JP7748829B2 - Automatic correction method - Google Patents

Description

The present invention relates to an automatic correction method that automatically corrects an encoder's output signal to reduce the interpolation error of the output signal.

The Laserscale interpolator interpolates position data into the count value of the "SIN" and "COS" waveforms from the sensor. The deviation from the true value of the interpolated position data is called interpolation error. Interpolation errors arise from 1) a deviation in the center position of the Lissajous waveform created from the "SIN" and "COS" waveforms, 2) a deviation in the amplitude of the "SIN" and "COS", and 3) a deviation from 90 degrees in the phase difference between the "SIN" and "COS". In conventional methods, all of 1) to 3) are obtained by calculation from the instantaneous values of the "SIN" and "COS" waveforms at the specified phase (see Patent Document 1).

Japanese Patent Application Publication No. 8-122097

In a typical interpolator, if the "COS" offset is off, the angle of one point on the circumference as viewed from the center of the Lissajous waveform will also be off. Since position data is calculated using this angle, this results in an interpolation error.

In conventional methods, for example, COS offset detection and correction is done as follows: The average of the positive peak value (COS+) and negative peak value (COS-) of the COS waveform is taken as the center value of the COS waveform, and a bias is added to the COS waveform to make this zero.

When a noise waveform is present on the "COS" waveform, the noise's +peak value (N+) is added to the detected +peak value. Similarly, the noise's -peak value (N-) is added to the detected -peak value. In this case, if the magnitudes of N+ and N- are equal, there is no effect on the average of the +peak value ("COS"+) and -peak value ("COS"-), but if the magnitudes of N+ and N- are different, the average of the +peak value ("COS"+) and -peak value ("COS"-) will shift by N+ + N- (≠ 0).

In other words, with conventional methods, when noise with an asymmetric waveform is added above and below, an error occurs in the detected offset value. If the amount of noise differs between detecting the +peak value and detecting the -peak value, an error occurs in the detected offset value. More specifically, with a laser scale, ideally the detected position changes in proportion to the movement distance of the detection head or scale, but when there is an offset, a waveform fluctuation is superimposed on this change. This fluctuation appears as an error in the detected position, in other words, an interpolation error. In addition to offset, other factors that can cause interpolation error include amplitude error and phase error, which will be discussed later.

In encoder systems that read periodic signals on a scale as "SIN" and "COS" signals and convert the periodic signals into position signals, it is desirable to provide an automatic correction method for the "SIN" and "COS" signals that is less susceptible to the effects of noise and waveform distortion contained in the signals.

An automatic correction method according to one embodiment of the present invention is an encoder system that reads periodic signals on a scale as "SIN" and "COS" signals and converts the periodic signals into position signals. To minimize interpolation errors during conversion, the system detects and feeds back to the "SIN" and "COS" signals offsets to reduce them. Components synchronized with sine and cosine, respectively, are extracted from the radius fluctuation waveform of a Lissajous waveform produced by the "SIN" and "COS" signals to obtain the offsets of the "SIN" and "COS" signals.

In this automatic correction method, in order to maintain a constant radius for the Lissajous waveform generated by the "SIN" and "COS" signals from the scale, the sin and cos components of that period, the sin and cos components of half the period, or the sin and cos components of a further 1/nth period (where n = 3, 4, ...) are extracted and fed back to the "SIN" and "COS" signals from the scale to cancel them out.

Another aspect of the automatic correction method of the present invention is an encoder system that reads periodic signals on a scale as "SIN" and "COS" signals and converts the periodic signals into position signals. To minimize interpolation errors during conversion, the system detects the amplitude and phase differences between the "SIN" and "COS" signals and feeds them back to the "SIN" and "COS" signals so that they approach 0 and 90 degrees, respectively. From the radius fluctuation waveform of the Lissajous waveform formed by the "SIN" and "COS" signals, components synchronized with half the period of sine and cosine are extracted, and the amplitude and phase differences between the "SIN" and "COS" signals are obtained.

This invention provides an automatic correction method for SIN and COS signals that is less susceptible to the effects of noise and waveform distortion contained in the signals in an encoder system that reads periodic signals on a scale as SIN and COS signals and converts the periodic signals into position signals.

FIG. 1 is a diagram schematically illustrating an encoder system according to an embodiment. 1 is a diagram illustrating the principle of an automatic correction method according to an embodiment. 1 is a diagram illustrating the principle of an automatic correction method according to an embodiment. 1 is a diagram illustrating the principle of an automatic correction method according to an embodiment. 1 is a diagram illustrating the principle of an automatic correction method according to an embodiment. 1 is a diagram illustrating the principle of an automatic correction method according to an embodiment. FIG. 2 is a functional block diagram of a correction device.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
The automatic correction method (correction device) of this embodiment is applied to an encoder system that calculates displacement information based on two sine wave signals with a 90-degree phase difference output from an encoder, and corrects the sine wave signals prior to calculating the displacement information. Hereinafter, the two sine wave signals will be referred to as a sine wave and a cosine wave, and will be abbreviated as "SIN" and "COS," respectively.

When an offset occurs in either "SIN" or "COS", the radius of the Lissajous waveform drawn with "SIN" and "COS" fluctuates. In this embodiment, this radius fluctuation is used to detect the amount of offset based on the degree of radius fluctuation. Furthermore, in the waveform representing the radius fluctuation (hereinafter also referred to as the "radius fluctuation waveform"), the component synchronized with "COS" and the component synchronized with "SIN" are proportional to the offset amounts of "COS" and "SIN", respectively. Similarly, if the radius fluctuation waveform changes twice per Lissajous period, the gain difference (amplitude difference) between "SIN" and "COS" and the phase difference between "SIN" and "COS" can be detected based on the phase.

In other words, a Lissajous waveform is obtained from the two sine wave signals output from the encoder, and a correction value can be calculated based on the variation in radius over one cycle of the obtained Lissajous waveform. The sine wave signal can then be corrected based on that correction value. The details are explained below.

FIG. 1 is a diagram schematically illustrating an encoder system according to an embodiment.
The encoder system includes an encoder 10, a correction device 12, an interpolation circuit 14 (interpolator), etc. The encoder 10 is, for example, an optical or magnetic linear encoder, and includes a scale 16 and a detection head 18.

The correction device 12 includes A/D converters 20 and 22, a correction unit 24, a correction value calculation unit 26, and a displacement information conversion unit 28. The "SIN" and "COS" signals from the detection head 18 are sampled at a predetermined frequency by the A/D converters 20 and 22 and converted into digital signals. The correction device 12 processes these digital signals.

The correction value calculation unit 26 calculates the correction error according to the automatic correction method described below to obtain a correction value. This correction error is the fluctuation of the radius fluctuation waveform, and includes offset error, amplitude error, phase error, correction error due to second-order distortion, correction error due to third-order distortion, etc. (described in detail below). The correction unit 24 corrects the correction error of the "SIN" and "COS" signals based on the correction value calculated by the correction value calculation unit 26. The displacement information conversion unit 28 converts the corrected "SIN" and "COS" signals into displacement information for the interpolation circuit 14 and the correction value calculation unit 26.

This displacement information includes information on the radius R(θ) for each phase angle θ of the Lissajous waveform obtained after correction. This displacement information is fed back to the correction value calculation unit 26. The interpolation circuit 14 interpolates and divides the Lissajous waveform of the corrected signal to improve the detection accuracy of the encoder 10, but a description of this will be omitted.

2 to 6 are diagrams illustrating the principle of the automatic correction method according to the embodiment.
Figures 2 and 3 show a method for detecting offset errors, Figure 4 shows a method for detecting amplitude errors, and Figure 5 shows a method for detecting phase errors. Each figure (A) shows a Lissajous waveform, and each figure (B) shows the waveform of a sine wave signal (solid line) and the radius fluctuation waveform (chain double-dashed line). Figure 6 shows a method for detecting second-order distortion. Figure 6 (A) shows the case where second-order distortion exists in "SIN," and Figure 6 (B) shows the case where second-order distortion exists in "COS." Here, θ represents the phase angle of the Lissajous waveform.

1) Offset Detection ("COS"): Figure 2
When the input waveform "COS" to the correction device 12 has an offset Bc, the Lissajous radius R = √("SIN" ² + "COS" ² ) fluctuates as shown by the two-dot chain line in Figure 2(B) (Figure 2(B) shows the case where Bc = 0.1). In other words, the Lissajous radius R(θ) becomes a radius fluctuation waveform that fluctuates with the same period as "COS", and has a correction error.

The offset Bc is calculated based on the radius R(θ) using the following equation (1).
Bc=∫(R(θ)・cosθ)dθ/2π…(1)
This is to find the cos θ component of R(θ). In other words, by extracting the component synchronized with the cosine wave corresponding to one of the sine wave signals from the fluctuation in the radius of the Lissajous waveform, the offset of one of the sine wave signals can be calculated.

2) Offset detection ("SIN"): Figure 3
When the input waveform "SIN" to the correction device 12 has an offset Bs, the Lissajous radius R = √("SIN" ² + "COS" ² ) fluctuates as shown by the two-dot chain line in Figure 3(B) (Figure 3(B) shows the case where Bs = 0.1). In other words, the Lissajous radius R(θ) becomes a radius fluctuation waveform that fluctuates with the same period as "SIN", and has a correction error.

The offset Bs is calculated based on the radius R(θ) using the following equation (2): Bs=∫(R(θ)·sin θ)dθ/2π (2)
This is to find the sin θ component of R(θ). In other words, by extracting the component synchronized with the sine wave corresponding to the other sine wave signal from the fluctuation in the radius of the Lissajous waveform, the offset of the other sine wave signal can be calculated.

3) Detection of amplitude difference: Figure 4
When the amplitude of the input waveform "COS" to the correction device is larger than the amplitude of "SIN" by G, the Lissajous radius R = √("SIN" ² + "COS" ² ) fluctuates as shown by the two-dot chain line in Figure 4(B) (Figure 4(B) shows the case where G = 0.1). In other words, the Lissajous radius R(θ) becomes a radius fluctuation waveform that fluctuates with a period half that of "COS", and a correction error occurs.

The amplitude difference G is calculated based on the radius R(θ) using the following equation (3).
G=∫R(θ)・cos(2θ)・dθ/2π…(3)
This is to find the cos(2θ) component of R(θ). In other words, by extracting a component synchronized with a half-cycle cosine wave corresponding to one of the sine wave signals from the fluctuation in the radius of the Lissajous waveform, the amplitude difference between the two sine wave signals can be calculated.

4) Phase difference detection: Figure 5
When the phase of the input waveform "COS" is significantly shifted by P, the Lissajous radius R = √("SIN" ² + "COS" ² ) fluctuates as shown by the two-dot chain line in Figure 5(B) (Figure 5(B) shows the case where P = 10°). In other words, the Lissajous radius R(θ) becomes a radius fluctuation waveform that fluctuates with a period half that of "SIN", and has a correction error.

The phase difference P is calculated based on the radius R(θ) using the following equation (4).
P=∫R(θ)・sin(2θ)・dθ/2π...(4)
This is to find the sin(2θ) component of R(θ). In other words, by extracting the component synchronized with the half-cycle sine wave corresponding to the other sine wave signal from the fluctuation of the radius of the Lissajous waveform, the phase difference between the two sine wave signals can be calculated.

5) Radius Detection In this embodiment, when detecting the Lissajous radius R=√(“SIN” ² +“COS” ² ) itself, the average magnitude of R can be calculated using the following equation (5).
R=∫R(θ)dθ/2π...(5)

6) Detection of second-order distortion: Figure 6
When the "sin" waveform has second-order distortion (Ds·(sin θ) ² ), the Lissajous radius R = √("sin" ² + "cos" ² ) fluctuates as shown by the two-dot chain line in Figure 6(A) (Figure 6(A) shows the case where Ds = 0.2). In other words, the Lissajous radius R(θ) becomes a radius fluctuation waveform that fluctuates with a period one-third that of "sin", and has a correction error.

Ds is the amplitude of the second-order distortion (referred to as "second-order distortion amplitude"), and is calculated based on the radius R(θ) using the following equation (6): Ds = ∫R(θ) sin(3θ) dθ/2π (6)
This finds the sin(3θ) component of R(θ). In other words, by extracting the component synchronized with one-third of the period of the sine wave corresponding to one sine wave signal from the fluctuation of the radius of the Lissajous waveform, the second-order distortion of one sine wave signal can be calculated.

On the other hand, when the "COS" waveform has second-order distortion (Dc·(cos θ) ² ), the Lissajous radius R = √("SIN" ² + "COS" ² ) fluctuates as shown by the two-dot chain line in Figure 6(B) (Figure 6(B) shows the case where Dc = 0.2). In other words, the Lissajous radius R(θ) becomes a radius fluctuation waveform that fluctuates with a period one-third that of "COS", and has a correction error.

Dc is the amplitude of the second-order distortion (referred to as "second-order distortion amplitude") and is calculated based on this radius R(θ) using the following equation (7).
Dc=∫R(θ)・cos(3θ)・dθ/2π...(7)
This is to find the cos(3θ) component of R(θ). In other words, by extracting the component synchronized with one-third of the period of the cosine wave corresponding to the other sine wave signal from the fluctuation of the radius of the Lissajous waveform, the second-order distortion of the other sine wave signal can be calculated.

Next, the configuration and specific operation of the correction device 12 will be described.
FIG. 7 is a functional block diagram of the correction device 12.
The components of the correction device 12 are implemented by hardware including arithmetic units such as FPGAs (Field Programmable Gate Arrays) and various computer processors, storage devices such as memories and storages, and wired or wireless communication lines connecting these devices, as well as software stored in the storage devices and supplying processing instructions to the arithmetic units. The computer programs may be configured by device drivers, an operating system, various application programs located at higher levels than these, and libraries that provide common functions to these programs. Each block described below represents a functional block, not a hardware configuration.

The correction device 12 includes an input/output interface unit 110, a data processing unit 112, and a data storage unit 114. The input/output interface unit 110 is responsible for processing related to the input/output interface, including the exchange of data with external devices. The data processing unit 112 performs various processes based on data acquired by the input/output interface unit 110 and data stored in the data storage unit 114. The data processing unit 112 also functions as an interface between the input/output interface unit 110 and the data storage unit 114. The data storage unit 114 stores various programs and setting data.

The input/output interface section 110 includes an input section 120 and an output section 122 .
The input unit 120 includes a sine wave signal acquisition unit 124. The sine wave signal acquisition unit 124 includes the functions of the A/D converters 20 and 22, and acquires sine wave signals (“SIN”, “COS”) from the encoder 10 as digital signals.

The data storage unit 114 includes a sampling data storage unit 140. The sampling data storage unit 140 stores sampling data of the Lissajous radius R acquired at a predetermined sampling period. The data storage unit 114 includes memory that functions as a working area when the data processing unit 112 performs calculations.

The data processing unit 112 includes the correction unit 24, correction value calculation unit 26, and displacement information conversion unit 28 described above. The correction value calculation unit 26 includes a radius fluctuation calculation unit 130, an offset calculation unit 132, an amplitude difference calculation unit 134, a phase difference calculation unit 136, and a second-order distortion calculation unit 138. The radius fluctuation calculation unit 130 calculates the fluctuation in the radius R(θ) of the Lissajous waveform (radius fluctuation waveform) by performing calculation processing based on the above equation (5) based on the sampling data stored in the sampling data storage unit 140.

The offset calculation unit 132 calculates the offset between the two sine wave signals by performing calculations based on the above equations (1) and (2) on the radius R(θ) stored in the data storage unit 140. The amplitude difference calculation unit 134 calculates the amplitude difference between the two sine wave signals by performing calculations based on the above equation (3) on the radius R(θ). The phase difference calculation unit 136 calculates the phase difference between the two sine wave signals by performing calculations based on the above equation (4) on the radius R(θ). The second-order distortion calculation unit 138 calculates the second-order distortion of the two sine wave signals by performing calculations based on the above equations (6) and (7) on the radius R(θ).

The correction unit 24 amplifies and integrates these calculated correction errors to create correction values, which are used to correct the sine wave signal. The displacement information conversion unit 28 converts the corrected sine wave signal into displacement information and feeds it back to the correction value calculation unit 26. This displacement information is stored in the sampling data storage unit 140 and is updated each time the correction process is repeated. Repeating the correction process reduces the correction error.

The output unit 122 includes a displacement information output unit 126. The displacement information output unit 126 outputs to the interpolation circuit 14 displacement information in which the correction error has been eliminated or reduced by correction.

In the above configuration, the correction value calculation process by the correction value calculation unit 26 is based on the automatic correction method using the above equations (1) to (7), but in this embodiment, the calculation process is performed discretely. Specifically, it is as follows:

(Calculation of Radius Rn)
First, the radius variation calculation unit 130 acquires a Lissajous waveform for one revolution from the two sine wave signals "SIN" and "COS." Then, the phase angle of the Lissajous waveform is divided into N regions (for example, N=8, each section being 45 degrees), and sample data Rn (n: an integer between 0 and N-1) of the Lissajous radius is acquired for the phase of each region. Specifically, a phase-specific accumulator may be used to store the radius Rn (n: an integer between 0 and N-1) of each region. At this time, the signal is divided into N phase regions between 0 and N-1.

The N phase regions may be equally divided, or may be regions that include the zero crossing points of "SIN" and "COS" and the crossing points of "SIN" and "COS", as described in Japanese Patent No. 3367226, for example.

The phase-specific accumulator may be implemented, for example, with a delay element that generates a delay equivalent to one sampling period, and an integration circuit including a feedback multiplier. This allows the data for each region to be smoothed each time new sampling data is input. In other words, the radius fluctuation waveform value of the Lissajous waveform can be a smoothed value obtained by passing the radius of the current section n through a low-pass filter. Smoothing removes or suppresses noise superimposed on the radius fluctuation, enabling stable detection of the cause of the interpolation error.

More specifically, data with noise superimposed on it is continuously acquired for each section of the phase region, and the noise is removed using a low-pass filter. The noise-removed data is then sampled at a predetermined point for each section (a stable point, such as immediately before the next section), and the sampled value can be used as the radius value of that phase region in the Lissajous waveform. Alternatively, this sampling can be performed for multiple cycles, the sampled values can be averaged for each section, and this average can be used as the radius value of that phase region in the Lissajous waveform. By repeating this smoothing and averaging, it is possible to understand the trend of the correction error, taking into account changes due to scale operation, and improve noise removal performance.

(Calculation of offset)
The offset calculation unit 132 calculates the offset Bc of "COS" based on the following equation (8).
That is, when extracting a component synchronized with [COS] from the radius fluctuation waveform of the Lissajous waveform, the radius fluctuation waveform is multiplied by cos(2πn/N) and averaged over the most recent cycle of the Lissajous waveform to obtain the offset Bc. Here, N is an even natural number. n is an integer between 0 and N-1, and is proportional to the phase of the Lissajous waveform during one rotation.

Similarly, the offset calculation unit 132 calculates the offset Bs of "sin" based on the following equation (9).
That is, when extracting a component synchronized with [SIN] from the radius fluctuation waveform of the Lissajous waveform, the radius fluctuation waveform is multiplied by sin (2πn/N) and averaged over the most recent cycle of the Lissajous waveform to obtain the offset Bs. Here, N is an even natural number. n is an integer between 0 and N-1, and is proportional to the phase of the Lissajous waveform during one rotation.

(Calculation of Amplitude Difference)
The amplitude difference calculation unit 134 calculates the amplitude difference G based on the following equation (10).
That is, when extracting a component synchronized with half the cycle of [SIN] from the radius fluctuation waveform of the Lissajous waveform, the radius fluctuation waveform is multiplied by cos(4πn/N) and averaged over the most recent cycle of the Lissajous waveform to obtain the amplitude difference G. Here, N is an even natural number, and n is an integer between 0 and N-1, proportional to the phase of the Lissajous waveform during one rotation.

(Calculation of phase difference)
The phase difference calculation unit 136 calculates the amplitude difference G based on the following equation (11).
That is, when extracting a component synchronized with half a cycle of [SIN] from the radius fluctuation waveform of the Lissajous waveform, the radius fluctuation waveform is multiplied by sin (4πn/N) and averaged over the most recent cycle of the Lissajous waveform to obtain the phase difference P. Here, N is an even natural number, and n is an integer between 0 and N-1, proportional to the phase of the Lissajous waveform during one rotation.

(Calculation of second-order distortion)
The second-order distortion calculating section 138 calculates the second-order distortion amplitude Ds of "SIN" based on the following equation (12), and calculates the second-order distortion amplitude Dc of "COS" based on the following equation (13).

In other words, when extracting a component synchronized with one-third of the [SIN] period from the radius fluctuation waveform of a Lissajous waveform regarding the second-order distortion of a "SIN" waveform, Ds may be obtained by multiplying the radius fluctuation waveform by sin(6πn/N) and averaging the result over the most recent period of the Lissajous waveform. When extracting a component synchronized with one-third of the [COS] period from the radius fluctuation waveform of a Lissajous waveform regarding the second-order distortion of a "COS" waveform, Dc may be obtained by multiplying the radius fluctuation waveform by cos(6πn/N) and averaging the result over the most recent period of the Lissajous waveform. Here, N is an even natural number. n is an integer between 0 and N-1, proportional to the phase of the Lissajous waveform during one rotation.

Note that, when considering implementation in an FPGA or the like, it is more convenient to have a smaller number of multiplication bits. Therefore, sin(2πn/N), cos(2πn/N), sin(4πn/N), cos(4πn/N), sin(6πn/N), and cos(6πn/N) in the discretization described above may be quantized and replaced with values whose number of bits is reduced to at least 1 bit. In other words, the weighting coefficients used when multiplying the Lissajous radius Rn by each value may be set to one of three values, for example, 0 and ±1. If quantization is not performed, the weighting coefficients may become decimals , which may complicate the calculation algorithm, but this can be avoided by quantization.

Note that, when taking into account the calculation of amplitude and phase differences, N=16 (i.e., 16 divisions) is preferable to N=8. Specifically, with regard to the offset weighting coefficients, if S1 to S16 are the values corresponding to n=0 to 15 for sin(2πn/N), then S1, S8, S9, S16 = 0, S2 to S7 = 1, and S10 to S15 = -1. With regard to cos(2πn/N), if C1 to C16 are the values corresponding to n=0 to 15, then C1 to C3, C14 to C16 = 1, C4, C5, C12, C13 = 0, and C6 to C11 = -1.

Furthermore, regarding the weighting coefficients for amplitude difference, if G1 to G16 are values corresponding to n = 0 to 15 for cos(4πn/N), then G1, G2, G8 to G10, G16 = 1, G3, G7, G11, G15 = 0, and G4 to G6, G12 to G14 = -1. Regarding the weighting coefficients for phase difference, if P1 to P16 are values corresponding to n = 0 to 15 for sin(4πn/N), then P1, P5, P9, P13 = 0, P2 to P4, P10 to P12 = 1, and P6 to P8, P14 to P16 = -1.

The automatic correction method has been described above based on the embodiment.
In conventional methods, the radius and offset are calculated from the peak values of the "SIN" and "COS" waveforms, which is easily affected by noise waveforms at the moment of the peak. However, in the method of this embodiment, the "SIN" and "COS" components are calculated by averaging processing such as integrating over one cycle of the "SIN" and "COS" waveforms, which makes them less susceptible to noise and stabilizes detected values such as offset. This reduces the residual interpolation error caused by noise.

Similarly, because detected values such as offset are stabilized, the stability of the system increases when they are fed back. This allows the feedback gain to be increased compared to conventional methods, resulting in faster response speeds.

In addition, the method of this embodiment can detect second- and third-order distortion contained in "SIN" and "COS" waveforms, which reduces interpolation errors caused by harmonic distortion, which was difficult to achieve with conventional methods.

More specifically, actual "SIN" waveforms, "COS" waveforms, and the radius fluctuation waveforms of the Lissajous waveforms created from them contain noise and irregular fluctuations. As already mentioned, in conventional methods, for example, when detecting the offset of a "COS" waveform, the average of the "COS" waveform's positive peak value ("COS"+) and negative peak value ("COS"-) is used as the center value of the "COS" waveform, and the deviation from 0 is used as the offset, this is susceptible to the effects of noise. The same is true for detecting not only offset but also phase difference and the amplitude difference between "SIN" and "COS," which causes the interpolation error contained in the displacement signal obtained by feeding this back to a certain level that cannot be reduced below that level. While it is possible to reduce noise using a low-pass filter with a certain cutoff frequency, the cutoff frequency cannot be lowered sufficiently to ensure peak value detection accuracy, and the effects of noise cannot be fully avoided.

In this regard, according to this embodiment, the peak values of the above-mentioned "SIN" and "COS" waveforms are not used, but rather the periodic components of the radius fluctuation waveform over at least one period (its frequency components and harmonic components) are used to detect offset and other interpolation error factors. This makes it possible to reduce noise at a cutoff frequency several times the "SIN" and "COS" frequencies. In other words, by combining the detection of the periodic components of the radius fluctuation waveform with the use of a low-pass filter with a sufficiently low cutoff frequency, it is possible to reduce interpolation errors to a level not previously possible.

Some of the noise and irregular fluctuations contained in the actual "SIN" and "COS" waveforms, as well as the radius fluctuation waveform of the Lissajous waveform created from them, are better averaged over several or more periods of the "SIN" and "COS" waveforms. For example, when the scale or detection head moves, the distance between the scale and detection head may fluctuate irregularly, causing corresponding fluctuations in the amplitude of the "SIN" and "COS" waveforms. In this case, if a low-pass filter with a long time constant spanning one period of the radius fluctuation waveform is inserted, the periodic components of the radius fluctuation waveform will also be attenuated, potentially making it difficult to detect offsets and other interpolation error factors. This is because the influence of noise in each phase region will extend to other phase regions. Therefore, in this embodiment, N registers are provided as low-pass filters for the radius fluctuation waveform, one for each of the N divided portions, and the waveform data for each portion is accumulated separately. This allows each portion of the radius fluctuation waveform to be averaged over multiple periods, preventing the periodic components of the radius fluctuation waveform from being attenuated.

[Modification]
Although not described in the above embodiment, third-order distortion may be detected and corrected. That is, the third-order distortion of one sine wave signal can be calculated by extracting a component synchronized with a quarter cycle of the sine wave [SIN] corresponding to one sine wave signal from the fluctuation in the radius of the Lissajous waveform. Furthermore, the third-order distortion of the other sine wave signal can be calculated by extracting a component synchronized with a quarter cycle of the cosine wave [COS] corresponding to the other sine wave signal from the fluctuation in the radius of the Lissajous waveform.

Furthermore, regarding the third-order distortion of a "SIN" waveform, when extracting a component synchronized with a quarter cycle of [SIN] from the radius fluctuation waveform of a Lissajous waveform, the third-order distortion may be obtained by multiplying the radius fluctuation waveform by sin(8πn/N) and averaging the result over the most recent cycle of the Lissajous waveform. Furthermore, regarding the third-order distortion of a "COS" waveform, when extracting a component synchronized with a quarter cycle of [COS] from the radius fluctuation waveform of a Lissajous waveform, the third-order distortion may be obtained by multiplying the radius fluctuation waveform by cos(8πn/N) and averaging the result over the most recent cycle of the Lissajous waveform.

The sin(8πn/N) and cos(8πn/N) used in the discretization above can also be quantized and replaced with values whose bit count is reduced to at least one bit. That is, if the radius fluctuation waveform for offset detection is represented using multiple bits (e.g., 20 bits) and multiplied by sin(8πn/N) and cos(8πn/N), two multipliers are required. Furthermore, one multiplier is required each for the amplitude difference and phase difference, resulting in a large circuit size. In this regard, if sin(8πn/N) and cos(8πn/N) are quantized to two values (±1) or three values (0 and ±1), multiplication becomes unnecessary, significantly reducing the circuit size. In this case, the multiplication result contains an error. However, in the above embodiment, when an offset or the like is detected, it is amplified and fed back, so the final detected value is infinitesimally close to zero. Therefore, errors in the multiplication along the way are not a problem.

In the above radius fluctuation waveform, "radius squared" may be used instead of "radius."

In the above embodiment, an example was shown in which the radius fluctuation waveform value of the Lissajous waveform is calculated by dividing one circumference of the Lissajous waveform into N sections (N is an integer) and passing the radius for each section through a low-pass filter. Examples of N = 8 and N = 16 were given. In a modified example, N = 24 may also be used. To accommodate a 90-degree phase shift between "SIN" and "COS", N should be set to a multiple of 4. Furthermore, when performing calculations using one-third period components, N should also be set to a multiple of 3. For accuracy reasons, it is also advisable to multiply the above by an integer of at least twice.

The present invention is not limited to the above-described embodiments and modifications, and the components can be modified within the scope of the gist of the invention. Various inventions may be created by appropriately combining multiple components disclosed in the above-described embodiments and modifications. Furthermore, some components may be omitted from all of the components shown in the above-described embodiments and modifications.

10 Encoder, 12 Correction device, 14 Interpolation circuit, 16 Scale, 18 Detection head, 24 Correction unit, 26 Correction value calculation unit, 28 Displacement information conversion unit, 110 Input/output interface unit, 112 Data processing unit, 114 Data storage unit, 120 Input unit, 122 Output unit, 124 Sine wave signal acquisition unit, 126 Displacement information output unit, 130 Radius variation calculation unit, 132 Offset calculation unit, 134 Amplitude difference calculation unit, 136 Phase difference calculation unit, 138 Second-order distortion calculation unit, 140 Sampling data storage unit.

Claims (7)

In an encoder system that reads a periodic signal on a scale as a "SIN" or "COS" signal and converts the periodic signal into a position signal, an automatic correction method is used to detect offsets in the "SIN" and "COS" signals and feed them back to the "SIN" and "COS" signals to reduce the interpolation error when converted,
The offsets of the "SIN" signal and the "COS" signal are obtained by extracting components synchronized with sine and cosine from the radius fluctuation waveform of the Lissajous waveform by the "SIN" and "COS" signals, respectively.
An automatic correction method characterized in that, in order to extract components synchronized with sin, cos, or half-cycle sin or cos from the radius fluctuation waveform of a Lissajous waveform, the radius fluctuation waveform of the Lissajous waveform is multiplied by sin(2πn/N), cos(2πn/N), or sin(4πn/N), cos(4πn/N), respectively, and the result is averaged over the most recent cycle of the Lissajous waveform .
where N is an even natural number, and n is an integer between 0 and N-1, proportional to the phase of the Lissajous waveform during one rotation. In an encoder system that reads a periodic signal on a scale as a "SIN" or "COS" signal and converts the periodic signal into a position signal, an automatic correction method is used to detect the amplitude difference and phase difference of the "SIN" and "COS" signals and feed them back to the "SIN" and "COS" signals so that the amplitude difference and phase difference approach 0 and 90 degrees, respectively, in order to reduce interpolation errors during conversion,
The present invention is characterized in that components synchronized with half-cycle sine and cosine are extracted from the radius fluctuation waveform of the Lissajous waveform by the "SIN" and "COS" signals, and the amplitude difference and phase difference between the "SIN" signal and the "COS" signal are obtained ,
An automatic correction method characterized in that, in order to extract components synchronized with sin, cos, or half-cycle sin or cos from the radius fluctuation waveform of a Lissajous waveform, the radius fluctuation waveform of the Lissajous waveform is multiplied by sin(2πn/N), cos(2πn/N), or sin(4πn/N), cos(4πn/N), respectively, and the result is averaged over the most recent cycle of the Lissajous waveform .
where N is an even natural number, and n is an integer between 0 and N-1, proportional to the phase of the Lissajous waveform during one rotation. In an encoder system that reads a periodic signal on a scale as a "SIN" or "COS" signal and converts the periodic signal into a position signal, an automatic correction method is used to detect second-order distortions of the "SIN" and "COS" signals and feed them back to the "SIN" and "COS" signals so that interpolation errors during conversion are reduced,
The device is characterized by extracting components synchronized with sine and cosine of one-third period from the radius fluctuation waveform of the Lissajous waveform by "SIN" and "COS" signals, and obtaining second-order distortion of the "SIN" signal and "COS" signal, respectively.
This automatic correction method is characterized in that, in order to extract components synchronized with sine and cosine of one-third of a cycle from the radius fluctuation waveform of a Lissajous waveform, the radius fluctuation waveform of the Lissajous waveform is multiplied by sin (6πn/N) and cos (6πn/N), respectively, and the results are averaged over the most recent cycle of the Lissajous waveform.
where N is an even natural number, and n is an integer between 0 and N-1, proportional to the phase of the Lissajous waveform during one rotation. In an encoder system that reads a periodic signal on a scale as a "SIN" or "COS" signal and converts the periodic signal into a position signal, an automatic correction method is used to detect third-order distortions of the "SIN" and "COS" signals and feed them back to the "SIN" and "COS" signals so that interpolation errors during conversion are reduced,
The device is characterized by extracting components synchronized with the sine and cosine of a quarter period from the radius fluctuation waveform of the Lissajous waveform using "SIN" and "COS" signals, and obtaining third-order distortions of the "SIN" and "COS" signals, respectively .
This automatic correction method is characterized in that, in order to extract components synchronized with the sine and cosine of a quarter cycle from the radius fluctuation waveform of a Lissajous waveform, the radius fluctuation waveform of the Lissajous waveform is multiplied by sin (8πn/N) and cos (8πn/N), respectively, and the results are averaged over the most recent cycle of the Lissajous waveform .
where N is an even natural number, and n is an integer between 0 and N-1, proportional to the phase of the Lissajous waveform during one rotation. In the automatic correction method according to any one of claims 1 to 4 ,
An automatic correction method characterized in that sin(2πn/N), cos(2πn/N), sin(4πn/N), cos(4πn/N), sin(6πn/N), cos(6πn/N), sin(8πn/N), and cos(8πn/N) are each quantized and expressed as two values of ±1 or three values of 0 and ±1. In the automatic correction method according to any one of claims 1 to 4 ,
The radius fluctuation waveform value of the Lissajous waveform is a value obtained by dividing one circumference of the Lissajous waveform into N sections (N is an integer) and passing the radius of each section through a low-pass filter .
An automatic correction method characterized in that N registers are prepared for each of the N divided parts, and waveform data for each part is accumulated separately . 7. The automatic correction method according to claim 1 , wherein the "radius" is replaced by the "square of the radius."