Apparatus adapted to provide an indication of an angular position of an input
member over multiple turns
The present invention relates to apparatus adapted to provide an
indication of an angular position of an input member over multiple turns.
Position-indicating devices, such as encoders, are used in many
applications, including detecting the position of mechanically-driven actuators
required for operating fluid valves, for example. An absolute encoder is an
encoder that can identify a position of an input member in an absolute sense,
e.g. as a specific angular position. Such encoders may be depowered and will
still normally be capable of indicating the position in absolute terms when power
is restored, even if the encoder moved whilst the power was off. A multi-turn
absolute encoder typically includes several gears in order to determine the
absolute position over multiple turns. However, if a sensing device of the
encoder happens to fall into an intermediate position, e.g. between two index
positions, then the measurement may be indeterminate. Further, if one of the
sensing elements fails then the entire encoder device will usually fail.
Embodiments of the present invention are intended to address at least
some of the issues discussed above.
According to a first aspect of the present invention there is provided
apparatus adapted to provide an indication of an angular position of an input
member over multiple turns, the apparatus including:
at least two rotatable members configured, in use, to rotate in accordance
with rotation of an input member;
at least one sensing device configured to measure and output an angular
position of at least one of the rotatable members, and
a device configured to use the angular position measurements from the at
least one sensing device to produce an indication of an angular position of the
input member over multiple turns,
wherein the rotatable members are configured to rotate simultaneously
but at different rates.
Each of the rotatable members can have a ratio of revolution with respect
to the other rotatable member(s). Each of the rotatable members may have any
unique (amongst the rotatable members) ratio of revolution and there may not
be any restriction on selection of the ratios, e.g. the ratios need not be selected
so that they satisfy a particular relationship, e.g. to work with a decoding
algorithm that requires the rotatable members to have indexing positions of an
integer base. Alternatively, the selection of the ratios may be made in
accordance with one or more design parameters. For instance, the rotatable
members may be arranged so that there is no common factor (other than one)
amongst the ratios of revolution. The rotatable members may comprise gears
with different numbers of teeth.
In use, the rotatable members normally move along with the input
member in a continuous, non-stepped manner. The sensing devices can
provide an absolute position measurement over 360° of one or more of the
rotatable members, e.g. using optical, magnetic or RF sensing technology. The
apparatus may include A to N said rotatable members, each having a respective
ratio of revolution RA to Rv, and wherein a said sensing device has a maximum
allowable peak error calculated as:
180°
Maximum Sensor Error = ,
In some embodiments, one of the plurality of rotatable members may be a
primary rotatable member that drives all remaining said rotatable members (i.e.
the remaining rotatable members do not drive any other of the rotatable
members).
The plurality of rotatable members may each have a notional
zero/initial/starting (rotation) position. The device may be configured to produce
the position indication by computing how far the rotatable members have passed
from their zero positions. The computing performed by the device can involve
virtually winding-back at least one, and typically all, of the rotatable members to
its respective zero position. The device may be configured to:
virtually wind back a first one of the rotatable members in a sequence to
its zero position; then
for each said rotatable member other than the first rotatable member in
the sequence:
computing a virtual position of the rotatable member based on an
angle through which a previous rotatable member in the sequence has
turned when being virtually wound back, and
virtually winding back the rotatable member and all the previous
rotatable member(s) in the sequence so that they are at their zero
positions.
The computing step may use stored data, such as a look-up table,
representing a position of a said rotatable member when the previous rotatable
member(s) in the sequence is/are all at their zero position(s), and data
representing how many times all of the previous rotatable members in the
sequence have passed through their zero positions together when the rotatable
member is in a corresponding said position.
The rotatable members may be arranged in a co-planar or co-axial
manner.
A set comprising more than one said sensing device may be provided for
measuring the position of a single said rotatable member, such that if one of the
sensing devices in the set develops a fault that another said sensing device in
the set is used instead.
The apparatus may not be provided with constant power. In some
embodiments, the sensing devices measure positions of the rotatable members
resulting from movement whilst power was not provided to the apparatus. A
switch arrangement may be included to enable the sensing devices to be
activated when the rotatable member(s) move.
According to another aspect of the present invention there is provided a
method of providing an indication of an angular position of an input member over
multiple turns, the method including:
measuring an angular position of at least two rotatable members
configured, in use, to rotate in accordance with rotation of an input member, the
rotatable members being configured to rotate simultaneously but at a different
rates, and
producing an indication of an angular position of the input member over
multiple turns using the angular position measurements.
The method may further include:
taking position measurements of a first one of the rotatable members over
a period of time;
taking position measurements of another of the rotatable members over
the period of time;
comparing the measured position of the other rotatable member with an
expected position of the other member, given the measured positions of the first
member and a known relationship between rotations of the members, and
if the measured position of the other rotatable member does not
correspond with the expected position then flagging a possible read error state.
The method may include checking for read errors by detecting lack of
angular rotation of at least one of the rotatable members.
The method may include determining an absolute position of the input
member by using data relating to an angular position of a single one of the
rotatable members, or a combination of positions of a subset of the rotatable
members. The method can include, upon detecting failure of one (or more) of
the rotatable members:
measuring an angular position of non-failed said rotatable member(s),
and
producing an indication of an angular position of the input member over a
reduced range of multiple turns using the angular position measurements of the
non-failed rotatable member(s).
The method may further include providing at least one incremental
counter associated with at least one of the rotatable members, and
using output from the at least one incremental counter to calculate the
position of the input member based on a count of multiple turns of the at least
one associated rotatable member.
According to an alternative aspect of the present invention there is
provided a method of providing an indication of an angular position of an input
member over multiple turns, the method including:
measuring a position of a plurality of rotatable members that, in use,
move in accordance with movement of an input member, each of the rotatable
members having a notional zero/initial/starting (rotation) position, and
computing how far the rotatable members have passed from their zero
positions to produce the input member multi-turn angular position indication.
A computer program product configured to perform at least part of a
method substantially as described herein may also be provided.
The invention extends to any feature, or any combination of features
described herein, whether or not that combination is explicitly described herein.
The invention can be put into effect in numerous ways, one example only
being described and illustrated with reference to the drawings, wherein:
Figure 1 is a plan view of a partially-assembled example embodiment;
Figure 2 is an exploded view of the example embodiment;
Figure 3 is a schematic drawing of sensing elements in an example
embodiment;
Figure 4 is a flowchart showing operation of the embodiment, including a
position calculation step and a gear "virtual wind back" step;
Figure 5 details the position calculation step;
Figure 6 details the gear "virtual wind back" step;
Figures 7A - 7D relate to a worked example of operation of the
embodiment;
Figures 8A and 8B illustrate an example of angular measurement of a
rotatable member of the device, and
Figure 9 is a flowchart showing operation of the apparatus involving an
incremental counter.
Referring to Figures 1 and 2, an example position-indicating device 100 is
shown. The device comprises a housing base plate 102 on which rotatable
members 104A - 104D can be fitted. In the example, there are four rotatable
members, which take the form of toothed gears 104A - 104D. However, it will
be appreciated that other types of rotatable members can be used and they
need not include formations such as teeth for directly driving each other.
A first gear 104A can be considered to comprise a primary gear that
drives the other three gears 104B - 104D, i.e. the three other gears 104B -
104D only engage with the teeth of the first gear 104A, and not each other. The
first gear is driven by an input gear 106, which rotates in accordance with an
input member, which may be a column of a valve actuator, for example (see
Figure 3). Thus, the first gear 104A is directly driven by the input gear 106,
whilst the other three gears 104B - 104D are indirectly driven by the input gear
(via the first gear 104A). It will be appreciated that in other embodiments more
than one, or all, of the gears could be directly driven by the input gear.
The example device 100 is assembled with the gears 104A - 104D in
their correct "zero positions". To assist with this, one or more of the gears
(and/or housing components) can include markings. In the example, the first
gear 104 includes markings in the form of three arrows. These are intended to
be aligned with correspondingly-marked arrows on the other gears 104B -
104D. Additionally or alternatively, "zero location holes" 107A - 107D may be
provided on the gears, which can be aligned with corresponding
markings/recesses 123B - 123D on the base plate 102 (some visible in Figure
2). A jig (not shown) can be used to assist with the assembly. A bar code label
111 can be fitted in the device 100 as well as upper foam padding 116 .
In the example device the gears and housing members comprise
moulded plastic parts. Due to the high tolerance for backlash in the device and
its low precision requirements, such low cost components can be used;
however, it will be understood that other materials can be used and that the
design and dimensions of the device can vary from the illustrated example.
Having a single drive arrangement in the form of gear 104A as in the illustrated
example has the benefit of reducing the effects of backlash; however, it will be
appreciated that other arrangements, e.g. gears in train driving one another, can
be used.
Returning to the example illustrated in Figures 1 and 2, each of the four
gears 104A - 104D is fitted with a respective sensor-compatible component
11OA - 110D, e.g. magnet for a magnetic sensing device. Spacers 108 are
fitted between the gears and the components 110 . An upper housing plate 113
is fitted above the components and this also houses a printed circuit board 114.
The board includes circuitry that functions as a sensing device (shown
schematically at 114A) that can sense the positions of the sensor-compatible
components 11OA - 110D and thereby provide an absolute position
measurement of the four gears over 360°. The sensing arrangement may be
based upon any angular sensing technology, e.g. optical, magnetic or RF. RF
and magnetic sensors can also offer in-built fault detection: if a magnet gets
detached from a gear or demagnetised, the sensor can detect this. It is also
possible to use sensor devices that provide an analogue output.
It will be understood that in alternative embodiments the rotatable
members may include integrated sensor-compatible components or their angular
positions may be determined by the sensor by in another way, such as by
visually identifying the angular position, e.g. by detecting a marking on a surface
of the rotatable member. It will be appreciated that in alternative embodiments
the sensing device and/or processor may be located remotely from other
components of the device with signals being transferred by means of
wireless/RF signals, for example. In other embodiments, sensing devices may
be built into the rotatable members/gears. Other variations to the example
device 100 shown in the Figures can be produced. For instance, the rotatable
members may be arranged in a co-axial rather than co-planar manner (or a
combination of co-axial/co-planar, or any other arrangement), which can result in
size reduction/design benefits.
The circuit 114 further includes a processor (shown schematically at
114B) configured to produce a position indication of the input member 106 over
multiple turns using the sensing device measurements, as will be described
below. Although a digital electronic processor is shown in the example, it will be
appreciated that the functionality it provides could be performed by suitable
analogue components/circuitry.
A mechanical switch can be included to enable the electronic sensing to
be activated after the sensing elements/gears have started to move. The switch
may be activated mechanically or magnetically and there is therefore no need to
supply constant power to the device, thus potentially reducing the overall power
consumption of the device. After power-up, the device can detect any
movement that occurred whilst the power was off and use that measurement to
provide a position indication. The example device includes a magnetic "wake
up" switch in the form of a gear 113 that rotates when the gear train rotates and
can be used to trigger the power supply to the device, but it will be appreciated
that alternative arrangements can be provided.
The rotatable members in the device are configured to rotate at different
rates. In the example device 100 this is achieved by having a different number
of teeth on all of the gears 104A - 104D. However, it will be appreciated that it
can be achieved by different means. For instance, providing inter-engaging
rotatable members that have different dimensions (e.g. radius or circumference)
can result in the members having different ratios of revolution. Further, rotatable
members that don't directly engage with/drive each other, e.g. disks, could be
used, the members being driven directly by the input member or connected
together by means of a belt or chain drive, or any other gearing mechanism.
The range of operation of the device will be determined by calculating the
maximum number of turns of the first analysed, or "primary", member before a
repeating pattern of the rotating members occurs, at which point the absolute
position can no longer be determined from the individual positions of the
members. This can be calculated as the number of turns required to rotate the
device from its reference position until it returns to this position. To improve the
range, minimum functions of the ratios of revolution can be increased.
For example, in a simple case where the device includes two rotatable
members A and B, if the ratios of revolution of the two members were 10 and 20,
respectively, then these ratios have a common factor of 10, and so can be
simplified to a 1:2 ratio. Such an assembly can only be used to give a position
indication over two turns. However, if gear B was selected so as to have 2 1
teeth instead of 20 then there would be no such common factor, and the
simplest expression would remain as 10:21 . Given this relationship, if A was to
be rotated twice then member B would move to a position 342.9° (720° x 10/21 )
from its zero position. As both members are not at their reference positions and
the device has not returned to its overall starting position the range of this device
can be calculated as the number of times member A must rotate for member B
to be zero at the same time, i.e. 2 1 turns. When member B has been rotated by
exactly 10 turns, both A and B will be zero and the device will be at its overall
starting position.
In a conventional multi-turn absolute encoder arrangement there is a
mechanical indexing mechanism between the members that increments each
one after a defined rotation of the previous members, i.e. if a member has been
indexed a certain number of times then the previous members will have rotated
by a known amount. However, no such indexing mechanism is present in the
continually moving (along with the input member), non-stepping positionproviding
device described herein; rather, embodiments of the device can have
a virtual indexing mechanism, which may take the form of the decoding
algorithm described below. Reference (or starting/initial/zero) positions are
defined for each rotatable member and these can lie anywhere in the 360° of
possible rotation, provided each member moves in such a way that all of the
members can be at their reference positions at the same time. This will usually
define the device's overall zero position.
Such a "virtual indexing mechanism" has the effect of taking almost
arbitrary looking position measurements and allowing them to be decoded to
provide an indication of an actual position. In conventional indexing
mechanisms, each member is measured in sequence, with each contributing a
proportion of the position depending on the mechanism. This information is
immediately available as the members are read; however, in embodiments of
the present device there may not be such an obvious relationship. In order to
decode the position, each member is sensed in sequence and then decoded.
The primary member must be used in order to find a "wound-back" position of
the next rotatable member. This calculation step provides information that is
directly relevant to the absolute position of the input member. To obtain the
absolute position the angular positions of the rotatable members are measured
and then a calculation is performed to find the rotation contribution so that the
next calculation can be performed, and so on.
Referring to Figure 3, examples of the number of teeth on the four gears
104A - 104D are given as follows: gear 104A: 22 teeth; gear 104B: 26 teeth;
gear 104C: 34 teeth, and gear 104D: 36 teeth, and so the minimum functions of
the ratios of revolution of the gears 104A - 104D are 11, 13, 17 and 18,
respectively. The example ratios mean that gear 104A will rotate 13/1 1 times
the speed of gear 104B, and gear 104D will rotate 17/1 8 times the speed of gear
104C, and so on. These ratios have no integer common factors; however, if the
ratios were 11, 13, 15, and 18 for instance, instead then there would be a
common factor of 3 (for 15 and 18), which would have an effect on the way the
gears move together and how far the input can rotate before a repeating position
sequence appears, thus the absolute range would be reduced.
When the first gear 104A is used as the primary measurement gear, the
maximum number of turns of that gear, from which the absolute position of the
mechanism can be derived, is equal to the product of the minimum functions of
the ratios of revolution of the other gears 104B - 104D. Thus, in general,
combinations of n gears (denoted A, B, C, n ) can be used to determine
absolute gear position over a range of X turns. If the first gear 104A is used as
the primary measurement gear then the absolute position range of the device
can be determined by the equation:
X = B x C x D .. . x n
where B, C, D, .. . n , are the minimum functions of the ratios of revolution
between each gear and there is no common multiple between the minimum
functions of the ratios of revolution of all sensed elements. Thus, for the
example of Figure 3, X = 3978 (13x1 7x1 8). In another example, the number of
teeth on the four gears are: gear 104A: 7 teeth; gear 104B: 11 teeth; gear 104C:
13 teeth, and gear 104D: 15 teeth, in which case X = 2145.
In order to find the total number of rotations of the input member 106,
each of the gears 104A - 104D are in effect "wound back" to their zero positions.
Whilst doing this, the total number of rotations is recorded so that it can be
related to the position of the input. In the example device the rotation of the first
gear 104A is first recorded. Thus, by recording how far all the gears in the
device have rotated it is possible to calculate the current multi-turn position of
the input member. It will be appreciated that the gears are not normally
physically wound back; for instance, a software/firmware simulation or other
process intended to calculate the number of movements/rotations can be used,
or any other electromechanical or electronic implementation of this function.
Referring to Figure 4, a flowchart of an example method of the calculation
of the total number of rotations is shown. The notation below shall be used in
the following description, with gears 104A - 104D being referred to as "A" - "D",
respectively:
A : Gear A Position
B : Gear B Position
pc : Gear C Position
D : Gear D Position
PR : Primary Ratio ( 1 :2 in example of Figure 3, i.e. two rotations of input column
= 1 rotation of primary gear A)
RA : Gear A Teeth Ratio ( 1 1 in the example, i.e. the number of gear teeth with
any common factor removed)
RB : Gear B Teeth Ratio ( 13 in the example)
Rc : Gear C Teeth Ratio ( 17 in the example)
R D : Gear D Teeth Ratio (18 in the example)
OTOTAL : Total Gear A Turns
: Gear A Turns from A Wind-Back
: Gear A Turns from B Wind-Back
ec : Gear A Turns from C Wind-Back
D : Gear A Turns from D Wind-Back
X B : Number of Zero Cycles from B
X C : Number of Zero Cycles from C
X D : Number of Zero Cycles from D
At step 402 the primary gear 104A is virtually "wound back" to its zero
position. The number of rotations required to achieve this corresponds to its
current position, , and so = -
At step 404 analysis relating to the next gear in the device commences.
The gears are considered in order, starting with the one with the fewest number
of teeth (after gear A) and working up in order to the gear with the greatest
number of teeth, although any ordering could be used. Thus, in the first iteration
of these steps, gear B will be analysed. In the formulae shown in the flow charts
and referred to below, the letter N is used to refer to refer to the "current" gear
being considered. At step 406 the position of gear B is calculated. Figure 5
illustrates the steps involved in this calculation. At step 502 the previous windback
distance and gear ratio is used to calculate the new position, , of gear B,
using the general formula shown in Figure 5, which applied to gear B becomes:
will be used with a look up table, as discussed below, to determine
how many zero sequence crossings have ocurred.
At step 504 is converted so that it falls between zero and the
maximum position. This conversion is a wrapping operation such that the
position is output between 0.0 and 360.0°. For example, if a measurement is
345.0° and the ratios were RA = 23 and RB = 2 1, then would be 345x23/21 =
377.86° and so could be wrapped by subtracting 360° to 17.86°. 0a may be
greater than 360°
Returning to Figure 4, at step 408 gear B is "wound back" to its zero
position. Figure 6 details steps involved in this operation. At step 602 the
rotation contribution index of the gear (i.e. how many times all of the previous
gears in the sequence have passed through their zero positions together) XN is
found using a look-up table (which will be discussed below). At step 604 the
movement of the primary gear (A) due to the winding back of the current gear to
its zero position is calculated using the general formula shown in Figure 6 .
Returning to Figure 4, at step 4 10 a check is performed as to whether all
the gears in the device have been analysed. If not then control passes back to
step 404, where the above steps 406 - 408 are performed for gear C (and then
gear D). If all the gears have been analysed then control passes from step 4 10
to step 4 12, where the total rotation of the input column is computed. This can
be achieved by calculating the absolute position of the input member using the
total primary gear (A) rotation and the primary ratio:
Position = Primary Ratio x (+ . . . + )
A worked example shall now be given, making reference to Figures 7A -
7D. Figure 7A shows the gears A - D in a given state 702A when measurement
of the absolute position is to be made. As discussed above, the first step (step
402 of Figure 4) is to "wind back" the primary gear A to zero, as shown
schematically/conceptually at 702B (with the circle on the gear disk indicating its
current positions and all the zero positions being at 0°). This can be achieved by
taking the current position of gear A and winding it back to zero, whilst recording
the value of rotation. For example, if gear A is at position 57.43°, winding it back
is equivalent to turning it through -57.43°. Knowing that gear A has been
conceptually rotated through this angle, it is possible to work out how far the
other gears would have turned based on the gear ratios.
Next, when gear A is back at its conceptual zero position, the positions of
the next gear (gear B) is calculated based on the gear ratios and the amount
that gear A was conceptually turned. Gear B is then conceptually wound back
so that both gears A and B are at their zero positions, as shown schematically at
704 in Figure 7B. This may involve several rotations of gear B because when
one gear in the system is moved, all of the gears rotate. This process continues
until the whole gear train is conceptually back at its zero position.
The new position of gear B is calculated (step 406) since gear A has been
rolled back. This can be done by taking the total number of rotations required to
roll back gear A to its zero position, and applying the gearing ratios to find the
change in gear B. Since this value represents rotational position, a passthrough-
zero or a full turn must be handled:
§B(New) ~ B Old) ,
R.
This is the number of times gear A must be rotated again to reach its zero
position at the same time as gear B reaches its zero position (step 408). This
value is recorded and added to the rotations from winding back gear A.
The procedure for rolling each subsequent gear is identical to that of
rolling back gear B: each is based on the position after winding back the
previous gear and the ratios between the various gears. The position of gear C
is first calculated from the total wind back rotations of the gear A thus far, and
from this the lookup table is used to find the number of times the previous gears
A and B have moved through zero at the same time. This number is then used
to find how much the first gear A must have rotated to have gears A, B and C at
their zero positions together (shown schematically at 706 in Figure 7C). This
movement value is the number of times gears A and B have been at zero
positions together, multiplied by the relative ratios of movement of gear B ( 13 in
the example). These numbers result from the need to have gears A and B at
their zero positions together; thus, gear A must move in multiples of 13 to keep
gear B at its zero position. Once this process is complete gears A, B and C
should all be at zero and gear D is then analysed:
The process for analysing gear D is identical to that of gear C, with the
exception that the first gear must also move in multiples of 17 to incorporate the
fact that gear C must remain at zero:
Once all the gears A - D are virtually wound back to their zero positions
(shown schematically at 708 in Figure 7D), the final step is to look back on the
total rotation of the first gear from each wound-back gear and combine them and
then relate this value to the input column using the primary ratio, i.e:
TOTAL = A + B + C +
Absolute Position =P.R.xQTOTAL
As mentioned above, look-up tables can be used to find the number of
times the previous gears in the device have passed their zero positions in order
to compute the current position for a particular gear. The "wound-back" position
of the gear being considered can then be compared to these stored positions,
and the corresponding position used to give the number of zero sequences of
the previous gears. It will be appreciated that this is optional and the zero
position passes could be determined by another mathematical method, or data
stored in another manner. Example look-up tables for the example device of
Figure 3 are given below:
n
on
0.923077 7 0.705882 9 0.666667 12
0.764706 14 0.722222 13
0.823529 2 0.777778 14
0.882353 7 0.833333 15
0.941 76 12 0.888889 16
0.944444 17
The left-hand column in the table for each gear B - D shows the position
of that gear when the previous gear(s) in the sequence/order (A, B and C) are all
at their zero position(s). The right-hand column in that gear's table shows how
many times all of the previous gears have passed through their zero positions
together, when the gear is in the corresponding position indicated in the lefthand
column. Using this table it is therefore possible to find the number of zero
positions that previous gears have passed through by comparing the calculated
gear position to the theoretical position, and matching the closest values. It is
important to note that if a gear is closer to a full rotation, i.e. 1, than the next
largest value, then this is treated as a full rotation, and thus zero position.
The look-up tables can be generated by moving the gears through their
zero positions and recording the position of the gear for each cycle until it
repeats. This could be done, for example, by performing a software simulation.
In detail, calculating the Gear B section of the lookup table is a matter of moving
through the zero cycles of gear A, as this is the only gear that precedes gear B.
Gear B will have 13 different positions for each zero position of A before it
repeats; after 13 full rotations of gear A, both gears A and B will be at their zero
positions together again. An example calculation is as follows:
Gear A Zero Cycles = 2,
» = 2—
13
= 1.6923 5 = 0.6923
In general:
= Zeros x-^
Calculating the table for gear C is similar to gear B; however, there are
now 17 total positions to calculate before the sequence loops. The zero cycle
must also include gear B. The general equation is:
< >c =Zeros x
c
Gear D follows the logical sequence and now has 18 unique positions,
and the zero cycles include C. The equation for gear D is:
R . R R
= Zeros BC x -
It will be understood that the rotatable members can be analysed in any
sequence (e.g. not necessarily starting with gear 104A). The analysis can be
performed in any order so long as the look-up tables have been calculated for
the appropriate sequence.
When the measurement error of the measurement of individual gear
positions becomes too great, the algorithm may return an incorrect value.
Before this point, the error in the device will be the measurement error of the first
measurement gear, multiplied by the primary ratio.
The maximum allowable peak error of the sensing device(s) before the
algorithm fails can be calculated as:
180°
Maximum Error = ,
Thus, for the gear ratios adopted in the example, high accuracy sensors
are not required. The sensors can be relatively coarse as they only need to
meet the 1807(RA+RN) maximum error requirement. The resolution of the
position is then only dependent on the chosen first gear. This can allow the use
of 'n-1 ' low-accuracy sensors and one very high accuracy sensor to provide a
high turn-count, multi-turn position-providing device with a very high accuracy.
The maximum error budget of the device can be found as the point at
which the rotatable member with the highest ratio, and thus the smallest
rotational distance between each zero sequence calculation position, can no
longer be known with confidence. This will occur when the total error on the
member is greater than the distance between two previous members' zero
sequence positions divided by two because at this point the calculated "windback"
position will be closer to the wrong look-up table position.
As a visualisation of this problem, each of the rotatable members (with
the exception of the primary member) can be conceptually split into segments.
Each of these segments represents the range of positions for which the "woundback"
position can be considered to relate to the corresponding number of
previous member zero sequences, as listed in the look-up tables. An example
rotatable member with 2 1 segments is shown in Figure 8A, with appropriate
segments labelled with the corresponding rotation contribution index. The
rotatable member is shown at its reference position, in the exact centre of the 0
segment. Theoretically, the member should always be measured in a position
such that the "calculate position" stage of the calculation with previous members
wound back will return a value in the exact centre of one of the segments, i.e.
corresponds to a position from the look-up table. However, when the error
causes this calculated position to drift into a neighbouring segment there will be
a significant increase in output error as the number of previous member zero
position sequences will change, potentially dramatically. This angular error is
illustrated in Figure 8B.
Considering the two major forms of error, measurement error and
mechanical error (backlash), the maximum allowable error can be found by
considering the point at which the returned value of zero sequences from the
analysis of the highest ratio member is incorrect. Backlash will have a direct
effect on the position relative to the primary member; however, the
measurement error affects the result in two ways. The first is by simply
corrupting the value measured on the member, and the second is by corrupting
the value of the primary member, which is then translated onto the analysed
member during the first stage of calculation, i.e. when the primary member is
"wound back". If this is incorrect then when the last known previous member
zero sequence position of the analysed member is calculated, there will be
further discrepancy between the calculation from the true value.
In mathematical terms the allowable error can be expressed as follows:
360° R
— > EN + E x - + N
where EN is the measurement error of the member under analysis, N, EA
is the measurement error of the primary member, and BAN is the backlash
between the primary member and N.
N + + AN can be thought of as the total error in the device, with
° as the error limit. If the total error is not less than the error limit for all
2xR N
members, then the decoding will not work.
Another way in which an error can be introduced in the device is time
delay between measuring each rotatable member. This will introduce another
term to the equation as follows:
360° R
> EN + E x - + B N + {TdN +Td )
XRN RN R
where TdN is the time delay to measuring n, TdA is the time delay to
measuring the primary member, vN is the rotational speed of N, and vA is the
time delay to measuring N. This type of error can be eliminated by using a
symmetrical reading algorithm, whereby the measurements of each member are
arranged around a single point in time. This removes the effects that cause the
decoding algorithm to fall over as each measurement is effectively at the same
point in time (assuming constant movement). This method can, however, still be
susceptible to error from acceleration, but the acceleration that would be
required to affect it would be extremely large as, in practice, the time delays are
very small.
In certain implementations of the device there is a possibility that all of the
rotatable members may not be capable of being positioned at the reference
position simultaneously. This can happen, for example, if there is no fixed way
of ensuring that the members are assembled in correct relative locations. There
are ways of trying to ensure that does not occur, such as using toothed gears
with the number of teeth equal to the minimum function of revolution (e.g. in a
three rotatable member example with ratios 10, 2 1 and 17) if these were toothed
gears with 10, 2 1 and 17 teeth, and properly aligned sensors for the driving
arrangement. This arrangement would ensure that the members could only ever
be placed in whole segment intervals, whereas if there were 20, 42, and 34 teeth
then it would be possible to place a gear out by half a segment. One possible
solution to this problem is to have an assembly process intended to prevent
incorrect assembly, for example a fixed mechanical structure can be used as a
guide that will not allow for improper assembly, or guides/formations, such as
105 or 107 mentioned above, can be used. An alternative solution is that the
first time the device is powered up, a measurement is taken and the
arrangement of the rotatable members can be calculated from this position using
the "wind-back" stage of the calculations, and the known assembled positions of
the rotatable members.
To deal with the risk of one or more sensor failing, one or more redundant
sensors can be provided, associated with one or more of the rotatable members
so that if a particular sensor fails then a redundant sensor can be used instead.
This can be achieved, for example, with RF and magnetic sensors as the
magnetic and RF fields permeate over a large area. The redundant sensor may
be positioned on an opposite side of the rotational axis of the rotatable member,
or may be positioned elsewhere, e.g. between the primary sensor and the
member, or behind the primary sensor. Should this technique not be feasible for
a particular embodiment then an alternative is to discard the faulty sensor's
readings and operate the device over a reduced absolute range. Using the
example of faults being detected on sensor 104B of Figures 1 and 2, the device
will use the readings based on the remaining three rotatable members 104A,
104C and 104D. This will mean that its range will reduce; however, this can be
sufficient for some applications. In this mode a new lookup table may need to be
generated/used, based on the relationships between the remaining rotatable
members.
In this reduced range mode it may be necessary to extend the range of
the device over more cycles than may normally be possible. In order to do this
an incremental counter (e.g. shown schematically at 114C in Figure 2) can be
used to keep track of the number of times that the system has passed through
its total range. In this failure mode, the apparatus can lose its ability to
confidently provide an accurate position reading when power is restored if the
device has moved a significant amount (e.g. more than half of the reduced
range) with the power off.
An example of the operation of the device in this reduced range mode is
shown in Figure 9 . In this mode if the incremental count is lost then the position
is not absolutely defined by the device. At step 902 the absolute measurement
of one of the rotatable members still in use is obtained. At step 904 the true
position of the member is calculated by adding the number of turns of the
member, as recorded by the incremental counter, to the absolute measurement
value. At step 906 a check is performed to see if the difference between the
value currently measured and the value measured at the previous iteration is
greater than the range of the device divided by two. If this is not the case then
control passes back to step 902; otherwise at step 908 the counter is
incremented and then control passes to step 902 again.
As all of the gears turn at the same time, but at unique rates, a reduced
range of absolute positions can be determined by looking at any combination of
gears. This provides built-in redundancy and/or the ability to check for read
errors. The ability to use incremental counters as described above can offer
increased measurement speed as the position of only one member is being
measured at any one time.
A disadvantage with conventional mechanical indexing systems is that if
there is a fault in one of the members and it does not update as it should then
this is almost impossible to detect. With a continually moving arrangement, as is
the case in embodiments of the present apparatus, it is relatively simple to
detect faults. As all of the movement ratios are known, it is possible over the
course of several samples to detect faults in parts of the device. After two
samples of each rotatable member it is possible to tell how far each member has
moved, provided that these samples are not taken after several rotations. From
this it is possible to derive how far each member should have moved as all of the
ratios of movement are defined.
Multiple subsets of gears can be used for self-checking over a reduced
range. In this configuration, the subsequently-sensed members/gears can be
used to verify the accuracy, and check for read errors in the chosen sensed
member. Further, in this configuration, the number of turns that the device can
record is unlimited. The ability to use a single sensed element with an
incremental counter offers the advantage of increased measurement speed as
only one gear is being read at any one time. The device may also be configured
to provide the option of taking measurements using all of the rotatable members
to confirm the input member's position.
The device may be configured to automatically switch into one of the
"redundancy modes" described above, or may allow a user to set the mode.
By increasing the resolution of a chosen first members (e.g. 104A in the
example, although the chosen sensing element doesn't have to be the first gear
in the train) the resolution of the device 100 can be increased. This can enable
the input ratio to the sensor to be manipulated in order to obtain the required
balance between the final number of turns required to be measured, and the
necessary accuracy that is desired. The range of the device can be increased
by changing the primary ratio leading to the device assembly at the cost of
reducing the resolution.
The device can continue computing the position beyond its absolute
range; it will wrap at that point, but in some applications that can be tolerated.
CLAIMS
1. Apparatus ( 100) adapted to provide an indication of an angular position of
an input member ( 06) over multiple turns, the apparatus including:
at least two rotatable members ( 104A - 104D) configured, in use, to
rotate in accordance with rotation of an input member ( 106);
at least one sensing device ( 1 14A) configured to measure and output an
angular position of at least one of the rotatable members, and
a device ( 1 14) configured to use the angular position measurements from
the at least one sensing device to produce an indication of an angular position of
the input member over multiple turns,
wherein the rotatable members are configured to rotate simultaneously
but at different rates.
2 . Apparatus according to claim 1, wherein each of the rotatable members
( 104A - 104D) has any unique ratio of revolution with respect to the other
rotatable members.
3 . Apparatus according to claim 1, wherein each of the rotatable members
( 104A - 104D) has a ratio of revolution with respect to the other rotatable
member(s), the rotatable members being arranged so that there is no common
factor (other than one) amongst the ratios of revolution.
4 . Apparatus according to claim 2 or 3, wherein the rotatable members
( 104A - 104D) comprise gears each having a different/unique number of teeth.
5 . Apparatus according to any one of the preceding claims, where, in use,
the rotatable members ( 104A - 104D) move along with the input member ( 106)
in a continuous, non-stepped manner.
6 . Apparatus according to any one of the preceding claims, wherein the at
least one sensing device ( 1 14A) provides an absolute position measurement
over 360° of one of the rotatable members ( 104A - 104D).
7 . Apparatus according to claim 6, wherein the at least one sensing device
( 1 14A) uses optical, magnetic or RF sensing technology.
8 . Apparatus according to claim 6, wherein the apparatus includes A to N
said rotatable members ( 104A - 104D), each having a respective ratio of
revolution RA to R v, and wherein a said sensing device ( 1 14A) has an accuracy
such that a maximum allowable peak error in the sensing device is satisfied by:
180°
Maximum Error = ,
9 . Apparatus according to any one of the preceding claims, wherein one of
the plurality of rotatable members ( 104A) directly or indirectly drives all
remaining said rotatable members ( 104B - 104D).
10 . Apparatus according to any one of the preceding claims, wherein the
plurality of rotatable members ( 104A - 104D) each has a notional zero position
and the device ( 1 14) is configured to produce the position indication by
computing how far the rotatable members have passed through from their zero
positions.
11. Apparatus according to claim 10, wherein the computing performed by
the device ( 1 14) involves virtually winding-back the rotatable members ( 104) to
their respective zero positions.
12 . Apparatus according to claim 11, wherein the device ( 1 14) is configured
to:
virtually wind back (402) a first one of the rotatable members ( 104A) in a
sequence to its zero position; then
for each said rotatable member ( 104B - 104D) other than the first
rotatable member in the sequence:
computing (406) a virtual position of the rotatable member ( 104B -
104D) based on an angle through which a previous rotatable member
( 104A) in the sequence has turned when being virtually wound back, and
virtually winding back (408) the rotatable member ( 104B - 104D)
and all the previous rotatable member(s) ( 104A) in the sequence so that
they are at their zero positions.
13 . Apparatus according to claim 12, wherein the computing (406) uses
stored data, such as a look-up table, representing a position of a said rotatable
member ( 104B - 104D) when the previous rotatable member(s) ( 104A) in the
sequence is/are all at their zero position(s), and data representing how many
times all of the previous rotatable members in the sequence have passed
through their zero positions together when the rotatable member is in a
corresponding said position.
14. Apparatus according to any one of the preceding claims, wherein the
rotatable members ( 104A - 104D) are arranged in a co-planar manner.
15 . Apparatus according to any one of claims 1 to 13, wherein the rotatable
members ( 104A - 104D) are arranged in a co-axial manner.
16 . Apparatus according to any one of the preceding claims, wherein the
apparatus is not provided with constant power.
17 . Apparatus according to claim 16, wherein the at least one sensing device
( 1 14A) measures positions of the rotatable members ( 104) resulting from
movement that took place whilst power was not provided to the apparatus.
18 . Apparatus according to claim 16 or 17, including a switch arrangement
( 1 13) for enabling the at least one sensing device ( 1 14A) to be activated when
the rotatable member(s) ( 104) move.
19 . Apparatus according to claim 1, wherein a set comprising more than one
said sensing device ( 1 14A) is provided for measuring the position of a single
said rotatable member (104), such that if one of the sensing devices in the set
develops a fault then another said sensing device in the set is used instead.
20. A method of providing an indication of an angular position of an input
member (106) over multiple turns, the method including:
measuring an angular position of at least two rotatable members ( 104A -
104D) configured, in use, to rotate in accordance with rotation of an input
member ( 106), the rotatable members being configured to rotate simultaneously
but at a different rates, and
producing an indication of an angular position of the input member over
multiple turns using the angular position measurements.
2 1. A method according to claim 20, further including:
taking position measurements of a first one (104A) of the rotatable
members over a period of time;
taking position measurements of another ( 104B) of the rotatable
members over the period of time;
comparing the measured position of the other rotatable member with an
expected position of the other member, given the measured positions of the first
member and a known relationship between rotations of the members, and
if the measured position of the other rotatable member does not
correspond with the expected position then flagging a possible read error state.
22. A method according to claim 20 or 2 1, further including checking for read
errors by detecting lack of angular rotation of at least one of the rotatable
members ( 104).
23. A method according to claim 20, further including providing at least one
incremental counter ( 114C) associated with at least one of the rotatable
members ( 104), and
using output from the at least one incremental counter to calculate the
position of the input member based on a count of multiple turns of the at least
one associated rotatable member.
24. A method according to any one of claims 20 to 23, wherein the method
includes, upon detecting failure of one (or more) of the rotatable members
( 104B):
measuring an angular position of non-failed said rotatable member(s)
( 104A, 104C, 104D), and
producing an indication of an angular position of the input member ( 106)
over a reduced range of multiple turns using the angular position measurements
of the non-failed rotatable member(s).