JPH04301927A - Rotary absolute encoder with error detecting function - Google Patents

Abstract

PURPOSE:To comparatively simplify a circuit to execute error detection by detecting read error at plural sensors by using the relation of specified source polynomials. CONSTITUTION:Exclusive OR C1 and C2 are calculated concerning the two pairs of sensor outputs extracted in the position relation making the coefficient of the prescribed source polynomial correspondent to '1' and when the both outputs are '0', an alarm signal (w) is turned to '0' as well since there is no error in reading. When one of the C1 and C2 is turned to '1', however, the alarm signal (w) is turned to '1' since it is discriminated that read error is generated at any one of sensors s1-s9. In such a case, when a detection part B is moved relatively to an encoding panel A, an address to discriminate 30 cyclic absolute positions along a pattern T is assembled by using the code of 6 digits in the binary number of 9 digits read from the pattern T by the sensors s1-s9. Simultaneously, a discrimination circuit C confirms that there is no read error at the sensors s1-s9 and when the read error is generated even at one of the sensors s1-s9, the alarm signal (w) is immediately generated.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention relates to a one-track absolute encoder capable of detecting sensor reading errors, and more specifically, to a rotary encoder having a cycle of 10, 20, 60, 2n, etc., which is highly practical in industry. -Tally absolute encoder
The present invention relates to a technique for detecting or correcting sensor reading errors with a simple configuration in a computer.

0002

2. Description of the Related Art An absolute encoder is a measuring instrument in which each relative position of a detecting section with respect to a code plate is output as a different address (absolute position signal). Of the code plate and detection unit that are combined to be relatively movable, the code plate side has an absolute pattern in which the numbers at the address are replaced with physical information such as magnetic and optical properties and are arranged continuously. A plurality of sensors for determining physical information of the pattern are successively arranged on the detecting section side. Then, the detection section directly reads the unique address recorded as physical information on the pattern. The absolute encoder is
In appearance, they are roughly divided into linear types, in which the detection unit moves linearly along a band-shaped code plate, and rotary types, in which the detection unit moves angularly relative to a disc or cylindrical code plate. is,
In the latter case, the original phase position is reached after one cycle, and of course the original absolute position (address) is output at that time.

Conventionally, absolute encoders have generally been of the multi-track type. This is a multi-track absolute system consisting of many parallel tracks.
A track pattern is formed on a code plate, and a sensor is arranged in the detection section corresponding to each track. ) Read the correct address. For example, 4 digit BCD
4-track absolute encoder for reading codes
In da, there are four parallel tracks on the code board,
23 Track …000000001111111
1 22 Track …00001111
00001111 21 Track …0
011001100110011 20 tracks …0101010101010101
Absolute position abcdef
ghijklmnop is formed and the same phase positions of the four tracks are read in parallel (vertically), a=0000,
From b=0001, . . . to o=1110, p=1111, 16 addresses a to p of binary numbers in the correct order are obtained one after another.

On the other hand, in recent years, one-track type absolute encoders that use a one-track type absolute pattern instead of a multi-track type absolute pattern have been actively researched. A one-track absolute pattern consists of replacing the 1s and 0s of a special binary number sequence, such as a full-cycle sequence or an M-sequence, with two types of minimum reading units with different physical properties and arranging them in a single line. It is something. In the detection section of the 1-track type absolute encoder, a plurality of sensors are arranged in a row along the 1-track type absolute pattern, and these sensors ``binary addresses, each different but in random order.'' A one-track type absolute encoder requires only one absolute pattern, has sensors arranged in a row at equal intervals, and has no space between tracks like a multi-track type. This is extremely advantageous in miniaturizing the absolute encoder and simplifying the overall structure including wiring, for reasons such as eliminating the need for phase adjustment of the sensor arrangement.

For example, the M sequence obtained from the primitive polynomial g(x) =
a, f=c+b, g=d+c, h=e+d, i=f+e
, j=g+f, k=h+g, etc., and calculate a=m+l, b=n+m, c=o+n, d=a+
000100110101111 abcdefghijklmno, which is a periodic sequence consisting of 15 codes a to o (returns to a after o), and is a 4-digit sequence consisting of any four consecutive codes on this sequence. Binary numbers are 0001, 0010, 0
All 15 are different: 100, ...0111, 1111, ...1000. Therefore, a one-track absolute pattern in which the 1's and 0's of the sequence are replaced with two types of minimum reading units is formed on the code plate, and the detection section receives four consecutive pieces on the pattern. A one-track absolute encoder is constructed by arranging the sensor to detect the minimum reading unit of n=1 from a=0001, b=0010,...
100, up to o=1000, 15 different binary numbers are read in random order. If there is a practical problem with this haphazard binary address, instead of outputting the address read by the sensor as it is, use a semiconductor memory element (comparison table) whose cost has been significantly reduced in recent years.
It is output after performing a one-to-one conversion into "addresses in the correct order" like a CD code.

By the way, one track type absolute
If even one sensor in the encoder misreads the minimum reading unit, an abnormal address that is far from the actual address will be output. Robots and machine tools may instantly become uncontrollable or run out of control, causing serious accidents. Therefore, regarding the problem of sensor reading errors in one-track absolute encoders, the conventional methods have been to (1) reduce the possibility that the sensor will read the pattern incorrectly, or (2) prevent the occurrence of reading errors. Countermeasures are being considered, including detecting and issuing a warning, and (3) identifying the sensor with a reading error and reversing the sensor output (correcting the reading error). Here, (1) relates to a technique for improving reading reliability, and (2) and (3) relate to a technique for detecting reading errors.

On the other hand, sensor reading errors are also an important problem in the case of multi-track absolute encoders; It can improve the reliability of reading, and ■B
Since it is possible to identify a sensor with a reading error using the regularity of the CD code or Gray code, countermeasures can be said to be relatively easy.

For example, in the case of the above-mentioned 4-track absolute encoder that reads the BCD code, (1) the track that is misread is almost limited to the track with the smallest pitch of 20 digits, (2) each track is a simple Since it is a repeating pattern with equal intervals, a method of detecting multiple pitches at once by overlapping masks with the same pitch (introduced in Japanese Patent Application No. 2-201313) can be adopted.
The read address can be easily checked in real time by comparing it with the result of addition or subtraction.

In contrast, in the case of a one-track type absolute encoder, even if a semiconductor memory element is used to convert addresses into the correct order as in a multi-track type absolute encoder, The addresses that the sensors in the detection section actually read from the code plate are "binary numbers that are different but in a random order," and there is no regularity in the pattern that can be used for reading, and each sensor is equally Since there is a possibility of reading errors, countermeasures against sensor reading errors are considerably more difficult than in the case of a multi-track type.

[0010] Such a one-track type absolute
In the encoder, the above-mentioned (2) measures to detect the occurrence of reading error and issue a warning, and (3) identify the sensor with reading error and invert the sensor output (correct reading error), that is, detect reading error. Considering the necessity and difficulty of technology,
The applicant of the present application previously proposed Japanese Patent Application No. 2-73537 and Japanese Patent Application No. 2-251730. In the former, the read address is confirmed using the relationship of the primitive polynomial that generated the absolute pattern, and in the latter, the misread sensor is identified by combining multiple relationships of the primitive polynomial in a matrix. .

That is, the M series is a primitive polynomial, g(x
)=ayxy+ay-1xy-1 +...+a1x1+
Among the binary number sequences determined by successive calculations from an appropriate initial value using a0, the one with the maximum period, (a)
The y-digit binary numbers extracted at any phase position on the sequence are all different, and all the y-digit binary numbers (
(b) ay=1 of the primitive polynomial for a binary number of y+1 digits extracted at an arbitrary phase position.
It has the property that the exclusive OR of all the codes at the positions corresponding to is always 0. Therefore, patent application No. 2-7353
In No. 7, the address read from the relationship (b) is confirmed, and in Japanese Patent Application No. 2-251730, the phase is shifted by one (b).
) are arranged vertically up to the required number and combined in a matrix to mechanically identify sensors with reading errors.

For example, in the case of the above M sequence 000100110101111 abcdefghijklmno, (a) 4 consisting of four arbitrary consecutive codes
Binary digits are 0001, 0010, 0100,...011
1, 1111, ...1000, each having 15 different values, and from the initial value 0001 (abcd), e=b+a, f=
c+b, g=d+c, h=e+d, i=f+e, j=g
As is clear from the process of obtaining the sequence by calculating +f, k=h+f,... and the exclusive OR of the first digit value is always 0.

Therefore, in a one-track absolute pattern created from the number sequence, seven consecutive minimum reading units (for example, abcdefg) on the number sequence are read,
(a) Assemble the address (abcd) from the minimum reading unit of four consecutive pieces (abcd) from the left, and (b) check the above calculation relationships (e=b+a and g=d+c) for each of the five pieces at both ends. Assembled address (abcd)
You can check that there are no misreadings.

[0014]

[Problem to be solved by the invention] Now, as mentioned above,
(a), (b) An M sequence with two properties is (a)
It is adopted as a one-track absolute pattern due to the property (b), and sensor reading errors can be detected due to the property (b), but the period is always limited to 2y -1 and cannot be arbitrarily determined. It has the disadvantage that it cannot. In other words, practically and industrially, 60, 360, 10n,
Periods such as 2y are useful, but M sequences of these periods do not exist. Therefore, when designing a rotary absolute encoder with these periods in which it is essential to return to the original address after one rotation, Deleting the middle and splicing it together,
In addition, especially for a period of 2y, various methods are employed in which one 0 is inserted into the continuous 0 part of the M sequence with a period of 2y -1 and one address with all 0s is added. Both methods ■ and ■ yield a periodic sequence that returns once and returns to the original address, and the above property (a) is satisfied, i.e., ``The y-digit binary numbers extracted at any phase position on the sequence are all different.''nature"
is saved.

[0015] According to these methods (2) and (2), it is possible to easily obtain a one-track absolute pattern with an arbitrary period, but each method has two methods: (1) deleted joint parts, and (2) added addresses with all zeros. In part (b), the property "the exclusive OR of all the signs at the position corresponding to ay=1 of the primitive polynomial for the y+1-digit binary number extracted at any phase position is always 0" is shown. There is a problem that reading error detection using this property cannot be performed at addresses that are lost and sandwich the part.

For example, by truncating the above M sequence of 15 codes, 000100110101111 abcdefghijklmno, at i, a new number sequence with period 9, 0001001
When creating 10 abcdefghi, the 4-digit binary number consisting of any four consecutive codes on the sequence is 0001, 0010,...011
All nine are different, 0, 1100, 1000, and the address a~
i can be mutually distinguished, but among the combinations of five consecutive codes including seams, in fghia, 0+1+0≠0, hia
1+0+0≠0 in bc, 0+0+1≠ in iabcd
0, and the property (b) is lost.

[0017] Also, for example, by adding 0 to the end of the M sequence 000100110101111 abcdefghijklmno, which is the M sequence consisting of the 15 codes mentioned above, a new number sequence with a period of 24 = 16 is created.
When creating 0001001101011110 abcdefghijklmnop, address p=0000 and addresses a to o
However, among the combinations of five consecutive codes including p, 1+0+0≠0 for opabc and 0+0+1≠0 for pabcd, so the property (b) is lost.

That is, in the one-track absolute pattern created by the above-mentioned methods (1) and (2), sensor reading error detection and correction depending on the above-mentioned property (b) cannot be performed over the entire circumference of the pattern. can't be done, and the above (
It was necessary to adopt a special method that does not depend on the nature of b). For example, regarding the part where the property (b) is lost, a huge comparison table (
A semiconductor memory (semiconductor memory) was prepared in advance and the binary numbers read by the sensor were compared and determined one by one. In this method, (b
) A complex circuit containing elements with huge capacitance and high-speed response is added to the sensor reading error detection circuit that depends on the nature of the sensor reading error.

[0019]

[Means for Solving the Problems] The present invention provides a rotary absolute encoder with an arbitrary period, which can perform reading error detection using the property (b) described above over the entire circumference. The purpose is to provide a.

[0020] The error detection function according to claim 1 of the present invention
A tally absolute encoder consists of a code plate provided with a one-track absolute pattern, and a code plate provided with a one-track absolute pattern.
In a rotary absolute encoder comprising a plurality of sensors arranged along a code plate and a detecting section movable relative to a code plate, the one-track absolute pattern The number of codes in one period is (
The pattern is generated from a primitive polynomial selected to be a specific number (industrially useful), and testing means is provided for detecting misreading of the plurality of sensors using the relationship of the primitive polynomial. It is something.

[0021] The error detection function according to claim 2 of the present invention
The rotary absolute encoder according to claim 1 is a rotary absolute encoder according to claim 1, wherein the inspection means is used to combine a plurality of relationships of the primitive polynomials to identify a misread sensor. That is.

[0022]

[Operation] In the rotary absolute encoder with an error detection function according to claim 1 of the present invention, an M sequence to which a primitive polynomial is given in advance is converted into an arbitrary cycle using the methods of (1) and (2) above. Instead, one primitive polynomial is selected from a set of countless primitive polynomials that do not constitute an M sequence but have various periods, giving priority to that period. In other words, here we have a 1-track absolute putter.
The periodic series used as a sequence is defined as ``the y-digit binary numbers extracted at any phase position on the sequence are all different and 1
Rather than an M sequence in which all y-digit binary numbers (excluding all digits 0) are included in the period, the y-digit binary numbers extracted at any phase position on the sequence are all different, but at a certain initial stage. Before exhausting all binary digits of y digits starting from the value (=2y
It is simply a periodic series that returns to its initial value (with a period less than -1), and exhibits various miscellaneous periods depending on the combination of primitive polynomials and initial values. Therefore, a list of combinations of primitive polynomials and initial values with the required periods (industrially useful 60, 360, 10n, 2y, etc.) is created, and from the table, the degree of the primitive polynomial is as low as possible ( (requires fewer sensors).

The primitive polynomial g(x)=
ayxy+ay-1xy-1 + ... +a1x1+a
When a periodic sequence is generated from the combination of 0 and the initial value of a y-digit binary number, it follows that (a) the y-digit binary numbers extracted at any phase position on the sequence are all different, and (b) For the y+1-digit binary number extracted at any phase position, the exclusive OR of all the codes at the position corresponding to ay=1 of the primitive polynomial g(x) is always 0. Therefore, the one-track absolute signal that becomes the periodic series is
In the pattern, the absolute position can be determined based on the property (a), and reading error detection can be performed using the property (b) without any discontinuity over the entire circumference.

For example, the primitive polynomial g(x)=x6+x4
Periodic sequence generated from initial value 000001 for +x3+x2+x+1: 00000110001001
Since 0111110011101101 closes one cycle with 30 operations, it is applied to a one-track type absolute pattern with 30 pulses per cycle. On this pattern, due to the property (a), it is possible to obtain an absolute position signal that mutually discriminates 30 absolute positions in one revolution by reading six consecutive minimum reading units, and due to the property (b), an absolute position signal is obtained that mutually discriminates 30 absolute positions in one revolution. Reading errors can be detected seamlessly.

[0025] Also, the primitive polynomial g(x)=x7+x2+
Periodic series generated from initial value 0000001 for x+1: 00000011011010001010
1100011110111111001001 is 6
Since one cycle is closed with 0 calculations, 1 of 60 pulses per cycle
Applies to track-type absolute patterns. On this pattern, due to the property (a), an absolute position signal that mutually discriminates 60 absolute positions in one revolution is obtained by reading seven consecutive minimum reading units, and due to the property (b), the absolute position signal Reading errors can be detected seamlessly.

Similarly, the primitive polynomial g(x)=x9+x8
Initial value 0000000 for +x4+x3+x+1
From the periodic series generated from 01, a 1-track absolute pattern of 120 pulses per cycle is obtained, which allows reading errors to be detected without interruption throughout the cycle.

Similarly, the primitive polynomial g(x)=x6+x4
From the periodic sequence generated from the initial value 001000 for +x2+x+1, a one-track absolute pattern of 23 (8) pulses per cycle is obtained without any discontinuity throughout the cycle, allowing detection of reading errors.

Similarly, the primitive polynomial g(x)=x9+x8
From the periodic sequence generated from the initial value 000000001 for +x+1, a one-track absolute pattern of 24 (16) pulses per cycle is obtained without any discontinuity throughout the cycle, allowing detection of reading errors.

[0029] The error detection function according to claim 2 of the present invention
In a tally-absolute encoder, a misread sensor is identified by combining a plurality of primitive polynomial relationships. In other words, for the outputs of a large number of consecutively installed sensors, a set of codes corresponding to ay = 1 of the primitive polynomial g(x) is given a phase of 1.
A plurality of sets are taken out in a shifted manner, and a reading error sensor is identified using a comparison table in which the combination of the calculation results of each set is used as an address.

[0030]

Embodiments An embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic diagram of a rotary absolute encoder according to an embodiment of the present invention. here,
A reading error alarm is issued all around the 1-track type absolute pattern of 30 pulses per cycle.

In FIG. 1, the one-track absolute pattern T formed on the code plate A is a primitive polynomial,
A periodic sequence with 1 period of 30 generated from the initial value 000001 using g(x)=x6+x4+x3+x2+x+1, 0000011000100101111100111
01101 is replaced with the minimum reading unit of transparent part: 0 and light shielding part: 1 and arranged with ▲ as the starting point. Further, in the detection unit B which is movable relative to the code plate A, nine sensors S1 to S9 are arranged along a pattern T so as to detect nine consecutive minimum reading units. . The discrimination circuit C converts the binary numbers read by the sensors S1 to S9 into the formula x6+x4+x3+
The determination is made by comparing the relationship x2+x+1 = 0, and a warning signal w is output if there is a reading error.

To explain in more detail, the primitive polynomial g(
x)=x6+x4+x3+x2+x+1 Two sets of sensor outputs, sensors S1, S3, S4, S5, S6, S7, S8 and S, taken out in a positional relationship where the coefficient of
2, S4, S5, S6, S7, S8, and S9, respectively, find exclusive ORs c1 and c2, and if both are 0, there is no error in reading, so the warning signal w will also be 0, but c1
, c2 becomes 1, the sensors S1 to S
It is determined that a reading error has occurred in any one of 9, and the warning signal w becomes 1.

In the rotary absolute encoder configured as described above, when the detection section moves relative to the code plate A, the sensors S1 to S9 read the nine digits from the pattern T. Addresses (absolute position signals) for determining each of the 30 absolute positions along the pattern T are assembled using predetermined consecutive six-digit codes of the binary number. At the same time, the discrimination circuit C confirms that there is no reading error in the sensors S1 to S9, and if a reading error occurs in even one of the sensors S1 to S9, a warning signal w is immediately issued.

Furthermore, in this embodiment, the presence or absence of a reading error is basically checked, but if the combination of 1 and 0 of the two exclusive ORs is examined, it can be determined that there is a reading error in one sensor. Misreading occurs in sensors S4 to S8, sensors S2 and S
9, it can be determined in which group of sensors S1 and S3 the problem is occurring. This is because if any one of the sensors S4 to S8 that is common to both exclusive ORs is misread, the combination of outputs c1 and c2 will be 1, 1, and the combination of outputs c1 and c2 will be 1, 1, and the outputs of sensors S2 and S9 that belong to only one exclusive OR will be incorrectly read. Any one
If the sensor is misread, only the output c1 will be output from the sensor S1, S
If any one of 3 is misread, only the output c2 will be 1.

Furthermore, as proposed in Japanese Patent Application No. 2-251730, if more sensors are arranged in the detection section B and the number of such exclusive OR sets is increased, the combination of the outputs of each set can be increased. Based on this, it is possible to identify a sensor with a reading error in one sensor, or to identify a sensor with a reading error or issue a warning of the occurrence of a reading error in the event of a reading error in two or more sensors.

[0037]

Effects of the Invention In the rotary absolute encoder with an error detection function according to claim 1 of the present invention, the one-track absolute pattern has an industrially useful 60,
Even though the period is 360, 10n, 2y, etc., the pattern is continuous without any exception or break throughout the entire circumference.
Regarding the y+1 digit binary number extracted at any phase position,
Since the error detection can be performed using the relationship that "the exclusive OR of all the codes at the position corresponding to ay=1 of the primitive polynomial g(x) is always 0," the error detection circuit (
The inspection means) can be relatively simple.

[0038] The error detection function according to claim 2 of the present invention
For tally/absolute encoders, the one-track absolute pattern is industrially useful.
Despite having a period of 60, 10n, 2y, etc., without exception, "a of the primitive polynomial g(x) for a binary number of y+1 digits extracted at any phase position"
The exclusive OR of all the codes at the position corresponding to y=1 is always 0.'' Since error detection can be performed using the relationship, a relatively simple circuit is required to identify a sensor with a reading error. .

[Brief explanation of the drawing]

FIG. 1 is a schematic diagram of a rotary absolute encoder according to an embodiment.

[Explanation of symbols]

A Code plate B Detection section C Discrimination circuit T 1-track type absolute pattern S1
Sensor S2 Sensor S3 Sensor S4 Sensor S5 Sensor S6 Sensor S7 Sensor S8 Sensor S9 Sensor

Claims (2)

[Claims] [Claim 1] One-track absolute putter
In a rotary absolute encoder, the rotary absolute encoder consists of a code plate provided with a pattern, and a detecting section that includes a plurality of sensors arranged along the pattern and is movable relative to the code plate. , the one-track absolute pattern is a pattern generated from a primitive polynomial selected such that the number of codes in one period is a specific number, and the plurality of patterns are generated using the relationship of the primitive polynomial. 1. A rotary absolute encoder with an error detection function, characterized in that it is provided with a test means for detecting reading errors of individual sensors. [Claim 2] The rotary absolute according to Claim 1
A rotary absolute encoder with an error detection function, characterized in that the inspection means is a means for identifying a reading error sensor by combining a plurality of relations of the primitive polynomials. .