Control of a brushless electrical machine

Control of a Brushless Electrical Machine

<p>The present invention relates to the control of switched brushless electrical machines. The present invention relates in particular, but not exclusively, to switched reluctance machines.</p>

<p>A switched reluctance appliance is a type of brushless electric appliance. It has a rotor, defined rotor poles, a stator, defined stator poles, and in relation to the stator poles to define one or more independent energisable phases. It contains a set of windings that are aligned. In a reluctance machine, the energization of one or more phase windings produces a magnetic flux in a circuit comprising associated stator poles, which brings the rotor to a position of minimal reluctance. forcing Timing the sequential excitation of the windings according to the rotor position causes rotor motion. The general handling of electrical drives incorporating switched reluctance devices can be found in various textbooks. For example, "Electronic Control of Switched Reluctance Machines" by TJE Miller, Newnes, 2001, which is incorporated herein by reference. For more details, see The Characteristics, Design and Applications of Switched Reluctance Motors and Drives' by Stephenson and Blake, presented at the PCIM'93 Conference and Exhibition at Nurnberg, Germany, June 21-24, 1993" paper (Stephenson) thesis), which is incorporated herein by reference. As is well known in the art, these devices can be operated as motors or generators by taking care to match the timing of the application of excitation to the phase windings.</p><p>Unlike conventional induction and simultaneous 'electromagnetic' machines, there are, for example, so-called brushless DC machines, in which the current is in the stator coils and the field is the rotor. made by permanent magnets on the Switched reluctance devices are purely 'magnetic' devices. When the reluctance of a magnetic circuit changes, a torque is produced only by the magnetic field (magnetic field). As a result, the methods of controlling the two types of devices are quite different. Because this control is related to how the torque is generated. In general, the control method used for conventional devices with sinusoidal input is not suitable for switched reluctance devices.</p><p>1 shows a typical switched reluctance device in a cross section. In this example, the magnetizable stator 10 has six stator poles 12 . The magnetizable rotor 14 has four rotor poles 16 . Each stator pole has one coil (18). The coils at completely opposite poles are connected in series to feed the three phase windings. Only one phase winding is shown for clarity. Control of the switched reluctance device may be accomplished by a variety of methods well known to those skilled in the art. If information on the angular position of the rotor is available, for example from a position transducer, an excitation can be applied as a function of position. Such machines are often referred to as "rotor position switched machines".</p><p>A typical switched reluctance drive is shown in FIG. 2 . In this example, device 36 corresponds to that shown in FIG. 1 . The three phase windings A, B and C are alternately switched in the DC supply V by a set of power electronic switches 48 . The moment at which the switches operate (ie, the rotor position) is determined by the controller 38 , which may be implemented in hardware or software of a processing device such as a microcontroller or digital signal processor. Control signals are sent to the switches via a data bus 46 . Closed circuit current feedback uses the current sensor 44 and the required current (i)<sb>D</sb>) (42) and feedback a signal proportional to the phase current compared to and provided by sensing the phase currents. Control algorithms often include proportional (P), proportional-plus-integral (P+I), time optimal, feedback linearised, proportional/integral/derivative (PID) ) function, or one of many others as well understood in the art. It is also common to the outer control loop of position or velocity supplied by feedback of the rotor position signal from the position sensor 40 .</p><p>In operation, a signal corresponding to the required current 42 is supplied from the controller 38 . It regulates the current in the windings to produce the desired output from the appliance according to the particular control scheme employed.</p><p>The performance of a switched reluctance machine depends, in part, on the precise timing of the phase energisation with respect to the rotor position. Sensing of rotor position can be conventionally accomplished by using a physical rotor position transducer (RPT) 40 , as shown in FIG. 2 , with rotating cogs mounted on the machine rotor. It is like a toothed disk and works with optical or magnetic sensors mounted on the stator. A pulse train indicative of the rotor position relative to the stator is supplied and generated to the processing device, allowing accurate phase excitation. Alternative methods of position sensing include so-called "sensorless" methods, in which there are no physical position transducers and the position is inferred from measurements of one or more other parameters of the device.</p><p>Because the current in the windings is relatively easy to measure, closed circuit control of the appliance can typically be achieved by monitoring and controlling the energizing current in the windings. However, the desired output of a device is usually torque, position or speed, and current has a very non-linear relationship to all of these. The result is that current control techniques generally result in output inaccuracies such as torque ripple, position error and/or speed error. Many current control schemes have been devised to address these shortcomings. Many are adding complexity.</p><p>It is supported by some researchers that the more fundamental control variable in switched reluctance machines is the magnetic flux set up in the magnetic circuit in the machine when the phase windings are energized. The magnetic flux is directly responsible for the force acting on the rotor forcing it to a position of minimal reluctance with respect to the excited stator poles. EP-A-1109307 and EP-A-1109308 disclose methods using magnetic flux as the main control variable. The drive merges with the flux controller.</p><p>Many different power converter topologies are known, some of which are discussed in the above-mentioned Stephenson paper. The most general configuration in FIG. 3 is shown for a single phase of a polyphase system, in which the phase windings of the device have two It is connected in series with the switching devices 21 , 22 . The busbars 26 and 27 are collectively described as the "DC link" of the converter. Energy recovery diodes 23 , 24 are connected to the winding such that when switches 21 , 22 are open the winding current flows in the opposite direction of the DC link. A capacitor 25, known as a "DC link capacitor", is any alternating component of the DC link current that cannot be pulled from or returned to the supply (i.e., the so-called "ripple current"). It is connected across the DC link to generate or sink (alternating components). In practical terms, capacitor 25 may include several capacitors connected in series and/or parallel. In case a parallel connection is used, some of the elements may be distributed through the converter. Polyphase systems typically use several "phase legs" of FIG. 3 connected in parallel to independently excite the phases of the electrical machine.</p><p>The phase inductance cycle of a switched reluctance device is the period of change in inductance for each phase or phase between common points in successive cycles. Successive cycles are, for example, between inductance maxima when the rotor poles and their respective stator poles are perfectly aligned. As described in the Stephenson paper mentioned above, the maximum inductance region, L<sb>max</sb>, is centered on the rotor position where the pair of rotor poles is fully aligned with the pair of stator poles. Similarly, as shown in Figure 1, the minimum inductance region (minimum inductance region), L<sb>min</sb>, corresponds to the position where the axis between the poles on the rotor aligns with the axis of the stator poles. </p><p>At low speeds, switched reluctance systems typically operate in current-controlled or "chopping" mode. As explained in the above-mentioned Stephenson paper, hysteresis current controllers using "hard" chopping are often used. An alternative regime is "soft" chopping, where only one switch opens when the current reaches an upper level. The current then decays more and more slowly through the winding, the second switch and one diode. Other types of current controllers are well known in the art, for example those described in EP-A-0769844, which is hereby incorporated by reference and are referred to as off-time controllers, constant frequency controllers. controllers), etc., and will not be further described here.</p><p>At high speeds, switched reluctance systems typically operate in "single pulse" mode of excitation instead of chopping mode. This is also explained in the Stephenson paper mentioned above.</p><p>So, the system generally uses chopping mode at low speed and single pulse mode at high speed. The upper and lower chopping current levels are typically set to values higher than the expected peak current of the single pulse mode, so these parameters do not interfere with the single pulse operation. It is known to set the upper current level to a value that acts as a "safety net" and if a fault condition develops in the drive the current exceeds this upper level and causes one or more switching devices to open thereby The current is limited to a safe value.</p><p>It is common knowledge that more output cannot be obtained in single pulse mode, and there are limits reached when the phase flux and current are only zeroed at the end of the electrical circuit. To increase the torque under this assumed constraint, it would be necessary to either increase the power supply voltage or decrease the number of turns in the phase windings of the machine, or both. For a variety of reasons they are either impractical or undesirable.</p><p>A further mode of operation is the so-called "continuous current" mode, which makes it possible to create mechanical output torques and powers in excess of those obtained up to the limiting state in single pulse mode. It is "US Patent No. 5469039 (Stephenson), incorporated herein by reference. This patent discloses a method of operating in a stable manner in these modes to allow for steady state operation. The phase windings therefore manipulate the current flowing continuously through them, unlike in chopping and single pulse modes, and are always linked by magnetic flux. The current change is apparently sinusoidal beyond its "standing" value. This becomes an important mode for systems that must produce high levels of overload output at some point in their operating cycle. Although the efficiency of the drive drops in this mode, it allows for what would otherwise be required in the design to be achieved which would otherwise require a larger appliance or increased volt-ampere (VA) capacity in power switching devices. . (Although the efficiency of the drive falls in this mode, it allows specifications to be achieved which might otherwise require a larger machine or increased volt-ampere (VA) capacity in the power switching devices.)</p><p>When motoring operation is described in the discussion mentioned, it is well known that switched reluctance devices work equally well in generating mode, and in generating mode the current waveforms ) are generally mirror images of motoring waveforms. </p><p>The parameters of angle and current for controlling the device in different modes are usually functions of speed, and are calculated in real time or, more generally, stored in the form of a data table so that they can be read at a suitable time. The parameter values are carefully chosen to obtain a smooth output from the instrument when the speed is changed. If the stored values are relatively sparse, some form of interpolation is used to give appropriate parameter values at intermediate speeds. There is a particular difficulty in selecting values at the transition points between chopping and single-pulse modes. And even between single-pulse and continuous current modes, a smooth transition is what was desired regardless of the torque level required. </p><p>One solution to this problem in chopping and changing single pulse mode is to use current control parameters as the main variable. The current level is generally held constant throughout the chopping range with the speed for a given torque demand. However, as the speed rises into the single-pulse mode region, the mode gradually changes from chopping to single pulse at the upper current level (and sometimes by a corresponding amount of lower current level). It is known to raise the (lower current level) gradually. Once the device reaches a speed at which the phase current never reaches the upper current level, the current control parameters are usually set to high values so that they do not cause chopping for the rest of the single pulse speed range.</p><p>The question of sensitivity to control the angle at entry, the continuous current mode, remains. Rapid changes in control are difficult to achieve and risk an unstable system for the reasons I will now describe.</p><p>The effect of "dead time" that occurs between cycles of flux and current during normal operation is that flux and current always start from zero at the beginning of each electrical cycle. When operating in continuous current mode, the steady-state flux, current and torque are not the current cycle, but the time derivative of the net applied winding voltage during all previous electrical cycles ( time integral), which is the case in a typical "single pulse mode". Therefore, the instrument output in continuous current mode is very sensitive to very small changes in the control angle. This creates difficulties in linearising the torque output with respect to the torque demand signal, and also the winding resistance and voltage drops in semiconductor switches. ) leads to a very high level of sensitivity to Attempts to control torque in continuous current mode only by controlling the angle, especially since winding resistance is a strong function of temperature, often result in changes in machine temperature, e.g. load or ambient changes. variations), resulting in unacceptable variations in output.</p><p>The problem of sensitivity to switching angles can be improved by using the aforementioned controller to control peak current and magnetic flux. In this case, the conduction angles are made slightly wider than required under the worst conditions (eg at the highest required winding temperature). The chopping action of the current controller then automatically provides stability of the operating point. At steady state, one "chop" is required to clip the peak current and the control flux. So excessive switching losses in power electronics are not an issue. It is described in US 6972533, incorporated herein by reference.</p><p>However, major difficulties remain. The use of current control to regulate torque in continuous current mode does not address the difficulty of achieving smooth transitions into and out of mode. The exact point at which onset of continuous current occurs will depend on the winding temperature and will also be sensitive to supply voltage and changes from one particular instrument sample to another. A single control parameter "map" can affect smooth changes in only one supply voltage and winding temperature. ; Given the high sensitivity of the instrument output to parameter variables while operating in continuous current mode, the difficulty of compensating for voltage and winding temperature will be evident.</p>

<p>According to a disclosed embodiment in which there is a controller provided for a brushless electric machine having at least one phase winding, said controller comprising a control means usable for making a control signal for excitation of the phase winding in response to an input signal. means), and can be used to modify the input signal using a flux stabilization signal, thereby changing the magnetic flux produced by the phase winding to stabilize the output of the device. Also post the equivalent method. The input signal is changed by the flux stabilization signal. This affects the output of the control means for changing the magnetic flux of the device.</p><p>Any of the disclosed embodiments may be used with respect to sensorless forms of rotor position detection. Other embodiments may be used with respect to rotor position detector hardware. In various embodiments, the flux stabilization signal is the instant at which flux decay in the phase winding of the machine is initiated by changing the input signal in accordance with the control means according to the instant in which the flux in the phase is arranged to decay. can be used to influence</p><p>In one form, the flux excitation signal has a sawtooth waveform with a generally linear slope. It is combined with the flux demand signal with a negative slope and compared with the flux feedback signal, which means the instantaneous flux in the device itself. Instead, the sawtooth waveform of the flux excitation signal has a positive slope and instead combines with the flux feedback signal. This combined signal is compared to the flux demand signal. In either case, the effect of the stabilization signal causes a change from an increasing flux to a falling flux. Even in the presence of any disturbance there is an adequate period (in angle and time) for the magnetic flux to be stably controlled before the end of the period of excitation of the phase. In general, the stabilization signal is adapted to keep the rise and fall of magnetic flux practically the same over the electrical cycle.</p><p>The combination of the flux excitation signal and the flux feedback signal or flux demand signal can be derived by adding them together in an adder. </p><p>The controller may comprise means for generating the magnetic flux excitation signal as a time-varying signal having a duration that coincides with a duration of repetition of the excitation of the phase winding. </p><p>The slope of the sawtooth waveform of the stabilization signal can change with the speed of the device. For example, the slope may increase as the speed increases and as the period of time of the electrical cycle shortens. However, in some embodiments, the peak magnitude of the sawtooth waveform may remain the same regardless of the instrument speed.</p><p>In another embodiment, the stabilization signal is used to change the signal indication of magnetic flux increase in the phase winding. In this form, the rate of increase in magnetic flux indicated by the flux feedback signal is changed so that the apparent rate of increase in magnetic flux increases. By predicting depending on the changing magnetic flux feedback signal, the input signal is changed thereby allowing the switch-off point to be determined.</p><p>Another embodiment disclosed is a controller for a switched reluctance machine having a rotor, a stator having at least one phase winding and a switch means for energizing the at least one phase winding, the controller being the switch means can be used to generate actuation signals for driving the controller, wherein the controller switches off to control the switch means when the magnetic flux in at least one phase winding exceeds a flux demand signal. first switch control means for making a switch-off signal and second switch control means for making drive signals for controlling the switch means according to the position of the rotor relative to the stator ) is included. </p><p>The controller is arranged to receive a first flux signal indicative of a magnetic flux in the at least one phase winding and a second flux signal indicative of a demanded flux, the first flux signal indicative and a comparator arranged to produce a switch-off signal when the first flux signal exceeds the second flux signal. </p><p>The second switch control means can be used to generate a switch-on signal when the rotor has reached a predetermined position with respect to the stator. Thereby, the magnetic flux in the phase winding increases.</p><p>The second switch control means may be used to generate a freewheel signal for freewheeling the current in the at least one phase winding when the rotor has reached the freewheeling position with respect to the stator. Thereby, the magnetic flux in the phase winding decays.</p><p>The second switch control means can be used to generate a switch-off signal when the rotor has reached a switch-off position with respect to the stator. Thereby, the magnetic flux in the phase winding is suppressed.</p><p>The present invention can be applied to various brushless electrical appliances such as switched reluctance appliances or brushless DC appliances.</p>

<p>The invention may be practiced in various ways, some of which are now described by way of illustration with reference to the drawings that follow. </p><p>Referring to FIG. 4 , a first embodiment of a switch reluctance drive system is a thin plate magnetizable stator 10 having phase windings A, B, C and a thin plate magnetizable stator 10 . and a switched reluctance machine (36) having a rotor (14) that is located there. Each is made of this embodiment of suitable steel laminations well known to those skilled in the art. A rotor position transducer (RPT) 40 is aligned relative to the rotor 14 for sensing position. As mentioned above, the RPT 40 can take a variety of forms, in which the movement of the first element past the sensing device is an output indication of the relative movement of the rotor relative to the stator. indicative). The flux controller 90 is configured to control the excitation of the phase windings A, B and C associated with the stator 10, for example in the phase legs as shown in FIG. 3 ; It makes control signals for driving the configured power electronic switches 48 . The illustrated connection of only one phase winding is shown for clarity, but each phase is energized independently by actuation of the switches. In an alternative embodiment, the excitation of the windings is provided by an amplifier of a well-known type.</p><p>The flux controller 90 is provided with three feedback signals. First, it receives rotor position signals (θ) from the RPT 40 . In an alternative embodiment, the rotor position signals are provided by an algorithm that infers the rotor position from other parameters of the machine. In other cases, signal [theta] provides positional information (and hence timing) to magnetic flux controller 90 . Second, the flux controller 90 receives phase current signals i from a current transducer 44 . Each phase winding will typically be associated with a current transducer 44 to produce a current signal i. However, only one transducer is shown here for clarity. It should be noted that the current feedback signal is not used in the manner of conventional current controllers: rather, it is used to provide overall monitoring of the current to ensure it stays within a predetermined safe level. Third, the flux controller 90 receives flux signals (?) proportional to the flux in each phase winding. These signals are derived either by direct measurement of magnetic flux or by one of several methods of magnetic flux determination well known to those skilled in the art.</p><p>The input 92 to the flux controller 90 is the desired peak flux (Ψ) in the instrument to produce the desired output (desired output).<sb>D</sb>) is a demand signal representing When operating a device as a motor, the desired output is torque, speed or position. In addition, the flux controller 90 further has a flux stabilization input 94 . It is explained in more detail below. The magnetic flux controller 90 produces control signals. These control signals are passed from a voltage source (V) via a data bus (46) to power electronic switches (48) which control the excitation of the phase windings.</p><p>The magnetic flux associated with the inputs provides a means of terminating the phase voltage pulse. The phase flux changes before the rotor reaches the end of the conduction period, where the phase switches are normally open in single-pulse mode, or before the phase currents in chopping mode. A limited value must be reached before the determined value is reached. This is shown in FIG. 5 . 5 shows a waveform for a device operating in a continuous current mode. The demanded peak flux level (Ψ)<sb>D</sb>) is the expected conduction angle (θ)<sb>c</sb>) is reduced. So the magnetic flux is Ψ<sb>D</sb>The value of and the standing value of Ψ<sb>s </sb>cycle between value Ψ<sb>s</sb>is not directly controlled and is the demanded peak flux (Ψ<sb>D</sb>), the machine speed, and the ratio of the rise and fall of the magnetic flux (determined alternately by the supply voltage and the number of turns in the phase winding). </p><p>In general, it is recognized by the inventors that the developed torque is largely independent of the winding resistance in single pulse mode. As the machine experiences a change in the load, the winding temperature (and hence the winding resistance) changes. However, for most systems, the output of the instrument is essentially independent. This allows the use of control parameters that are not affected by the winding resistance. However, a difficulty encountered in the more volatile continuous current mode is the very high increasing torque obtained for a given increase in conduction angle. When the flux and current finally settle to a controlled steady-state value, small increments in conduction angle will cause very large changes in the standing flux, current and average torque. This is especially true if the phase winding resistance is small. In a practical drive system, this will cause problems with torque control resolution when attempting to use commutation angles to adjust torque in continuous current mode. </p><p>Although acceptable torque resolution can be maintained, the system level control must be able to cope with sudden increases in the torque "gain" of the machine at the point where it enters the continuous current mode from the single pulse mode. Furthermore, the steady-state values of the average flux, current and, therefore, torque now depend critically on the phase winding resistance and also on the voltage drop within the semiconductor switch. All these parameters show strong temperature dependence. This means not only that the actual torque developed in a continuous current mode is strongly sensitive to the temperature it composes, but that the exact conduction angle required to reach that mode will be temperature dependent. This latter point is in particular using a conventional control angle "look-up" table, or other static means of determining the commutation angle (eg, a polynomial function). It makes it very difficult to linearize the torque-control discontinuity.</p><p>A preferred embodiment of a commutator for an SR device that overcomes the problems noted above is shown in FIG. 6 . This forms part of the controller 90 in FIG. 4 making the switch outputs on the bus 46 . The output θ from RPT 40 is the switch-on θ at each of comparators 100 , 102 and 104 .<sb>on</sb>, freewheeling θ<sb>fw</sb> and switch-off θ<sb>off</sb> is compared with predetermined values of the angle for . The flux feedback signal Ψ (which may have been derived from direct feedback or an evaluation of the flux) is the demanded flux Ψ at comparator 106<sb>D</sb>compared with </p><p>The output of the switch-off comparator 106 together with the output of the switch-off comparator 104 is applied to an OR gate 108 . The output of the magnetic flux comparator 106 together with the output of the freewheel comparator 102 is applied to the OR gate 110 . The output of the switch-on comparator 100 supplies the SET input to the SR set/reset latch 120 . The RESET input to the first latch 120 is the output of the OR gate 110 . The output of the switch-on comparator 100 also feeds the SET input to the second SR latch 122. The RESET input to the second latch 122 is the output of the OR gate 108. The outputs from the latches are busses to monitor the current at the phase winding level and supply control signals to the power electronic switches 48 in FIG. 4 . inputs of the controller 124 leading to 46. As mentioned above, the controller is part of the magnetic flux controller 90 . As such, it is a processing device configured by sub-routines or separated from a single processing device. In either case, it is, for example, a microprocessor or digital signal processor running software executing the recognized steps to perform the control techniques described herein. The controller includes features that monitor the current and take appropriate action if the current exceeds a predetermined threshold.</p><p>The circuit of FIG. 6 may operate as follows. The rotor angle (θ) is the switch-on angle (θ)<sb>on</sb>), the SR latches 120 and 122 are both "set". This then activates the phase switching devices (subject to the overall agreement of the controller 124 ). θ<sb>on</sb>After the onset of conduction at<sb>S</sb>is made from If the rotor position estimate (or measurement) is the phase flux limiting value Ψ<sb>D</sb>before reaching the freewheel angle (θ)<sb>fw</sb>) or turn-off angle (θ)<sb>off</sb>), the output of comparator 102 or 104 changes state. </p><p>One or both SR latches are then cleared or "reset" by the output from one or two OR gates 180 and 110, which are R inputs to latches 120/122. This opens one or two phase switches 21/22 ( FIG. 3 ) depending on the state, ie freewheeling or turn-off. The phase current and flux will then decay slowly (if the freewheeling angle is reached and only one of switches 21/22 open) or rapidly (if the off-angle is reached and both switches open). Calculated in real time or stored in some known form of memory, θ<sb>on</sb>, θ<sb>off</sb> and θ<sb>fw</sb>The value of is set as a threshold value. This is an angle-based commutation.</p><p>As explained in Fig. 5, however, the phase flux is limited by the flux limit Ψ before the off-angle is reached.<sb>D</sb>may reach At this point, both latches 120 and 122 will be reset immediately. So the phase flux is pulled down irrespective of the expected conduction angle. The instrument is now being controlled as a function of its peak flux. When the flux limit is reached, the alternative way is to open only one of the phase switches 21/22 - this will initiate freewheeling and the phase flux will then only slowly drop until the off-angle is reached.</p><p>The system of FIG. 6 can be operated in single pulse or continuous current modes depending on the inputs of the comparators 100 , 102 , 104 and 106 . In each mode, the system can operate as a conventional chopping current controller, angle controller or peak flux controller. The created torque is Ψ<sb>D</sb> All these modes of operation of the controller 90 are useful, especially in the continuous current mode, as they are smoothly and progressively controlled by , and less sensitive to the precise values of winding resistance, supply voltage or control angles. It also provides a means for smooth and continuous change between single-pulse and continuous current modes, even as the operating state of the device changes.</p><p>However, the best operation of the controller described above, especially such as operation in continuous current mode, is dependent on the absence of disturbance of the input and feedback parameters. These disturbances can come from many sources. In drive operation in electrically unsuitable environments, electrical noise may be present, for example in current or magnetic flux measurements. In drives with low inertia, the rotor may change speed or direction abruptly. In a drive with "sensorless" rotor position detection, the evaluation may suffer from random errors, creating jitter in the control angles fed to the controller. may be Drives using digital circuitry in control systems may suffer from quantisation effects. Any or combination of these factors can lead to instability in the flux controller. </p><p>An example of such instability is shown in FIG. 7 . This is θ<sb>off</sb>before the arrival of Ψ<sb>D</sb>Describe the magnetic flux that has reached So the first cycle of flux is the flux demand parameter Ψ<sb>D</sb>controlled by However, the transient disturbance ΔΨ is ΔΨ during the second cycle<sb>D</sb>, which causes early termination of the increasing magnetic flux. Even if the flux demand is stored back to its intended value, subsequent cycles of flux become unstable, leading to an erratic output of the instrument. In general, very small values of disturbance are sufficient to cause instability, since the error introduced by disturbance will not decay by itself.</p><p>The slope of the (linearized) magnetic flux waveform is m as shown in FIG.<sb>u</sb><sb></sb>and m<sb>d</sb>are each designed with The unit of the slope is [Weber/degree], and if the diagram is drawn in the time domain, it is [Weber/sec]. Units are interchangeable because the machine rotates at a constant speed. m<sb>u</sb>The numerical value of m is positive<sb>d</sb>The numerical value of is negative. By using linear algebra, if a disturbance ΔΨ is embedded in the flux waveform, then this disturbance is only m<sb>u</sb> &gt; |m<sb>d</sb>If |, it can be shown that it will decay from cycle to cycle. If this condition is not met, any error embedded in the waveform will grow from cycle to cycle. resulting in a chaotic sub-harmonic oscillation of the magnetic flux waveform. A person skilled in the art will recognize that disturbances are equally well generated at the control angle or the measured magnetic flux. However, regardless of the source of disturbance, the gradients of the magnetic flux are fixed by the device design and supply voltage.</p><p>The problem of instability is addressed in this embodiment by adding a stabilization signal in the form of a negative-going ramp, which is synchronized with the period of the rpt signal as shown in FIG. 8 . . ramp m<sb>stab</sb>The gradient of also has units [Weber/degree] in the angular domain and its numerical value is negative. By linear algebra, the error in the waveform decays to zero if the following equation is satisfied.</p><p>m<sb>u</sb> - m<sb>stab</sb> &gt; | m<sb>d</sb> |</p><p>i.e. m<sb>u</sb> - m<sb>stab</sb> +m<sb>d </sb>&gt; 0</p><p>Fig. 9 shows how this can be implemented in the flux controller of Fig. 6, the reference numerals being used the same. The controller 124 of FIG. 9 has the inputs of the flux controller 90 of FIG. 4 . flux demand signal (Ψ)<sb>D</sb>) is in the form of a negative-going saw-tooth waveform shown in FIG. 8 and is a flux stabilization signal (Ψ) that is repeated while coincident with each conduction cycle<sb>stab</sb>) and is supplied to an adder 126 that is combined. The effect of combining the two signals is independent of time (i.e. rotor position) and includes signal disturbances and/or parameter variations that change the balance between the increasing and decreasing portions of the magnetic flux waveform. to give a peak flux demand that automatically takes into account the effect of Adding this flux to stabilize the signal makes the system stable.</p><p>the flux stabilization signal (Ψ)<sb>stab</sb>) can be equally applied to the other terminals of the flux comparator 106 . That is, the flux stabilization signal is added to the flux feedback instead of the demand reference. In this case, the flux stabilization waveform, m<sb>stab</sb>A positive-going ramp for ) would be required to achieve the same effect. Other combinations using suitable combinations of ramp waveforms (negative-going or positive-going) are of course possible again with equivalent results. Although the flux stabilization signal is referred to above as a linear sawtooth waveform, other forms of the flux stabilization signal are possible within each electrical cycle. For example, the slope of the flux stabilization signal will probably be exponential or some other non-linear function. The choice of waveform will depend on the application.</p><p>In large and/or high voltage SR devices, the resistive and semi-conductor voltage drops will typically be relatively small compared to the DC link voltage. gradient of magnetic flux m<sb>d</sb>is the slope of the magnetic flux due to the high value of the voltage (which is the determining factor of the rate of flux change per unit time), m<sb>u</sb>It will only get very slightly steeper than it is. So, in this state, only a relatively small amount of stabilization results in a combined slope (m<sb>u</sb> - m<sb>stab</sb>) is m<sb>d</sb>is required for device stabilization by ensuring that it is steeper than the slope of In practice, this causes the rate of increase in the measured magnetic flux to be greater than the fall. The result of adding the flux stabilization signal is that the measured magnetic flux reaches its threshold relatively quickly, thereby leaving time for the decaying flux to drop sufficiently. In this way, the effect of the flux stabilization signal maintains the rise and fall of the magnetic flux produced by the phase windings during the electrical cycle, which is substantially equalized by ensuring that any errors embedded in the waveform decay from cycle to cycle. will do</p><p>Another consideration is speed variation, as it will be clear that the period of the conduction cycle decreases the increasing rate. To maintain the effect of stabilization, m<sb>stab</sb>needs to be a suitable function of velocity. </p><p>If the unit resistance of the device is not high, the degree of stabilization required for a given device is directly tied to the level of magnetic flux required. It makes a larger difference between mu and md. As a practical matter, the low resistance of typical devices effectively results in a slope m<sb>u</sb> and m<sb>d</sb>Slope m as the load on the instrument increases to make it practically constant irrespective of the load.<sb>u</sb> and m<sb>d</sb>each will change by only a small amount. However, ripple in the DC link voltage creates, in some cases, an uncertainty that needs to be addressed by stabilization. That is, the stabilization is m due to ripple<sb>u</sb> and m<sb>d</sb>It needs to be increased to prepare for the chaos in </p><p>For example, a DC link voltage of 750V, a voltage drop across the power semi-conductor switches in the controller of 3V, a diode voltage drop of 1.7V and a winding resistance (iR) voltage of 5V. Consider a switched reluctance device with a winding resistance voltage drop. </p><p>To boost the magnetic flux in the electrical cycle, the available voltage is:</p><p>750 - (2x3) - 5 = 739V</p><p>That is, the supply voltage is less than the voltage drops across the windings and the switches being driven. </p><p>To pull the magnetic flux down in the electrical cycle, the available voltage is: </p><p>750 + (2x1.7) + 5 = 758.4V</p><p>That is, the supply voltage boosted by the recirculating diodes and the voltage induced in the circuit across the winding. </p><p>at that time,</p><p>- The maximum rate of flux increase is 739/60 = 12.32 [Weber/sec]. </p><p>- The maximum rate of flux reduction is 758.4/60 = 12.64 [Weber/sec]. </p><p>So, the stabilization needs to be large enough to boost the distinct rate of flux increase by the difference between these two figures, i.e. 0.32 [Weber/sec]. </p><p>For illustrative purposes, using a machine operating at 5000 rpm and having an eight-pole rotor, the electrical duration of the conduction cycle is 1.5 msec. From this, m<sb>stab</sb>The size of is required to be at least 0.48 mWebers (= 0.32 x 0.015), which is 2% of the reference value. It is necessary to further include an amount of magnitude to prepare for other disturbances in the system, such as the ripple voltage discussed above.</p><p>In practical situations, the speed range over which continuous current mode operation is required is often quite limited. So, one possibility is to use a fixed initial magnitude of stabilization across the speed range. At higher speeds, the dynamic response of the device will theoretically be adversely affected. However, this is likely to be unacceptable for many applications.</p><p>The most difficult part for stabilization occurs at maximum phase current, where the ohmic voltage drop across the winding is greatest and the resulting flux waveform becomes non-linear. An ideal solution for non-linear flux increase and decrease would be a stabilization curve with a non-linear profile harmonized to the flux waveforms, but still with a period equal to the period of repetition of excitation of the phase winding as mentioned above. (stabilization curve). This may be implemented if necessary at the expense of more complex circuitry and/or algorithms.</p><p>The nature of the flux stabilization signal can change dynamically because it manipulates the state for device changes. It should also be noted that the above-mentioned magnetic flux control scheme should be limited to be used only in a mode, which is particularly advantageous for the control of a continuous current mode. It is equally applicable to discontinuous flux operation at lower torques and speeds. When it is used for the control of the continuous current mode, it may be found beneficial to innovatively address the flux limit prior to the initiation of the continuous current mode. So, there is no break between discontinuous and continuous current/flux modes.</p><p>In general, the disclosed embodiments use a flux stabilization signal that periodically coincides with the conduction cycle of the device. The flux waveform is superimposed on the value of the flux demand signal or measured flux signal so that the effective value of changing the flux dropped in one phase falls during the phase state cycle. In fact, any disturbance in the system that will cause innovative magnetic flux growth or erratic output is such that an appropriate portion of the conduction cycle (and therefore a larger period of time) will ensure that the magnetic flux is stabilized by actuation of the switches. It can be compensated for by a changing value that changes the rising flux to the falling flux, so as to be useful to be controlled.</p><p>The embodiment shown in Figures 8 and 9 uses the measurement or evaluation of the continuously available magnetic flux. In many systems operating with sensorless position detection, this signal is already present and thus can be used further for flux stabilisation. If, however, the system derives the position information from the physical position transducer described above, the flux signal may not be available and needs to be specially generated for flux stabilization. Typically, this will be done by integrating the phase voltages, requiring sufficient processing power to obtain a reliable signal. Another embodiment is shown in FIG. 10 as implemented in the controller 90 of FIG. 4 when aligned to control the machine 10 according to a signal from a physical rotor position transducer. This avoids the need for continuous flux information with a consequent reduction in processing power required. </p><p>Rather than using successive derivation of the phase flux, this embodiment provides a snapshot of the flux at a particular moment and then estimates to find the appropriate turn-off time. In Figure 10, the signal from the rotational position transducer (or some functional equivalent) indicates that the phase winding is t<sb>o</sb>It is hypothesized that it can be used to make it excitable in The current can be recorded and its value stored, the value being zero if the instrument is in single-pulse mode, and the instrument is in continuous current mode If it is at , it is not a spirit. From knowledge of this current and turn-on point to, the magnetic flux can be determined from the magnetic properties of the device. This value of magnetic flux is stored. The timer is t<sb>o</sb>The sequence of steps that are set to be driven from and can be performed according to this embodiment is shown in FIG. 11 . </p><p>When Lmin is reached as signaled by the rotor position transducer, the current is measured again and the magnetic flux is evaluated again. L<sb>min</sb>Since the current in V is in fact linear with the magnetic flux, simple calculations can be used if desired, or rather a look-up table of magnetic properties. This makes the magnetic flux value fLmin. magnetic flux stabilization value f<sb>s</sb>is the point f<sb>1</sb>is added to this value to give point f<sb>1</sb>t through<sb>0</sb>By predicting from , the intersection with the flux demand is t<sb>2</sb>is determined in This is taken as the turn-off point for the phase.</p><p>Any other predetermined position of the rotor may be used to the same effect, as long as this embodiment uses the convenient position of . For example, as a practical matter, it is sometimes necessary to "skew" the RPT by leading it or lagging it to account for an operation or device characteristic. So for example the RPT output may not actually match an event like Lmin. This will still be a convenient predetermined rotor position for the purposes of the present invention.</p><p>In situations where the flux increase is linear (ie, the voltage (iR) drop across the windings is small, typically large and for high-voltage devices), a linear extrapolation is appropriate. For other cases, an appropriate non-linear interpolation routine may be used. For example, there is a quadratic interpolation.</p><p>Fig. 10 shows this method in addition to the method shown in Fig. 8 and illustrates the equivalent of the two embodiments. In this embodiment, a stabilization term is added to the calculated flux rather than subtracted from the flux demand signal. As noted above, an advantage of this embodiment is the elimination of the need for successive integration of the applied phase voltage to make a flux estimate. Only one timer is needed with a simple storage and counting process.</p><p>11 shows a preferred flowchart for executing the process of this embodiment. The process is performed in an appropriately programmed controller such as labeled 90 in FIG. 4 where signals from a hardware rotor position transducer (RPT) are supplied in an essentially conventional manner. The process is followed by a phase energization point t according to the signal from the RPT.<sb>0</sb>Start at step 110 by testing to see if n is reached. If not reached, control will loop around the beginning of the process. If the phase energization point t<sb>0</sb>When n is reached, step 112 already samples the current in phase, for example using the current transducer 44 shown in FIG. 4 . its value, and t<sb>0</sb>The corresponding rotor angle is used to evaluate the magnetic flux. The timer function in the controller is started and the phase is excited by connecting it to the voltage supply by means of switches in the power converter 48 . In step 113, a test is set up to determine if a minimum inductance position Lmin has been reached when the rotor position transducer (or equivalent) is signaling. If it has been reached, in step 114 the current is sampled again and stored and the flux is evaluated. The flux stabilization term is the flux f<sb>1</sb>It is also added to the calculated flux to give a value of . </p><p>Step 115 is t<sb>0</sb>the value of the magnetic flux at and f<sb>1</sb>Take a value and time t at which the flux is expected to reach the required flux level<sb>2</sb>Estimate according to the chosen estimation method to find The timer continues to run and step 116 indicates that the timer is t<sb>2</sb>decide when to reach Step 117 switches off the phase so that the magnetic flux (and hence the current) decays towards zero. Control loops to start the next cycle at 118.</p><p>The invention in various embodiments provides a magnetic flux control technique for various types of electronically switched brushless machines, and is particularly suitable for switched reluctance machines. It creates a real-time signal of the magnetic flux in the magnetic circuit of the device, in which the reference value of the magnetic flux can be compared to control the device output in the presence of electrical noise and other disturbances in a stable manner. The disclosed embodiment demonstrates an important advantage, in which additional control schemes can be incorporated into an existing SR control system with minimal changes and without redefining existing sets of control data, It remains a valid outside of continuous current mode. The advantages of magnetic flux control discussed above are that only two commutation events per electrical cycle (one switch-on event and one switch-off event) - as in the case of conventional angle control, are Likewise - available now with the added benefit of being still. It will be apparent to those skilled in the art that various modifications and variations are possible without departing from the invention. The invention should be limited only by the claims that follow.</p>