GB2587771A

GB2587771A - Improvements relating to rotary encoders

Info

Publication number: GB2587771A
Application number: GB1905669.6A
Authority: GB
Inventors: David Dixon Christopher; James Robert Huxford
Original assignee: TRW Ltd
Current assignee: TRW Ltd
Priority date: 2019-04-23
Filing date: 2019-04-23
Publication date: 2021-04-14
Anticipated expiration: 2039-04-23
Also published as: DE102020205005A1; GB201905669D0; GB2587771B

Abstract

A rotary encoder (12, fig.1) comprises an annular track with an alternating pattern of at least two different encoding regions. A first detector 16 produces a first alternating output signal as the track of encoding regions rotates about its axis past the detector. The state of the output signal depends on which encoding region is facing the first detector. A second detector 17, offset from the first detector, produces a second alternating output signal as the track rotates about its axis past the detector. The state of the output signal depends on which encoding region is facing the second detector. The combined values of the output signals assume a plurality of unique states, changing as the encoder rotates. Post-processing apparatus comprises a memory which stores first and second sets of compensation values, each associated with a possible state change. Each compensation value is indicative of an average error in the position at which the corresponding change of state occurs. A processing means is configured to generate a position signal that is updated when there is a change in the combined states of the output signals, the position signal being corrected based on a compensation value of one of the first set of compensation values associated with the latest change in state. The processing means compares the first set of compensation values with the second set of compensation values to determine a change in the sensor. The first and second set of compensation values is generated during and after an initial calibration respectively.

Description

IMPROVEMENTS RELATING TO ROTARY ENCODERS

The present invention relates to encoders, especially rotary encoders, to methods of determining the position of an object using a rotary encoder, to post processing apparatus for use with rotary encoders and to steering systems for vehicles that incorporate rotary encoders It is known to provide an encoder which comprises a track of magnetic elements arranged in an alternating sequence of North and South poles, and a detector which produces an output signal having a first state when proximal to one of the North poles and a second state when opposite one of the South poles. Thus, as the track moves past the detector the output of the detector will be a modulated signal which alternates between the first and second states.

An encoder with one detector is limited in use as it is not possible to tell which direction the track is moving. This can be overcome by using two detectors, offset from one another by an amount that is less than the spacing between centre of a North pole and the centre of an adjacent South pole. This is shown in Figure 4 of the accompanying drawings. The two detectors are typically identical, each producing an alternating sequence of first and second states as the track moves but with the two patterns offset from one another.

The combined values of the two outputs from the detectors will pass through four states, as shown by the state machine in Figure 7 of the drawings, and by identifying the states before and after the latest change of state it is possible to identify the direction in which the track is rotating. Each change of state will occur as the detector crosses an edge where two adjacent poles meet. If the magnetic poles are all equal length these edges will be evenly spaced apart around the track, and if the detectors are spaced apart by an angle equal to half of the spacing between pole centres the states will change at regular, equally spaced, time intervals, when the track rotates at a constant velocity. The velocity can therefore be determined from the elapsed time between each change in state.

The change of state will not, by itself, uniquely identify the position of the encoder track if the track has many poles, which will always be the case in a practical encoder.

Over one full rotation, a given change of state will occur multiple times and this will be repeated on further revolutions of the encoder. However, by counting the changes in state, it is also possible to produce a position signal relative to a known datum position.

It is known that encoders of this form suffer from inaccuracies if the spacing between the magnets is not ideal or if external influences, such as other magnets, cause distortion in the magnetic fields emitted by the magnets as seen by the detectors. Variations in the switching threshold of the detectors can also lead to inaccuracies.

This can lead to small shifts in the position at which the combined output signals change state away from the expected positions. For example, in the case where the changes should occur at equal time intervals when the track is rotating at a constant velocity as described above, the error can lead to different timings between the changes in state as the changesOccur at positions that are offset from the ideal expected positions.

Such shifts in the position of the changes of state of a rotating encoder track can lead to the presence of unwanted harmonic frequencies in the position signal output from the encoder. For example, where the encoder track comprises an annular disc of 36 magnets, rotating at a constant angular velocity, a noise component of the 36th order may be observed. When the position signal is being used in a sensitive application, such as within the control loop of a motor control circuit for a motor of an electric power assisted steering system, this noise can cause acoustic noise where the harmonic frequency interacts with the resonant frequency of a part of the motor or other part of the steering system.

The applicant's own prior publication, WO 2015 004 472 A2, discloses a post-processing assembly for a rotary encoder, the post-processing apparatus having a memory that stores compensation values associated with the changes of state, or transitions. These stored values are used to correct a position signal when it is updated on the occurrence of a respect one of the changes of state, or transition, of the combined output signals. Using compensation values that represent an average error in the position of a change it has been found that a position signal can be produced in which the level of harmonics is greatly reduced or even completely eliminated.

However, whilst adapting the position values over time is helpful for the operation of the rotary encoder, by compensating for inaccuracies of the encoder assembly, it has been noted that it would be useful to be able to predict for potential failure of such an assembly. The present invention aims to provide such a predictive assembly.

According to a first aspect, there is provided a rotary encoder post-processing apparatus for use with a rotary encoder comprising: an annular track comprising an alternating pattern of at least two different encoding regions; a first detector at a first fixed position which is arranged to produce a first alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the first detector; a second detector at a second fixed position which is offset from the position of the first detector and is arranged to produce a second alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the second detector; in which the combined values of the two output signals is capable of assuming a plurality of unique states with the states changing from one to another as the encoder rotates, in which the post-processing apparatus comprises: a memory which stores a first set of compensation values and a second set of compensation values, each compensation value of the first and second sets of compensation values being associated with a respective one of the possible state changes that can occur when the combined values of the two output signals change upon movement of the encoder track, and in which each compensation value is indicative of an average error in the position at which the corresponding change of state occurs; and a processing means which is configured to generate a position signal that is updated when there is a change in the combined states of the output signals of the two detectors, the position signal being corrected by an amount indicated by a compensation value of one of the first set of compensation values and the second set of compensation values associated with the latest change in state; wherein the processing means is configured additionally to compare the first set of compensation values with the second set of compensation values to determine a change in the sensor; wherein the first set of compensation values has been generated during an initial calibration of the rotary encoder and the second set of compensation values is generated after the initial calibration of the rotary encoder.

It is therefore possible to compare a current set of compensation values, generated after an initial calibration of the rotary encoder, to a historic set of compensation values generated during the initial calibration of the rotary encoder. By comparing these sets of compensation values, it can be determined whether or not the compensation values are drifting or changing over time, for example over the lifetime of the rotary encoder. It has been noted that changes in the compensation values can be indicative of a change in the encoder, or indicative of an impending failure of the encoder, for example in the sensor switching times or changes in the magnetic track.

Thus, by comparing the sets of compensation values, potential problems may be spotted prior to the eventual failure of the assembly or encoder system.

The first and second alternating output signals provided by the sensors are binary, i.e. they are indicative of either a North or a South, and therefore usual methods of determining sensor drift are not usable. By using the compensation values, which are generated by relative changes in the sensors and are calculated over the course of a rotation of the encoder, a change in the sensor is possible to measure indirectly.

The memory may be configured to store the first set of compensation values in a calibration compensation table and the second set of compensation values in a runtime compensation table. The runtime compensation table may therefore be updated during the use of the encoder assembly, whilst the calibration compensation table may keep the same compensation values during the lifetime of the encoder assembly.

Comparisons of the two compensation tables can then give an overall indication of changes in the encoder over time.

The difference between the first set of compensation values and the second set of compensation values may be compared to a predetermined amount. if greater than this predetermined amount, it may be determined that a change in the sensor has occurred.

The predetermined amount may be selected to correspond to a difference that is large enough that it cannot be due to random errors, for example. Otherwise, the predetermined amount may be selected to correspond to a difference known to be indicative of an impending failure of the encoder.

The difference may be calculated by comparing each pair of compensation values. By "pair of compensation values" we mean each compensation value of the first set of compensation values and its corresponding compensation value of the second set of compensation values -e.g. the compensation values from each set that correspond to a 0<>1 change of state. Any absolute difference between any pair of compensation values, or a difference above a predetermined amount, may be indicative of an issue, error, or drift of the underlying sensors system. The compensation values may also be compared using percentage differences rather than absolute differences.

The difference may be calculated by summing the differences between each pair of compensation values. The total of the differences may be calculated by summing the moduli of the differences between each pair of compensation values, or the sum of the square of each pair of compensation values.

The difference may be calculated by comparing plots of the sets of compensation values. For example, a shape of a plot of the first set of compensation values may be compared to a shape of a plot of the second set of compensation values. The plots may be compared by comparing symmetry, skewness, peak values, spread variance, or any other feature of the plots or statistical analysis.

The processing means may be configured to correct the position signal by an amount indicated by the compensation value of the first set of compensation values associated with the latest change in state.

Alternatively, the processing means may be configured to correct the position signal by an amount indicated by the compensation value of the second set of compensation values associated with the latest change in state.

The system may therefore actively respond to changes in the compensation values during use or may simply monitor for changes whilst using the compensation values determined during initial calibration of the encoder. The latter may provide for simpler operation of the system whilst retaining the advantage of being able to identify changes in the encoder.

The initial calibration may comprise an offline calibration. This means that the initial calibration may be performed once assembled but prior to use of the rotary encoder assembly.

Alternatively, the initial calibration may comprise an online calibration This means that the initial calibration may be performed during use of the rotary encoder as The encoder may generate a count signal from the combined values of the two output signals, and the post-processing apparatus may receive this count signal and generate the position signal by processing the count signal.

Alternatively, the encoder may output directly the two output signals and the post-processing means may generate a count signal when generating the position signal.

According to a second aspect, there is provided a rotary encoder assembly comprising: an annular track comprising an alternating pattern of at least two different encoding regions; a first detector at a first fixed position which is arranged to produce a first alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the first detector; a second detector at a second fixed position which is offset from the position of the first detector and is arranged to produce a second alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the second detector; in which the combined values of the two output signals is capable of assuming a plurality of unique states with the states changing from one to another as the encoder rotates; and a memory which stores a first set of compensation values and a second set of compensation values, each compensation value of the first and second sets of compensation values being associated with a respective one of the possible state changes that can occur when the combined values of the two output signals change upon movement of the encoder track, and in which each compensation value is indicative of an average error in the position at which the corresponding change of state occurs; and a processing means which is configured to generate a position signal that is updated when there is a change in the combined states of the output signals of the two detectors, the position signal being corrected by an amount indicated by a compensation value of one of the first set of compensation values and the second set of compensation values associated with the latest change in state; wherein the processing means is configured additionally to compare the first set of compensation values with the second set of compensation values to determine a change in the sensor; wherein the first set of compensation values has been generated during an initial calibration of the rotary encoder and the second set of compensation values is generated after the initial calibration of the rotary encoder.

Each compensation value of the second set of compensation values, and optionally of the first set of compensation values, may be derived from observing multiple instances of the corresponding change of state of the combined output signals from the two detectors, the multiple observations being combined to provide an arithmetic average value that is stored as the compensation value of the respective set of compensation values.

Where the first set of compensation values is generated this way, this may be considered to be the initial calibration of the encoder.

Each stored compensation value of the second set of compensation values, and optionally of the first set of compensation values, may be indicative of the error in the relative position of a state change to the position of one state change which is chosen as a reference state change The error may therefore be expressed in terms of a mechanical angular value, or an electrical angular error value Any one of the changes in state may be selected as the reference change in state.

The processing means may include means adapted to update the stored compensation values of the second set of compensation values by performing the steps of: recording the relative timing at which each change in state in a complete cycle of changes of state occurs, a cycle corresponding to each state changing at least once, determining the total time taken for all changes to occur starting with a first change in state and ending when that first change in state occurs again, determining the fraction of that total time between each change in state, deriving a compensation value of the second set of compensation values for each change in state apart from the first change in state based on the fraction and the total elapsed time; and subsequently determining an intermediate compensation value for the first change in state which results in the sum of all of the compensation values of the second set of compensation values being equal to zero.

The encoder may be arranged to update the stored compensation values by combining them with the intermediate compensation value. For instance, the update may comprise forming a weighted sum of the stored value and the intermediate value.

The average may therefore be an average for several cycles of the state machine, i.e. one repeat of the transitions.

The encoder may be arranged to perform the update when the encoder is rotating above a minimum threshold angular velocity. In order to function correctly, the encoder must be rotating at a constant velocity during the time of capture The processor may increment the position signal in the event that a change of state occurs which is associated with rotation of the encoder in a first direction, and may decrement the position signal when a change of state occurs that is associated with rotation is a second, opposite. direction.

The encoder may comprise No, encoding regions, and they may be spaced equally from adjacent encoding regions. There may be two types of encoding regions arranged in an alternating pattern, each of the detectors producing a different response depending on which type it sees. For instance one type may be a North pole and the other a South pole, the detector being perhaps a Hall Effect sensor.

The centre of each region may be spaced from the next region by 360/N0p degrees. Where Nop is greater than two, then each change in state of the combined output signals of the detectors will occur at least twice in each full rotation, allowing a mean value to be calculated and stored in the memory.

The two detectors may be offset angularly by 1/4*360/Nep degrees.

The encoder may be associated with a motor and used to produce a motor position signal indicating the mechanical or electrical position of the motor. The processing means may then generate the position signal from the count signal by the following equation: position = M0D360(count signal value x 3600 x Np / (4 x Nei)) + stored compensation value; where position in the electrical position, stored compensation value is the value in the memory associated with the latest change in state. No" is the number of encoder poles, and Np is the number of rotor magnet pole pairs which defines the number of electrical cycles per mechanical revolution of the motor.

Where there are two detectors, the value of the two output signals will adopt four unique states, giving four corresponding unique changes of state that each occur once over a machine state cycle but occur many times during each rotation of the encoder.

The memory may therefore store only four mean position error values, one for each transition. Storing only four values, one per change in state, uses far less memory than would be required if a position error value was to be stored for every possible position of the encoder. For instance, with 36 poles and 2 sensors with 2 states per sensor there would need to be 36 x 2 x 2 = 144 values stored, rather than 4, giving a saving of a factor of 36. The time to learn all of these values would also be greater.

The encoding regions of the encoder may each be defined by a pole of a magnet, the poles being arranged as an alternating pattern of North and South poles. The encoder detector may then comprise a Hall Effect sensor or other device that is sensitive to magnetic fields.

Other types of encoding region can be used, but in each case it is preferred that there are two different basic encoding regions that are arranged in an alternating pattern around the encoder track. For instance, the encoding regions may comprise two different colour patches, such as red and green, arranged in an alternating pattern or red-green-red-green-etc., for use with an optical detector that produces an output signal having a first value when it sees red and a second value when it sees green.

The detector may produce an output signal of a first value when facing a North pole and a second value when facing a South pole The state will therefore change as the magnets move past the detector.

The encoder may include more than two detectors, for instance three or four detectors, and in such as case the combined output signal may have more than the four possible state changes that are present with an encoder having two detectors. For instance, with three detectors there will be eight different changes of state. The memory may store one calibration value for each of the different changes of state The encoder calibration and compensation described in this application may therefore be extended out to the calibration and compensation of any position sensor which uses two or more sensors together with a track of encoding regions. The term encoder should be interpreted broadly to cover any sensor which converts a position into an electrical signal, and may be extended to include linear encoders as well as rotary encoders.

According to a third aspect, there is provided a method of determining a change in a rotary encoder comprising: an annular track comprising an alternating pattern of at least two different encoding regions; a first detector at a first fixed position which is arranged to produce a first alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the first detector; a second detector at a second fixed position which is offset from the position of the first detector and is arranged to produce a second alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the second detector; in which the combined values of the two output signals is capable of assuming a plurality of unique states with the states changing from one to another as the encoder rotates; and a post-processing apparatus comprising: a memory which stores a first set of compensation values and a second set of compensation values, each compensation value of the first and second sets of compensation values being associated with a respective one of the possible state changes that can occur when the combined values of the two output signals change upon movement of the encoder track, and in which each compensation value is indicative of an average error in the position at which the corresponding change of state occurs; and a processing means which is configured to generate a position signal that is updated when there is a change in the combined states of the output signals of the two detectors, the position signal being corrected by an amount indicated by a compensation value of one of the first set of compensation values and the second set of compensation values associated with the latest change in state; wherein the first set of compensation values has been generated during an initial calibration of the rotary encoder; the method comprising the step of: comparing the first set of compensation values with the second set of compensation values to determine a change in the sensor.

The method may further comprise updating the second set of compensation values by performing the steps of: observing multiple instances of the corresponding change of state of the combined output signals from the two detectors; combining the observations to provide an arithmetic average value that is stored as the compensation value of the second set of compensation values; and repeating the step of comparing the first set of compensation values with the second set of compensation values to determine a change in the sensor.

A change in the sensor may be determined by the first set of compensation values and the second set of compensation values differing by more than a predetermined amount.

The step of comparing the first set of compensation values with the second set of compensation values to determine a change in the sensor may be repeated throughout a lifetime of the rotary encoder.

According to a fourth aspect, there is provided a steering system for vehicle, comprising a rotary encoder assembly according to the second aspect of the invention.

The invention will now be described in detail with reference to the accompanying drawings, in which: Figure I is a schematic view of a steering assembly which includes an encoder in accordance with an aspect of the invention; Figure 2 is a block diagram showing the key functional components of the encoder which provide a position measurement signal to the motor controller of Figure 1; Figure 3 is a plan view of the encoder disk of the encoder and the relative positions of the detectors; Figure 4 is a detailed plan view of a part of the encoder of Figure 3 showing the relative positions of the magnet poles and the detectors; I3' Figure 5 is a timing diagram showing the values of the output signals from the two detectors P and Q and the corresponding four encoder states 0. I, 2, and 3 represented by the combined output signals; Figure 6 shows the variation in the count signal due to the change of states, represented by the arrows, that occurs as the encoder rotates in one direction and as it rotates in the other direction; Figure 7 is a state diagram showing by arrows the four possible changes of state as the encoder rotates; Figure 8 is a diagram showing the measured errors in position for each position of the encoder at which a change in state occurs, along with a mean error value for each set of changes of state, i.e. a mean value for all errors corresponding to change of state 0-2, a mean value for all errors corresponding to change of state 1-0 and so on; Figure 9 is a timing diagram showing the ideal positions of encoder state changes, the actual positions of encoder edges in an imperfect encoder, and the effect of errors in the relative positions on the timings at which the states change over a full cycle of encoded state change; Figure 10 is an overview of the method used to compare the first set of compensation values with the second set of compensation values, where the count value is being combined with the second set of compensation values; and Figure 11 is an overview of the method used to compare the first set of compensation values with the second set of compensation values, where the count value is being combined with the first set of compensation values.

An electric power assisted steering system in shown in Figure 1. The system 10 comprises an electric motor 11 that is connected to part of a steering shaft (not shown) through a gearbox or drive belt. The motor 1 I. in use, applies an assistance torque to the steering shaft that helps a driver to turn the steering wheel. To determine how much assistance torque is required, a torque sensor (not shown) is attached to the steering shaft and provides a torque signal to the motor controller indicative of the torque carried by the steering shaft.

The motor I I is a three phase AC motor which is driven by applying pulse width modulated voltage waveforms 22 from a motor controller 21 to each of the three phases. Such motors 11 and PWM schemes are well known in the art and will not be described here in detail. Unless the motor drive scheme is one of the well-known position sensorless schemes, a position sensor 12 must be provided, which feeds a position signal into the motor controller 21 indicative of the angular position of the motor rotor. As shown, the sensor 12 comprises a rotary encoder that is connected to the output shaft 13 of the motor 11, but could be connected to the steering shaft.

It comprises an encoder disk 14 attached to the output shaft 13 of the motor I I. The disk 14, which can also be seen in Figures 3 and 4, comprises Nei, encoder pole magnets, in this case 36 magnet encoder poles, arranged as an alternating sequence of North and South poles. Each magnet has the same width, and the centres of each magnet are spaced from the adjacent magnets centre by 360/N," degrees.

The encoder 12 includes a support bracket 15 that is fixed in position relative to the motor casing so that it does not move as the encoder rotates. The bracket 15 supports two sensors 16, 17 (referred to here as detectors P and Q, for "phase" and "quadrature"), each one in this embodiment comprising a Hall Effect sensor. The active part of each sensor 16, 17 faces the magnets so that the output signal from each Hall Effect sensor will be one of two states depending on whether it 'sees' a North pole or a South pole on the magnetic encoder track. For convenience, the states are defined here as 1 and 0. The two sensors P and Q are offset around the circumference of the encoder track by an angular distance of Ve360/N,, degrees.

As the motor rotates, the output signal from each of the P and Q sensors changes state according to the polarity of the magnetic track facing the magnets, producing the repetitive pattern pulse train as shown in Figure 5, where the encoder changes state (the combined state of P and Q) every 1/4*360/N," degrees.

Depending on the direction of rotation, the sensor that first sees the edge of the magnetic track changes. This can be used to determine the direction of rotation. This can also be seen in Figure 6. The combined values of the two output signals will change between one of four possible states, the sequence of states depending on the direction of rotation of the encoder disk as shown in Figures 6 and 7. For convenience the states are referred to as state 0, I, 2 and 3.

The position sensor is shown schematically in Figure 2. In this example the encoder includes a counter 18 and this can be incremented/decremented according to the direction of rotation of the encoder as identified from the change of state as shown in Figure 6. The value of the counter is used by a processor 19 as the basis for a measurement of the position of the encoder and it is this position measurement that is fed to the motor controller 21. In this example, as shown in Figures 6 the position count increments for the sequence of states 0,2,3,1,0,... and decrements for the sequence 0,1,3,2,0,.., but this is arbitrary and may be swapped.

In reality, the state transitions will not occur every 1/4*360/Nep degrees apart, with errors present due to, amongst other things, imperfections in the encoder track, magnetic interference from nearby magnetic fields, sensor switching levels, and variations in the air gap between the track and the detectors. An example of this potential error can be seen in Figure 8.

To reduce the effect of these errors, the encoder includes a memory 20 and a processor 19. This may, of course, be provided separate from the encoder as part of a post-processing apparatus, the encoder simply providing the P and Q signals and optionally producing the count. The counting could also be done by a discrete post-processing circuit.

The memory 20 stores a first set of calibration values and a second set of calibration values, each set of calibration values including one value for each transition in the state machine. The values are stored in two look-up table which are indexed by the state changes. Each calibration value is representative of the mean error in the position of the change of state in the machine over at least a part of a complete revolution of the encoder track compared with the expected position. The processing means, upon detecting a change in state of the machine, produces a position signal that is corrected using the stored calibration value. For example, if the mean error value indicates that the location of the edge is not at the ideal position, the time at which the change occurs will be delayed by an amount equal to the store value.

The angular amount by which the change in state position is moved depends on the value stored in the memory.

Two possible ways of generating the calibration values are set out below.

Offline calibration An example measurement can be seen in Figure 8, with the individual transitions in state for the 4 encoder states over one mechanical revolution shown, along with their mean values. As we are looking at the relative change-change error any DC bias can be removed, i.e. the mean value for all states should be removed to bias around zero degrees such that: errorO<>1 + error2<>0 + error3<>2 + error I <>3 = 0 where -error x<>y" is a stored calibration value corresponding to a change from state y to state x or a change from state x to state y.

The net effect over one mechanical revolution is 00. meaning that any alignment offset used to ensure the position sensor is aligned with the back ENIF is not affected. For example if this algorithm was introduced into an existing system any position sensor alignment correction would remain valid and not require re-calibrating which would be a costly and time consuming exercise.

With a set of 4 mean state transition error values the encoder position calculation can be modified to include the compensation term as shown in Figures 10 and I I. Note that the compensation values may be expressed in mechanical degrees or electrical degrees, e.g. for a 4 pole pair machine a mechanical error of 10 is a 4° error in the electrical reference frame. Calculation of the encoder position, where the compensation is in °electrical, for a motor with Np pole pairs is: encoder position = M0D360(encoder count x 3600 x N / (4 x Nop)) + encoder state compensation where encoder state compensation is extracted from CALIBRATION COMPENSATION TABLE based on the latest change in state. This is based on the set-up of Figure 11.

Online calibration An alternative to offline calibration against a second sensor is to calculate the mean compensation during the normal operation of the motor. This is achieved by modifying the off-line compensation algorithm to include an adaptive calculation to learn the offsets during normal operation rather than off-line. Rather than observing the absolute positions at which changes occur, the relative positions of changes in a sequence are used.

The adaption algorithm may be executed once per cycle of the encoder state machine. One state transition is chosen as the reference transition, the choice is arbitrary but in this case Oolhas been chosen. It may be executed for every cycle (a cycle comprising a full sequence in which every state change occurs once only), or may be performed periodically (either at pre-set time intervals, or when pre-set conditions are met or at random or pseudo-random intervals of time).

The adaption principle is summarised in Figure 10 and is used as the basis of the

description of the algorithm.

At the end of a cycle, on the reference change of state, transition (0<>1), the timing information for the previous 4 encoder state transitions are used to estimate the errors in position for each of the state transitions. An ideal encoder has equidistant edges; in reality this is not the case and edge-to-edge errors occur where the states are not uniform in size. Assuming for one cycle of the encoder states the velocity is constant (which over the small angular distance of one encoder cycles is probable) then the edges should occur uniformly at 0, 0.25, 0.50, 0.75 and 1.0 over the cycle, where 0 and 1.0 are the same reference state transition By capturing the transition times over tins period the equivalent edge positions can be estimated and the edge-to-edge error (compensation) terms calculated, i.e. the timing error can be expressed as a ratio of the overall time (t0 -t4). This ratio can then be expressed as a position error scaled to the overall distance rotated (N,,, x 4). Depending on the direction of rotation the state after the 0<> I reference transition will be either 0 or 1. The ratio values for the 3 states are different depending on direction of rotation (see Figure 3), and are summarised in Table 1 where EST is encoder state transition 2<>0, 3<>2 or 1<>3.

errorLEST] = encoder adaption ratioLEST] x (Nei, x 4) These error terms are stored in either CALIBRATION COMPENSATION TABLE. Once generated, the calibration compensation terms, i.e. the first set of compensation values, for a particular encoder will be fixed. For robustness, the update of the CALIBRATION COMPENSATION TABLE may be filtered to reduce the effects of measurement error in the edge timing, e.g. CALIBRATION COMPENSATION TABLELEST] = ((1 -Kf) x CALIBRATION COMPENSATION TABLE LEST]) + (Kf x error[ESTD Where Kf is the filter constant (0 < Kf <= I) As with the offline calibration, the net effect of the compensation should be zero to avoid the introduction of a position offset. This information can be used to calculate the correct value for the reference transition: tit'Ittt,- 2 <,; -t-t, 2 0:50, ilti - .0,75.

- t if kb 1 ct 3 encocior state trarisMon it:six:oder adaption rao[E: The adaption ratio can then be used to calculate the error (compensation) values: CALIBRATION COMPENSATION TABLE [0°11 = -(CALIBRATION COMPENSATION TABLE 12001 + CALIBRATION COMPENSATION TABLE [3<>2] + CALIBRATION COMPENSATION TABLE [1<>3]) As with the offline algorithm, the compensation is applied based on the latest encoder state transition: encoder position = M0D360(encoder count x 360 x N / 4xN) + encoder state compensation where encoder state compensation is extracted from CALIBRATION COMPENSATION TABLE, based on the latest state trans t on This is based on the set-up of Figure I I: Limit Adaption Conditions To successfully adapt the motor must be rotating; at zero speed it is not possible to adapt. It is advantageous to limit the speed range over which the adaption algorithm operates, disabling the adaption when the speed drops below a specified threshold.

Similarly, it may be advantageous to disable the adaption above an upper speed threshold: Hysteresis may be applied to these thresholds to minimise jittering in and out of adaption.

Comparison of compensation values As well as the first set of compensation values, stored in the CALIBRATION COMPENSATION TABLE, the systems shown in Figures 10 and 11 include a second set of compensation values, which are stored in a RUNTIME COMPENSATION TABLE: The second set of compensation values are generated in the same way as the online calibration, but during normal use of the encoder assembly. The processing included is identical to that used to generate the first set of compensation values during online calibration, except that the values are stored in the RUNTIME COMPENSATION TABLE.

However, the generation of the second set of compensation values differs from the generation of the first set of compensation values in that, whilst the first set of compensation values is static once generated and stored in the memory, the second set of compensation values is updated throughout the lifetime of the encoder assembly. The updating may be at a regular interval, or may be upon start-up of the encoder assembly, or may be updated at any other time during its lifetime.

The generation of a second set of compensation values that is updated throughout the lifetime of the encoder assembly allows a change in compensation values over time to be determined by comparing this second set of compensation values with the first set of compensation values that is generated during the initial calibration (either online or offline). Any change in compensation values relative to the first set of compensation values will indicate some sort of change in the sensor, whether this is a physical change of the components of the sensor assembly, such as a change of the magnetic strips, or degradation of any sensing electronics, such as a change in switching time of the Hall Effect sensors.

The comparison of the first and second sets of compensation values may result in an indication that there is an error or likely future failure by comparison with a predetermined value, range, or difference that is indicative of such an error or likely future failure.

The memory may, in addition to the storing of the first and second sets of compensation values, store the differences that are calculated each time the first and second sets of compensation values are compared. In this manner, the differences can be tracked over a period of time in order to provide a view of the rate of change of the sensor.

Two different methods of operating the invention are shown in Figures 10 and I I. The methods of operation are generally very similar, differing only in which set of compensation values is used to generate the position signal of the encoder.

In Figure 10, the second set of compensation values is used in conjunction with the encoder count to generate the compensated encoder position. This second set of compensation values, stored in the RUNTIME COMPENSATION TABLE, is updated throughout the lifetime of the sensor assembly. Each time it is updated, the compensation values of the second set of compensation values is compared with the compensation values of the first set of compensation values, stored in the CALIBRATION COMPENSATION TABLE. The health of the encoder assembly can therefore be monitored by monitoring the difference between the two sets of compensation values The difference between the sets may be calculated in a number of different ways, and one or more ways of measuring the difference may be used to spot differences that could be indicative of an error. One way to calculate a difference may be compare each pair of compensation values, i.e. compare each compensation value of the first set of compensation values with a corresponding compensation value of the second set of compensation values. Any absolute difference between each pair of compensation values may be indicative of an error or change in the sensor assembly, or alternatively the difference may be compared to a predetermined amount. The compensation values may also be compared using percentage differences rather than absolute differences.

Whilst comparing individual differences between pairs of compensation values may be useful, it may also be useful to compare the differences of the sets as a whole. For example, a total difference may be calculated by summing the differences between each pair of compensation values. In the present embodiment, this sum should always equal 1, so instead of taking a simple sum of the differences it may be more instructive to take a sum of the moduli of the differences between the pairs of compensation values or a sum of the squares of the differences between each pair of compensation values.

As well as comparing the pairs of compensation values, it may be useful to provide a plot of each set of compensation values and then compare the plots using one or more measures. For example a shape of a plot of the first set of compensation values may be compared to a shape of a plot of the second set of compensation values. Such a comparison could be or include a comparison by symmetry, skewness, peak values, spread, variance, or any other feature of the plots or statistical analysis.

By generating the compensated encoder position using the second set of compensation values, which are, by definition, more up-to-date than the first set of compensation values, the compensated encoder position will be more accurate than would otherwise be the case.

In Figure I I. the first set of compensation values is used in conjunction with the encoder count to generate the compensated encoder position. The first set of compensation values is static once the initial calibration has been completed and therefore each encoder count will be adapted in the same way, no matter whether or not the second set of compensation values, which is updated throughout the lifetime of the encoder, now differs from the first set of calibration values. This provides some stability to the system.

As in the system of Figure 10, in the system of Figure 11 the first set of compensation values is compared to the second set of compensation values every time the second set of compensation values is updated. As such, the change over time of the compensation values can be determined and used to determine a change in the sensor.

Whilst the invention has been described in detail based on an encoder with encoding regions formed by alternating North and South poles, other encoding regions could be used with the scope of the invention. For example, the encoding regions could comprise regions of varying transmittance or reflectivity, perhaps of varying colour, and the detectors may comprise optical detectors. A transition will occur as the regions move past the detector and the detector secs a change in reflectance of colour. A light source may be provided as part of the encoder that directs light onto the encoding regions where it can be reflected back towards the detector, or direct light onto the back of the encoding regions where it may be partially transmitted through the encoding regions onto the detector or blocked by the encoding regions. Significantly, there must be at least two different types of encoding regions in order for the full set of four state changes to be identified with two detectors

Claims

CLAIMS1. A rotary encoder post-processing apparatus for use with a rotary encoder cornprising: an annular track comprising an alternating pattern of at least two different encoding regions; a first detector at a first fixed position which is arranged to produce a first alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the first detector; a second detector at a second fixed position which is offset from the position of the first detector and is arranged to produce a second alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the second detector; in which the combined values of the two output signals is capable of assuming a plurality of unique states with the states changing from one to another as the encoder rotates, in which the post-processing apparatus comprises: a memory which stores a first set of compensation values and a second set of compensation values, each compensation value of the first and second sets of compensation values being associated with a respective one of the possible state changes that can occur when the combined values of the two output signals change upon movement of the encoder track, and in which each compensation value is indicative of an average error in the position at which the corresponding change of state occurs; and a processing means which is configured to generate a position signal that is updated when there is a change in the combined states of the output signals of the two detectors, the position signal being corrected by an amount indicated by a compensation value of one of the first set of compensation values and the second set of compensation values associated with the latest change in state; wherein the processing means is configured additionally to compare the first set of compensation values with the second set of compensation values to determine a change in the sensor; wherein the first set of compensation values has been generated during an initial calibration of the rotary encoder and the second set of compensation values is generated after the initial calibration of the rotary encoder.
2. A rotary encoder post-processing apparatus according to claim I, wherein the memory is configured to store the first set of compensation values in a calibration compensation table and the second set of compensation values in a runtime compensation table.
3. A rotary encoder post-processing apparatus according to claim 1 or claim 2, wherein the processing means is configured to correct the position signal by an amount indicated by the compensation value of the first set of compensation values associated with the latest change in state.
4. A rotary encoder post-processing apparatus according to claim 1 or claim 2, wherein the processing means is configured to correct the position signal by an amount indicated by the compensation value of the second set of compensation values associated with the latest change in state.
A rotary encoder post-processing apparatus according to any preceding claim, wherein the initial calibration comprises an offline calibration or wherein the initial calibration comprises an online calibration.
6. A rotary encoder post-processing apparatus according to any preceding claim, wherein the encoder generates a count signal from the combined values of the two output signals, and the post-processing apparatus receives this count signal and generate the position signal by processing the count signal.
7. A rotary encoder post-processing apparatus according to any of claims 1 to 5, wherein the encoder outputs directly the two output signals and the post-processing means generates a count signal when generating the position signal.
S. A rotary encoder assembly comprising: an annular track comprising an alternating pattern of at least two different encoding regions; a first detector at a first fixed position which is arranged to produce a first alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the first detector; a second detector at a second fixed position which is offset from the position of the first detector and is arranged to produce a second alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the second detector; in which the combined values of the two output signals is capable of assuming a plurality of unique states with the states changing from one to another as the encoder rotates; a memory which stores a first set of compensation values and a second set of compensation values, each compensation value of the first and second sets of compensation values being associated with a respective one of the possible state changes that can occur when the combined values of the two output signals change upon movement of the encoder track, and in which each compensation value is indicative of an average error in the position at which the corresponding change of state occurs; and a processing means which is configured to generate a position signal that is updated when there is a change in the combined states of the output signals of the two detectors, the position signal being corrected by an amount indicated by a compensation value of one of the first set of compensation values and the second set of compensation values associated with the latest change in state; wherein the processing means is configured additionally to compare the first set of compensation values with the second set of compensation values to determine a change in the sensor; wherein the first set of compensation values has been generated during an initial calibration of the rotary encoder and the second set of compensation values is generated after the initial calibration of the rotary encoder.
9. A rotary encoder post-processing apparatus according to claim 8, wherein the memory is configured to store the first set of compensation values in a calibration compensation table and the second set of compensation values in a runtimecompensation table.
10. A rotary encoder post-processing apparatus according to claim 8 or claim 9, wherein the processing means is configured to correct the position signal by an amount indicated by the compensation value of the first set of compensation values associated with the latest change in state.
11. A rotary encoder post-processing apparatus according to claim 8 or claim 9, wherein the processing means is configured to correct the position signal by an amount indicated by the compensation value of the second set of compensation values associated with the latest change in state.
12. A rotary encoder post-processing apparatus according to any of claims 8 to 11, wherein the initial calibration comprises an offline calibration or wherein the initial calibration comprises an online calibration.
13. A rotary encoder post-processing apparatus according to any of claims 8 to 12, wherein the encoder generates a count signal from the combined values of the two output signals, and the post-processing apparatus receives this count signal and generate the position signal by processing the count signal.
14. A rotary encoder post-processing apparatus according to any of claims 8 to 12, wherein the encoder outputs directly the two output signals and the post-processing means generates a count signal when generating the position signal
15. A rotary encoder assembly according to any of claims 8 to 14, in which each compensation value of the second set of compensation values, and optionally of the first set of compensation values, is derived from observing multiple instances of the corresponding change of state of the combined output signals from the two detectors, the multiple observations being combined to provide an arithmetic average value that is stored as the compensation value of the respective set of compensation values
16. A rotary encoder assembly according to any of claims 8 to 15, in which each stored compensation value of the second set of compensation values, and optionally of the first set of compensation values, is indicative of the error in the relative position of a state change to the position of one state change which is chosen as a reference state change.
17. A rotary encoder assembly according to any of claims 8 to 16, in which the processing means includes means adapted to update the stored compensation values of the second set of compensation values by performing the steps of: recording the relative timing at which each change in state in a complete cycle of changes of state occurs, a cycle corresponding to each state changing at least once, determining the total time taken for all changes to occur starting with a first change in state and ending when that first change in state occurs again, determining the fraction of that total time between each change in state, deriving a compensation value of the second set of compensation values for each change in state apart from the first change in state based on the fraction and the total elapsed time; and subsequently determining an intermediate compensation value for the first change in state which results in the sum of all of the compensation values of the second set of compensation values being equal to zero.
18. A rotary encoder assembly according to claim 17, in which the processing means is arranged to update the stored compensation values of the second set of compensation values by combining them with the intermediate compensation value, optionally in which the processing means is arranged to perform the update when the encoder is rotating above a minimum threshold angular velocity.
19. A rotary encoder assembly according to any of claims 8 to 18, in which the processing means is adapted to increment the position signal in the event that a change of state occurs which is associated with rotation of the encoder in a first direction, and adapted to decrement the position signal when a change of state occurs that is associated with rotation is a second, opposite, direction.
20. A rotary encoder assembly according to any of claims 8 to 19, in which the encoding regions of the encoder are each defined by a pole of a magnet, the poles being arranged as an alternating pattern of North and South poles.
21. A method of determining a change in a rotary encoder comprising: an annular track comprising an alternating pattern of at least two different encoding regions; a first detector at a first fixed position which is arranged to produce a first alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the first detector; a second detector at a second fixed position which is offset from the position of the first detector and is arranged to produce a second alternating output signal as the track of encoding regions rotates about its axis past the detector, the state of the output signal depending on which of the two different encoding regions is facing the second detector; in which the combined values of the two output signals is capable of assuming a plurality of unique states with the states changing from one to another as the encoder rotates; and a post-processing apparatus comprising: a memory which stores a first set of compensation values and a second set of compensation values, each compensation value of the first and second sets of compensation values being associated with a respective one of the possible state changes that can occur when the combined values of the two output signals change upon movement of the encoder track, and in which each compensation value is indicative of an average error in the position at which the corresponding change of state occurs; and a processing means which is configured to generate a position signal that is updated when there is a change in the combined states of the output signals of the two detectors, the position signal being corrected by an amount indicated by a compensation value of one of the first set of compensation values and the second set of compensation values associated with the latest change in state; wherein the first set of compensation values has been generated during an initial calibration of the rotary encoder; the method comprising the step of: comparing the first set of compensation values with the second set of compensation values to determine a change in the sensor.
22. The method of claim 21, further comprising updating the second set of compensation values by performing the steps of: observing multiple instances of the corresponding change of state of the combined output signals from the two detectors; combining the observations to provide an arithmetic average value that is stored as the compensation value of the second set of compensation values; and repeating the step of comparing the first set of compensation values with the second set of compensation values to determine a change in the sensor.
23. The method of claim 21 or claim 22, wherein a change in the sensor is determined by the first set of compensation values and the second set of compensation values differing by more than a predetermined amount.
24. The method of any of claims 21 to 23, wherein the step of comparing the first set of compensation values with the second set of compensation values to determine a change in the sensor is repeated throughout a lifetime of the rotary encoder.
25. A steering system for a vehicle, comprising a rotary encoder assembly according to any of claims 8 to 20.