably comprising sensor sections 21 (Fig.2) having output signals 22 representing movement of the carriage 13 along the track 11.

The gratings 17, 18 constitute a spatial filter F. The relative position of the scale 10 and the filter F is such that the light from the source 15 is reflected at the marks 14 to illuminate the filter F.

The scale 10 (Fig.3.) comprises a body 10B to which the marks 14 are applied at a given periodicity, i.e. at given periods or pitches. The marks may have a single periodicity indicated by periods PI. Alternatively, the marks may be arranged in accordance with a number of periodicities defining a band, the "scale band", and including said single periodicity as a dominant periodicity among a range of secondary periodicities, all as defined later herein. The scale band may be produced by a random variation in the periods of the markings along the length of the scale.

Said random variations are indicated in Fig.3. as a surface structure 23 having substantially randomly distributed reflective regions 24 including regions having the period PI. Such a scale profile can be more economical to produce than a scale in which only a single periodicity is present. As will be explained in detail later herein, the apparatus has a sampling region 16 spanning a portion of the length of the scale and determining the pass band of the filter F. The dominant periodicity of the light source defined by the scale marks 14 lies within the pass band of the filter. The filter responds to the latter sources and acts on the sensor 20 to produce the signals 22.

Filter Geometry.

The regions 24 define light sources and the grating 17 is spaced from the scale 10 to be illuminated by said sources and, by diffraction, to produce fringes 30 in a fringe plane 30A located at the side of the grating 17 remote from

the scale 10. Referring to Fig. 4, the grating 17 is an amplitude grating, typically a Ronchi grating, and use is made of the diffraction phenomenon known as "self-imaging" or "Fourier imaging" of periodic transmission masks.

This phenomenon requires for this type of grating that the following expressions are satisfied:

1/u + 1/v = λ/(n x D2 ² ) ⁽ 1)

D2 / D3 = u/(u + v) ( 2 )

D2 / Dl = v/(u + v) (3) wherein: u = the distance between a generating plane 31A and the grating 17, the plane 31A lying in the XY directions and containing a notional point source 31 being of substantially monochromatic light and giving rise to the fringes 30 which are represented by a square wave, as shown, since these fringes are a self-image of the grating 17; v = the distance between the grating's, 17,18; λ = the wave length of the light;

Dl = the pitch of a plurality of said point sources lying in the plane 31A and co-operating to re-inforce the fringe pattern:

D2 = the pitch of the grating 17;

D3 = the pitch of the grating 18;

D4 = the pitch of the fringes 30 formed at the plane 30A: n = a positive integer.

The head 12 and the scale 10 are matched by making the pitch Dl of the head and the pitch PI of the scale the same, and the head is so positioned relative to the scale that the plane 31A of the light sources 31 is substantially coincident with the plane, 10A, of the scale. The notional light sources 31 are then actual sources defined by light

reflected from surface features of the scale forming the dominant periodicity PI and associated secondary periodicities .

During relative movement of the head 12 and the scale 10, the resulting movement of the light sources 31 in the generating plane 31A in the direction X produces a corresponding movement of the fringes 30, also in the direction X, relative read head 12. If u and v are equal, the amount of the movement of the fringes 30 relative to the read head 12, is the same as that of said relative movement of the head and the scale. A hypothetical point sensor 21X situated in the plane 30A of the fringes will detect fluctuations in light intensity as the fringes pass across it. The grating 18 has a pitch D3 equal to the fringe pitch D4 and is arranged for its plane 18A to coincide with the plane 30A.

The sensor sections 21 are provided for sensing sub-divisions of the fringe pitch conveniently generated by dividing the grating 18 into sections 18B ^; 'Fig.2) whose grating marks are mutually offset. Four such grating sections 18B,and correspondingly four said sensor sections21, may be provided to divide the fringe pitch by four. Alternatively, a similar effect is achieved by placing the grating 18 in a tilted position relative to the grating 17 thereby to produce at the plane 30A moire fringes sensed in phase quadrature by the sensor assembly in a manner known per se.

It will be clear that, in this example, the performance of the head 12 is governed by relatively strict adherence to the relationships given by equation (1). Notably, the formation of the fringes is dependent on wave length. Departure from an ideally single wave length causes reduction in contrast of the fringes. This reduction is

made worse as the value of n is increased although a high value of n may be desirable for practical reasons e.g. so as not to be restricted to too small a spacing of the head 12 and the scale 10. However, a practical head can be constructed by using values n between 2 and 16, together with a value of 20 microns for Dl and a wave length of 900 nanometers.

Inevitably small variations in the spacing between the head and the scale can produce reading errors. The lens 19 which is introduced to overcome this difficulty is a telecentric device having a front focal plane preferably lying at the scale plane 10A and a rear focal plane lying at the plane, 21A, of the sensor assembly, and the lens 19 allows said variations without necessarily invalidating equation (2) and/or (3) .

In the second embodiment iFig.5) the parameters of the head 12 are given wholly by:

D2/D3 :u / u + v) (4) D2./D1 2v/ (u + v (5)

1/u + 1/v = λ/[ (n + *fc ) x D2 ² J ( S )

The restriction of equation (1) does not apply at all in this case. However, equation (6) should be applied when n is low and/or the light is substatially monochromatic. Otherwise, the fringe contrast is substantially independent of wave length and broad-band light, e.g. white light, may be used. Further, in this embodiment, fringes of a given pitch are formed dependent only on the ratio u v and not on the absolute values u and v. There is some loss of fringe contrast associated with the formation of fringes in this case, but this is overcome by using a phase grating for the grating 17. Generally, this embodiment would be the prefered embodiment of the invention.

The pitch Dl is also referred to as the "nominal periodicity" of the filter, and the filter may be said to be .tuned to read only those marks 14 of the scale 10 which have the nominal periodicity of the filter or as will be explained, which lie within the pass band of the filter.

The filter as such, denoted F, comprises a housing 12B (Fig.l) supporting the gratings 17,18 at the spacing v and a support means for supporting the housing 12B relative to the scale 10 at the distance u between the scale 10 and the grating 17. In the present example said support means is defined by the track 11 and the carriage 13.

Convolution

It can be shown on the basis of Fourier theory, that an optical convolution is performed between the two patterns, being respectively the scale patterns 24 and the fringe pattern 30, due to a single light source 31 illuminating the grating 17 (Figs. 3,4,5, ^' ;. Since the fringe pattern is substantially sinusoidal, it can be shown that said ^" convolution represents a spatial filtering of the light distribution of the scale in favour of the spatial periodicity of the fringe pattern produced by said single light source. In other words, the head 12 is a tuned spatial filter. The filtering action is strengthened by a second convolution, in this case between the fringe pattern 30 and the grating 10.

The convolutional character of the read head 12 has the advantage that the read head 12 is substantially independent of angular misalignment, particularly about the Z axis, between the read head 12 and the scale 10, thus rendering the read head substantially immune to quadrature phase error due to such misalignment. This is due to the fact that the gratings 17,18 are fixed one relative to the

other and the fringes 30 have a fixed alignment with the lines of the grating 17. Therefore, the head 12 can be set up _» relative to the scale, by simple mechanical methods, such as setting gauges and it is not normally necessary, during setting up, to monitor the phase of the signals 22 and make adjustments in the head position to eliminate phase errors as between the respective signals 22.

Relationship of Filter and Scale.

The periodicity to which the filter is tuned lies within a band of periodicities constituting the pass band of the filter or the "filter band" which is defined as the inverse of the length of the illuminated or sampling region 16 (Fig. 1) insofar as that region lies within the optical aperture of the grating 17.

The region 16 may be illuminated over a length less than the greatest possible aperture of the grating 17 in which case the effective aperture is less than said greatest possible aperture. In any case, the filter band is the inverse of the region 16. In practice, given that the scale has the periodicity PI , the filter F is designed to be tuned to the periodicity PI and the pass band of the filter F is chosen in terms of said region 16. To cope with a given tolerance in the periodicity of the scale, i.e. in the spacing of the marks 14, due to manufacturing tolerances, the pass band of the filter is made sufficiently wide to include that tolerance. So long as the dominant scale periodicity PI is detectably present on the scale in the sense of lying within said sampling region 16 and within the pass band of the filter. It will be clear that the greater the length of the region 16 the narrower is the filter band and vice versa.

Fig. 6 is a diagram showing the relationship between a given pass band FBI of the filter and the dominant periodicity PI as the only periodicity of the scale. The curve Fa represents the whole response of the filter F in terms of the contrast FC of the fringes 30 for different scale periodicities SP. A fringe contrast above a line FC1 is sufficient to produce a signal 22 (Fig.2).

So long as the periodicity PI lies within the band FBI, the filter F can respond to it and produce a signal 22 of acceptable amplitude. While being uniform within the sampling region, in any one position of the read head along the scale, the periodicity PI may vary, as between different positions of the sampling region along the scale, over a range P1A and produce a signal 22 so long as the range P1A lies within the filter band FBI. The filter responds in sympathy with any changes in the periodicity within the range P1A.

This is acceptable for a given error tolerance. However, the arrangement has the advantage of relatively good freedom from phase quadrature error. In a typical example, the nominal periodicity is 20 mm and the width of the pass band is 0.1 micron for a sample period 16 of 10mm. If the range P1A is 0.05 micron, the error tolerance would have to be 0.25%, i.e. 2.5 mm per m. However, this can be compensated for and be reduced, typically, to 20 microns per .

Fig. 7 shows a band of scale periodicities P1B present within the sampling region 16 and including the dominant periodicity PI substant ally at the centre of the band. If the dominant periodicity coincides with the nominal periodicity of the filter, the filter response is in accordance with the nominal periodicity. However, if, as - shown as (c) the position of the scale band relative to the

pass band of the filter is such that the dominant periodicity lies to one side of the nominal periodicity of the filter, the filter tends to respond to a scale periodicity Px closer to the nominal periodicity of the filter than the dominant periodicity; the dominant periodicity must still lie within the pass band of the filter. A consequence of this arrangement is that the accuracy of the apparatus is higher than in the example of Fig. 6. In other words, the introduction of a band of periodicities about the dominant periodicity, leads to improved accuracy while still maintaining said relatively good freedom from phase quadrature errors.

Fig. 8 shows a scale band PIC exceeding the filter band FBI. In this case, even though the dominant frequency still remains within the pass band FBI, the filter can see scale periodicities, not only in the pass band, but over the entire range, FB2, ^* of the filter curve Fa. This contributes to building up improved accuracy by virtue of allowing the filter to respond even more closely to its nominal periodicity.

Scale Geometry

Fig.9 represents a part length of the scale 10 showing positions P spaced along the scale at the dominant periodicity defined by the periods PI. A pair of reflective marks 14/1 are provided at two adjacent positions P at regular intervals I along the scale. The intervals I are each an integer multiple of the period PI and the sampling region 16 substantially extends over a distance equal to one ^" such interval I. In an example, the period PI is 20 micron, the interval I is 8mm and the sampling region is 10mm.

Further reflective marks 14/2 are provided on the scale in positions offset from the positions P by departures or distances D, thus giving rise to secondary periodicities defined by periods P2 which _j in this example, vary in accordance with a ramp-shaped characteristic . Alternatively, the characteristic may be sinusoidal (Fig. 10a) within each sampling region with corresponding sinusoidal variation in the secondary periodicities. Alternatively, the distances D may vary so that the characteristic is random (Fig. 10b).

In most cases it is desirable that the maximum departure D from any one position P is less than one half, preferably one quarter, of the period PI because any greater such departures could result in destructive interference in the filter F such that certain periodicities, including the dominant periodicity, are not detectable. This would lead to a condition that all or some periodicities are no longer detectably present in the apparatus with consequent failure of the reading.

The foregoing departures D may be described as phase or frequency modulation of the scale marks. Amplitude modulation may be provided (Fig. 11) by arranging the marks 14 at selected groups of positions P while leaving the remaining positions P unmarked as shown at PO. The unmarked positions may vary in any appropriate, regular or random pattern.

It would not be appropriate from the modulation point of view if, for example, every second or third position P were unmarked i.e. if the period between any two marks 14 were the same integer multiple of the nominal frequency Dl of the filter, but this may infact be done to provide what is in effect a coarse-pitch scale.