GB2581200A - Commutation timing derived from partial back-EMF measurements - Google Patents

Abstract

Commutation timing for a permanent magnet motor is determined by defining, for each of one or more monitored phase voltage, discrete time intervals within an electrical cycle where the phase voltage is expected to comprise only an induced back emf (electromotive force). The phases voltages are monitored only during the defined time interval whilst disregarding other time intervals and used to obtain commutation timing. Using the monitored phase windings may comprise estimating the back emf signal using phase voltage samples associated with time periods for which there is no distortion, and discounting phase voltages associated with time periods where there is distortion. Back emf for periods where there is distortion may be estimated by interpolation or extrapolation from phase voltage measurements taken when there is no distortion, may be based on phase voltage measurements taken for a second set of phase windings, or may be based on previous back emf estimates. An open circuit measurement operation may be carried out where phase voltages are sampled over a period when no current is being driven and used to estimate a back emf for the period.

Description

COMMUTATION TIMING DERIVED FROM PARTIAL BACK-EMF MEASUREMENTS

Technical Field

The present invention relates to a method and apparatus for determining commutation timing for an electric induction motor and in particular to such a method and apparatus that derives commutation timing from partial back-EMF measurements.

Background

One of the tasks that must be performed by the controller of a brushless electric permanent magnet motor (an "induction motor") is to apply current at any time to the motor's phase windings in the manner best calculated to produce torque in the motor. Whatever the desired torque output of the motor and the amplitude and waveform of the current being applied to the phase windings, there is, in general at any time, one "commutation state" (one direction of current flow though the various windings) that allows the electromagnetism of the motor's stator to best interact with the magnetism or electromagnetism of the motor's shaft to produce a torque in the desired direction of rotation with a higher efficiency than any other commutation state. The above statement is true except at a few finite positions where one or more commutation states are equivalent. These are the "crossover points" where ideally the controller should implement a "commutation event" by changing the commutation state.

An example of a motor with sensorless controller is illustrated in Figure 1. A "sensorless" controller is a controller for a brushless motor that achieves commutation with no sensors except the following which are internal to itself: sensors for phase-tophase or phase-to-neutral voltage On general, "phase voltage") and (in some instances) sensors for phase current. In the example illustrated in Figure 1, there are three phase windings 106 in the motor 102 connected to the controller 101 by phase terminals 103, and there is one connection to the controller 104 that is linked to the motor's internal "neutral point" 105 connection between the three phases. The three phase terminals 103 are designated A, B, C. The controller is capable of detecting the potentials between the phase terminals 103, as in the voltage potential from A to B, the voltage potential from C to A, for example, and the voltage potentials between each phase terminal 103 and the neutral connection 104.

These voltage readings described above, collectively called "phase voltages" (and, if available, the readings of the current passing through each phase winding 106, collectively called "phase currents") give an imperfect view of the rotational position of the motor. In an ideal case, a motor with nothing connected to its phase windings ("open circuit' operation), rotating because of some external motive force (for example, a motor being driven mechanically as a generator but not connected electrically to any load) will give phase-to-phase or phase-to-neutral voltage readings that are free of distortion and give a very good indication of which phase windings represent the strongest magnetic interaction between the rotor and the stator at any particular time.

In general, the phase-to-phase voltage that is measured to be the largest is identically the optimal path for current to be applied within the motor to produce a torque, because this particular permutation of phase connections is interacting strongly with the magnetism in the motor. For example, if, during open-circuit rotation of the motor illustrated in Figure 1, the voltage potential from phase terminal 103C to 103B was measured to be the largest positive voltage potential between any two phase terminals 103, then it would mean that the controller's optimal course of action, to produce torque within motor in an efficient manner, would be to send a current into the phase terminal connection 103C and draw the same current out of the phase terminal connection 103B. The controller would wish to maintain this status until the condition of the voltage potential from 103C to 103B being the largest positive observed voltage potential, when operating the motor "open circuit", was no longer true. At that point, the another observed voltage potential would be higher and the controller's optimal course of action would be to provide current through a different path within the motor. This would be called a "commutation event". The controller's problem, in this example, is that the motor is not operating "open circuit" but is being driven by current supplied by the controller. Therefore, the controller has imperfect knowledge of when a commutation event" should take place.

The phase voltages measured when a rotating motor is operating "open circuit" are the motor's "back-EMP voltages. Figure 2 shows and example of the idealised back-EMF voltage signals from a three-phase motor with phases named 'A', 'B', and 'C'. Voltage potential is plotted on the y-axis 201 versus angle of rotation of the motor on the x-axis 202. Positive values of the voltage from phase terminal A to B are shown as 203, from B to C as 204, from C to A as 205, from B to A as 206, from C to B as 207, and from A to C as 208. If the actual phase voltage readings from a motor in service were the idealised back-EMF signals pictured in Figure 2, then a sensorless motor's optimal commutation pattern could be determined simply by comparing back-EMF voltages and affecting a commutation event whenever a different permutation of phase windings came to exhibit the largest phase-to-phase voltage, in other words, at the angles of rotation of the motor labelled 209 in Figure 2. In practice, back-EMF voltage cannot be measured in isolation from other phenomena when the motor is operating, and the phase voltage signals of a motor operating normally can be considered as the backEMF signals of the motor with distortions. Furthermore, as motor designers strive to make motors more efficient, more sparing of material and lower in impedance, the result is a trend away from clear and discernible back-EMF signals and towards a situation where confounding signals represent an ever-more-significant component of any voltages measured at the motor's phase windings. In general, it can be said that efforts to improve the efficacy of a motor as a means of producing torque from applied current reduce its ability to exhibit the undisturbed, open-circuit phenomenon of back-EMF, not because back-EMF is absent within the motor but because we lack the means to measure and interpret it amid confounding signals.

In some conditions, a sensorless controller might apply an arbitrary commutation pattern and completely ignore any voltage or current sensors to inform the commutation. For example, when starting a motor, a sensorless controller might simply apply current to an arbitrary choice of phase windings and then switch the current with a pattern and a timing that correspond to the motor spinning at five revolutions per minute in a chosen direction. A motor produces no back-EMF when it is at rest, so this type of approach is necessary when starting a motor. However, this means of determining commutation is non-optimal and leads to a mis-match between the motor's torque and speed. Depending on temperature, lubrication and other conditions, the current applied to the motor by the controller will produce a certain net output torque, which will cause the motor to accelerate against any applied mechanical load according to the inertia of the system. If at any time in this example, five revolutions per minute is very different from the speed at which the motor would naturally rotate under the prevailing conditions, then the motor will either "slip" (fail to keep up with the commutation pattern and therefore spin much more slowly) or "jog" (accelerate and then decelerate, waiting for the next commutation event so that the motor can continue its rotation). At the very least, even if not noticeably slipping or jogging, the motor might be subjected to counterproductive magnetic fields internally. At any given time, the magnetic fields arising in the stator as a result of the commutation might generate less torque then could be generated with a different commutation, or the magnetic fields may actually resist the rotation of the shaft in the desired direction at various angular positions throughout the rotation of the rotor. A sub-optimal commutation pattern, such as an arbitrary pattern, is therefore detrimental to the motor's efficiency and performance, and it will generally destroy the motor if attempted at a significant proportlon of full rated power or for prolonged periods.

In general, we desire that a sensorless controller be effective in detecting back-EMF and determining the optimal commutation state of the motor at any time. Various methods of filtering can produce a better estimate of back-EMF than the raw phase voltage reading, and those can be combined with this invention. However, this invention proposes a different approach. In this invention, parts of the phase voltage signal are treated as more important than other parts in estimating back-EMF. The parts of the phase voltage signal that are deemed important correspond to times during the operation of the motor when the phase voltage is less distorted by the action of the controller.

Figure 3 shows an example of one measured phase voltage corresponding to the voltage potential from phase terminal wire A to phase terminal wire B. Voltage potential is plotted on the y-axis 301 versus angle of rotation of the motor on the x-axis 302. In the vicinity of 303, the controller is driving current through the phase AB circuit (into the phase terminal wire A and out of the phase terminal wire B), and the phase AB voltage reading (the voltage from the phase terminal wire A to the phase terminal wire B) is similar to (or biased towards) the voltage of the controller's power supply, since the power supply is effectively connected to wires A and B through the controller. Efforts to make the motor more efficient may have led to a reduction in the motor's impedance, which makes the measured phase AB voltage in this condition more similar to the voltage of the controller's power supply than it would have been had the impedance of the motor been greater. In the vicinity of 304, the controller is driving current into the phase BA connection (into the phase terminal wire B and out of the phase terminal wire A). In this situation, the measured phase AB voltage is similar to the reversed polarity of the controller's power supply voltage. Furthermore, in the vicinity of 305 and 306, the controller has recently disengaged a circuit that was passing current through the phase AB connection in the forward and reverse directions, respectively. The circuit having been disconnected only recently, a current continues to flow through the phase A and phase B windings due to the inductance of the motor's windings. Diodes within the controller allow this current to flow and prevent damage to the controller. During this time, the phase A wire and the phase B wire are effectively short-circuited through a diode that is "open" and allowing current to flow. Therefore, the phase AB voltage is very small (near zero), equal only to the bias of the diode and any small resistance within the controller's internal circuitry. This condition is called 'freewheeling'.

There will be other instances when the phase AB voltage is distorted by one of the wires A or B being connected in a freewheeling circuit with wire C. In these instances, the reading of the phase AB voltage will either be (1) near to the controller's power supply voltage, (2) near to the inverse of the power supply voltage or (3) undefined (normally near zero in practice), depending, respectively, upon whether the wire not involved in the freewheeling circuit is connected is (1) being used to drive current though another of the motor's circuits in the positive direction, (2) being used to drive current through another of the motor's circuits in the negative direction or (3) not being used. For example, if current has recently passed through the phase C-to-B circuit (into the phase terminal wire C and out of the phase terminal wire B), then the wire B is involved in a freewheeling circuit that includes the controller's supply voltage. Current continues to pass from the controller's power supply into wire C and out of wire B. Since the controller has ceased to drive current through this circuit, wire B is no longer connected to the negative terminal of the controller's power supply. Instead, a protective diode within the controller allows current from wire B to pass back to the positive connection of the controller's power supply and then into wire C, creating a loop that will eventually dissipate once resistance overcomes the inductance within the motor's windings. Meanwhile, wire A is initially disconnected and has no defined value. Therefore, the reading of phase AB voltage is essentially zero (imperfections within the controller's measurement devices allow the voltage at the disconnected wire A to equal or nearly equal the voltage at wire B. In Figure 3, this condition is labelled as 307. A similar situation in respect of the phase B-to-C circuit is labelled 309.

A short time later, when the controller switches to accept the current passing through wire C into wire A and thus form a new circuit driving current through the motor from wire C to wire A, the reading of phase AB voltage changes. Current continues to pass through the phase B winding until resistance overcomes the motor's inductance, and wire B continues to be connected to the positive terminal of the controller's power supply through a protective diode, but now wire A is connected to the negative terminal of the controller's power supply and no longer able to float. This new connection causes the phase AB voltage to change from undefined or zero to strongly negative, equal to the inverse of the controller's power supply voltage. In Figure 3, this condition is labelled as 308. A similar situation with reverse polarity (the controller activating the A-to-C connection after deactivating the B-to-C connection) is labelled as 310.

Additionally, phase AB voltage will appear undefined and therefore close to zero in practice, when measured with real equipment, when the motor is freewheeling just after the controller disconnects the A-to-C circuit and just after it disconnects the C-to-A circuit. These are labelled as 311 and 312, respectively, in Figure 3.

For a three-phase motor, one could create charts similar to that shown in Figure 3 that represent the potential from phase A to phase C, from phase B to phase C, and likewise representing the reversed polarity of each of the above charts. In some motors, there are also several charts representing the current detected in each of the phase-to-phase circuits. The total number of graphs will increase analogously if the motor contains more than three phase windings and decrease if it contains less than three. In all cases, the controller is working to compare the values of the respective phase voltages represented by these charts, based on the assumption that they are tolerably good indications of open-circuit voltages and therefore tolerably good indications of when to enact a commutation event (to switch the flow of current). The distortions described here limit the ability of the controller to estimate when commutation events should be enacted, because the signals being compared do periodically become undefined in value or else held close to the voltage or inverse voltage of the controller's power supply, as described above, and they do not always represent the motor's open-circuit voltage tolerably well.

In general, freewheeling, as in the examples described above, occurs only for short times within one electrical period of a motor's rotation. In many motors, distortions in the back-EMF estimate that would arise from freewheeling events can be removed simply by applying strong low-pass filtering that removes any short-duration disturbances. In general, when freewheeling is not taking place, the measured phase voltage is more nearly equal to the motor's back-EMF. Other distortions, such as phase winding resistance and inductance, will distort the amplitude or phase angle of the phase voltage, but the signals will at least be sinusoidal in character. But in very high-performance motors, all such distortions, including and especially freewheeling distortions, may upset the efficient operation of the motor sufficiently to demand a new solution.

Summary

Various aspect of the invention are set out in the appended claims.

Brief Description of the drawings

Figure 1 illustrates schematically an induction electrical motor having three phase windings; Figure 2 illustrates idealised back-EMF voltage signals for the three phases of the motor of Figure 1; Figure 3 illustrates an actual monitored back-EMF voltage signal for the motor of Figure 1 when in use; Figure 4 illustrates portions of the actual back-EMF voltage signal of Figure 3 which correspond to the idealised back-EMF voltage signal; Figure 5 illustrates the phase voltage in a region where it crosses the horizontal axis, based on incomplete data remaining after disregarding the data received during freewheeling; Figure 6 is a flow diagram illustrating a method of using partial monitored back-EMF to determine commutation timing; Figure 7 is a flow diagram illustrating a method of determining commutation timing by discarding certain back EMF signal data and using the remaining data to determine a zero volt point; Figure 8 is a flow diagram illustrating a method of obtaining commutation timing by monitoring back-EMF with the motor phase windings temporarily disconnected from the current supply; and Figure 9 is a flow diagram illustrating a method of determining a commutaion timing for a given motor using data collected from a network of motor's within which the given motor resides.

Detailed Description

In the following description a controller is proposed that is programmed with some awareness of when distortions occur in the measured phase voltages and which responds by discounting measurements of phase voltage that are obtained at those times. Distortions are likely to occur when the phase being measured is "energised" (connected to the controller's power supply) or when a commutation event has recently occurred elsewhere (involving other than the particular two phases whose voltages are being compared. In general, when a controller is at all times measuring the relative voltages of all permutations of all phases and determining commutation timing, then it will periodically happen that some (but typically not all at the same time) phase voltages are heavily distorted. The controller, at all times directing all commutation events, can be programmed to exclude measured data from particular phase voltage sensors at particular times corresponding to particular commutation events. This pre-programming is possible because the a-priori known properties of the motor and controller must dictate a pattern of particular commutation events distorting particular phase voltages whenever said commutation events occur during the operation of the motor. The amount of time when, in response to a commutation event, a phase voltage signal is distorted must also be an a-priori known property, whether known based on theoretical calculations and properties of the motor and controller or else known empirically though experimentation on this or a similar motor and controller. The amount of time during which a controller should disregard a phase voltage reading in response to a commutation event can therefore be pre-programmed into the controller as a mathematical function or a lookup table based on motor speed, temperature, load, or other known or estimated conditions of the motor's operation.

In one embodiment, the controller calculates a distortion time as described above. In another embodiment, a controller disregards all data that are near in value either to zero or to the controller's input power supply voltage (or its inverse voltage), "near being pre-programmed as "falling within a pre-set range of".

In disregarding portions of the measured phase voltages, the resulting picture that the controller is able to paint of the phase voltages is incomplete, having gaps in the data, but it is free from the grossest distortions. In a typical conventional controller, measured phase voltage is at all times taken as the only and best estimate of the motor's back-EMF. According to the controller proposed here however, the controller could sometimes take measured phase voltage as a good estimate of back-EMF and at other times largely or entirely discount it.

For example, in one embodiment, the controller uses measured phase voltage to estimate back-EMF only at times when the measurement is likely to be sinusoidal. At other times, when the signal is likely to be held at zero or held at the voltage of the controller's input power source (or its inverse) due to the switching action of the controller, the controller estimates back-EMF by interpolating historical data with the foreknowledge that the curve should be sinusoidal, or the controller uses data from one or more other phases that are assumed to be free from distortion at that particular time.

In the example of the three-phase motor with one phase voltage A-to-B shown in Figure 3, at the times denoted 303 and 304, there are likely to be good (distortion-free) readings of at least one other phase voltage, B-to-C or C-to-A. Knowing the 120-degree phase separation between these phases in a three-phase motor, the controller could estimate the voltage A-to-B as being a phase-shifted derivative of one of the other, measurable phase voltages, during these times 303 and 304, when the voltage A-to-B not a good representation of the motor's back-EMF. Continuing the example, at the times indicated by 305 and 306, the same controller could interpolate based on a sinusoidal curve fit to the past one or several periods of the voltage A-to-B, there being no alternative measurement which may be taken to be a good representation of the motor's back-EMF at those times.

Figure 4 illustrates the incomplete data, corresponding to one given phase voltage measurement, that remain available to the controller after disregarding the distorted portions of the measured phase voltage. The gaps (shown by broken lines) in the graph 401 correspond to times when the phase in question is connected (or inversely connected) to the controller's power supply. The gaps 402 correspond to times when a different permutation of the motor's phase windings are connected to the power supply. The remaining incomplete data 403, and the similar incomplete data available corresponding to the other phase voltages, may be used by the controller to determine the optimal commutation timing.

In one embodiment, the controller reconstructs a sine wave by applying a curve fit to the available data 403 where the curve fit parameters define a sinusoidal waveform with a known amplitude. It is an intrinsic property of any collection of phase-shifted sinusoidal waveforms that they will cross each other at the same positions along the horizontal (in Figure 4) axis if their amplitudes are scaled in unison. A controller would typically not attempt to calculate the amplitude of any phase voltage but merely enforce the same arbitrary amplitude on all phase voltages as part of a curve fitting algorithm, knowing that idealised measurement of back-EMF would produce curves of the same amplitude for all phases in a typical, symmetrically-wound motor.

The points where a given phase voltage crosses other phase voltages are typically the points when the given phase voltage begins or ceases to be connected to the controller's power supply. Therefore, distortions are likely near these crossover points. But in a symmetrically-wound motor, the same crossover points correspond in time to other phase voltages crossing the horizontal axis (taking on a value of zero). Figure 2 shows this, whereby, for example, the phase voltage 204 crosses the phase voltage 206 and becomes larger in amplitude than the phase voltage 206 at precisely the same time (position along the horizontal axis) when the phase voltage 203 and the phase voltage 207 each cross the horizontal axis (take on a value of zero). Because of this property of typical, symmetrically-wound motors, a complete knowledge of when any phase voltages cross the horizontal axis is also a complete knowledge of when any crossover points occur and when different phases should be energised in commutation events. The correspondence from one data set to another is a simple matter for a lookup table that can be pre-programmed based on the designed properties of the motor. This can be useful, because the calculation when a phase voltage crosses the horizontal axis is often more reliable than the calculation of when said curve crosses other phase voltages. Even a linear extrapolation or interpolation of the data may serve to provide a good estimate of when the phase voltage crosses the horizontal axis.

Figure 5 illustrates the phase voltage in the region where it crosses the horizontal axis 505, based on incomplete data remaining after disregarding the data received during freewheeling. The data 501 can be linearly extrapolated 502 to identify a crossing point 506. If the controller is examining historical data from one or more complete electrical periods in the past, then it also has data 504 from after the horizontal axis was crossed. The controller can interpolate 503 between these data to achieve another estimate 507 of the crossing point. The estimate 507 may be better than the estimate 506 in the sense that it is based upon a greater quantity of data. However, the estimate 506 may be better than the estimate 507 in the sense that it can be based upon more recent data (data from the present period of commutation, rather than a previous period of commutation). Both estimates 506 and 507 could be improved if the controller were equipped with sufficient processing power to apply a sinusoidal curve fit to the data rather than a linear extrapolation or interpolation.

In one embodiment, the controller (1) monitors and records the values of phase voltages during a pre-set amount of time or number of periods. In recording this data, the controller records the value of each phase voltage at as many discrete times as it is capable of recording, according to the characteristics of the measurement apparatus and the controller's internal memory. The controller then (2) excludes certain of the recorded data based upon either (a) the data occurring too near in time to a commutation event or (b) the data being too near in value to zero or to a value corresponding to the controller's power supply voltage (or its inverse). The controller then (3) estimates the point when the phase voltage crosses the horizontal axis using either (a) linear extrapolation, (b) sinusoidal extrapolation, (c) linear interpolation based on one or more prior commutation periods, (d) sinusoidal interpolation based on one or more prior commutation periods, or (e) other extrapolation or interpolation appropriate to the capabilities of the controller and the quality of the data available. The controller (4) consults a lookup table to determine which commutation event to enact (which permutation of phases to connect to the controller's power supply) in response to the given phase voltage crossing the horizontal axis.

Figure 6 is a flowchart illustrating the steps of a method according to an embodiment. For each of one or more phase voltages to be monitored, one or more discrete time intervals are defined, 601, within each electrical cycle of the phase voltage, during which the monitored phase voltage is expected to comprise substantially only the induced back electromotive force, EMF. The or each said phase voltage is monitored, 602, during the defined time intervals whilst disregarding other time intervals. The phase voltage or voltages monitored during the defined time intervals are used to obtain said commutation timing, 603.

Figure 7 is a flowchart illustrating the steps of a method according to another embodiment. The controller measures and records values of a phase voltage, 701, over a predetermined number of discrete time intervals, to create a data set of phase voltages. The controller then identifies, 702, readings that are distorted by a commutation event. In an embodiment, the identification of a distortion is achieved by determining whether a reading is within a predetermined time of a commutation event. In an alternative embodiment, readings which are within a predetermined range of a zero value or within a predetermined range of the controller's power supply are deemed to be distorted. The distorted readings are then removed from the data set, 703. An estimate is then made of a point in time at which the phase voltage is at zero volts, 704. In an embodiment, this is achieved by linear interpolation. In another embodiment, this is achieved by linear extrapolation. In another embodiment, this is achieved by sinusoidal interpolation. In another embodiment, this is achieved by sinusoidal extrapolation. The timing of a commutation event is then determined, 705, based on the obtained estimate of the zero volt point. In an embodiment, this is achieved by the use of a look-up table.

In one embodiment, the controller switches off (ceases to drive current through the motor) temporarily for a sufficiently long time that current ceases to flow through the motor (the controller being programmed with an estimate of the resistance and inductance within the motor's phase windings and thus an estimate of how long will be required, under various circumstances, for current to cease to flow). When the motor exists in this condition, having little or no current flowing through its phase windings, it behaves as in an open circuit condition, and the phase voltages measured at its terminal connections become equal to or nearly equal to the motor's back-EMF. Even if carried out for only a short time within each revolution of the motor, or only once per multiple revolutions of the motor, this procedure may dramatically improve the controller's estimate of back-EMF, the improved estimate being useful subsequently to correct or to phase-offset measurements of phase voltage that are observed by the controller during the motor's normal operation. The above corrections allow the controller to operate more efficiently, particularly if it is equipped with sufficient mathematical capabilities to apply machine learning or fuzzy logic to problem of reconciling open-circuit measured phase voltages with more recently-measured phase voltages occuring during normal operation.

For example, a motor may complete one whole electrical period with the motor in an open-circuit condition, record the measured and filtered phase voltage at every point during that period, and then use this recorded data as the estimator for back-EMF through several subsequent revolutions. Through these subsequent revolutions, the controller observes "marker data", such as a phase voltage signal reaching a certain value that is known a-priori by the controller's programmer to be a reliable measurement, not occurring near a commutation event. When a "marker" is observed, the controller recalibrates its estimate of its position, counting forwards again from the latest observed "marker to maintain an estimate of its position. The position is identically the phase angle of the motor with respect to the previously recorded back-EMF estimate (the phase voltage recorded during the most recent instance of open-circuit operation). The controller determines when to enact a commutation of the motor by: (a) maintaining an estimate of phase angle based on a recently-observed "marker incremented by a counter, (b) maintaining an estimate of each phase back-EMF function recorded during a previous instance of operating the motor "open-circuit', (c) deriving an instantaneous estimate of the back-EMF of each phase by looking-up the value of the back-EMF function (from b, above) with the phase angle estimate (from a, above). The value of the back-EMF function (b) at the phase angle (a) is, for each phase, compared to the same values for all other phases to determine commutation timing. Thus, the back-EMF estimate derived from occasional "open-circuit" operation is used, in preference to instantaneously-measured phase voltage, to determine commutation timing.

By the methods of using partial back-EMF data described above, even a very poor or distorted phase voltage signal can provide occasional good data for use as a "marker".

Therefore, a motor applying this method of occasional open-circuit operation combined with "marker" data can produce more reliable and correct commutation timing than a motor relying only on contemporary measurements of phase voltage to estimate backEMF and to determine commutation timing.

Figure 8 is a flow chart illustrating a method according to an embodiment in which the controller stops the driving current to enable readings of phase voltage to be taken. During operation of the motor, the controller stops the driving of current, 801, until no current flows through the motor. Sampling is then performed in a period where no current is being driven, 802. A back EMF is estimated, 803, based on samples measured during this period 803.

In one embodiment, a motor is intentionally designed with less inductance than would otherwise be the case so that open-circuit operation can occur more quickly when the controller executes the recalibration procedure above. In another embodiment, the motor or the mechanical system to which it is connected is additionally or alternatively created with more inertia than would otherwise be the case so that open-circuit operation has less immediate impact on the behaviour of the motor's output, and the motor is able to continue spinning in open-circuit condition with less reduction in its speed In another embodiment, multiple motors are connected together in one mechanical system, for example by a system of gears, and each motor controller times its period of open-circuit operation and recalibration in such a way that no two motors do this at the same time. Thus, the mechanical system as a whole has motive power at all times, provided by the motor or motors that are not at that time running open-circuit.

In another embodiment, the controller is equipped with additional internal switches for rendering the motor's open-circuit condition more absolute (as in a breaker or separate switch for the purpose).

Figure 9 is a flow chart illustrating the steps of a method of implementing a multiple motor system. During operation of a plurality of motors, the motors being interlinked such as to act as a single motor system, the controller stops the driving current for a first motor of the plurality of motors, 901, until no current flows through the first motor, and over a first period in which no current is driven in the first motor, sampling phase voltages, 902, during the period, and estimating, 903, a back EMF based on the phase voltages. The controller recommences driving a current through the first motor, and immediately or after a time interval, stops driving the current through a second motor, 904. Phase voltages are then sampled 905 during a second period in which no current is driven in the second motor. An estimate 906 of the back EMF is then obtained from the voltage samples during the second period.

In another embodiment, the controller is programmed with knowledge of the operating requirements of the system such that it can avoid implementing open-circuit operation during a time when this would be relatively more inconvenient to the operation of the system powered by the motor or motors. For example, a controller may be programmed with the knowledge that, once a motor has begun to accelerate due to an external command, then it is likely to receive commands instructing it to go on accelerating the motor until it reaches maximum speed. The same controller, armed with this knowledge, could avoid implementing open-circuit operation immediately after receiving a command to accelerate, since open-circuit operation would delay the process of accelerating the motor.

In an embodiment, the controller implements an open-circuit operation for back EMF measurement by disconnecting all phase windings from the controller's power supply, on a periodic basis. In an embodiment, an override mechanism is provided which stops the implementation of the open circuit operation in response to a certain control commands received, for example an instruction to accelerate, an instruction to accelerate where the acceleration is above a threshold, or an instruction to accelerate to a maximum speed.

In another embodiment, a fuzzy-logic-style controller could weigh the relative importance of recalibrating its estimate of back-EMF (a metric that is proportional or indirectly proportional to the time elapsed since the most recent instance of open-circuit operation) against the relative importance of maintaining full motor functionality (for example, by avoiding an instance of open-circuit operation when the controller receives a command to accelerate). The fuzzy logic controller assigns a score (between 0 and 1) to the relevance of its back-EMF estimate. The assigned score reduces over time, as the estimate of back-EMF becomes older and potentially less relevant to the contemporary operation of the motor. Meanwhile, the fuzzy logic controller maintains at all times an estimate (between 0 and 1) of how much the motor's performance would be affected if its next revolution were carried out "open-circuit" (with the motor's phase windings disconnected from the controller's power supply). The latter estimate of performance impact would be based upon the motor's operating state (torque and speed), requested operating state (requested torque and speed), and a-priori preprogrammed tabulated data such as a lookup table. When the ratio of the two estimators -(a) the relevance of the previously-estimated back-EMF versus (b) the impact of open-circuit operation -falls below a threshold value, then the fuzzy logic controller determines that it will be appropriate to operate the motor "open-circuit" during the next revolution of the motor. After operating "open circuit" for one revolution, the controller updates its estimate of back-EMF and assigns a score close to 1 to the relevance of that estimate The score then deteriorates over time by the controller repeating the above procedure.

In another embodiment, the controller or the motor are additionally provided with further inductance or resistance, either being connected at all times or connected (for example by a switch) only when the controller is operating in an open-circuit condition, for the purpose of more rapidly dissipating any freewheeling current in the motor's windings after the controller's power supply is disconnected, so that the motor's operating condition can be taken to be an open-circuit operating condition sooner after the controller's power supply is disconnected, and a greater quantify of data useful for estimating the motor's back-EMF can be harvested by the controller in a shorter time than would be possible without the further inductance or resistance.

In another embodiment, several similar machines connected to a network share their recorded back-EMF waveform data, so that each motor can benefit from the back-EMF waveform data recorded by the most recent motor in its network to have implemented open-circuit operation. This is likely to entail some post-processing of the data obtained from the network so as to ensure its appropriateness to the local motor in question. For example, a controller for a motor may be operating with an estimate of back-EMF (derived from an earlier instance of "open-circuit" operation), and the controller may have lately assigned a value of 0.35 to the relvance of that back-EMF estimate to the motor's present operation. Then, the controller receives a new back-EMF estimate from another controller on the network, the new back-EMF estimate being communicated as a series of data points relating back-EMF voltage as a function of phase angle. The new back-EMF estimate carries a relevance score close to 1 for the controller that sent it, being recently measured. However, our local controller receiving the new estimate over the network assigns a score of 0.55, having been pre-programmed to de-value incoming back-EMF estimates from the network by a factor of 0.55. Nevertheless, our local controller observes that the estimate of back-EMF incoming from the network having a relevance score of 0.55 is nevertheless more relevant than its own contemporary estimate of back-EMF having a relevance score of only 0.35. The controller then determines to use the incoming estimate of back-EMF as its own estimate, despite it having come from another controller across the network. When our local controller does have the opportunity to execute "open-circuit" operation and to recalibrate its estimate of back-EMF again, then it will assign a relevance of 0.1 to its own new estimate and broadcast the same back-EMF estimate across the network for other controllers to use or disregard as appropriate according to the relevance scores that those other networked motor controllers assign to the incoming data.

By any of the methods describe above, the process of the controller switching off and recalibrating its estimate of back-EMF becomes more feasible than if the procedure were implemented on a general motor.

Claims (18)

1. CLAIMS: 1. A method of obtaining a commutation timing for an electrical motor having a stator with a plurality of phase windings and a rotor with one or more permanent magnets, the method comprising: defining, for each of one or more phase voltages to be monitored, one or more discrete time intervals within each electrical cycle of the phase voltage, during which the monitored phase voltage is expected to comprise substantially only the induced back electromotive force, EMF; monitoring the or each said phase voltage during the defined time intervals whilst disregarding other time intervals; and using the phase voltage(s) monitored during the defined time intervals to obtain said commutation timing.
2. 2. A method according to claim 1, wherein using the phase voltage(s) monitored during the defined time intervals to obtain said commutation timing comprises estimating the back EMF signal using phase voltage samples associated with time period for which there is no distortion, and discounting phase voltages associated with time periods where there is distortion.
3. 3. A method according to claim 1, further comprising estimating the back EMF for periods where there is distortion by interpolation or extrapolation from phase voltage measurements taken when there is no distortion.
4. 4 A method according to Claim 3, wherein if the estimation is sinusoidal.
5. 5. A method according to claim 2, further comprising estimating the back EMF for periods where there is distortion based on phase voltages taken for a second set of phase windings.
6. 6. A method according to claim 2, further comprising estimating the back EMF for periods where there is distortion based on previous back EMF estimates.
7. 7. A method according to claim 1, further comprising performing an open circuit measurement operation during the operation of the motor, wherein the open circuit measurement operation comprises the steps: stopping, by the controller, the driving of current over a given time interval so that no current flows through the motor; sampling phase voltages over a first period in which no current is being driven; and estimating a back EMF for the period based on samples measured whilst there is no current being driven.
8. 8. A method according to claim 7, wherein open circuit measurement operations are performed periodically.
9. 9. A method according to claim 8, further comprising overriding the performance of an open circuit measurement operation in response to receipt by the controller of an operational command.
10. 10. A method according to claim 9, wherein the operational command comprises one or more of: a command to accelerate, a command to accelerate where the acceleration is above a threshold, and a command to accelerate to a maximum speed.
11. 11. A method according to any one of the preceding claims, wherein said step of using the phase voltage(s) monitored during the defined time intervals to obtain said commutation timing comprises updating a previous commutation.
12. 12. A method according to claim 11 when appended to claim to claim 7, wherein a commutation timing is determined from the back EMF estimated for the period based on samples measured whilst there is no current being driven and that commutation timing is updated using the phase voltage(s) monitored during said defined time intervals.
13. 13. A method of operating a system comprising a plurality of coupled motors, the method comprising performing the method of claim 7 to operate each motor where the given time intervals for each motor are not aligned or overlapping.
14. 14. A method of obtaining a commutation timing for an electrical motor having a stator with a plurality of phase windings and a rotor with one or more permanent magnets, the method comprising: during operation of the motor, disconnecting all phase windings from the controller's power supply for a given time interval; monitoring one or more phase voltages during said given time interval; and determining a commutation timing from the monitored phase voltage(s).
15. 15. A method according to claim 14 and comprising, subsequent to said given time interval, monitoring one or more phase voltages and using a result to update said commutation timing.
16. 16. A method according to claim 15, wherein said step of monitoring one or more phase voltages comprises monitoring said one or more phase voltages during one or more discrete time intervals within each electrical cycle of the phase voltage(s), during which the monitored phase voltage is expected to comprise substantially only the induced back electromotive force, EMF.
17. 17. A method of obtaining a commutation timing for a first electrical motor having a stator with a plurality of phase windings and a rotor with one or more permanent magnets, the first motor being one of a network of similar motors, the method comprising, at a controller of the first motor: monitoring the phase windings of the first motor to obtain first commutation timing data and applying a first reliability score to that data; obtaining second commutation data from a monitoring of one or more of the other motors of the network and applying a second reliability score to that data; and obtaining a commutation timing for the first motor by selecting one or other of the first and second commutation data based upon the relative reliability scores.
18. 18. A method according to claim 17, wherein said motors of the network are synchronised.