TW201504598A - Method for self-calibrating a rotary encoder - Google Patents

Abstract

A method for self-calibrating a rotary encoder including a single read-head and a circular scale, comprises the steps of: acquiring calibration samples by the read-head for rotational angles of the circular scale; and estimating spatial frequency and spatial distortion parameters of the encoder from the calibration samples for self-calibrating the rotary encoder.

Description

Self-correcting method for rotary encoder

The present invention relates to measuring devices, and more particularly to absolute rotary encoders for measuring absolute rotational angles.

Accurate estimation of linear position and rotation angle is an important task in industrial automation and similar applications. Accurate measurements are required for applications such as numerical control (CNC) machines, drill bits, robotic arms or laser cutting machines and assembly lines. Feedback control is often used for precision measurements.

A typical encoder includes a scale and a read-head. Optical encoders are commonly used to measure absolute or relative linear position or angle of rotation. The relative encoder measures the relative position or angle over the scale period and counts the number of scale cycles traversed to determine the absolute position or angle. Absolute encoders do not require memory or power to store the current position or angle, and are available at all times, especially at startup.

The optical encoder can be linear or rotary. A linear encoder measures the position and a rotary encoder measures the angle. Conventional absolute rotary encoders typically use multiple tracks and apply sine-based cosine interpolation To achieve high resolution.

A single track absolute linear encoder using a single scale and a single CCD/CMOS sensor is described in this patent application. The encoder does not use conventional sine-based cosine interpolation. Instead, the encoder detects edges or zero-crossing in the scan line and fits the model to the edge positions to obtain high resolution absolute position information. The encoder uses a linear read head to capture a one-dimensional image of a linear scale.

Precise machining and manufacturing equipment requires highly accurate rotary encoders. However, rotary encoders may introduce several types of errors during manufacturing. These errors include the error of the scale pattern, the error of the scale on the spindle, the head alignment error, and the circuit noise error.

For rotary encoders, the spacing of the scale lines changes due to the circular nature of the scale. Another source of error is the eccentricity induced when the scale on the turntable is placed on the shaft. In addition, out-of-plane motion (wobble), as well as mounting misalignment, can also result in variations in the distance between the read head and the scale. These factors affect the overall accuracy of the rotary encoder. The encoder corrects manufacturing variations, scale pattern errors, mounting scale errors on the spindle, read head alignment errors, and circuit noise errors. Temperature variability and mechanical vibration can cause further distortion during operation, further reducing accuracy.

Because of the proximity to the light source, the center of the sensor receives more light than the side. This causes vignetting, where the one-dimensional image captured is brighter at the center and darker at the side. Vignetting leads to zero crossing There is an error in the cross point (edge), thus reducing the overall accuracy.

Several prior art methods require multiple additional read heads to counteract errors due to eccentricity. For example, reference is made to U.S. Patent No. 6,215,119 and U.S. Patent No. 7,143,518. An equal mean (EDA) method is described by Masuda et al. in "High accuracy calibration system for angular encoders" by J. Robotics and Mechatronics, 5(5), 448-452, 1993. However, rotary encoders that use multiple read heads to reduce eccentricity errors add cost to the system and make the system design large and cumbersome.

Conventional methods also require precise movement of the rotating components for self-correction. For example, U.S. Patent No. 5,138,564 discloses a method of making the encoder move at a low speed and at a high speed. U.S. Patent No. 6,598,196 drives a servo system on a predetermined trajectory such that encoder errors occur at frequencies outside of the servo feedback loop. Such requirements increase the strength and time of correction.

U.S. Patent No. 7,825,367 describes a self-correcting rotary encoder in which the angular difference is defined as a Fourier series. The correction of the rotary encoder based on sine and cosine interpolation can be as described in U.S. Patent No. 8,250,901. The voltage data corresponding to the sine and cosine of the rotation angle is fitted to the ellipse. Linear correction parameters are obtained by converting the ellipse into a circle.

A rotary encoder capable of self-correction is described in U.S. Patent No. 7,825,367. The rotary encoder includes a turntable having an angle code, a light source, and a line sensor (CCD) that can read the angle code. The processing unit retrieves the read value f(θ) of the predetermined angle. The difference between the read values f(θ+□) and f(θ) in the reading range of the line sensor is g(θ, □). The difference is determined as a Fourier series. Here, the rotation angle θ at a position is Obtained by analyzing the CCD image. The self-calibration is based on finding the angle of rotation at two different locations and analyzing the difference for self-correction.

DETAILED DESCRIPTION OF THE INVENTION A self-correcting, single track, single read head absolute rotary encoder is provided. The encoder takes a full circle (360°) or a measurement of one part of the full circle. Thus, the encoder compensates for any errors or distortions introduced during manufacturing and subsequently introduced during use due to changes in environmental or mechanical conditions. The encoder can also compensate for vignetting caused by illumination variations.

Embodiments of the invention do not require multiple read heads to counteract eccentricity errors. This can greatly reduce the cost and complexity of the encoder. Moreover, embodiments of the present invention do not require the motor to move at multiple speeds or predetermined correction trajectories. In addition, the present invention can also correct other mounting errors, such as gap changes and shaft swings.

100‧‧‧round scale

101‧‧‧Alternating light reflection

102‧‧‧Do not reflect

103‧‧ Dibun sequence

110‧‧‧Read head

111‧‧‧Sensor

112‧‧‧(LED) light source

114‧‧‧Sensor linear array

115‧‧‧Digital Signal Processor

116‧‧‧Rotation center point

120‧‧‧High resolution phase P

130‧‧‧ Turntable

150‧‧‧ Calibration sample

151‧‧‧ test sample

160‧‧‧ Estimate

161‧‧‧Frequency F and distortion parameters α and β

170‧‧‧Modeling

171‧‧‧Parameter function

190‧‧‧determined

195‧‧‧ phase

200‧‧‧Example

300‧‧‧frequency variation

301‧‧‧ High frequency variation

302‧‧‧Low frequency variation

400‧‧‧α(θ)

500‧‧‧variation

600‧‧‧fourth polynomial

700‧‧‧ measurements

800‧‧‧Scale factor

900‧‧‧ offset factor

F‧‧‧frequency

α, β‧‧‧ distortion parameters

P‧‧‧ phase

1A is a schematic diagram of a rotary encoder according to an embodiment of the present invention; FIG. 1B is a schematic diagram of a circular scale of a sector according to an embodiment of the present invention; and FIG. 1C is a circular scale according to an embodiment of the present invention; Schematic diagram of a ruler and a linear read head; FIG. 1D is a block diagram of an encoder for correcting FIG. 1A according to an embodiment of the present invention; 2 is a graph of spatial frequency F(θ) and rotation angle according to an embodiment of the present invention; FIG. 3 is a graph of spatial frequency variation caused by noise according to an embodiment of the present invention; The figure is a graph of spatial distortion parameter α(θ) variation and rotation angle according to a specific embodiment of the present invention; and FIG. 5 is a graph of spatial distortion parameter β(θ) variation and rotation angle according to an embodiment of the present invention; Figure 6 is a graph of a fourth-order polynomial fitted to α(θ) according to an embodiment of the present invention; Figure 7 is a plot of a scan line obtained by a one-dimensional sensor and depicting vignetting; Figure is a graph of scaling factors in accordance with an embodiment of the present invention; Figure 9 is a graph of an offset factor in accordance with an embodiment of the present invention; and Figure 10 is a diagram of a vignetting correction according to an embodiment of the present invention Correct the sensory value graph.

A specific embodiment of the present invention provides a single track type absolute rotary encoder. The read head can be a linear charge coupled device (CCD) or a complementary metal oxide semiconductor (CMOS) to capture a one-dimensional image of a rotating circular scale. The one-dimensional image includes a linear pixel array. The scale includes reflections and non-reflections configured according to the de Bruijn sequence region. The Dibun sequence is suitable because the style itself is actually circular.

Absolute circular scale

Figure 1 shows a small portion of a circular scale 100 of an absolute encoder in accordance with an embodiment of the present invention. The details of the scale are described in U.S. Patent Application Serial No. 13/100,092. This scale is used to determine the high resolution phase P 120.

The scale can include alternating light reflections 101 and non-reflection 102 marks or bits. The markers can also alternate between transparent and opaque depending on the relative position of the light source to the read head. Each mark has a B micron width which is the scale resolution. The width B of each mark is half-pitch. In a specific embodiment, B is equal to 20 microns. Since the dimensions of the indicia are relatively small, the exemplary indicia shown in the figures are not drawn to scale.

The read head 110 is mounted at a distance from and parallel to the scale. The read head includes a sensor 111, a (LED) light source 112, and an optional lens. The sensor can be an array of detectors consisting of N sensors, for example, N can be equal to 512. The array can be a complementary metal oxide semiconductor (CMOS) or a charge coupled device (CCD). The read head can also be associated with a digital signal processor 115 and memory connected to the sensor. It should be appreciated that other types of processors can be used.

Marks or bits on the exemplary scale 100 can be configured on the turntable 130 or the shaft. The only requirement is that the tag be configured sequentially for a particular code or non-periodic sequence.

As shown in FIGS. 1B and 1C, the marks are arranged on the scale 130 as sectors in a circle. Read head 111 includes a linear array of sensors 114. Here, the read head is tangentially centered at an offset 115 relative to the center point of rotation 116. Therefore, it should be noted that the sensor pixels near the ends of the linear read head can observe a sector portion that is wider than the sensor near the center of the read head. This causes the signal of the one-dimensional sensor to be distorted.

Correction

The DSP performs the correction of the encoder as shown in Fig. 1D. This correction can be performed offline or continuously during the operation of the encoder.

During the correction, for a 360 degree full turn or a portion of the rotation, the calibration sample 150 for the rotation angle of the circular scale 101 is captured by the read head 111. This part of the rotation can be used when the scale oscillates circularly instead of going through a full circle. It should be noted that calibration samples for multiple rotations can also be retrieved.

From these corrected samples, 160 frequencies F and distortion parameters α and β 161 can be estimated. The frequency F and distortion parameters α and β can be stored directly in the memory, for example, as a lookup table, and are sufficient to accurately determine the phase of the encoder during operation. This lookup table may be advantageous if the lookup fast or lookup time and memory used are less than the evaluation parameter function.

For convenience, we modeled the variation of the frequency F and the distortion parameters α and β with the parameter function 171, and stored variations in the memory for use during immediate operation of the encoder on-line.

Immediate operation

During immediate operation, the phase 195 of the encoder can be determined 190 by the test sample 151 and the modeled variations of the frequency F and the distortion parameters α and β. It will be appreciated that such variations can be obtained from the raw parameters of the lookup table when stored in memory during operation. It should be appreciated that these parameters can also be retrieved during immediate operation of the encoder.

Here, the details of the encoder structure and correction are described in detail.

Dibun sequence

In order to achieve 100% information density on the scale, a sequence of bits is used. Each subsequence has a finite length and is unique, for example, the Dibun sequence 103. Each unique sequence corresponds to an approximate phase angle. The object of the invention is to self-correct the encoder thereby obtaining a fine or precise angle.

The N-th order k-element Dibun sequence B(k,n) is a cyclic sequence consisting of a given letter (the number of angles) and is of size k, so each possible sub-sequence of the letter length n appears to be exactly one time by successive A sequence of characters. If each B(k,n) has a length k ⁿ , then there are (k! ^k(n-1) )/k ⁿ different Dibun sequences B(k,n). When the sequence is truncated before or after, the resulting sequence also has the same uniqueness of n values. It should be noted that any non-periodic sequence with a non-repetitive subsequence can be used.

In order to be able to decode, the detector array requires a field of view (FOV) consisting of at least n bits. With a half-pitch B = 20 microns and a 16-order Dibun sequence, a FOV of 16x20 = 320 microns is required on the scale. In a specific embodiment, the field of view is designed to have 1 to 2 mm to have The accuracy of the desire.

In the case of Nyquist sampling, each bit of the sequence, i.e., each half-pitch of the scale, is mapped to at least two pixels of the linear detector array. This requires at least 16 x 2 = 32 pixels, which is much lower than the number of pixels of a conventional sensor. To handle optical aberrations, such as defocus blur or diffraction, the number of pixels per half-pitch can be increased.

Since the scale is circular, the angles of the reflective and non-reflective areas are equal, and are not equidistant when using the line sensor, as shown in Figure 1C. Due to the use of the circular scale, the width of the reflective/non-reflective regions increases at both ends of the sensor. Therefore, the spatial frequency F is not constant on the sensor.

Let z(i) be the zero crossing point (edge position) detected, P be the phase angle, and F be the frequency. Let k(i) be the number of bits between two consecutive zero crossings z(i) and z(i+1). If we define: Then the i-th zero-crossing point of the rotary encoder can be written as a cubic model of c(i): z ( i )= P + Fc ( i ) + αc ( i ) ² + βc ( i ) ³ , where the cubic model The parameters include phase P, spatial frequency F, and spatial distortion parameters α and β. This model takes into account errors caused by uneven spacing of the scale lines on the disk 130. Using N zero crossing points, N equations can be obtained. For example, if there are N zero crossing points z(1), ..., z(20), the corresponding c(1), ..., c(20) are known. These equations describe linear systems with unknowns P, F, α, and β. The linear system is solved to obtain values of P, F, α, and β.

The rotation angle θ is obtained by using the equation θ=P/F *360/K+ approximate position (Coarse_Position), where K is the number of gradations of the scale, and the approximate position is based on the image-based code. The phase angle of the sequence. For example, K can be 1024.

Self-correction

The estimated parameters F, α, and β 161 are expressed as F(θ), α(θ), and β(θ) by the actual rotation angle θ. These particular embodiments take into account variations in the three parameters F(θ), α(θ), β(θ). These parameters have small variations (normal variations) due to imaging noise.

In the absence of any mechanical errors and any eccentricity errors installed, the variation of these parameters shall be a normal variation of one full circle or a portion thereof when θ is changed from 0 to 360 degrees. However, due to eccentricity, sway, or gap variation (distance between the read head and the scale), the spatial parameters F(θ), α(θ), β(θ) show greater variation than attributable to noise.

Figure 2 illustrates an example 200 of the estimated F(θ) versus a full turn rotation angle.

Figure 3 illustrates the frequency variation 300 in more detail. The high frequency variation 301 is caused by noise. The low frequency variation 302 is caused by eccentric sway and gap changes. The goal of the invention is to amend these variations.

In the course of this self-correction procedure, these variations are beneficial Modeled with a parameter function. The scaled shaft rotates one full turn (360°) or one part (less than 360°), and the encoder scale is sampled at several positions. For example, the scale can be rotated 2 degrees each time, and the sensor image corresponding to the angle is stored in the memory. The estimates of the frequency and distortion parameters for all of these angles are stored along with the estimated encoder phase P.

Curve Fitting

Appropriate parametric functions or splines are used to model variations in frequency and distortion parameters using least squares fitting.

Figure 4 is a graph showing the variation of α(θ) 400, which can be modeled by a fourth-order polynomial model of the rotation angle θ: α ( θ ) = t ₁ + t ₂ θ + t ₃ θ ² + t ₄ θ ³ + t ₅ θ ⁴ , where t ₁ , t ₂ , t ₃ , t ₄ and t ₅ are model parameters. These model parameters are estimated using the least squares fit of the estimated α(θ). Figure 5 is a graph showing the variation 500 of β(θ). It should be noted that the number and form of the models of these three parameters need not be the same. For example, the frequency F(θ) can be modeled using a spline basis function, and α(θ) and β(θ) can be modeled using polynomial functions. Figure 6 is a graphical representation of the estimated α(θ), which is fitted to the fourth degree polynomial 600 of the full circle. After the curve is fitted, the model parameters are stored in the memory of the DSP 115.

operating

During encoder operation, the last encoder position can be used to determine the frequency and spatial distortion parameters of the current position. These values are used to determine the phase P. Alternatively, the frequency and distortion parameters can be determined with the phase in an iterative manner. This can be used at startup because the last encoder phase is unknown or invalid in this case. As described above, the current rotation angle can be obtained, The first estimate of the frequency and distortion parameters. Using the estimated rotation angle θ, the parameters F, α and β of the current position can be re-determined with the respective models. The new values for these parameters are then used to re-determine the phase P.

In U.S. Patent No. 7,825,367 issued to Nakamura, the self-calibration is based on the angle of rotation at two different locations, and the difference for self-correction is analyzed and corrected. Nakamura does not describe spatial frequency and distortion parameters. The encoder according to the invention is not based on the actual angle of rotation as revealed by Nakamura, but based on the fundamental frequency and distortion parameters used to model the zero crossings at a particular angle of rotation.

Vignetting correction

As shown in Fig. 7, the vignetting correction can also be accomplished by taking the measured value 700 during the scale rotation.

As shown in FIG. 8, for each pixel p of the sensor, the maximum pixel value m ₁ (p) is a scaling factor 800, and as shown in FIG. 9, the minimum pixel value m ₂ (p) It is an offset factor of 900. These factors are used for vignetting correction as follows.

As shown in Fig. 10, the sensed value i(p) of each position of 1000 is modified by the following formula: i(p)←255*(i(p)-m ₂ (p))/(m ₁ (p )-m ₂ (p))

This modification ensures that the minimum intensity of each pixel is set to zero and the maximum intensity of each pixel is set to 255 as the encoder rotates. This removes the vignetting effect.

Self-correction and vignetting correction as known in the art The steps of the method can be performed in a DSP or similar microprocessor connected to the memory and input/output interface.

100‧‧‧round scale

101‧‧‧Alternating light reflection

102‧‧‧Do not reflect

103‧‧ Dibun sequence

110‧‧‧Read head

111‧‧‧Sensor

112‧‧‧(LED) light source

115‧‧‧Digital Signal Processor

120‧‧‧High resolution phase P

130‧‧‧ Turntable

Claims (20)

A self-correcting method for a rotary encoder, the rotary encoder comprising a single read head and a circular scale, the self-correcting method comprising the steps of: using the read head to capture a corrected sample of the rotation angle of the circular scale; And estimating, by the corrected samples, the spatial frequency and spatial distortion parameters of the encoder for self-correcting the rotary encoder. The self-calibration method of claim 1, further comprising: taking a test sample of the scale; determining the phase of the encoder with the frequency and distortion parameters. The self-correcting method of claim 1, further comprising: modeling the frequency and the variation of the distortion parameters by using a parameter function; taking a test sample of the scale; and using the modeled The frequency and distortion parameters determine the phase of the encoder. The self-calibration method of claim 1, wherein the mark on the scale is configured as a sector, and the read head is tangentially centered at a deviation from a center of rotation of the scale . The self-correcting method of claim 1, wherein the read head data is obtained for a rotation angle of 360 degrees or less. The self-correcting method of claim 1, wherein the parameter function is a spline. The self-correcting method of claim 1, wherein the parameter function uses a least squares fitting. The self-correcting method of claim 4, wherein the parameter function is a fourth-order polynomial of the rotation angle. The self-correction method of claim 1, further comprising: storing the frequency and distortion parameters in a memory as a lookup table. The self-calibration method of claim 1, wherein the frequency and distortion parameters correct the eccentricity of the circular scale. The self-calibration method of claim 1, wherein the frequency and distortion parameters correct the sway of the circular scale. The self-calibration method of claim 1, wherein the frequency and distortion parameters correct a change in distance between the read head and the circular scale. The self-calibration method of claim 1, wherein the frequency and distortion parameters correct temperature or mechanical vibration of the encoder during operation. The self-correcting method of claim 1, wherein the frequency and distortion parameters are captured during immediate operation of the encoder. The self-correction method of claim 1, wherein the read head comprises a linear pixel array, and the self-correction method further comprises: measuring the intensity of the pixels to obtain a maximum intensity as a scaling factor, and as an offset The minimum strength of the factor. The self-correcting method of claim 15, wherein the pixel intensity is modified by the scaling and offset factors. The self-correction method according to claim 15, wherein the pixel intensity i(p) is modified according to the following formula: i(p)←255*(i(p)-m ₂ (p))/( m ₁ (p)-m ₂ (p)) wherein m ₁ (p) is the maximum intensity and m ₂ (p) is the minimum intensity. The self-correcting method of claim 1, wherein the scale is in the form of a de Bruijn sequence. The self-correction method according to claim 7, wherein the spatial frequency parameter is F(θ), and the spatial distortion parameters are α and β, and wherein the fourth-order polynomial is: α ( θ )= t ₁ + t ₂ θ + t ₃ θ ² + t ₄ θ ³ + t ₅ θ ⁴ , where t ₁ , t ₂ , t ₃ , t ₄ and t ₅ are the ones estimated by the least squares fit method The parameters of the fourth degree polynomial. The self-correcting method of claim 19, further comprising: measuring c bits between two consecutive zero-crossing points on the scale; and modeling the zero-crossing points into: z ( i )= P + Fc ( i )+ αc ( i ) ² + βc ( i ) ³ , where P is the phase value, F is the spatial frequency, and α and β are the spatial distortion parameters.