WO2019021138A2

WO2019021138A2 - A threshold detector circuit for lossless switching

Info

Publication number: WO2019021138A2
Application number: PCT/IB2018/055433
Authority: WO
Inventors: Raymond Peto
Original assignee: Quepal Limited
Priority date: 2017-07-25
Filing date: 2018-07-20
Publication date: 2019-01-31
Also published as: GB202002624D0; GB2579931A; GB201711925D0; WO2019021138A3; GB2565760A

Abstract

A threshold detector circuit is described which is suitable for lossless switching of a transistor. The circuit senses a voltage across the transistor and is operative such that when the voltage across the transistor exceeds a user defined HIGH threshold, the threshold detector circuit remains in a quiescent state with a low input impedance to a signal to the detector circuit. When the voltage across the transistor (collector-emitter or drain-source) is less than a user defined LOW threshold, the threshold detector circuit switches to an active state, so that the transistor is switched on thereby creating a low input impedance to a signal to the detector circuit. The continual isolation of the detector circuit from the transistor ensures that there is minimal noise passing to/from the transistor.

Description

A Threshold Detector Circuit For Lossless Switching

Field

The present invention relates to a threshold detector circuit for lossless switching and more particularly but not exclusively to a circuit whose operation and timing control strategies, and their implementation in resonant power conversion topologies, achieves maximum efficiency and reliable operation.

Background

With the ever increasing desire to improve the efficiency of electric power conversion equipment, every aspect of the process needs to be scrutinised. Conventional electric induction motors are said to consume approximately 70% of all the electricity used in industry and about 45% of all the electricity used globally.

Prior Art

In many electric induction motor drive systems, power step-down converters are used as a variable voltage element. Conventional motor drives rarely incorporate a variable voltage element prior to a 3 phase pulse width modulation (PWM) half bridge circuit directly connected to the motor being controlled. Instead they rely on the mark space ratio of the PWM and the inductance of the motor itself to generate a current waveform of a particular amplitude and frequency in the motor. This kind of pulse width modulation is known as 'hard switching'. In other words, the switching device is turned off with current still flowing in it. Switching devices are also turned on, when a current may be flowing elsewhere, for example the reverse recovery time of a diode maintaining the inductor current during the off period of the switch.

If the timing is such that the inductor current drops to zero while the switching device is still off (discontinuous operation), then resonance occurs. However this resonance is unlikely to be of sufficient amplitude for the voltage to reach a negative value required for threshold detection to come into operation. Its amplitude depends on several variables all being met or exceeded.

US201 1/0181210 (Sanyo Electric Co) teaches a signal generating unit that generates a drive signal to deliver a positive current and a negative current to a coil. A non-conducting period occurs during conducting periods. A current is generated in response to the drive signal and supplies drive current to a coil. A zero cross over detection unit sets a detection window for avoiding detection of zero cross over of voltages other than the induced voltage. US 4525638 (Motorola) teaches a zener voltage threshold detector. The voltage at which a switching transistor turns on is determined by the breakdown voltage of a zener diode coupled between ground and the base of the switching transistor in conjunction with the base-emitter voltage of the switching transistor itself. In order to render the threshold detector circuit immune from noise at a trip point, a portion of the collector current of the switching transistor is supplied to a second transistor which when turned on reduces the voltage at the base of the switching transistor.

JPH05335908 (Mitsubishi) teaches a zero cross detector which eliminates noise in an output waveform of a trailing edge detection circuit and a leading edge detection circuit and outputs the result as a zero cross signal to the detector.

At best the nearest point to the zero, at which the voltage minimum can be detected, (the so- called valley point) and therefore the voltage across the device (at switch on) is at a minimum. This results in switching, ideally with less losses, than hard switching but it still has significant losses due to the repeat charging/discharging of stray or intentional capacitances.

The invention arose to provide an improved threshold detector that detects the exact moment at which a resonant waveform causes a switching device in its non-conducting state to go from a positive voltage across the switching device to a negative voltage and to switch this device on.

Summary of invention

According to a first aspect of the invention there is provided a threshold detector circuit for lossless switching of a transistor which senses a voltage across the transistor (collector- emitter or drain-source) and is operative such that when the voltage across the transistor exceeds a user defined HIGH threshold, the threshold detector circuit remains in a quiescent state with a low input impedance to a signal to the detector circuit; and when the voltage across the transistor (collector-emitter or drain-source) is less than a user defined LOW threshold, the threshold detector circuit switches to an active state, so that the transistor is switched on thereby creating a low input impedance to a signal to the detector circuit, thereby continually isolating the detector circuit from noise from its connection to the transistor.

Ideally the transistor remains switched on until it receives a reset command from a controller. An advantage of this is simplicity of control and the ability to use cheaper components with wider tolerance. During the quiescent state the low input impedance to the signal to the detector circuit from the transistor, is provided by a switchable low impedance circuit; and in the active state, the low input impedance to a signal to the detector circuit from the transistor, is provided by the switch when the switchable low impedance circuit is disabled. A benefit of this is that the detector circuit consumes very small amounts of current in either state.

As the circuit device is already at zero potential at the moment of switch on, there are no 'Miller' capacitances that need to be dealt with, thus reducing the amount of gate charging power required per cycle.

Ideally the threshold detector includes a switch control circuit connected across input terminals of a motor.

The detection circuit is ideally operative at the instant at which, in a resonant waveform, the natural resonance causes a switching device in its non-conducting state, to switch from a positive voltage across the switching device to a negative voltage. In resonant control topologies, this state is detected and the switching device is then turned on.

Preferably the circuit is included in a high frequency driver circuit, for electrical energy conversion apparatus used in either a quasi-sine or pure sine motor drive.

The threshold detector has a variable input impedance to resolve the dual problem of avoiding nuisance detection of noise and ultra-low current consumption. This is done with a view to migrating this threshold detector and switching device driver circuit to an integrated circuit. The high voltages to be monitored are isolated by a high voltage breakdown series diode.

When the threshold detection circuit is in its off mode, the input impedance of the threshold detector is very low and so is immune to transient noise causing an unexpected turn on event. As the voltage across the threshold detector goes negative, the input series diode becomes forward biased and the low impedance of the threshold detector is overcome by an even lower impedance of the voltage across the switching transistor.

At this instant the switching device is turned on and the input impedance of the threshold detector is set high to conserve energy. The circuit maintains the switching device in its on condition so that any noise activity on the threshold detector cannot turn the switching device off. When the switching device is to be turned off, the switching device is turned off quickly; the low impedance on the threshold detector is re-established and the possibility of repeat turn on for the next short period, while the switching device is turned off, is masked out typically by the width of a turn off command pulse itself. Ideally the threshold detector and its associated switch driving circuitry are now switched to its original state and awaits a next turn on resonance event.

Preferred embodiments of the invention will now be described with reference to the Figures in which:

Brief Description of the Drawings

Figure 1 illustrates a block diagram of a quasi-sine resonant drive;

Figure 2 illustrates a more detailed circuitry of power components of the quasi-resonant drive of Figure 1 ;

Figure 3 shows one phase of a waveform applied to a motor running at high speed with retarded turn on of a complementary switch;

Figure 4 shows one phase of a waveform applied to a motor running at high speed with minor complementary switch turn on retardation;

Figure 5 shows one phase of waveform applied to a motor running at high speed with complementary switch turn at ideal instant;

Figure 6 shows one phase of a waveform applied to a motor running at high speed with premature complementary switch turn on;

Figure 7 shows a block diagram of a quasi-sine motor drive;

Figure 8A shows a starter circuit for the self-adjusting drive circuit;

Figure 8B shows spare components and inter-wiring information for the self-adjusting drive circuit;

Figure 8C shows a reset circuit for the self-adjusting drive circuit;

Figure 8D shows a threshold detector and logic circuit for the self-adjusting drive circuit;

Figure 8E shows an auxiliary gate driver output circuit (hard and soft) for the self-adjusting drive circuit;

Figure 8F shows a main gate driver output circuit (hard and soft) for the self-adjusting drive circuit;

Figure 9 is an overall view of a turn on circuit shown in Figures 8C and 8D; Figure 10 shows a circuit diagram of a starter isolation circuit for the self-adjusting drive circuit;

Figure 1 1 shows an auxiliary gate driver output circuit in its low output state for the self- adjusting drive circuit;

Figure 12 is a circuit diagram of part of the self-adjusting drive circuit for enabling soft turn- on or hard turn-on of a driven switch;

Figure 13A shows an under-energised resonant waveform and block diagram of inputs required to enable oscillation;

Figure 13B is an overall diagrammatical representation showing a sequence of events that enables oscillation to occur;

Figures 14A to 14H show portions of a circuit diagram of a self-adjusting control circuit;

Figure 15 illustrates a sequence of events that enables oscillation to occur where output voltage is 50% or less of input voltage;

Figure 16 illustrates current excursions for operation of a control circuit that allows net current in or out of drive connections of a driven circuit;

Figure 17 is a table showing current and efficiency values for different output loads;

Figure 18A is an example of a circuit used for sensor-less current measurement;

Figure 18B shows a theoretical waveform output obtained with the circuit in Figure 18A;

Figure 18C shows a theoretical waveform output obtained with the circuit in Figure 18A;

Figure 19 shows an actual waveform output obtained over more than one cycle, with the circuit in Figure 18A;

Figure 20 shows an actual waveform output obtained within one cycle, with the circuit in Figure 18A;

Figure 21 shows an actual waveform output obtained for a single cycle, with the circuit in Figure 18A;

Figure 22 is a diagrammatical overall view of a whole motor system including: a drive module, a power drive system; and a motor system;

Figure 23 is a circuit diagram of a synchronous Buck converter; Figure 24 is signal timing diagram for the synchronous Buck converter in Figure 23;

Figure 25 is a circuit diagram of a first embodiment of a switching supply including a feedback controller;

Figure 26 is a circuit diagram of another embodiment of the switching supply without a feedback processor and with capacitor (C2);

Figure 27 is a circuit diagram of the switching supply without the feedback processor to illustrate operation of components with power switch Q1 switched on;

Figure 28 is a circuit diagram of the switching supply without the feedback processor to illustrate operation of components with power switch Q2 switched on;

Figure 29 shows signal timing diagrams for the switching supply;

Figure 30 is signal timing diagram for the start-up phase;

Figure 31 is signal timing diagram for the start-up phase which indicates a prohibited or indeterminate condition which is over-ruled during a first start-up cycle;

Figure 32 illustrates a voltage waveform of a conventional transistor at switch on;

Figure 33 shows an idealised voltage waveform of a conventional transistor at switch on;

Figure 34 is a diagrammatical view of a conventional power switch drive circuit arrangement;

Figure 35 is a diagrammatical view of an example of a power switch input circuit with hard switching;

Figure 36 is a diagrammatical view of an example of a power switch input circuit controlling a resonant switching device;

Figure 37 is a diagrammatical view of an example of an energy recovery power switch drive circuit (with its mid-point switched);

Figure 38 is an example of a circuit for achieving soft start-up;

Figures 39A to Figure 39D are graphs showing current and voltage profiles against time for a power switch input current using an energy scavenging circuit;

Figure 40 shows in block diagram form, an example of a resonant voltage conversion circuit; Figure 41 shows in block diagram form, an example of a resonant frequency conversion circuit;

Figure 42 shows in block diagram form, an example of a split function voltage control/frequency control quasi sine motor drive;

Figure 43 is an example of a three phase bi-directional (AC to DC or DC to AC) power supply;

Figures 44A to 44E are functional diagrams showing alternative system capabilities, handling energy flows, operating in different modes;

Figure 45 is a circuit for limiting in-rush current at phase input;

Figure 46 is an example of a circuit for limiting in-rush current between an active front end and a DC connection; and

Figure 47 is another example of a circuit for limiting in-rush current between an active front end and a DC connection.

Detailed Description of the Drawings

Figure 1 illustrates a block diagram of a quasi-sine resonant drive. Here it is shown that the output part of the circuit, consisting of the variable frequency stage, the slew rate capacitors and the motor itself forms a resonant circuit. In order for the system to operate correctly a self-adjusting turn on of the appropriate switch is required which occurs in this quasi sine form of output. Sensors may be shown connected to the motor giving an indication of speed. This can also give an indication of torque ripple if differentiated. Alternatively motor information can be calculated or derived from other measurable parameters. The variable voltage part of the circuit, shown at figure 1 , is typically also a resonant voltage conversion topology. By using these two techniques together, extremely high efficiencies can be obtained.

Figure 2 illustrates a more detailed circuit showing power components of a quasi-resonant drive circuit. Figure 2 shows a three phase half bridge frequency determining circuit with slew rate capacitors C6, C7 and C8 arranged in parallel with their outputs connected to the motor. The voltage amplitude of the generated waveforms at the output is determined by the variable voltage part of the circuit.

In operation, at the appropriate time determined by control circuitry (shown in Figure 1 ), one of the output drive transistors, Q3, (shown in Figure 2), is turned off quickly. The current that was flowing prior to switch off of Q3 transfers to charging or discharging slew rate capacitors C6 and C8 until the voltage across switching device Q4 becomes reverse biased, at which instant diode D4 switches to conduct. Diode Q4 may be either intrinsic or external to the now reverse biased switching device.

Control circuitry (shown in Figure 1 ), now turns on switch Q4, (shown in Figure 2) and maintains it on until it is switched off quickly. This repeats the resonant switching process. This same resonant process occurs on both of the other phases of the output; or as many phases that are appropriate for the motor/generator that is being controlled (Figure 2).

The operation of output circuit, the variable frequency circuit part of Figure 2, is essentially determined by a controller (Figure 1 ) which acts to force outputs to go off at a predetermined instant. Referring to Figure 2 switches Q3 to Q8 are switched on again by detecting the instant when the voltage across a switch is at zero potential, thereby ensuring no "shoot through" currents can occur. Therefore switch on occurs with no voltage potential across a switch. This ensures that there are no transient (voltage x current x dt) losses.

This type of operation, where the devices are turned off by the waveform frequency control mechanism (figure 1 ) and turned back on again by the natural resonance, ensures that all component values and tolerances are automatically taken into account in order to derive optimum input parameters to drive a system (motor), for every switching transition that occurs. Further, in one embodiment, this can be achieved without the need for a microprocessor type hardware or software burden.

The variable voltage element, shown in Figure 2 of the circuit, includes switches Q1 and Q2 and associated other components which are also operated in a resonant mode. As configured the variable Voltage circuit provides a voltage step down function from the supply across C1 .

Figure 3 shows one phase of waveform applied to a motor running at relatively high speed with turn on using an opposite (complementary) switch for example Q4 in figure 2 where the turn on signal to Q4 is too slow. Figures 3 to 6 show the importance and effect of correct timing of the turn on point of the opposite switch in relation to the turn off of the first switch. The Figures also show the slew rate clearly and, because the motor speed is fast, the opposite transition occurs and the cycle repeats itself. This sequence of events, of the transition from one voltage state to the other, is identical at different voltages, currents and frequencies. The sequence is also the same sequence that occurs in the variable voltage part of the circuit shown in figure 2 as it operates to maintain a given output voltage. In Figure 3 in particular, the timing of switching of the opposite switch Q4 in Figure 2 is shown as having been delayed from switch on at the right time.

Figures 3 to 6 show the effect of the resonance of the shunt capacitors C6, C7, C8 in Figure 2 during operation. Considering one of the 3 phase outputs to the motor in Figure 2, when Q3 is turned off, the motor phase current is being maintained by a combination of the displacement current in C6 and C8 and the forward conduction of the parallel diode D4 across the opposite switch Q4. When this current is reversed and flows into the motor and if switch Q4 is not closed then resonance occurs in the opposite direction of the voltage applied to the motor. When the switch Q4 conducts either before D4 is in conduction or after the motor phase current has changed direction then there is a very rapid change of voltage across switch Q4 that occurs at the same time with a large current pulse (not shown) in switch Q4. This current pulse can reach destructive levels if switch turn on time is very short and the switch impedance is low enough.

Even if when this potentially destructive switching is not a problem, the switching devices experience significant repetitive transient switching losses that are proportional to: volts x current x switching time. There are therefore significant issues with sharp edge of such voltage transitions with cable resonances, EMC radiation and dV/dt stress applied to motor windings.

Figure 4 shows one phase of waveform applied to a motor running at high speed with turn on of opposite switch, for example Q4 in figure 2, very slightly too slow. As the opposite device Q4 is turned on closer to an optimum instant, undesirable voltage transitions become smaller (lower amplitude) with consequent smaller unwanted harmonics.

Figure 5 shows one phase of waveform applied to a motor running at high speed with turn on of opposite switch Q4 at an ideal instant. The waveform is the ideal voltage waveform. It is important that the turning on of the opposite device Q4 occurs at some point when the motor current is still flowing through the forward conduction of the diode connected across the opposite device. Ideally if the switching device is also capable of conducting current in the reverse direction (for example a field affect device) then it is advantageous to turn the switching device on as soon after the forward conduction of the diode has taken place. This is beneficial from a losses point of view if the value of (reverse current x device on resistance) is less than the voltage drop across the diode at this current level. Also switching in this manner allows for ease of optimising the on switching time of the opposite switch Q4. Figure 6 shows one phase of waveform applied to a motor running at high speed with turn on of opposite switch Q4 very slightly too early before the resonance has forward biased D4. Here the opposite device has been switched on slightly in advance of the optimum turn on time. The resonance has not yet delivered the voltage to opposite device Q4 to drive it negative. When this occurs it gives rise to a high transient current in the opposite device Q4; causes excessive device switching losses: and is the source of problems associated with a fast edge, rather than the relatively smooth and slow edges associated with the resonant circuit operation. Note that the correct timing of the on switching event is not related directly to the frequency of the drive. The correct time to turn the switch on is at instant shown in Figure 5.

Figure 7 shows optimisation of the operation of controlling power in or out of a synchronous or non-synchronous motor/generator/alternator in order to achieve maximum overall efficiency (least losses) of the combination of the drive and motor/generator/alternator consistent with other desired parameters.

Figure 7 shows a block diagram of a quasi-sine motor drive. This shows the position of a self-adjusting switching device driver described as a self-triggering turn On' circuit. In Figure 7 there are 3 motor phases so there are 6 switching devices shown as 1 to 6. Each of these switches is controlled by 6 self-triggering turn on circuits which are more fully described with reference to the circuit in Figures 8A to 8F. It is possible to drive switching devices 1 to 6 directly on and off by calculating switching criteria. However the self-triggering drive circuit described inherently compensates for turn on timing for each switching event and thereby automatically takes into account variables that would make a calculation based decision too complicated and therefore too time consuming to perform. These variables include: coil, motor, shunt/resonance capacitor, speed, load, voltage, current, temperature or any combination thereof.

Where the individual switching device is shown, there may in fact be several devices in parallel. Under these conditions it may be possible to have (within the switching device drive circuit) one part that detects the instant to switch on the devices and one or more driver circuits, for example one driver circuit for each switch in a parallel arrangement. Furthermore some of the drive circuits need to be floating while others have a common connection and so in some configurations it may be possible to employ circuit redundancy and so save components, cost and weight. Also an overall control microprocessor identified as 'μ' may optionally be referenced to the low voltage common terminal connecting switches 2, 4 and 6 of the power circuitry thus eliminating a significant amount of unnecessary signal isolating components. Figures 8A to 8F depict automatic turn on circuitry to achieve optimum turn on timing of its associated power switching device. The circuit is used in an example of an automatic self- adjusting motor drive system.

Figure 8A to 8F shows a circuit diagram of a complete self-adjusting drive circuit. This design incorporates fundamental aspects of the turn on detection circuitry. It has one input (reset) that is basically a 'must turn off and stay off command input. The circuit in Figures 8 has one input (drain/collector) that measures the voltage across the switching device to be controlled. It has an optional input (starter) that allows the switching device associated with it to be turned on slowly irrespective of any reset and drain/collector status. The circuit in Figure 8 has power supply pins which are nominally at 12 volts. It has one or more outputs to enable the switching devices to be switched on or off.

Figure 9 is an example of a reset circuit of the type that may be used in the circuit detailed in Figures 8A to 8F. An important feature is the bi-stable element U1 a. A reset on pin 4 under normal running conditions overrides any status of the switching device itself. The wiring and polarity of the connection to the reset opto-coupler U6 is failsafe. The voltage detecting circuit for the drain/collector ideally has variable impedance. The advantage of this is for the detection circuit to present very low impedance to the switching device drain/collector while diode D8 is reverse biased. This eliminates the possibility of high frequency noise leaking through the reverse biased diode and causing a false zero volt detection to occur. This low impedance can only be overcome when the diode is properly forward biased which can only occur when the voltage across the switching device is about zero volts. The circuit itself operates at low voltage except for the cathode of diode D8.

Figure 10, Figure 11 and Figure 12 show detailed views of the starter circuit and gate driver used in Figures 8A to 8F. The way these work is that in mode A, gate of associated switching transistor is turned on slowly so that a current spike (from charging or discharging shunt (C6, C7, C8 in Figure 2 for example) or other resonant capacitors) does not cause a significant current spike in the associated switching device of such an amplitude that it stresses or in a worst case destroys the associated switching device. Actual one shot energy loss here is not detrimental to the associated switching device. With motors, all three phases can be started separately at reduced voltage (and at the same output voltage). Different conditions apply for quasi sine and pure sine drives.

The starter circuit in Figure 10 is operable even if the voltage conditions across the device are considered inappropriate for normal operation of the turn on circuit as shown in Figure 31 . Care must be taken to ensure that inappropriate operation of this circuit cannot occur. Figures 11 and Figure 12 are detailed views of a power transistor drive circuit used in Figure 7. In this particular implementation, the circuitry in Figure 7 is capable of operating two independent power switching transistors. The circuits in Figures 11 and Figure 12 have the capability of either mode a: soft high, or mode b: hard high, operation. In both cases Figures 11 and Figure 12 have a hard/low capability.

In the circuit of Figure 11 , the switch driver U5 has two outputs. One output is connected to pin 6 and pulls current out of the associated switching device to turn it off, thereby effectively driving the gate/base low. This output has a very low impedance and thus switches the associated switch very quickly, typically within around a few 10s of nanoseconds. High output from pin 7 is effectively off so no positive current can flow through R5 or R6.

For stability and control reasons it may be beneficial for the soft turn on, mode a, to only use one transistor in an output switch consisting of multiple parallel connected power devices. In this implementation both the circuits shown in Figure 11 and Figure 12 are capable of doing this. A simple logical modification allows only the circuit in Figure 12 to have a soft turn on mode a.

Figure 12 in mode a operation is a detailed view of a power transistor drive circuit used in Figure 12 with particular attention being paid to its soft turn on capability when there is a high voltage present across the device being switched on. This part of the circuit is used for the initial starting phase of resonant operation. A problem it overcomes is how to start a resonance system that has no inherent mechanism to do this, as turn on pulses are generated by the action of resonance itself once resonance has been established. Therefore simply enabling the transistors does not switch them on as none of the switches Q3 to Q8 (Figure 2) are unlikely to be sitting in a suitable quiescent state with no voltage across them.

Under these conditions, the voltage of the switching transistor is unknown and so the zero voltage detecting circuit is inhibited. When the soft start input is enabled, mode an operation, there are two switches in a totem pole like output configuration. The other switching device in the totem pole is turned off so there is no possibility of a shoot through condition occurring. D4 inhibits a so-called 'strong pull up'. This leaves R20 to limit the input current into the combined capacitance of both the input capacitance and the Miller or reverse transfer capacitance of the switching device as it turns on.

This slow turn on minimises the peak current that results from discharging the shunt capacitance across the switching device. In the sequence of operation of the start-up, the system firstly applies off pulses to both top and bottom transistors of the totem pole like output configuration, selects a transistor to turn on and applies a switch on current to the soft start input which bypasses latch U1 a in Figure 8D and applies a soft start pulse to the selected transistor. This soft on pulse bypasses the latch U1 a in Figure 8D and 'hard on' driver so as to apply a soft pulse in order to avoid capacitive current from the shunt capacitor from destroying the selected transistor.

To start the three phase quasi-sine system it may be necessary to initialise the system and set it up for commutation by connecting one phase of the motor to the positive rail and the other two phases to the negative rail in order to inject current into the motor to enable resonant commutation to commence once the appropriate 'off pulses are initiated. For the quasi sine implementation, for minimum components it is convenient to turn on one of the pull up output transistors as they tend to be at a high voltage potential and so require isolated drive capability. The other two legs of the 3 phase half bridge are switched to the common negative rail and therefore do not require extra isolation components.

Figure 12 in mode b operation, is a detailed view of a power transistor drive circuit used in Figure 12 with particular attention being paid to its hard (fast) turn on capability when there is a negligible voltage present across the device being switched on. Here pull up transistor Q3 is enabled, D4 is reverse biased, and R1 1 is switched high so a strong turn on current to the switching device is provided via R17. This is the normal turn on mechanism that is enabled by the voltage across an appropriate switching device reaching zero volts. This soft or hard turn on option is required for both the circuitry involved in driving and controlling the variable voltage stage Q1 and Q2 in Figure 2 as well as the circuitry involved in driving and controlling the variable frequency stage Q3 to Q8 also depicted in Figure 2.

Figure 13A and Figure 13B depict diagrammatically a sequence of events that enables oscillation to occur. When the circuit is at rest, all switching devices have their resets enabled. To start the resonant circuit it is initially required to generate pulses of a suitable duration and apply them to the appropriate switching devices while other switching devices not required for the initialising process are still held in their reset conditions. This tends to charge up inductors in the circuit with sufficient current to enable a positive voltage rail to negative rail excursion to be able to occur with the associated resonant capacitors shunted across the switching devices. At this moment the opposite switches have to be enabled so that when the original switching devices are turned off, voltage detection circuitry operates correctly by detecting a very near zero state and turn the opposite switching device on. From this point circuits (shown for example in Figures 2, 8A to 8F and 12 and Figure 13A) continue to resonate. The opportunity to allow for the turn on time of the drive circuit can be allowed for by triggering its inception at a voltage point in advance so allowing for delays. The strategy adopted is always to be turning off the appropriate devices when a target current is achieved. This is decided by the control system shown in Figures 14A and Figure 14B. The fact that devices are resonating gets it turned on again.

This method of commutation can be controlled by software. In such an embodiment there is little chance of 'shoot through' caused by uncertainty of device switching speeds and tolerances. Therefore this method of commutation eliminates all overlap and dead band timing issues that conventional switching systems suffer from.

Because of the way the resonant circuit (in Figure 13B) operates, any stray or inherent capacitance of the switching devices, motor, cable or inductors or any other components connected to node, in this case the junction of the common connection between Qt and Qb in Figure 13B that is being switched, is in parallel with an additional capacitance component required to make the circuit function correctly. A normal switching topology finds these stray and inherent capacitances to be significantly detrimental to idealised operation and so introduces a significant power loss as well as circulating currents and EMC issues.

The turn off pulse to switching device driver circuit Figure 8C reset on PL5 has to be of a duration that is sufficient for an associated switching transistor connected to PL2 in Figure 8E to be switched off (cease conducting) and so that voltage across the collector/drain of this switching transistor to have risen sufficiently so that the voltage zero detection circuit connected to Q drain/collector on Figure 8C does not allow this switching transistor to be turned on again when turn off pulse (Figure 8C) is removed. The duration of this pulse is very small compared to the pulse repetition frequency of consecutive reset pulses applied to reset PL5 in Figure 8C so this is relatively straightforward to implement. Additional blanking gating or status feedback of latch U1 a in Figure 8D could be reported back to control circuitry in Figure 14A and Figure 14B and could be used here if necessary for ultra safety critical requirements such as aerospace.

Figure 13A shows an under energised resonant waveform and block diagram of inputs required to enable oscillation. It is a requirement for correct operation of the resonant circuit shown in Figure 13B that at all times there is sufficient stored energy in the Inductance shown in Figure 13B to ensure correct rail to rail commutation. Occasional use of the soft input PL1 in Figure 8A could be used to provide the soft voltage change shown as Vadd in Figure 13A.

Figure 13B shows the power stage that is controlled by the control circuit Figure 14A-14H. Figure 14A to 14H shows a circuit diagram of self-adjusting voltage control circuit. This is used in the variable voltage section as shown in Figure 2. In order to derive the desired voltage, variable volts output, from the circuit in Figure 2, the control circuit (shown in Figures 14A to 14H) compares an output voltage with the required target voltage shown in Figure 14A. From this it ensures that off pulses, applied to the two switching devices Q1 and Q2, (shown in Figure 2) in order to maintain the output voltage at a desired level.

A reset circuit as shown in Figures 8A to Figure 8F for a resonating circuit (of the type shown in Figure 2) includes a control circuit operative to continually reset a latch in order to force the latch to an off state, whereby the condition of an output transistor, for example Q1 in figure 2, which is controlled by latch U1 a in Figure 8D is switched off. Shortly after switch off of transistor Q1 , the control circuit in Figure 8A to 8F isolates the reset signal so that transistor Q1 is not triggered on again until the voltage across device Q1 reaches zero again.

The circuit shown in Figure 14A and Figure 14B is analogue but it could be converted partly or completely to a digital domain if required. The fundamental aspects of operation would be unchanged. In Figure 2, the variable voltage supply output is capable of operation from one voltage rail to the other and has a full power bandwidth of several kHz in this configuration. The full power bandwidth can be increased to many kHz. The resonant operation used has no inherent high frequency limitation. The high frequency range is limited primarily by turn off speed of the switching devices.

The switching devices have to turn off completely, within a few percent of their rise time, which is dictated by slew rate capacitance and operating current. Switch off times slower than this tend to waste power in the switching devices as they have to handle a repetitive switching loss where there is both voltage and current present for a period of time in the switching device. The resonant operation overcomes this under normal conditions, effectively by bypassing the current that is present as the device turns off, into becoming the charging or displacement current of the resonant shunt capacitors both deliberate and parasitic.

The circuit in Figure 14A to 14H has several important features and functions. Controlling a resonant circuit so that it always resonates under all conditions of applied input power voltage, desired power output voltage and desired output current requires a radically different kind of control strategy. This is especially so if extremely high levels of conversion efficiency are to be achieved. Loop gain stability under all conditions has been one of the most difficult issues to control. This is particularly so where a high full power bandwidth is required as near to critical damping as possible whilst still maintaining operation of a resonant circuit.

Assuming that the desired target voltage is presented to the circuit (in Figure 14A) as a 0 to 5 volt value with a midpoint of 2.5 volts. Knowing what the power input mean voltage value is, then the centre point of the voltage output (which is half the voltage input) can be set by adjusting an 8 bit attenuator pad U4 and U13. Any error between the attenuated output voltage and the target input voltage is developed as an error signal from the differential amplifier U10b as shown in Figure 14A. This error is developed across C6 after modification for loop gain and the response time by the variable gain elements controlled by U14 a, b and c. This error voltage is amplified by amplifier U7a which nominally sits at 2.5 volts if the input and attenuated output voltages are identical. Any deviation from this tends to cause the voltage at U7a to vary from 0 to 5.0 volts.

It is important to consider the idling state of variable voltage resonant circuit shown in Figure 2 while it is running at a particular voltage output but where no net current is being drawn from its output terminal. The control circuit in Figure 14A to Figure 14H gives an output voltage that is exactly equivalent to the current flowing in the coils of the output inductor L2a,b in Figure 2. The circuit shown in Figure 14G is discussed in greater detail below and in combination with Figure 18A and Figures 18B and 18C. Under these conditions the timing of the off pulses to the switching devices Q1 and Q2 is arranged so that the current in the inductor builds up to a certain positive level at which point outputs from circuit U3a and b in Figure 14H toggle and coil current drops to zero and then increases to an equal and opposite value to the current on the previous half cycle.

At the corresponding negative point the outputs of U3 a and b again toggle and the coil current in L2a,b in Figure 2 becomes less negative, crosses zero and increases to its original positive value again. This cycle then repeats. Any minor errors in currents and timing result in the variable volts output voltage drifting up or down and the voltage discrepancy causes the off times of each switching device Q1 and Q2 to be altered slightly. This tends to force the variable volts output voltage to its correct value. This continual oscillation and shuttling of current back and forth has sometimes been considered wasteful but the inherent losses are so small in this kind of resonant topology. Even with relatively small output powers, in relation to full power capability, the overall efficiency is extremely high. The offset circuit comprises two comparators U3a and U3b in Figure 14H which compare two adjustable voltages so that when power is required, a feedback controller U7a offsets the voltages by a predetermined amount in order to derive more/less power which is proportional to the offset. Because of the unusual topology and the control strategy adopted the variable volts output circuit shown in Figure 2 operates in all four quadrants. In the circuit in Figures 14A and Figure 14B, a careful analysis of the voltage outputs of U5a and b, U7a and b, and the networks on pins 2, 3, 5 and 6 on comparators U3a and U3b identify that the outputs 1 and 7 of U3 provide the correct off pulses when required.

Figure 15 depicts the sequence of events that enables oscillation to occur where output voltage is 50% or less of input voltage. A particular problem that has to be overcome in this resonant topology is that when current flows from the output, say the variable volts output in Figure 2 and the output voltage is at about 50% or less of V_max, where V_max is the voltage at A, there is required an injection of negative current (l_rev) introduced into the resonant inductor L2a,b for resonant commutation to occur. Without this negative current, there is not enough energy to resonate shunt capacitors C4 and C5 so that the opposite switching device is reverse biased sufficiently to trigger the on pulse and ensure lossless commutation to continue.

This negative current v needs to be increased as the output voltage decreases. This negative current is derived from the specific operation of amplifier U5a and depends on the output polarity of U7b and the network values at input of comparator U3. The configuration in Figure 14A and 14B also allows for normal operation where output voltage is 50% or higher and the current flows into the output terminal.

Figure 16 shows the coil L2a,b current excursions for operation of the control circuit Figure 14A and Figure 14B for allowing net current in or out of variable voltage output connections in Figure 2. When the variable voltage circuit in Figure 2 is in its resonant idling mode, the current flowing into and out of resonant inductor L2a,b is equal and opposite, thus giving no net current flow either into or out of the variable volts output terminal. As current is drawn from the variable volts output terminal, the output voltage tends to fall and feedback circuit U10b, U7a and associated components acts to offset the switching points at turn off for comparators U3a and U3b. In turn these results in a higher value of positive current, and a lower value of negative current, flowing thus giving the desired net current outflow. Initially the control circuit in Figure 14A and Figure 14B, attempts to maintain the same overall switching frequency, typically up to the point that the coil current doubles in the positive direction and drops to zero on the negative phase of the cycle.

If more current is required then the positive current increases to its maximum of, say 4 times peak idle current, while still maintaining its minimum current at zero. Consequently when more power is required it is necessary to reduce the input frequency of this system. This is helpful as the coil losses increase due to the higher currents but are reduced due to the lower frequency of operation.

As the circuit operates in all four quadrants, for full current in the negative direction the triangular wave is displaced below the zero current line.

Figure 17 shows the current and efficiency calculations for the level of loading on an output. Figure 17 shows how this unusual effect of efficiency versus loading characteristics that this resonant topology achieves. This efficiency is aided by altering the internal current ramp offset and the frequency change under load to achieve the efficiency range.

Figure 18A is an example of a circuit used for sensor-less current measurement. It overcomes the use of conventional current measurement techniques which introduce series resistance and amplifiers. Alternatively the circuit in Figure 18A may be used to measure the passage of current by the use of a Hall effect principles. It can be difficult to get at the current to be measured itself because of voltage offsets or high dV/dt issues superimposed on the current. As a consequence there can be significant power losses as well as measurement time delays which impinge on loop stability and limit the speed of operation of the circuitry.

The use of physical current measuring techniques can also cause problems in layout of the circuit and some of the sensors themselves are sensitive to stray electrical and magnetic fields. From this evolved the necessity to develop an improved current measuring technique.

The block diagram in Figure 18A illustrates where a voltage at the coil terminal A is significantly higher than the voltage at B, then the current into the capacitor C1 at B is closely approximated to (VoltsA/R1 ). From this it can be calculated that the change in coil current dl is equal to ((C1 x R1 )/L) x dVd where Vc1 is the voltage change across C1 .

So ignoring any offset issues, the voltage waveform across C1 is identical to the current waveform in the inductor L. However it is necessary to include a voltage clamp on this waveform at every ½ cycle, as an integrator has no absolute DC value.

The next stage in the sensor-less current measuring method is to determine the absolute values of the current and these can be determined at each time the coil current in L2a,b in Figure 2 resonates shunt capacitors C4 and C5 across switching devices Q1 and Q2. At this point the inductor current can be determined from the rate of change of voltage at the instant of commutation.

(I inductor/C shunt) = dV/dt at point A. Cshunt = Cstray + Cdevices + Cdeliberate

Passing the switching edge at point A in Figure 18A through differentiator C2, R2 and associated amplifier, the voltage across C2 and R2 gives a voltage at C which is proportional to the absolute current in inductor L2a,b at the point of switching off. This voltage is then used to restore the direct current (DC) current waveform at B by operating switch D at the zero cross of voltage A.

Figure 18B shows a theoretical waveform output obtained with the circuit in Figure 18A. Here inductor L2a,b has two different current values. The first is at negative going transition and is at a relatively low current. Therefore the slew rate of shunt capacitors C4 and C5 is relatively long. This lengthy slew rate translates to a relatively small voltage at C. The second transition occurs at a high inductor current and so the slew rate is relatively fast. This faster slew rate translates into a much narrower, but higher amplitude, pulse at C.

Figure 18C shows a theoretical waveform output obtained with the circuit in Figure 18A. Here the relationship between the voltage at B and the inductor L2a,b current related voltage C is DC restored at the clamping point D.

Figure 19 shows an actual waveform output obtained with the circuit in Figure 18A. It shows the voltage at C derived from the inductor current at the switching point superimposed on the voltage waveform at B.

Figure 20 shows an actual waveform output obtained with the circuit in Figure 18A. Figure 20 shows an expanded trace of the voltage waveform at B superimposed with inductor voltage at A as it crosses the zero volts axis. It can be seen that a DC restore clamping point, shown as two vertical lines in Figure 20, derived from this zero point is the optimum point to perform DC restoring as it occurs at the peak of voltage waveform at B.

Figure 21 shows an actual waveform output obtained using the circuit in Figure 18A. Referring to Figure 19C this illustrates the desirability of performing DC restore at the inductor zero voltage point. Here a very non symmetrical inductor voltage A is used and the zero crossing point very closely identifies the peak of the current waveform at B.

Figure 22 In order to understand how a motor drive system is considered, convention has arranged boundaries for the motor in context with the power connection to supply the power for it. Figure 22 shows the boundaries of a complete drive module (CDM), a power drive system (PDS) and a motor system comprising the motor itself and the attached mechanical load. This is included as a requirement of "CE Marking and Technical Standardisation Guidelines" for application to electrical power drive systems. The relevance here is that the overall efficiency of the techniques described is to be read and understood in the context of the 'motor system' in this guide.

It is recognised that further development of an existing pulse width modulator (PWM) drive for motors, which already encounters and creates significant technical obstacles, results in even greater problems to be overcome in order to get it all to work correctly and these further developments may have other undesirable effects as well. In order to introduce the advantages of newer transistor materials, such as silicon carbide (SiC) and gallium nitride (GaN) switching speeds are increasing and associated switching edges are becoming sharper. This more rapid switching imposes greater constraints on design and requires drives to be more complex, mainly due to greater parasitic impedances; as well as subjecting the motor to even more aggressive waveforms than existing ones (that already cause considerable problems in motor design and installation) as a consequence of EMC.

By adopting a drive based on the fundamental principles outlined herein it is possible to revert to lower cost motor materials and also materials that give superior performance as they are only subjected to a fundamental frequency. Motor design is intended to ensure the motor runs on its fundamental frequency without having to compromise its design to cope with the issues of pulse width modulation. Improved design also allows the use of lower quality (and therefore cheaper components) and switched or variable reluctance motors (which would allow for the fundamentally cheaper and physically toughest motor design that switched or variable reluctance motors offer compared to induction or permanent magnet motors) as existing torque ripple problems are overcome.

Figure 23 shows an example of a synchronous Buck converter which comprises a power switch illustrated by transistor Q1 and an auxiliary switch illustrated by transistor Q2. A DC supply provides a constant voltage Vin. Output stage consists of an inductor shown as coil L and an output capacitor C4 in series. A load impedance, Zi_oad, is connected in parallel with the output capacitor C4. The junction between Vin positive and the power switch Q1 is referred to herein as the top rail. The voltage of the top rail is Vin. Junction at the output to the auxiliary switch Q2 and the negative of Vin is referred to herein as the bottom rail. The voltage of the bottom rail is ground in many, but not all, applications. For the purposes of the present embodiment the voltage of the bottom rail is zero.

Referring again to Figure 23, the mutual junction of switch Q1 , switch Q2, and coil L is designated junction Q and the voltage at this junction is VQ. The junction of coil L and output capacitor C4 and load impedance Zload is designated the output junction. The voltage at this junction is designated Vout. The current passing from junction Q through coil L to the output junction is designated IL.

Connected in parallel across switches Q1 and Q2 are protection diodes D1 and D2. Protection diode D1 is in parallel with switch Q1 and protection diode D2 is in parallel with switch Q2. Protection diode D1 is arranged to block current if the voltage at the top rail is higher than the voltage at junction Q. Current only flows through diode D1 if the output voltage VQ is greater than Vin + D1 diode forward voltage drop across Q1 .

Protection diode D2 is arranged to block current if the voltage at junction Q is higher than the voltage of the bottom rail. Current only flows through diode D2 if VQ is less than the bottom rail voltage less the D2 diode forward voltage drop voltage across Q2.

As illustrated in Figures 23 to 31 inclusive, the current IL flowing through the inductor L is considered positive when it flows from junction Q to the output junction. That is inductor current IL is said to be 'forward' when it is flowing from junction Q to the output junction. The current IL flowing through the inductor L is considered negative when it flows from the output junction Q to junction Q. That is inductor current IL is said to be 'reversed' when it is flowing from the output junction to junction Q. If the inductor current IL is said to be increasing positively, it means that its magnitude is increasing while it is flowing forward. If the inductor current IL is said to be "increasing negatively", it means that its magnitude is increasing while it is flowing reversed.

Figure 24 shows a signal timing diagram for the synchronous Buck converter. It shows the way that voltages and currents change in this circuit over time. Using Figure 23 for reference, initially the voltage VQ and Vout are zero; the top rail voltage is Vin; the bottom rail voltage is zero; current IL is zero; and switches Q1 and Q2 are both off. In the first mode the power switch Q1 is turned on and the auxiliary switch Q2 is off. Then voltage VQ is equal to the top rail voltage Vin. If switch Q1 is a transistor, then voltage VQ is not exactly equal to Vin due to semiconductor effects. The current IL, through the inductor, rises. This rising current charges the output capacitor C4. The voltage V_out rises. Upon reaching an upper desired level for IL, power switch Q1 is turned off and auxiliary switch Q2 is turned on.

In the second mode the power switch Q1 is off and auxiliary switch Q2 is switched on. Voltage VQ is equal to the bottom rail voltage which is zero. If switch Q2 is a transistor, then voltage VQ is not exactly equal to zero due to semiconductor effects. The current IL through the inductor falls because the voltage Vout is higher than VQ. Although the current IL through the inductor is falling, it is still flowing into output capacitor C4 through output junction. Therefore the voltage on the capacitor C4 continues to rise initially. However if auxiliary switch Q2 is left on long enough, the current through the inductor L eventually drops to zero. Therefore the voltage at the output junction Vout keeps rising until the current through the inductor L reaches zero, at which instant voltage Vout stops rising.

If continuous operation of the circuit in Figure 23 is required, it is important that the coil current is not allowed to fall to zero, and under these conditions auxiliary transistor Q2 is turned off and Q1 is switched on again while current is flowing through the coil L, thereby enabling the cycle to repeat. Note that this means there are problems associated with this such as reverse recovery losses in D2 and switching losses in Q1 due to the simultaneous presence of voltage and current in Q1 as Q1 is switched.

If discontinuous operation of the circuit in Figure 23 is required, the coil current is allowed to fall to zero; at which point in time the auxiliary transistor Q2 is turned off and transistor Q1 is turned on thereby enabling the process to repeat. The first mode is then repeated with the power switch Q1 on and auxiliary switch Q2 switched off.

By continuous repeated operation of the first and second mode of operation the output voltage Vout rises to the desired voltage and is maintained around the desired voltage by the controller in Figure 23 adjusting the drive timing to transistors Q1 and Q2 on and off thereby effectively adjusting the inductor current value IL.

An example of a prior art drive circuit is described by Panda, Pattnaik, and Mohapatra in the International Journal of Power Management Electronics, Volume 2008, Article ID 862510, in the article entitled "A Novel Soft-Switching Synchronous Buck Converter for Portable Applications". Such prior art switching converters suffered from: auxiliary switches being turned off whilst they conducted current. This resulted in switching losses and EMI. The power switch that is described with reference to Figure 23 and is one of the preferred embodiments described herein.

It operates with higher peak current stress and more circulating current, as well as active and passive circuits that are more complex than existing power circuits.

Although developments in switch mode supplies have resulted in designs of considerable ingenuity, they normally suffered from increased complexity, cost or exotic components.

Figure 25 shows an embodiment of the switching supply in addition to the elements and connections of the synchronous Buck controller. The circuit also comprises a first switch capacitor C1 connected in parallel across the terminals of the first switch Q1 ; a second switch capacitor C2 connected in parallel across the terminals of the second switch Q2; and a rail capacitor C3 connected between the top rail and the bottom rail.

The switching supply also comprises a feedback controller. The feedback controller receives inputs. The switching supply sends a control output, that is based on the values and timing of the inputs, which turns the switch Q1 on or off or and sends a control output signal which turns the switch Q2 on or off. Figure 26 shows an embodiment of the switching supply, similar to that shown in Figure 25, however there is no first switch capacitor C1 present. Therefore there is no first capacitor C1 connected in parallel across the top rail and junction Q.

The operation of the switching supply according to the invention is described below. The circuitry has to operate in several different modes. There are a defined set of principles that need to be followed to start the circuit correctly. There is also a second set of principles that are required to operate the circuit at steady state with an output voltage less than half the input voltage. Furthermore there is a third set of principles to operate the circuit at steady state with an output voltage greater than half the input voltage. These principles are related to the overall current flow in L.

For the sake of simplicity, the circuit shown in Figure 26 is discussed in detail below. Capacitors C1 and C2 both are effectively in parallel one with another and are connected across either switching transistor Q1 and Q2. They are represented as two capacitances so that there is effectively a capacitor connected to each switching device Q1 and Q2 so as to minimize the region and physical area of circulating currents during device switching events. For minimum electromagnetic interference issues the path taken and consequent area of this path are important. Assuming the value of C2 is the parallel value of C1 and C2 in Figure 25.

The startup of circuit in Figure 26 is now described with reference to Figure 30 and Figures

31 for the first cycle of startup. Initially the voltage VQ and Vout are zero; the top rail voltage is Vin; the bottom rail voltage is zero, and current IL is zero. Switches Q1 and Q2 are off. The voltage across C2, which is effectively in parallel to both Q1 and Q2, is also zero. In the first mode, mode 1 , the power switch Q1 is turned on and the auxiliary switch Q2 is switched off.

In a preferred first mode the power switch Q1 is turned on 'softly'. That is power switch Q1 is partially opened to let current slowly seep through at a rate of typically a small fraction of the rated current of Q1 . Note that consideration of the second breakdown characteristics of Q1 need to be allowed for during this slow turn on transition. If Q1 is a transistor, turning it on softly means that its resistance is gradually decreased. The advantage of the soft start is low in-rush currents and less stress on switch Q1 . It can be seen that Q1 has to provide a charging current for C2 to charge from zero volts to the top rail voltage. A fast turn on here may lead to a potentially destructive high peak current in switch Q1 and this circuit arrangement prevents this from occurring. Note that the transient heat dissipation that occurs in this relatively inefficient switching action is not an issue as it ideally only occurs once with subsequent switching transitions being in a much more efficient mode.

Referring again to Figure 25, after the power switch Q1 is turned on, capacitor C2 is charged to the top rail voltage Vin almost immediately and voltage VQ is equal to the top rail voltage Vin. If switch Q1 is a transistor, then voltage VQ may be just slightly less than Vin due to semiconductor effects.

The circuit in Figure 25 is now described in its second mode of operation. Current IL through the inductor rises. This rising current charges the output capacitor C4. The voltage Vout rises. The near step increase in voltage at VQ at the start of the first mode causes the output voltage Vout to rise. The shape of the wave forms over time, including voltage VQ, voltage Vout, and current IL are shown in the signal timing diagram of Figure 30 and Figure 31 . Upon the first to occur of either of the following events power switch Q1 is turned off. The first event is the current through the inductor rises to a predetermined maximum, ILmax. The second event is an upper desired level for Vout is reached.

The maximum current in inductor L (ILmax) is limited by magnetic saturation, overheating of the inductor L, exceeding the peak current rating of Q1 or any other limiting parameter chosen. The second mode ends and the third mode begins when power switch Q1 is turned off. Preferably Q1 is turned off quickly. When this occurs Q1 is turned off and the resistance of switch Q1 increases quickly. At this instant inductor current IL, that was flowing through Q1 , is transferred to flow through capacitor C2. In the third mode the power switch Q1 is off and the auxiliary switch Q2 is also off.

At the beginning of the third mode the inductor current IL continues to flow which draws down the voltage VQ by draining charge off capacitor C2. Voltage VQ decreases according the relation between voltage and a resonating series circuit of the inductor L and capacitors C2 and C4. An advantage of drawing charge from capacitor C2 is that there is no resistive power loss. If this function were to be provided by the turn off of Q1 , in the usual manner, the simultaneous application of voltage across and current through the switching device would result in a substantial power loss only limited by the speed of the switching event itself. When the voltage VQ decreases to a relatively small level, below the bottom rail voltage, the protection diode D2 in switch Q2 is suddenly forward biased. This small relatively level is about 0.7 V and depends on the particular specification of diode D2. Therefore the voltage at VQ drops to a minimum voltage of about -0.7 V and can fall no further. Detection of this predetermined minimum level of voltage VQ is a criterion for turning Q2 on.

Referring again to Figure 26 and Figure 29 upon detection of the predetermined minimum level of voltage VQ, auxiliary switch Q2 is turned on quickly. This is now the beginning of the fourth mode. Advantageously by turning auxiliary switch Q2 on at this time the voltage drop across switch Q2 is minimal because the current that was flowing through D2 can now be routed through Q2 if its impedance to this current is less than that presented by diode D2.

The action of the circuitry at the end of mode 2 and during mode 3 has therefore effected a lossless transition of VQ from the top rail voltage to the bottom rail voltage with a waveform shape dictated by the resonant values of C2, L2 and C4, and by the rail voltage and coil current at the moment of switching. Switching losses are substantially reduced and are limited to losses in the equivalent series resistances (ESR) of the components involved. Radio frequency interference is considerably reduced due to the lower dV/dt of the switching edge. All stray and parasitic capacitances in the circuit are additive to the effect of the shunt capacitance C2. This means that any stray capacitance that is effectively connected to the node at Vq such as the capacitance of Q1 in its off state for example.

The third mode ends and the fourth mode begins when auxiliary switch Q2 is turned on. At this time Q1 is off and Q2 is on. Because the voltage VQ is about zero volts, which is less than the output voltage, Vout, the current IL through the inductor L continues to decrease. Depending on the output voltage Vout, which is the voltage on capacitor Cout, switch Q2 stays on until either of the two criteria a) or b) below occurs. a) The point in time where the inductor current IL reaches zero and when the output voltage Vout is in the range of being greater than or equal to half the top rail voltage Vin.

In practice, to allow for resonance losses, Vout needs to be slightly greater than half the top rail voltage Vin to allow a successful rail to rail resonance to occur. b) When the output voltage Vout is in the range of less than half the top rail voltage Vin, to just slightly greater than half the top rail voltage Vin, the switching behaviour of Q2 is modified. In order to route enough energy into C2 so as to commutate VQ from the bottom rail to the top rail, it is necessary to inject energy into the inductor L to achieve this. By allowing switch Q2 to stay on past the point in time where IL drops to zero, the inductor current IL reverses and increases to a predetermined negative value. The stored energy in the inductor begins to increase again. This is the extra energy required to commutate C2 from the bottom rail to the top rail.

At the beginning of the next mode, which is the fifth mode, switch Q2 is turned off quickly. The advantage of applying criterion a or b to the turn off time of Q2 is that some of the additional energy stored in the inductor L is available to be transferred to capacitor C2 when switch Q2 is turned off. In many cases this additional energy is sufficient to eventually raise the voltage VQ to the value of the top rail voltage a certain amount of time after switch Q2 is turned off.

By this additional delay in turning off switch Q2, a stronger reversal of current is achieved through the inductor than if switch Q2 is turned off immediately upon the current in the inductor reaching zero. This stronger reversal of current through the inductor L causes additional energy to be stored in the inductor.

Advantageously some of the current that flows through the inductor when the value of the current is negative may be drawn from not just capacitor C4 but also a load connected to the output.

The fourth mode ends and the fifth mode begins when switch Q2 is turned off. During mode five switch Q1 is off and switch Q2 is off.

Still referring to Figure 29 waveform is that of a damped sinusoid according to an equation corresponding to the series combination of the inductor L and the capacitor C4 and capacitor C2. Preferably the intrinsic resistance of these components is low enough for a waveform to be that of an under damped sinusoid.

Due the nature of the series LC circuit resonance, the current flowing through the inductor L continues to increase negatively, at the beginning of mode 5 because the voltage VQ is about zero volts, which is less than the output voltage Vout. This adds to the energy already stored in inductor L at the beginning of mode 5 by the negative pre-charge current already flowing.

Since the current IL flowing through the inductor L is zero or negative at the start of mode four depending on whether criteria a) or b) is used to switch off Q2 at the end of mode 4, the current IL flowing through the inductor is negative immediately after mode 5 starts. This negative "reversed" current IL charges capacitor C2 and raises the voltage VQ. If criterion a) in mode 4 triggers switch Q2 off, the voltage VQ eventually rises to the top rail voltage Vin plus an additional small voltage that is enough to forward bias protection power diode D1 . This additional small voltage is about 0.7 V above the top rail voltage depending on the particular diode D1 . Therefore the voltage VQ is limited to rising to the top rail voltage plus this additional small voltage.

This aspect of the invention therefore detects when protection power diode D1 becomes forward biased. By turning Q1 on fast at this time there is a very low switching loss since voltage drop across power switch Q1 is less than the power diode D1 voltage drop. This is the beginning of mode 2 again.

If criterion b) in mode 4 triggers switch Q2 off, the voltage VQ may or may not eventually rise to the top rail voltage Vin in addition to the additional small voltage which is sufficient to forward bias protection power diode D1 .

If voltage VQ does reach the top rail voltage in addition to the forward voltage drop of the protection power diode D1 , then the circuit in Figure 13A detects the peak of the resonance of the voltage QV and when this occurs it turns switch Q1 softly on at this instant. The diagram illustrates the situation that occurs during the start-up phase when a prohibited or indeterminate condition is over-ruled for a first start-up cycle in order to commence oscillation.

It is possible that the voltage VQ peaks below the top rail voltage Vin, depending on such factors as: the value of L, the value of C2 and C4, the amount of delay imposed by criterion b), the energy present in the inductor L when Q2 is switched off, the relative values of Vout and the top rail voltage, and the current drawn by any load connected to the output. The circuit in Figure 13A detects if voltage VQ peaks below the top rail voltage plus the forward voltage drop of power protection diode D1 and turns switch Q1 on at this time. Advantageously this is when the voltage drop across switch Q1 is minimized and therefore the switching power loss (and RFI) are also minimized at this time.

Preferably to further minimize any switching loss and RFI, if the voltage VQ peaks below the top rail voltage in addition to the forward voltage drop of diode D1 , switch Q1 is turned on "softly". In this event, the resistance across switch Q1 is reduced gradually. When the voltage VQ rises to the top rail voltage, switch Q1 is then fully turned on fast.

If there is voltage undershoot of the rail target voltage, the circuit (shown in Figure 13A) can vary the pre-charge current in the inductor L by increasing the current slightly so as to ensure sufficient energy is available from the inductor to achieve correct commutation for the next cycle. This active monitoring of the resonant voltage at mode 5 allows for the control circuitry in Figure 13A to adjust the reverse or pre-charge current in the inductor L to be just enough or in excess of what is required to ensure rail to rail commutation.

At the end of the fifth mode switch Q1 is on and Q2 is off. Then the cycling process is repeated beginning with the second mode. This is shown in Figure 29. Referring to Figure 25, the total value of capacitance that is required, in parallel with the switching devices Q1 and Q2, may be either one capacitor across one of the switches such as C1 or C2; or two smaller capacitors C1 and C2 each connected across each switching device Q1 and Q2. The desired capacitance value is the sum of these two smaller capacitors.

The input capacitor C3 connecting the top rail to the bottom rail, has a much larger value than the switch shunt capacitor(s) C1 or C2. The choice of how to split the capacitors and capacitance values made in order to minimize circulating RFI currents due to the transfer of inductor current from the transistors Q1 , Q2 to the capacitors C1 and C2 and back again during each switching transient.

A single capacitor C1 could be connected across Q1 as the highest diverted current normally occurs here. Further embodiments of the switching supply circuit that operate according to the abovementioned sequential method of control will be apparent to those skilled in the art. For example capacitor C1 , illustrated in Figure 25 as a single element, could be replaced by two or more capacitors in series or parallel, inductor L could be replaced by two or more inductors in series or parallel and so forth.

Figure 26 shows another embodiment of the switching supply with the feedback controller and the connections to the feedback controller removed. Figure 26 shows a simplified overview of circuit components and connections without feedback controller and its connections.

Figure 27 is a circuit diagram of an embodiment of the switching supply with the feedback processor removed to illustrate its operating components with the power switch on.

Figure 28 is a circuit diagram of the switching supply with the feedback processor removed to illustrate the components with the auxiliary switch on. Each embodiment has its advantages in terms of optimizing current flows between the switching devices and associated capacitors depending on voltage transfer ratios and net current flow directions. These have implications on stray inductance, circuitry and component resistance and electromagnetic interference (EMI), both from a perspective of EMI generation and EMI susceptibility. Figure 29 is a signal timing diagram for the switching supply when the system is running.

Figure 30 is a signal timing diagram for the switching supply when the system is initiating its startup phase.

Figure 31 illustrates the situation that occurs during the start-up phase when a prohibited or indeterminate condition is over-ruled for a first start-up cycle in order to commence oscillation.

Figure 32 shows a typical switching waveform and the background behind each transition. a) the first transition.

If a circuit, as shown in Figure 23 for example, is running in continuous mode operation, when the device Q1 turns on there is a very high likelihood that another device D2 is exhibiting reverse recovery and that there is a significant current build up at the instant of switching, leading on to the initial transition, which may lead to significant RFI being generated. Figure 32 shows a typical waveform of the circuit in Figure 23. b) The slope.

The slope of the waveform here represents, in a normal switching system, a point where significant currents are flowing at the same time as there are voltages across the switching device Q1 . (This is especially so in the case when a switching device is turning off while supplying a significant current to inductive load). In the turn on situation of Q1 there may be significant currents flowing in parasitic inductances and capacitances as well. These circulating currents are prolific generators of RFI. The sharpness of the slope at Vq in Figure 23 indicates how many harmonics which may be present and what level of harmonics are required to create this waveform.

There is an optimisation problem at this instant as a sharp edge in the slope indicates that there will be less losses in the switching device, but at the same time tends to create more harmonics and RFI problems. c) The final transition.

A sudden arrival of 'full switch on' condition of the switching device Q1 in itself therefore creates another problem. Because of stray inductances associated with the switching device, as well as associated components, there is a tendency for voltages to overshoot. This introduces risks of instability (or even catastrophic) operation as a result of so-called 'ground bounce'. rail softness.

This is a measure of how much supply rails are decoupled one from another. If there is a change in current draw from the supply, then the voltage rail connected to Vin experiences transient voltage changes. If there is a certain amount of inductance in series with these components, then this transient voltage change creates a voltage ring that continues for a certain amount of time while decaying depending on the degree of damping. Conversely, depending on the circuitry, a low inductance supply may be problematical if sudden currents occur, such as shoot through, due to incorrect timing of switching devices or as switching devices overcome reverse recovery of an associated component.

Both these disturbances are potential sources of RFI. The continual development and use of switching devices, capable of ever faster transitions, is exacerbating this problem. Having understood the sources of RFI creation by switching devices there is a solution that can be adopted so as to minimise these deleterious effects. The waveforms and problems discussed herein are manifested in a circuit running in continuous operation or hard switching. They can be reduced to a certain extent by operating in a discontinuous mode.

This discontinuous mode normally significantly reduces RFI generation at point Vq in Figure 23 due to reduced reverse recovery transients but it can lead to uncertainty of circuit operation.

Another way in which RFI can be reduced is to adopt a resonant mode of operation. Unfortunately virtually every topology of resonant operation requires components to have a voltage rating to almost twice the highest voltage that is expected to be handled by the circuit or device. This demand places a burden on the cost of the product or a loss in efficiency, due to higher voltage drops of higher voltage rated components.

The resonant topology described below with reference to Figure 25 and Figure 40, have the advantage of only requiring a voltage rating for components being used at the highest voltage present when the circuit is operating; not twice that voltage which was previously the case.

Figure 33 shows an idealised switching waveform. In order to improve RFI performance further the incorporation of non-linear devices (where the capacitance of the device varies according to the potential difference across the device itself) is shown so as to modify the switching wave shape at its corners a and c on the switching waveform as shown in Figure 33. Modification of wave shape still allows for a fast transition, b, from one switching state to the other and by judicious choice of component values, it is possible to reduce total RFI emission, while still achieving quicker switching times.

Referring to switching waveforms shown in Figure 32 which are associated with a resonant topology, see Figure 25 and Figure 40, there is an opportunity to modify the wave shape of the switching transient. Firstly slope b) can be adjusted using the total shunt capacitance across the appropriate switching device. Secondly the smoothness or 'roundness' of initial transition a) and final transition c), can be adjusted using components that have a 'variable capacitance related to their voltage' characteristic.

Figure 34 is a block diagram of a conventional power switch gate drive circuit, Q1 and Q2, connected to a power switching device Q3. Figure 34 explains the general principles of conventional gate drive circuit operation. The interface and logic elements of a gate drive circuit (not shown) receives an 'on' or 'off command and this then switches Q1 and Q2 so that for 'on', Q1 is on and Q2 is off. Conversely, when the command is for 'off then it switches Q1 off and switches Q2 on. Assuming the power switch device Q3 is off, Q2 is on and potential Ve is nominally at Vb. Vf is at some significantly higher potential than Vb

At the 'on' command, Q2 switches off and Q1 switches on. Input current flows from C1 via R1 and charges input capacitance Ciss until switch Q3 starts to conduct. At this point there is an output voltage transient change of Q3. Vf, via the reverse transfer capacitance Crss, now absorbs nearly all of the available input current. When the output voltage Vf is nominally equal to Vb, the voltage Ve can continue to rise in order to turn Q3 fully on.

It can be deduced from the operation of Q1 and R1 , and the voltages Va and Ve, that the current profile of this method of operation is not ideal. At the instant that the switch device Q3 is turning on, the input current is already reduced. At the point that the switch device Q3 is now turned on, the voltage at Ve is rising but at a slower and slower rate as the current to charge Ciss is rapidly tailing off. This can lead to unnecessary switching device resistive losses. The operation of the drive circuit transistor Q1 in Figure 34 requires a considerable amount of available energy to drive the input capacitances of the switch device Q3 in order to change the voltage Ve.

At the 'off command, Q2 and R2 take the energy from input terminal of Q3. All of this energy is wasted in Q2 and R2. Again the current profile of the input current is not ideal for efficient switching of Q3 especially as the effect of Crss occurs at a relatively low value of Ve where the available current from Q2 and R2 is significantly reduced. Figure 35 shows a circuit power switch input current flowing with the power switch hard switching. This is the normal operation of a power switch device Q3. The operation of the circuit is as described in Figure 34. However there is another issue to be considered.

Referring to Figure 35, the circuit includes an inductor L1 and a diode D10. Current IL flows from the inductor L1 into D10. As Q3 is turned on this current has to be diverted into Q3 away from flowing into D10. Two problems occur here. The first is that the full current from L1 flows through Q3, around the same time as the full voltage Vf is present. Secondly there is a so-called reverse recovery characteristic of D10 to overcome. This excess current requirement can cause RFI issues as the reverse recovery current is overcome and the voltage Vf suddenly starts to fall.

During 'hard switching' conditions of Q3 the effects of the input capacitance Ciss and the reverse transfer capacitance Crss both occur around the same time. The input current is therefore a combination of currents generated by these two parasitic capacitances. This is the same for both turn on and turn off of Q3.

Figure 36 shows an example of a power switch input current that flows with power switch resonant switching. The current input current flows are different for the power switch device Q3 when it is operated in a resonant mode. The input capacitance charging and discharging currents occur at a different time to the currents associated with the reverse transfer capacitance Crss in conjunction with the voltage change across Q3.

When Q3 is in its turned off condition, Q2 is on and R2 is connected across the input of Q3 and the voltage Ve is Vb. At the same time inductor L1 and diode D10 are also connected in the circuit (shown in Figure 36) and a resonant capacitor C5 is positioned across Q3. Assuming current IL flows from inductor L1 into D10. Assuming the voltage Ve is Vb (Q2 on) and that after a period of time that the current IL ceases flowing into D10, then a resonant circuit (comprising L1 and C5) now causes voltage Vf, across Q3 and C5, to reduce until the voltage Vf is equal to Vb. The current, caused by the reverse transfer capacitance, while voltage Vf is reducing to Vb, flows through Q2 and R2 and so maintains voltage at Ve nominally at the level of Vb. Power switch Q3 is now turned on by a current flowing via Q1 and R1 . It is evident that the current flowing out of the junction of R1 and R2 now only has to charge the input capacitance Ciss of Q3. This therefore reduces the requirement of power from the external power supply (not shown) to C1 .

At the point in time where it is required to turn off switch Q3, to facilitate lossless commutation of the resonant circuit, it is important that Q3 is switched off as quickly as possible. Under these conditions the current, from both the discharging of the input capacitance Ciss and the dynamic effects of the current through the reverse transfer capacitance Crss, are routed to the ground terminal of Q3 at Vb via R2. The stored capacitive energy in both cases is wasted in R2.

Figure 37 shows an example of an energy recovery power switch drive circuit arrangement (mid-point switched). Figure 37 shows the basic concept of an energy recovery circuit that both assists with the speed with which the current can be injected or taken out of the junction of R1 and R2 and the input connected to the switching device Q3, while at the same time significantly minimising the external power required into C1 . Circuit shown in Figure 37 achieves this by efficiently delivering energy from the midpoint Vg on C6 of the circuit in Figure 37 into the input capacitances of Q3 when the device Q3 is turned on; and returns this energy from the input capacitances of Q3 when the device Q3 is turned off.

The circuit, shown in Figure 37 reduces the overall net power requirement of the gate drive by approximately a factor of 10. There are still some losses that cannot be overcome, for example track resistances and effective internal input terminal resistances due to the physical design of the power switch and the resistance of the inductor L2 and S1 . Other than these losses the circuit is almost lossless.

Referring to Figure 37 and assuming Q3 is switched off, and the midpoint circuit output voltage and voltage across C6, is Vg which is approximately ½ Va. Q2 is then switched on, thereby setting Ve to Vd. Gate control signal command then switches from Off to On'. Switch Q2 is now turned off and S1 is closed. The current Igate in Figure 39D increases through L2 into the input of the power switch Q3. The voltage Ve transits from Vb to close to Va, as L2 resonates with the input capacitance of Q3. At the appropriate instant of resonance, where the voltage Ve comes closest to Va, S1 is turned off and Q1 is turned on. This maintains the voltage Ve at Va. Any shortfall in voltage at the point of Q1 switching on, is made up by extracting energy from C1 .

The turn off of Q3 is done in a similar manner. The timing of the control signals to Q1 , Q2 and S1 are handled by the section called gate timing control circuitry.

Figure 38 shows a circuit depicting a soft start capability. In order to be able to start the resonant operation of Q3 in a resonant power control application it is necessary to be able to slowly turn on Q3 once at the beginning of the cycle, see Figure 30 and Figure 31. A fast turn on, with a significant voltage across the drain-source junction of Q3, is problematical because a large current flows from the resonant shunt capacitance C5, in position as shown in Figure 36, into the switch Q3. This large current can cause damage to Q3. The solution to this is to turn Q3 on in a current limited mode. One way this is done is by introducing a limited current, set by R3, (with Q1 off) into the input terminal of Q3 and using the reverse transfer capacitance Crss to limit the dV/dt across the shunt capacitor C5.

Figure 39A to Figure 39D show graphs of current profile against time for power switch input current using energy recovery techniques described above.

Figure 40 shows an example of a high efficiency resonant converter device configured for voltage conversion.

The diagrams in Figure 40 show a high efficiency resonant converter circuit and its main power conversion components. In addition to components shown it is understood that an electronic drive circuit, as well as control and auxiliary power components are required to enable the circuit to operate. Due to the symmetrical nature of the circuit (Figure 40) it can be operated in all four quadrants if required. It is appreciated that the circuit can operate in four main modes.

Step down

Here the DC voltage is applied between D and C where D is the more positive terminal. The operation of the switching devices, depicted as NPN transistors, in a resonant mode allows a voltage between zero and the voltage applied at D to be available at terminal A.

Step up

Here the DC voltage is applied between A and C, where A is the more positive terminal. The operation of the switching devices, depicted as NPN transistors, in a resonant mode allow a voltage between the voltage applied at A and a voltage higher than A depending on the operating regime of the resonant circuit to be available at terminal D.

DC to AC conversion

Here the DC voltage is applied between D and C where D is the more positive terminal. If the load connected to A has its other connection connected to the midpoint of the voltage between C and D, then the operation of the switching devices, depicted as NPN transistors, in a resonant mode allows the voltage at A to be taken above or below the midpoint voltage and this impresses an AC waveform on this load with a maximum amplitude swing between zero and the voltage applied at D.

AC to DC conversion

Here the AC voltage is applied between A and returned to a point between C and a value that is equal to the peak to peak value of the AC signal present on A, where A is always the more positive terminal relative to C. The load is connected between C and D. The operation of the switching devices, depicted as NPN transistors, in a resonant mode allows the voltage at A to be taken above or below the midpoint voltage and this is converted to a DC voltage between C and D.

Figure 41 shows an example of a resonant frequency conversion block. This is a block diagram of a standard resonant frequency conversion block and shows its main power conversion components, including the resonant dV/dt limiting capacitors and a 3-phase motor connected. It is appreciated that ancillary electronic drive circuit, control and auxiliary power components required to make the circuit operate correctly are not depicted. Due to the symmetrical nature the circuit shown in Figure 41 can be operated in all four quadrants if required.

Figure 42 shows a split function voltage control/frequency control circuit for a quasi-sine motor drive. The block diagram is of a complete high efficiency resonant conversion quasi sine drive. Combined voltage control and frequency control can be achieved with a conversion efficiency in the region of around 99%. The 3 phase rectifiers, depicted as diodes in Figure 42, introduce diode drop losses of approximately 1 .4 volts at the operating current of the drive. In percentage terms this represents a loss of around 0.3%. For a single phase drive, this is closer to 0.6%. These drive losses are comparable with PWM drive losses where the output stage is hard switching at about 4 to 8 kHz directly into the motor.

Figure 43 shows a block diagram of an active front end driver. This takes the place of the 6 diode rectifiers shown in Figure 42. The block diagram in Figure 43 shows how three voltage control blocks are used so as to provide a versatile active front end power conversion system. In this example, a three phase AC to DC conversion is detailed so at least one of the three phases is always positive at any given time and likewise one of the three phases is always negative. During a full mains cycle there are periods where two of the three phases are positive and likewise there are times when two of the three phases are negative. The topology in AC to DC mode, is a composite of synchronous rectification and boost voltage conversion. The circuit in Figure also ensures that the current drawn from the three phase connection L1 , L2 and L3 is power factor corrected, has a low crest factor and is supportive of the voltage waveform. In its DC to AC mode the circuit is also capable of full four quadrant operation.

The active front end also can be used to isolate the three phase supply from the DC connection point so that the integrity of the DC supply to the connection point can be maintained during power outage or disconnect conditions of a mains supply. Figures 44A to 44E shows diagrammatically front end current paths under different modes. Figure 44A to Figure 44E show the way that the active front end, in conjunction with some energy storage capability and the motor drive, can perform several functions. Some of the functions can occur at the same time. The size of the energy storage device (not shown) can be adjusted to suit the functions required.

Figure 45 shows a circuit diagram for limiting inrush current at phase input of active front end shown in Figure 43. The circuit in Figure 43 could be the circuit as shown in Figure 40 for example. This conducts current in one direction even if the switching devices, shown as NPN transistors in Figure 40, are off. This is because of the parallel diodes placed across the switching devices in the circuit, which behave either as part of the substrate construction (in the case of a MOSFET) or are added in parallel with the switches, shown as NPN transistors, (in the case of IGBTs).

To avoid inrush current at the instant of connection or reconnection of a supply voltage to input 01 at an active front end of the circuit, it is necessary to incorporate some extra means to eliminate, or at least significantly reduce, the inrush current.

Here an extra switching device is put in series with the line input terminal shown as 01 . This extra switching device is connected in the opposite polarity and sense to the existing components in the active front part of the circuit, so if this switching device is turned off, no current can flow into the input of the active front end of the circuit in Figure 45. However it can perform another function in this position. Careful analysis shows that if the switch transistor at the top right of Figure 45 is permanently on and the switch transistor at the bottom right of Figure 45 is permanently off, this topology operates in a similar manner to a boost converter (using the extra diode in parallel with the extra switching device connected to 01 ) so that power and voltage can be increased in a controllable manner on DC side. Once the DC output is at the peak level of the AC input, the extra switching device connected to 01 can be switched on permanently and the active front end is able to operate in its usual step up (boost) mode.

Figures 46 and 47 shows examples of current inrush limited circuitry added at the junction of the +ve connection, between the active front end and DC link. To avoid inrush current at the instant of connection (or reconnection) of the supply voltage to the inputs of the active front end, it is necessary to incorporate some additional means to eliminate, or at least significantly reduce, the inrush current. The circuits (Figure 46 and Figure 47) behave in very different manner to each other. The circuit in Figure 46 works very similarly to the switch circuit connected to 01 in Figure 45. However 3 of these switch circuits, one per phase, (it is possible to just use 2 switch circuits for economy purposes) are required to prevent inrush current. This is because in Figure 46 and Figure 47 the extra circuitry is on the DC side of the 3 phase active front end only one extra switching circuit is required. This extra switching circuit in conjunction with the characteristics of the 3 phase active front end ensures complete isolation of current flow under all voltage conditions. In Figure 46 the current limiting means is placed between the output of the active Front ends and the DC link connection. Figure 46 shows a modified Buck converter that provides good control of the amount of energy that can be transferred from the active front ends to the DC link. This feature of the invention is very important as it allows reconnection and continuation during 'brown out' and blackout conditions without nuisance tripping and therefore offers significant benefits in areas where surety of supply is not always guaranteed.

Figure 47 shows an example of using a conventional resistor and relay to protect against inrush current. A disadvantage with this arrangement is that the resistance value is high in order to limit current inrush when the DC link is at a zero to low voltage at the point of application of the utility to the three phase inputs. This value is too high to allow for normal motor operation so a state exists where the motor has to be disconnected until the DC link has been established at the correct voltage. Also the resistance is present when the voltages are higher on the three phase input side than the voltage present on the DC link even when the switching devices in the 3 phase active front end are off.

Aspects of the invention have been described by way of a number of exemplary embodiments, each exhibiting different advantageous features or benefits; and it is understood that features, components or circuits from two or more of the aforementioned embodiments may be combined together to overcome specific problems or to provide a bespoke solution to a particular problem. Variation may be made to the invention by including a connection to the load dump from an induction motor or incorporating the load dump in or on the body of a motor or its housing. The threshold detector may be incorporated in a permanent magnet motor or a switched/variable reluctance motor. Such modified motors may be included in electrical appliances, such as pumps, tools, drive systems or other machines or vehicles, such as electric cars, busses or trains.

Claims

1 . A threshold detector circuit (Fig 9) for lossless switching of a transistor (Qx) senses a voltage across the transistor (collector-emitter or drain-source) and is operative such that when the voltage across the transistor exceeds a user defined HIGH threshold, the threshold detector circuit remains in a quiescent state with a low input impedance to a signal to the detector circuit; and when the voltage across the transistor (collector-emitter or drain-source) is less than a user defined LOW threshold, the threshold detector circuit (Fig 9) switches to an active state, so that the transistor (QX) is switched on thereby creating a low input impedance to a signal to the detector circuit, thereby continually isolating the detector circuit from noise from its connection to the transistor (QX).

2. A threshold detector circuit according to claim 1 wherein the transistor (Qx) remains switched on until it receives a reset command from a controller (Figs 14A to 14H).

3. A threshold detector circuit according to any preceding claim wherein during the quiescent state, the low input impedance to the signal to the detector circuit from the transistor, is provided by a switchable low impedance circuit (ZD1 , D5, R9 and Q2); and in the active state, the low input impedance to a signal to the detector circuit from the transistor, is provided by the switch (Qx) when the switchable low impedance circuit (ZD1 , D5, R9 and Q2) is disabled.

4. A threshold detector circuit according to any preceding claim includes a soft start circuit.

5. A threshold detector circuit according to any preceding claim wherein a decision to apply a quasi-sine input drive signal or a pure sine input drive signal is determined in accordance with a self-adjusting detection circuit.

6. A threshold detector circuit according to any preceding claim wherein an over-ride switch is configured to switch on/off when a negative drive current is sensed.

7. A threshold detector circuit according to any preceding claim includes a means for over current detection.

8. A threshold detector circuit according to any preceding claim includes a driver module for a resonant power supply.

9. A threshold detector circuit according to claim 8 includes a driver chip for the resonant power supply.

10. A threshold detector circuit according to any preceding claim wherein a feed-forward input drive signal is generated using a signal derived from an input current signal.

1 1 . A threshold detector circuit according to any preceding claim includes a switch control circuit connected across input terminals of a motor.

12. A threshold detector circuit according to claim 1 1 wherein the feed-forward input drive signal is used to control a means for adjusting a supply voltage in order to vary motor torque at low frequencies.

13. A threshold detector circuit according to either claim 1 1 or 12 wherein a microprocessor generates the feed-forward signal.

14. A threshold detector circuit according to any preceding claim includes at least one shunt capacitor operable to modify the drive voltage from a quasi-sine to approximate to a sine wave.

15. A threshold detector circuit according to any of claims 1 1 to 14 includes a resonant half bridge circuit for converting an applied input voltage and/or an applied input current to the motor in order to supply a turn on current for the motor, at a selected instant, independent of the supply voltage and/or the supply current and/or load of the motor.

16. A threshold detector circuit according to any of claims 1 1 to 15 includes a feedback circuit with a resonant half bridge operable to switch on independently of at least one of the following: motor input voltage, motor load, load current, shunt capacitance, inductance and resistance of the motor.

17. An electrical appliance, such as a pump, tool, drive system or machine, including the threshold detector circuit according to any of claims 1 to 16.

18. A vehicle including the threshold detector circuit and motor of any of claims 1 1 to 16.