CA1275692C - Electric power inverter with adaptive third harmonic auxiliary impulse commutation - Google Patents

Abstract

ELECTRIC POWER INVERTER
WITH ADAPTIVE THIRD HARMONIC
AUXILIARY IMPULSE COMMUTATION

ABSTRACT OF THE DISCLOSURE
In a 3-phase third harmonic auxiliary impulse commutated electric power inverter, during a commutation interval the firing signal for the oncoming main valve of the inverter is delayed for a programmed interval of fixed duration after the commutation capacitor voltage changes polarity, whereby the peak voltage on the capacitor can automatically vary with the magnitude of load current.

Description

:~75~

ELECTRIC POWER INVERTER WIT~ ADAPTIVE
T~lIRD ~IARMONIC AIJXILIARY IMPULSE COMMUTATION
Background of the Invention This invention relates to electric power inverters for converting direct current (d-c) to polyphase alternating current (a-c), and more particularly it relates to improvements in the old and well known current-fed "third harmonic"
auxiliary lmpulse commutated inverters. The principles of commutation and a typical practical application of such inverters ~ere described in a technical paper entitled: "Analysis of a Novel Forced-Commutation Starting Scheme for a Load-Commutated Synchronous Motor Drive," which paper was presented by R. L. Steigerwald and T.A. Lipo at the IEEE/IAS annual meeting held in Los Angeles, Calif. on Oct. 2-4, 1977. The Steigerwald and Lipo paper was reprinted in IEEE T~ANS. Vol. IA-15, No. 1~ Jan/Feb 1979, pgs. 14-24.
In essence, a third harmonic auxiliary impulse commutated inverter comprises six main unidirectional conduction controllable electric valves, such as thyristors, that are interconnected in pairs of series aiding, alternately conducting valves 20 TR-1~20 to form a conventional 3-phase, cdouble-way, 6-pulse bridge between a pair of d-c terminals and a set of three a-c terminals. The d-c terminals of the b~idge are adapted to be co~nected to a suitable source of relatively smooth direct current. A large, multicell, heavy duty electric storage battery is a suita~le current source, as is the combination of an electric power rectifier to which an alternatiny voltage source is cormected and a current smoothing reactor or choke in the d-c link between the d-c terminals of the rectifier and inverter, respectively~ The a-c terminals of the aforesaid bridge are respectively connected to the different phases of a 3-phase electric load circuit which typically comprises star-connected 3-phase stator windings of a dynamoelectric machine such as a large synchronous motor.
To supply the load circuit with 3-phase alternating current, the six main valves of the inverter are cyclically turned on (i.e., rendered conductive) in a predetermined sequence in response to a family of "firing" signals (gate pulses) that are periodically generated in a prescribed pattern and at desired moments of time by associated control means.
To periodically turn off the main valves by forced commutation, the inverter is provided with an auxiliary circuit comprising a precharged commutation capacitor and at least seventh and eighth alternately conducting unidirectional controllable electric valves that are arranged to connect the capacitor between the neutral or common point o-f the 3-phase a-c load circuit and either one of the d-c terminals of the bridge.
During each full cycle of steady state operation of a third harmonic inverter, each vf the 20-TR-1~20 valves in the auxiliary commutation circuit is briefly turne~ on three separate times. More particularly, the 7th valve is fired at intervals of approximatel~
120 electric degrees, and the 8th valve is fired at similar intervals that are staggered with respect to the intervals of the 7th valve, whereby one or the other auxiliary valve is fired every 60 electrical deyrees. When an auxiliary valve is turned on, the commutation capacitor is effectively placed in parallel with one phase of the load circuit and a first one of the two main valves which are then conducting load current. Initially, the capacitor voltage magnitude is higher than the amplitude of the line-to-neutral voltage that is developed across the inductive load, and its polarity is such that the capacitor starts discharging. Consequently current is forced to transfer (commutate) from the first main Yalve (i~e., the offgoing or relieved valve) to a parallel path including the turned-on auxiliary valve and capacitor. The rate of change of current du~ing commutation will be limit~d by the load inductance.
After current in the offgoing main valve decreases to zero, the magnitude of capacitor voltage is still sufficient to keep that valve reverse biased for longer than its "turn-off time." As soon as the commutation capacitor is fully discharged, load current begins recharying it with opposite polarity.
Once the commutation capacitor is recharged to a voltage magnitude exceeding that of the line-to-neutral load voltage, the next (oncoming) main valve in the bridge is forward biased and can be turned on, whereupon load current commutates from the turned-on auxiliary valve and commutation capacitor to the oncoming main valve. This causes the auxiliary valve to turn off and completes the commutation '5~

process. The capacitor is left with voltage of proper polarity and sufficient peak magnitude for successful commutation of the seconcl one of the first-mentioned two conducting main valves when the opposite auxiliary valve is turned on approximately 60 degrees later. It will be apparent that there are six intervals of commutation per cycle, the direction of current in the commutation capacitor during each interval is reversed compared to the preceding inter.val, and therefore the fundamental frequency of the alternating capacitor current equals the third harmonic component of load frequency.
As is pointed out in the referenced Steigerwald and Lipo paper, one practical application of a current-fed third harmonic auxiliary commutated inverter is in an adjustable speed a-c drive system where the 3-phase star-connected stator windings of a synchronous machine are supplied with variable frequency a-c power by the inverter which needs to be forced commutated in order to start the m~chine. In such an application, for reasons explained in that paper, a technique of "delayed gating" is used to ensure that at the end of each commutation interval the commutation capacitor has recharged to a sufficiently high level of voltage to guarantee successful commutation during the succeeding interval.
According to this technique, the sequential firing signals for the main valves are each delayed, after the oncoming valve is forward biased, while the capacitor continues accumulating and storing electrostatic charge until its voltage attains a threshold level required for extinguishing current in the next offgoing valve. Steigerwald and Lipo suggest that the threshold level can be proportional to the magnitude of source current so that the peak magnitude ~ 6~

of capacitor voltage (and hence the commutating ability of the inverter) will desirably track -the demands of the load. In other words, the magnitude of capacitor voltage is high when necessary to commutate high current and is relatively low when only light load needs -to be commutatec~. This "adaptive"
commutation capacitor voltage t:echnique advantageously reduces comrnutation time and power losses during the commutation intervals when the magnitude of load current is relatively low.
In his prior art U.S. Patent No.~,244,017 -Steiyerwald issued January 6, 19~1, discloses and claims a modified third harmonic auxiliary impulse commutated inverter having parallel commutation circuits which allow three different values of commutating capacitance to be actively selected as a function of the magnitude of load current. Assuming that the inverter is supplying a synchronous machine load, current tends to decrease as frequency (i.e., rotor speed) and hence machine back emf increase. By switching to a commutation capacitor of smaller size when the current magnitude falls below a preset level, the commutation time is desirably shortened at light loads. As a result, the maximum permissible fundamental frequency is increased, and the operating range of the inverter is extended.
A current-fed third harmonic inverter is well suited for supplying variable frequency alternating current to the 3-phase stator windings of a rotatable synchronous machine that is used to start or "crank" a prime mover such as a large internal-combustion engine. In such a system, the rotor of the machine is coupled to a mechanical load comprising the crankshaft of the engine. Initially the output torque of the rotor (and hence the magnitude of current in ~" ~d ~

the stator windings) needs to be relatively high in order to start turning the crankshaft. As the rotor accelerates from rest, less torque (and current) will be required, while the fundamental fre~uency of load current increases with speed (revolutions per Minute).
In its cranking mode of operation, the inverter supplies the machine with current of properly varyiny magnitude and frequency until t:he engine crankshaft is rotating at a rate that equals or exceeds the minimum speed at which normal runniny conditions of the engine can be sustained. It should be apparent that the above-mentioned adaptive commutation technique, ~lerein decreasing current is accompanied by lower commutation voltage and hence shorter commutation intervals, will desirably raise the upper limit of the permissible range of inverter operating frequency.
Sunmary of the Invention A general objective of the present invention is to provide a third harmonic auxiliary impulse commutated electric power inverter characterized by an improved adaptive commutation feature that is simple yet effective.
Another objective is the provision, for sucl an inverter, of an adaptive commutation -feature that does not require means ~or sensing or measuriny current magnitude.
In carryiny out the invention in one form, a source of relatively smooth direct current is connected to a 3-phase inductive load circuit by means of an electric power inverter comprising at least three pairs of alternately conducting main controllable electric valves arranged in a 3-phase, double-way, 6-pulse bridge configuration. For commutating the main valves, a precharged commutation capacitor is connected between the load circuit and 20-rrR--1~0 the juncture of first and second auxiliary controllable electric valves that are interconnected in series aiding fashion across the d-c source.
Bistable volta~e sensing means is coupled to the capacitor; it has one state whenever the electrical potential on one side o-f the capacitor is measurably positive with re~pect to the other side, and otherwise it has a different state. A zero-crossing detector is coupled to the 3-phase ]oad circuit for detectin~ all zero crossin~s of the fundamental phase-to-phase alternating voltages that are developed at line terminals of the respective phases of the loacl.
The above-summarized inverter also comprises control means coupled to both the voltage sensing means and to the zero-crossing detector for cyclically producing a ~amily of periodic firing siynals that cause the valves to turn on selectively. In a third harmonic commutation Mode of operation, the aforesaid family includes a first series of firing siynals respectively produced in response to consecutive zero crossings of the phase-to-phase voltages for alternately turning on the first and second auxiliary valves, wher~upon load current can immediately transfer from an offgoing main valve to a parallel path including the turned-on auxiliary valve and the commutation capacitor which is first discharged and then recharged with reverse polaxity by this pulse of current. The family also includes a second series of firing signals respectively produced in delayed response to successive state chanyes of the voltage sensing means for turning on the main valves in a predetermined sequence, whereupon load current can then transfer to the oncoming main valve from the turned-on auxiliary valve. In accordance with the present invention, the control means includes time 20-TR-1~20 delay means effective at least after a predetermined initial period of time, which starts when the third harmonie commutation mode of operation commenees, for delaying the production of each firin4 signal in the second seri~s until a programmed interval of time has elapsed following each state ehange of the voltage sensing means. ~uring -the pro~rammed delay interval, load eurrent will continue to reeharge the commutation eapaeitor at a rate that is a function o~ its magnitude. Preferably, eaeh delay interval has a fixed, relatively short duration. As a result, during eaeh commutation interval the eapaeitor is automatically eharged to a peak vo]tage level that varies with current magnitude. Thus, as load current decreases the eommutation intervals are desirably reduced and the upper frequeney limit of the inverter is correspondingly inereased.
The invention will be better understood and its various objects and advantages will be more fully appreciated from the following description taken in eonjunction with the accompanying drawings.
Brief_Description of the Drawin~s Figure 1 is a schematic diagram of a system eomprising a rotatable eleetrical machine of the ~5 synehronous type having a rotor which is meehanieally coupled to a variable speed prime mover and having 3-phase, star-eonneeted stator windings whieh are eonnected to an electrie storage battery via a plurality of controllable eleetrie valves that in turn are intereonneeted and arranged to form a variable frequeney third harmonie auxiliary impulse commutated inverter;
Figure lA is a simplified block diagram of the eontroller (shown as a single block in ~igure 1) which cyclically produees a family of periodic firing ~j ~ 7 5 ~
20~TR-1420 _ 9 _ signals for respectively turning on the various valves of the inverter;
Figure 2 is a time chart showing, for one full cycle of operation in a third harmonic commutation mode, the six poss:ible states of 3-phase fundamental stator voltages and the family of twelve firing signals produced by the controller;
Fiyure 3 is a larger scale time chart showing variations in voltage and current of the commutation capacitor during transitions from odd to even states and from even to odd states, and also showing the programMed intervals of the present invention; and Figures 4-11 are flow charts that explain the operation of the preferred embodiment of the Figure lA controller to produce the firing signals shown in Figures 2 and 3.
Description of the Preferred Embodiment The electric power system shown in Figure 1 comprises a third harmonic auxiliary impulse commutated inverter having a pair of d-c terminals lOp and lOn connected to a source of relatively smooth direct current and a set of three a-c terminals 11, 12, and 13 connected, respectively, to line terminals of three star-connected armature windings on the stator of a rotatable, variable speed, 3-phase a-c synchronous machine 1~ which has a rotor 15 that is mechanically coupled to a prime mover 16. In the illustrated embodiment of the invention, the current source for th~ inverter comprises the combination of a source of voltage, such as a heavy duty electric storage battery 17, in series with impedance means which has appreciable electrical inductance, preferably p.rovided by the d-c field winding 18 on the rotor lS of the macnine 14. By way of example, the battery 17 is a lead-acid or nickel-cadmium type having 32 cells and rated 6~ volts, and the average magnitude of voltage at its terminals normally does not exceed 76 volts. Its internal resistance is typically in the range of 16 to 37 milliohms. The battery is intended to supply electric energy for starting the prime mover, and the system showrl in Figure 1 can successfully perform -this fLInction with the battery voltage as low as 61 volts.
The prime mover 16 can be a conventional thermal or internal-combustion engine, and in one particular application of the invention it is a high-horsepower, 16-cylinder diesel engine that is used to provide the motive power on a large self-propelled diesel-electric traction vehicle such as a locomotive. The synchronous machine 14 has dual modes of operation: as a yenerator for supplying alternating current to an electric load circuit that is connected to its stator windinys, and as an a-c motor for cranking or starting the engine 16. In its generating mode, the rotor 15 oE the machine is driven by the crankshaft of the engine 16, and the field winding 18 is energized by a suitable excitation source 20 (e.g., the rectified output of auxiliary windings on the stator of the machine 14) to which it is connected by means of a suitable contactor K which is closed by a conventional actuating mechanism 21 on command. The machine 1~ now generates alternating voltages at the line terminals of its 3-phase stator windings. The rms magnitude of the fundamental sinusoidal components of these voltages depends on the angular velocity (rpm) of the rotor and on the amount of field excitation. The generated voltages are applied to a-c input terminals of at least one 3-phase, double-way rectifier bridge 22, and the 3~
20-T~-1420 -- 1]. --recti~ied el~ctric power at the output terminals of each such bridge i6 supplied via a d-c bus to one or more d-c traction motors (not shown~. As is shown in Figure 1, the bridge 22 comprises simple solid~state diodes, but alternatively it could be a controlled rectifier if desired. As is suggested by the broken lines 23a and 23b, an additional traction motor (not shown) could be connected between the d-c terminals lOp and lOn if desired.
In the motoring mode o~ operation, which is assumed throughout the remainder of this description, the rotor 15 of the synchronous machine 14 drives the crankshaft of the engine 16. Now electric eneryy is supplied from the battery 17 to the windings on both the rotor and the stator of the machine, and the rotor 15 exerts torque to turn the crankshaft and -thereby crank the engine. As the rotor accelerates from rest, both the frequency and the rms magnitude of the fundamental alternating voltage waveforms developed at the line terminals of the stator windings (i.e., the back emf) correspondingly increase, while load current (i.e., current in the field and armature windings) decreases in magnitude. Once the rotor is rotating faster than a predetermined rate, which typically is 240 rpm, the engine is presumed to be started and the motoring mode (i.e., engine cranking mode) of operation is discontinued. Assuming that the machine 14 has ten poles, 240 rpm corresponds to a fundamental frequency of 20 Hertz. Thus the fundamental frequency of alternating current supplied to the stator windings of the machine 14 needs to increase from 0 to approximately 20 Hz in order for the illustrated system to per;Eorm its engine starting Eunction.
The previously mentioned third harmonic au~iliary impulse commuta-ted inverter is operative to ~! ~7~
20-T~-1420 convert direct current from the battery 17 into variable frequency alternating currents in the three different phases A, B, and C of the 3-phase armature windings on the stator of -the machine 14. The inverter has at least three pairs of alternately conducting main controllable electric valves interconnected and arranged in a 3-phase, double-way bridge configuration between the set of three a-c terminals 11, 12, and 13 and the pair o-f d-c terminals lOp and lOn. More particularly: a first pair of valves Tl and T4 are connected in series aiding fashion from terminal lOp to terminal lOn, and their juncture, comprising terminal 11, is connected to phase A of the stator windings; a second paix of valves T3 and T6 are connected in series aiding fashion from lOp and lOn, and their juncture, comprising terminal 12, is connected to phase B of the stator windings; and a third pair of valves T5 and T2 are connected in series aiding fashion from lOp to lOn, and their juncture, comprising terminal 13, is connected to Phase C. Each valve preferably comprises at least one solid state unidirectional controlled rectifier popularly known as a th~ristor. It haq a turned on (conducting) state and a turned off (non-conducting) state. In prac~ice the valves are respectively shunted by conventional snubber circui-ts (not shown). The illustrated means for connecting the d-c terminals lOp and lOn of the inverter to the battery 17 will next be described in more de-tail.
The first d-c terminal lOp is connected to the relatively positive terminal of the battery 17 via a single pole contactor Klp, and the second d-c terminal lOn is connected to the other terminal of the battery by means of a conductor 25, one pole K3a of a double-pole contactor K3, the field winding 18, the 7 ~

- 13 ~
other pole K3b of the same contactor, and a conductor 26. The -field winding 18 typically has a resistance in the ran~e of 0.12 to 0.2~ ohm and an unsaturated inductance of more than 0.3 henry.
single-pole contactor Kln, a conductor 27, and resistance means comprising two resistors 28 and 29 are connected in parallel circuit relationship with the ~ield winding 18 in the load current path between conductors 25 and 26. The resistors 28 and 29 are in series, and both have very low ohmic values; for example, the resistance of resistor 2~ is approximately 14 milliohms and the resistance of resistor 29 is approximately 23 milliohms. The second resistor 29 is shunted by another single-pole contactor K2 which, when c]osed, reduces the ohmic value o~ the resistance means to that of the ~irst resistor 28 alone.
An inductor 30 of approximately one millihenry inductance is connected in series with resistors 28 and 29 between the second resistor 29 and the conductor ~6 to smooth the current flowiny in this branch of the load current path. The inductor 30 is shunted by a conventional overvoltage protective device 31 the resistance of which is normally very high but automatically decreases to a negligible amount in substantially instantaneous response to the magnitude of voltage across the inductor rising to a predetermined breakover level (e.g., 750 volts). A
similar protective device 32 with bidirectional response is connected across the field winding 1~.
resistor 36 of significant ohmic value (e.g., 100 ohms) is also connected across the winding 1~ to enable thyristor "latching" current to bypass the ~ield and the inductor 30 when battery current starts flowing to precharg~ the inverter's commutation capacitor as will later be explained.
~ S i5 shown in Figure 1, a battery charger 33, in series with a blocking diode 3~ and a circuit breaker 35, is connected across ~he combination of battery 17 and inductor 30. With the engine 16 running under steacly-~state conditions, the battery charger holds the battery voltage at approximately 74 volts. It can be energized from any suitable source, such as auxiliary windinys (not shown) on the stator of the synchronous machine 14.
With the field winding 18 in the load current path during engine cranking, the synchronous machine 14 will operate with a characteristic similar to that of a series d-c motor, namely, high current and hence desirably high starting torque at low speeds. The resistance means 28, 29 in parallel with the field reduces the ohmic value of resistance that the field winding alone would otherwise introduce in this path, thereby initially allowing a higher magnitude of armature current and later, as speed increases, providing automatic field weakening which permits the machine to run at a higher speed.
Initially load current is limited by the internal resistance of the battery 17 as well as other resistance in its path, and as speed increases it is limited by the back emf of the armature (i.e., stator~
windings. Thus load current and torque tend to decrease with increasing speed. A short time after cranking commences, the contactor K2 is closed to further reduce the amount of resistance in parallel with the field, thereby permitting more load current to flow and hence more torque to be developed at higher speeds compared to the quantities that wo~lld be achieved if the parallel resistance were not so reduced.

5 ~

~ hen the cranking mode of operation commences, the contactor K is open, and all o~ the contactors in the load current path between the battery 17 and the d-c terminals lOp and lOn are closed excep-t K~. In a manner that will soon be explained, contactor K2 is commanded to close upon the expiration of a predetermined length of time after cranking commences. Thereafter, in response to the speed of the engine attaining a threshold (e.g., 240 rpm) that marks the conclusion of cranking and therefore the succesful starting of the engine 16, all of the previously clo~ed contactors are opened. Upon opening contactor K3 the field winding 18 is disconnected from the load current path between the conductors 25 and 26, and the contactor K is then closed by its actuating mechanism 21 in order to reconnect the field to the normal excitation source 20.
Each of the four contactors Klp, Kln, K2, and K3 has an associated actuating mechanism that determines its closed or open status. All four such mechanisms are represented in Figure 1 by a single block 38 labeled "Contactor Drivers," and they respectively respond to opening/closing signals received over lines 40, 41, 42, and 43 -from another block 44 labeled "Controller." The controller 44 knows the actual status of each contactor by virtue of feedback signals that it receives from conventional position indicators (not shown) that are associated with the separable contact members of the respective contactors, as represented symbolically by broken lines in Figure 1.
In order to turn on each of the controllable valves Tl through ~6 in the inverter, an appro~riate signal is applied to the associated gate while the main electrodes of that valve are forward biased 20-~rR-l420 (i.e., anode potential is positive with respect to cathode). Such a signal is sometimes called a trigger or gating signal, and it is herein referred to generically as a "firing signal." In a manner soon to be described, the controller 44 cyclically produces a series of periodic firing signals for turning on the respective main valves Tl-T6 in numerical order. lIt is assumed that the alternating voltages developed at the line terminals of the 3-phase stator winding6 of the machine 14 have the conventional A-B-C phase rotation.) In order to quench or turn off each valve when desired, the inverter has a forced commutation circuit including at least first and second auxiliary controllable electric valves Tp and Tn interconnected in series aiding fashion between the d-c terminals lOp and lOn and connected via a commutation capacitor 45 to the stator windings of the machine 14. The capacitor 45 is shunted by a bleeder resistor 46 which effectively keeps the capacitor initially in a substantially discharged state prior to closing the contactors Klp and Kln and starting up the illustrated system. Preferably, the commutation capacitor is connected between the juncture M of the auxiliary valves and the neutral S of the three star-connected stator windings.
In the manner previously explained under the heading "Background of the Invention," the main valves Tl-T6 in turn are forced to turn off by the commutation action that is initiated each time one or the other of the auxiliary valves Tp and Tn is turned on. The controller 44 is arranged cyclically to produce a series o~ periodic firing signals for alternately turning on the two auxiliary valves in synchronism with the variable frequency ~undamental component o~ the sinusoidal phase-to-phase alternating 5~
20~TR-1420 voltages that are developed at the line terminals of the respective phases A B, and C of the stator windings as the field winding 18 rotates inside the stator of the machine 14~ L~ote that the peak magnitude of reverse voltage imposed on the auxiliary valves can be reduced, if desired, by respectively inserting simple diodes in series therewith.
To produce the valve firing signals at proper times, the controller 44 needs to receive fron the power s~stem information or data indicating when the fundamental waveforms of line-to-neutral magnetic flux in the three phases A, B and C of the machine 14 cross zero and change polarity, and indicating the status of the electrostatic charge or voltage on the commutation capacitor 45. Such data are supplied by means of a voltage processor 38 which, as can be seen in Figure 1, has a plurality of input wires respectively connected to the line terminals o-f the s-tator windings and to opposite sides of the capacitor 45. Inside the processor 48 there is bistable first means for sensing the electrical potential difference across the commutation capacitor. Whenever the potential at the juncture M
is measurably positive with respect to the neutral S, the first means is in one state and provides a discrete signal (VEl) that is high or "1," but when this potential is measurably negative with respect to neutral the first means is in a diffarent state in which the output signal VEl is low or "0." Voltage sensors suitable -for this purpose are well known and readily available to a person skilled in the art~ The signal VEl is supplied over an output bus 50 to the controller 44. ~n additional bistable voltage sensing means is provided in the voltage processor 48 for detecting whether or not the capacitor voltage has a 5 ~ ~ ~

2~-TR-142 magnitude exceeding a predetermined level, either positive or negative. In one practical application of the illustrated system, the predetermined level is 400 volts. The additional sensor produces a discrete signal (VE2) on the output bus 50. As the commutation capacitor charges or recharges to a voltage magnitude in excess of the predetermined leve, the signal VE2 changes from a "0" to a "1" state.
The voltage processor ~8 also includes suitable means for integrating the respective line-to-neutral voltages of the stator windings and for indicating whether the polarity of the integral is positive or negative. The latter means provides three discrete output signals XA, XB and XC which are respactively supplied over lines 51, 52 and 53 to the controller 44. The output signal XA is high or "1"
only during the half cycles that the time integral of the voltage between the line terminal of phase A and the neutral S is relatively positive. It will be apparent that up and down changes of XA coincide with successive zero crossings of both the magnitude of line-to-neutral flux in phase A and the magnitude of the fundamental phase to-phase alternating voltage developed at the stator line terminals of phases C
and B (i.e., the line terminals to which the a-c terminals 13 and 12 of the inverter are respectively connected). Similarly, the output signal XB is "1"
only during the half cycles that the integral of the phase B-to-neutral voltage is relatively positive, whereby up and down changes of XB coincide with successive zero crossings of both the magnitude of line-to-neutral flux in phase B and the magnitude of the fundamental phase-A-to-phase-C alternating voltage developed at the stator line terminals to which the a-c terminals 11 and 13 are connected. In a similar 5~
20-~rR-l420 lg -manner, the output signal XC is "1" only during the positive half cycles o-f the integral of the phase C-to neutral voltage, whereby the up and down changes of XC coincide with successive zero crossings of both the magnitude of line-to-neutral flux in p~ase C and the magnitude of the fundaMental phase 8-to-phase-A
alternating voltage developed at the stator line terminals to which a-c terminals 12 and 11 are respectively connected. By logically processing the resulting output signals XA, XE~, and XC, the si~
different combinations of relative polarities o~ the three phase-to-phase voltages are indicated duriny each cycle of operation. Each time -the magnitude of any of these voltages crosses zero, a different one of the output signals changes either from 0 to 1 or from 1 to 0~
The controller 44 also communicates with master controls 54 via input and output busses 55 and 56. A starting switch 57 is associated with the master controls 54. The starting switch 57 can be either a pushbutton type or a turn-and-hold type.
The presentl~ preferred embodiment of the controller 4~ is shown in more detail in Figure lA.
Its main component is a microcomputer 60. Persons skilled in the art will understand that the microcomputer 60 is actually a coordinated system of commercially available components and associated electrical circuits and elements that can be programmed to perform a variety of desired functions.
It typically comprises a central processing unit (CPU) which executes an operating program permanently stored in a read-only memory (ROM) which also stores tables and data utilized in the program. Contained within the CPU are conventional counters, registers, accumulators, flag flip flops, etc. along with a precision oscillator which provides a hiyh-fre~uency clock signal. The microcomputer also includes a random access memory (I~M) into which data may be temporarily stored and from which data may be read at various address locations determined by the program stored in the ROM. The CPU, ROM, and RAM are interconnected by appropriate address, data, and control busses. In one practical embodiment of the invention, an Intel 8031 microprocessor is used.
The other blocks shown in Fiyure lA
represent conventional peripheral and interface components that interconnect the microcomputer 60 and the external circuits of Figure 1. More particularly, block 61 is an input/output circuit (I/O) for connecting the output bus and lines 50~53 of the voltage processor 4~ to the microcomputer 60, and block 62 is another I/O for connecting the microcomputer 60 to the contactor drivers 3~O
Block 63 is suitable means for decoding the position indicators that are respectively associated with the contactors Klp, Kln, K2, and K3. Block 64 is a gate pulse generator (GPG) and bufer that produces, on command of the microcomputer ~0, properly shaped and isolated firing signals that turn on the respective valves Tl-T6, Tp, and Tn.
The operation of the controller 44 during engine cranking can best be understood with the aid of Figures 2 and 3. In Figure 2, the sinusoidal waveforms of the Eundamental components of the three line-to-neutral voltages VAs, VBs, and Vcs of the 3-phase stator windings of the machine 14 are depicted by solid-line traces for a full cycle of steady-state operation, and the integral of one such ~Javeform (i.e., phase A) is shown by the broken-line trace. This integral is known t~ be in phase with the 20-T~ L420 flux of phase A. Assuming a s~mmetrical 3-phase machine and balanced loading, the zero crossings of the integral of VAs are seen to coincide with the moments of equality between VBS and Vcs, that is, with the zero crossings of the instantaneous magnitude of the fundamental phase-to-phase alternating voltage between the line terminals of phases B and C.
Consequently the discrete signal XA on output line 51 of the zero crossing detecting means in the voltage processor 48 is "1" throughout each half cycle o~
relatively positive polarity of the phase C-to-phase B
voltage and is "0" throughout each relatively negative half cycle thereof, the signal XB on the output line 52 is "1" throughout each relatively positive half cycle of the phase A-to-phase C voltage but is otherwise "O,"
and the output signal XC is "1" only during each relatively positive half cycle of the phase B-to-phase A voltage. The six different states of these three signals are marked off and numbered consecutively in Figure ~. For example, state 1 exists so long as both XA and XB but not XC are "1", whereas state 2 exists while X~ alone is "1". A state change is experienced each time any one of the signals X~, XB, or XC changes up or down, and each state coincides with a different 60-degree segment of a full cycle (360 electrical degrees) of the fundamental component of alternating voltage.
In a manner that will soon be described, in its third harmonic commutation mode of operation the controller 44 automatically responds to consecutive state changes of the signals XA, XB, and XC by producing a series of firing signals (which are represented by the pointers 68 and 6~ in Figure 2) for alternately turning on the two auxiliary valves Tp and Tn. More particularly, the controller is 7s~
20-Tl~-1420 effective to produce a firing siynal 69 for turning on the auxiliary valve Tn in immediate response to each change from an odd numbered state to the succeeding even numbered state (i.e., from state 1 to state 2, from state 3 to state 4, and from state 5 to state 6), and it is effective to produce a firing signal 6~ for turning on the auxiliary valve Tp in immediate response to each change from an even numbered st~te to the succeeding odd numbered state (i.e., from state 6 to state 1, from state 2 to st~te 3, and from state 4 to state 5). By thus synchronizing the firing signals 68 and 69 with the state changes (which are determined by the angular location oE the rotor 15 in the machine 14), the angle between the field mmf and the stator mmf of the machine is controlled.
As was previously explained, turning Oll an auxiliary valve causes load current immediately to transfer Erom an offgoing main valve to a parallel path including the turned-on auxiliary valve and the precharged commutation capacitor 45 which is first discharged and then recharged with reverse polarity by such current. As is indicated in Figure 2, the polarity of the capacitor voltage will change from positive (i.e., the potential at the juncture M is positive with respect to t~e neutral S) to negative as a result of the auxiliary valve Tn being turned on by one of the ~iring signals 69, and it will change from negative to positive when the auxiliary valve Tp is turned on by one of the firing signals G8.
Following the production of each of the firing signals 68 and 69, the controller 44 selectively produces the next one of a series of six firing signals (represented in Figure 2 by the pointers 71-7~) which are applied to the gates of the main valves Tl-T6, respectively. The controller 20-Tl~ 20 selects the proper firing signal to turn on whichever one of the main valves is associated with the oncoming or relieving phase of the stator voltages, whereupon load current ~an then transfer to the oncominy valve from the turned-on auxiliary valve. More particularly, as indicated in Figure 2, the controller selects the firing signal 71 for turning on the main valve Tl if the preceding state change was from state 6 to state 1, it selects the firing signal 72 1~ for turning on the main valve T2 if the preceding state change was from state 1 to state 2, it selects the firing signal 73 for turning on the main valve T3 if the preceding state change was from 2 to 3, it selects the firing signal 74 for turning on the main valve T4 if the preceding state change was from 3 to 4, it selects the firing signal 75 for turning on the main valve TS if the preceding state change was from 4 to 5, and it selects the firing signal 76 for turning on the main valve T6 if the preceding state change was from 5 to 6. Whichever one of the firing signals 71-76 is selected, it is not actually produced until after the first-mentioned bistable capacitor voltage sensing means in the voltage processor 48 changes state, as indicated by an up or down change of the discrete signal VEl on the output bus 50 of the processor 48. This is best seen in Figure 3 which will now be described.
Figure 3 shows the instantaneous magnitudes of capacitor voltage (VMs) and current (icAp) during two consecutive commutation intervals. The first of these two intervals is initiated at time tl when the means for detecting the zero crossings of phase-to-phase voltages changes from an odd state to an even state and the controller responds by producing a firing signal 69 to turn on the auxiliary valve Tn, 7 ~., ~
20-rrR-l420 and the second commutation interval is initiated at time t6 when a firing signal 68 for the auxiliary valve Tp is next produced in response to the same detector changing from even to odd states. Once the first commutation interval is initiated, current begins to increase in the auxiliary valve Tn and in the commutation capacitor 45, while current in th0 offgoing main valve decreases to zero at time t2 which occurs when all of the load current has transferred to the parallel commutation circuit. The resulting pulse of current in the commutation capacitor first discharges it and then recharges it with reverse polarity. At time t3 the capacitor is fully discharged, whereupon the discrete signal VEl chanyes from its initial "1" state to a different state. The time from t2 to t3 is the circuit turn off time during which the offgoing main valve recovers its ability to withstand reapplied forward voltage. The next one (7X) of the series of firing signals 71-76 for the main valves Tl-T6 is produced in response to the 1-to-0 change of VE1.
In accordance with the present invention, the controller 44 includes time delay means effective at least after a predetermined initial period of time for delaying the production of the next firing signal 7X until a programmed interval of time has elapsed following the 1-to-0 change of VEl~ This interval is designated by the delta t- symbol in Figure 3, and it elapses at time t4. Now the oncoming main valve TX is turned on, and load current begins transferring to it from tha parallel commutation circuit while continuing to recharge the capacitor 45. At time t5 all of the load current has transferred to the oncoming main valve, the auxiliary valve Tn turns off, and the first commutation interval ) r;~3~
20-T~-1420 is finished. The peak magnitude of capacitor voltage (i.e., its magnitude at time t5) is a function of the magnitude of load current that recharges the capacitor during the delta t- delay interval. As load current decreases, the peak magnitude of capacitor voltage will decrease, and consequently the length of the commutation interval is desirably reduced. From t5 to t6 the commutation capacitor will retain a voltage of su~icient magnitude and proper polarity (neyative) to ensure successful commutation when the next zero crossing of phase-to-phase voltages initiates the second commutation interval.
As is apparent in Figure 3, the ~.econd commutation interval is essentially a dual o~ the first. In this case the programmed delay interval (i.e., the time from the 0-to-1 change of the discrete signal VEl to the production o~ the firing signal 7X
for turning on the next oncoming main valve) is designated delta t+, and its duration can either be the same as or differ from the duration of delta t-.
In practice, the programmer will select delay intervals that are compatible with the parameters of the power system and the ratings of its components.
In the preferred embodiment, after a predetermined initial period of time from the start of engine cranking, alternate delay intervals (delta t-) are programmed to have a shorter duration than intermediate delay intervals (delta t~).
Figures 4 through 11 display flow charts o-f the presently preferred programs that are executed by the microcomputer 60 in the controller 44 in order to produce ~iring signals that enable the inverter to operate in a third harmonic commutation mode for purposes of cranking the engine 16. The Main Routine is shown in Yigure 4. It begins at the entry poirlt S ~
20~T~-1420 labeled "Start". ~len commanded to start, the first step 80 oE the Main Routine is to initialize the various inputs to the microcomputer 60, to reset its timers to 0, to decrement its counters to 0, and to set the stack pointers, registers, latches, outputs, and variable values of the microcomputer to their respective quiescent states or normal levels at the start of the first pass through the Main Routine.
Upon completing this initializing step, the program determines, at a decision point ~31, whether or not the start switch 57 is "on". Assuming the start switch is turned on or closed (which will happen at a time when the rotor of the synchronous machine 14 i5 at rest and the commutation capacitor 45 is discharged), the control proceeds to a step 82 in which a first timer is started~ By way of example, this timer will run for approximately 90 seconds after being started.
Once timer ~1 has been started, a Set-Up Routine 83 is executed, and this is followed by the execution of a Normal Cranking Routine 84.
The Set-Up Routine 83 is shown in Figure 5.
Its purpose is to control the contactors and valves of the power system (Figure 1) so as to: (1) precharge the commutation capacitor 45, (2) initiate a pulse of excitation current from the battery 17 through the field winding of ~he machine 14 and find the initial angular position of the rotor 15, and (3) ensure that the capacitor voltage has the right polarity for successful third harmonic commutation once the ~ormal Cranking Routine is initiated. The Set-Up Routine is entered at a point labeled "Set-Up" and then proceeds to a step 85 which causes the controller 44 to issue signals, via lines 40 43, that command the contactor actuating mechanisms (38) to close the four contactors Klp, Kln, K2, and K3. This step is followed by an inquiry, at point 86, as to the open or closed status of the contactors. ~nce all ~our are actually closed, a "Contactor Error" flag is set in an "off" state, and the control is transferred to a Capacitor Ring-up Subroutine 87 which will soon be described. If all four contactors do not close in response to the closing commands of step 85, they are commanded to open (step 88), the Contactor ~rror flag is set in its "on" state, and the Set-up Routine is aborted at the stop point ~9.
The presently preferred embodiment of the Capacitor Ring-up Subroutine 87 is shown in Figure 6.
While this subroutine is being e~ecuted, the controller 44 will produce firing signals that cause the Figure 1 system to operate in a capacitor "ring-up" mode that precharges the commutation capacitor. It is entered at a point labeled "Ring-up"
and then proceeds to a step 91 which sets a counter in the microcomputer 60 at a predetermined maximum number of ring cycles (e.g., 20 cycles). Step 91 is followed by a step 92 in which a second timer is started. ~y way of example, this timer will run for an interval of approximately 10 milliseconds a~ter being started~
From step 92 the program proceeds to an inquiry point 93 where the state of the first bistable capacitor voltage sensor in the voltage processor 4 is tested. If the commutation capacitor has a mçasurably positive voltage (i~e., the juncture M has a positive potential with respect to the neutral S and the amount of potential difference exceeds a predetermined threshold such as 5 volts), the first voltage sensor is in one state (which is indicated by VEl being high) and the inquiry yields an a~firmative answer. On the other hand, if the capacitor voltage 3~ were measurably negative (i.e., the potential at 20-T~-1420 juncture M is more than 5 volts negative with respect to the neutral S), the voltage sensor is in a different state (as indicated by VEl being low) and the answer to the inquiry is "no".
In response to an affirmative answer at the inquiry point 93, the next step 94 in the program is to instruct the controller's gate pulse generator ~4 to generate a first pair of concurrent firing signals ~or the auxiliary valve Tn and for a preselected complementary one of the main valves (e.g., Tl). In response to a negative answer at 93, the Capacitor Ring~up Subroutine alternatively proceeds to a step 95 in which the gate pulse genera-tor is instructed to generate a second pair of concurrent firing signals for the auxiliary valve Tp and for another preselected complementary one of the main valves (e.g., T~). In practice, each firing signal that is generated in step 94 or 95 can actually comprise a burst of several high-frequency, short-duration discrete d-c signals having sufficient magnitude to turn on the associated valve.
Upon turning on either the complementary pair of valves Tn and Tl or the complementary pair Tp and T2, battery current will begin flowing through a path which in Figure 1 is seen to comprise: (1) the field windiny 18 in parallel with both the resistor 36 and the series combination of resistor 28, inductor 30 and the closed contactor ~2, (2) one phase of the armature windings of the machine 14, and (3) the commutation capacitor 45. Preferably this path has a sufficiently high Q so that current quickly oscillates from zero to a peak magnitude and back to zero, and in the process the capacitor is incrementally charged with reverse polarity. The conducting pair of valves will automatically turn off by self commutation when 75~i9~
20-TR-1~0 current oscillates to zero at the conclusion of ~ach cycle of this ringing action. The resulting pulse of current typically has a duration of less than two milliseconds.
In the Capacitor Riny-up Subroutine, the status of the second timer is tested immediately after either step 94 or step 95. This testing step is indicated in Figure 6 by the inquiry point 96.
Assuming that timer #2 is still running, the next step 97 in the program is to check the state o~ the second bistable capacitor voltage sensor in the voltage processor ~8. So long as the ma~nitude of capacitor voltage does not exceed a predetermined maximum (i.e., the level at which the output signal VE2 of the second voltage sensor changes from "0"
to "1"), step 97 yields a negative answer, and the control returns to the preceding step 96. Whenever timer ~2 stops running (i.e., its time delay interval is over), the control proceeds from step 96 to a step 98 in which the count stored in the cycle counter (see step 91) is reduced by one, and then to a step 99 determines whether or not the count has reached 0. I~
not, the control returns to step 92 t and the steps 92 through 99 are recycled. In this manner the controller repeatedly produces the aforesaid second pair of firing signals (for turning on Tp and T2) if it is determined in step 93 that the capacitor voltage is not positive, and the aforesaid first pair of firing signals (for turning on Tn and Tl) if the voltage is positive. The start of each such repeated cycle of operation is delayed by an interval of time determined by timer #2. This interval is sufficiently long to enable the pulse of battery current to first discharge the commutation capacitor and then incrementally recharge it with reverse polarity until 3~

20-TR~ 0 the current oscillates to zero. At the end oE each consecutive cycle, the electrostatic charge that is stored in the capacitor (and hence the capacitor voltage) will be progressively increased in magnitude due to the ringin~ nature of the charging current path (the inductance of which is provided by the fiel~ and armature windings of the machine 14), and it will have alternately positive and negative polarity. This action continues for a su~-ficient number of cycles to enable the magnitude of capacitor voltage to attain the aforesaid maximum at which the in~uiry step 97 yields an affirmative answer, whereupon the second timer is stopped (step 101) and, after waiting a very short, fixed period of time (step 102), the control proceeds to execute one final ring-up cycle of operation.
As is shown in Figure 6, the final cycle of the Capacitor Ring-up Subroutine is carried out by steps 103, 104, and 105 ~hich are duplicates of the previously described steps 93, 94, and 95, respectively. Follo~ing the generation of the last pair of concurrent firing signals (either for the complementary valves Tn and Tl or ~or the complementary valves Tp and T2), timer #2 is cleared (step 106), and then the control returns to the Set-up Routine (Figure 5). I'he total number of ring-up cycles that are carried out by the Capacitor Ring-up Subroutine is sufficient for the commutation capacitor to charge to a voltage magnitude many times (i.e., more than approximately five times) higher than the average magnitude of voltage at the terminals of the battery 17. In one practical embodiment, ~he fully charged voltage rating of the battery 17 is 74 volts, and in four or five cycles the commutation capacitor can be charged to a voltage magnitude exceeding 400 s~

voltsO If for any reason the capacitor were not charged to the desired level within the maximum number of ring cycles that was set at step 91, the count in the cycle counter would reach 0 (step 99) before the second capacitor voltage sensor detects maximuM
voltage (step 97), and in this abnormal event all of the contactors are commanded to open (s-tep 107) and a "No Ring-up" signal is issued (step 108).
After execution of the Capacitor Ring-up Subroutine 87, and with the commutation capacitor now precharged, the Set-up Routine 83 continues as shown in Figure 5. The next step 110 is to start a third timer. By way of example, this timer will run for an interval of approximately 250 milliseconds after being started. From step 110 the control proceeds to a step 111 which causes the controller ~4 to command the contactor actuating mechanisms to open both o-f the contactors Kln and K2. X'his is followed by testing, at point 112, the status of the third timer and by testing, at point 113, the open or closed status of the contactors ~ln and K2. As soon as both of these contactors actually open, but no later than the time at which timer #3 stops running, the control proceeds to a step 114 in which the gate pulse generator is instructed to generate concurrent firing signals for turning on a preselected pair of valves that will provide a path for excitation current from the battery 17 through the field winding 1~ of the synchronous machine 14. Any appropriate pair of valves can be turned on for this purpose. The auxiliary valves Tp and Tn were selected in the illustrated embodiment. Once these valves are turned on, field current starts increasing or ramping up from zero.

S~

20-TR-1~20 After generating the firing signals for the auxiliary valves Tp and Tn (step 114), the Set-up Routine proceeds to an inquiry point llS where the status of timer #3 is tested again~ If the time delay interval of this timer is overl the control proceeds Erom step 115 to a step 116 in which the contactor Kln is commanded to reclose~ When Kln recloses, the series resistors 28 and 29 are reconnected in parallel with the field winding 18 (Figure 1), thereby permitting a more rapid increase of current in the auxiliary valves. Step 116 is followed b~ a step 117 which introduces an additional delay (e.g., a fixed period of approximately 300 milliseconds) to allow excitation current in the field windiny 18 to continue increasing. Upon the expiration o~ this additional delay (at which time current in Tp and Tn may have attained a magnitude as high as 1,300 amperes), the control is transferred to a subroutine 118 which will soon be explained. Executing the subroutine 118 completes the Set-up Routine (Figure 5), and the control will then return to the Main Routine (Figure 4).
rrhe presently preferred embodiment of the subroutine 11~ is shown in Figure 7. While this subroutine is being executed, the controller 4~ finds the initial or at-rest position of the rotor 15, and it produces firing signals that will turn on the proper main valve(s) for commutating (turning o~f) the auxiliary valves rrp and Tn and for obtaining the right polarity of voltage on the commutation capacitor.
Note that at the time this subroutine is entered, excitation current is rising in the field winding 18 on the rotor 15 of the machine 14. Consequently the field winding (rotor) generates magnetic flux of increasing magnitude, and this changing flux in turn s~j~3~

20-T~-1420 interacts with the three phases A, B, and C of the armature windinys (stator) to induce therein line to-neutral voltages that can be sensed and integrated by the voltaye processor 48. The initial angular position of the rotor can be deduced from knowledge o~ the relative polarities oE the kime integrals of these three voltages.
As is indicated in Fiyure 7, the subroutine 118 is entered at a point :Labeled "Find Rotor Position.O.." and then proceeds to a step 120 in which the microcomputer 60: (1) reads data representing the high or low conditions of the three phase-to-phase stator voltage zero-crossing detecting signals XA, XB, and XC on the output lines 51, 52, and 53 of the voltage processor 48, (2) uses a loyical combination of this data to find, in an appropriately encoded look-up table, whichever one of the six different states 1 through 6 (see Figure 2) is extant, and (3) places the number of the extant or actual state in a selected register of its memory where this nu~ber is saved as "oldstate". From step 120 the subroutine proceeds to a decision point 121 which determines whether or not the oldstate is even (i.ea, 2, 4, or ~). If the oldstate is even, the next step 122 in the program inquires as to the polarity of the voltage on the precharged commutation capacitor (as indicated by the high or low state of the discrete output signal VEl from the first voltage sensor in the voltage processor 48), and if the polarity is positive the control proceeds to a step 123. On the other hand, if it is determined that the oldstate is not even (i.e., i9 1, 3, or 5), the subroutine alternatively proc~eds from step 121 to another inquiry point 124 which is a duplicate of the inquiry 122, and if the polarity is not positive the 7 r~
20-~rR-l~20 control proceeds to the same step 123.
In s~ep 123 the microcomputer reads the oldstate, increments it by 1 to find the number o-f the succeeding state, and identifies the proper one ~TX) of the six main valves that is scheduled to be turned on in response to the next state change. For example, if oldstate = 2, the succeeding state is 3 and TX
is T3. Step 123 is immediately followed by a ste~ 125 in which the controller 44 is i:nstructed to generate a firing signal for the identified valve TX. Once TX is turned on by this firing signal, it completes a path for battery current in parallel with a first one of the two auxi.liary valves that were turned on by step 114 oE the Set-Up Routine 83. The parallel path includes one phase of the stator windings and the precharged commutation capacitor, and the capacitor voltage will have the proper polarity to provide commutating action that quickly forces the -first auxiliary valve to turn off. I'he current flowing in this path and through the other auxiliary valve xises to a peak magnitude and then decays to zero (in the process of which the commutation capacitor will be discharged and then recharged with reverse polarity), and both TX and the second auxiliary valve will automatically turn off by self commutation when the current oscillates to zero at the conclusion of this ringing action. Now the capacitor voltage has the correct polarity for successful commutation of the main valves at the next zero crossing of the phase-to-phase stator voltages during the engine cranking mode o~ operation. Figure 2 illustrates that the correct polarity is negative duriny even numbered states and positive during odd numbered states.
After step 125 causes the controller to produce a firing signal for turning on TX, and while s~

the above-described rinying action i5 taking place, the field winding 1~ of the machine 14 continues to be excited by current which now "freewheels" through the contactor Kln and the branch of the load current path joining conductors 26 and 27 (E'igure 1). At the same time, the subroutine shown in Figure 7 proceeds to in~uire, at point 126, as to whether or not the capacitor voltage has changed polarity (as indicated by a state change of V~l). As soon as the answer is affirmative, the control returns to the Main Routine, and the Normal Cranking Routine 84 can begin.
As is shown in Figure 7, there is an alternative way to get to the last inquiry point 126 of the subroutine 118. It includes two steps 127 and 128 that are similar to steps 123 and 125, respectively, except that step 127 increments the oldstate by 2 and id~ntifies the main valve that is scheduled to be turned on in response to the second state change to come. For example, if oldstate = 2, oldstate ~ 2 - 4, and TX is T4. Step 127 is init~ated if the answer to inquiry 122 indicates that the capacitor voltaye is not positive, or if the answer to inquiry 124 indicates that the capacitor voltage is positive. It is immediately followed by the step 128 which causes the controller to generate a firing signal for the main valve TX. When turned on, TX
connects the commutation capacitor across a first one of the conducting auxiliary valves Tp and Tn, whereupon the first auxiliary valve is forced to turn off, the commutation capacitor is discharged then recharyed wi~h reverse polarity, and both TX and the other auxiliary valve will turn off as soon as the battery current oscillates to zero. At this time the polarity of the capacitor voltaye will not be correct for successful commutation of the main valvPs when ~ ~7~

- 3~ -cranking starts unless prior thereto a series o~
additional steps 130 through 136 are followed to reverse this polarity from wrong to right.
Step 130, which immediately follows step 128, performs the same inquiry as point 126, and as soon as the capacitor voltage changes polarity the control proceeeds to step 131. In step 131 the microcomputer reads the oldstate, decrements it by 1 to ~ind the number of the precediny state, and identifies the proper pair (TX, TY) of the six main valves that are normally conducting load current during such preceding state. For example, if oldstate = 2, the preceding state is 1 and the valve pair TX, TY are T6 and Tl, respectively. Step 131 is immediately followed by the step 132 in ~hich the controller is instructed to yenerate concurrent firing signals for the valves TX
and TY. Wherl the valves TX and TY are turned Oll by these firing signals, they enable battery current to resume flowing in the armature windings of the machine 14. After the firing signals are generated (step 132), the control waits for a period of approximately 30 milliseconds to allow excitation current in the field winding 18 (step 133) to increase in magnitude, and it then proceeds to a decision point 134 which detPrmines whether or not the ol~state is an even number. If the oldstats is even, the next step 135 is to generate a firing siynal for turniny on the auxiliary valve Tn, if the oldstate is not even, an alternative step 136 is implemented to generate a firing signal for turning on the auxiliary valve Tp, and in either case the control proceeds from step 135 or 136 to the previously described inquiry point 126.
Once valve Tn (or Tp) is turned on by step 135 (or 136), it connects the precharged commutation capacitor across the main valve TX which is thereby ~''7~~

forced commutated off. The current that transfers from TX to the auxiliary valve first discharges and then recharges the commutation capacitor with opposite polarity, and both TY and the auxiliary valve will automatically turn off by self commutation as this current oscillates to zero. Now the capacitor voltaye has the right polarity for successful commutation of the main valves when cranking starts. As was previously explained, the control transfers to the ~ormal Cranking Routine 84 as soon as the inquiry point 126 detects a change in the polarity of the voltage on the commutation capacitor.
The Normal Cranking Routine ~4 will cause the controller 44 to produce the proper pa~tern of properly synchronized firing signals as required for the inverter to operate in a third harmonic commutation mode, whereby mechanical torque is developed in the rotor of the machine 1~ to start turning the crankshaft of the engine 16 and to accelerate it ~rom rest to a predetermined speed (e.g., 240 rpm) well above the minimum "firing speed"
of the engine. The presently preferred embodiment of this routine is shown in Figure ~. In its first step 140, the oldstate that was saved in the memory of the microcomputer 60 is used to find, in an appropriately encoded look-up table, the identity of the pair of main valves (TX, TY) that normally would be conducting load current during such state. For example, if oldstate = 2, TX is Tl, and TY is T2.
Step 140 is followed immediately by a step 141 in which the controller is instructed to generate the proper pair of firing signals (e~g., 71 and 72) to turn on both of the valves TX and TY. In practice, each firing si~nal can actually comprise a 35-microsecond burst of from five to ten ~7S~
20-1'R-142() - 3~ -short-duration (1~5 microseconds) discrete d-c signals having sufficient magnitude to turn on the associated valve.
Once the main valves TX an~ TY are turned on by step 141, they complete a path for load current to flow from the batter~ 17 through two phases of the armature (stator) windings of the machine 14, and through the section of the path that interconnects conductors 25 and 26 (Figure 1)~ The latter section comprises the field winding 18 (which was inserted therein when the two poles K3a and K3b of contactor K3 were closed by step 85 of the Set-Up ~outine 83) arld the parallel branch that includes conductor 27, resistor 28, resistor 29 shunted b~ contactor K2, and inductor 30. Since K2 opened at step 111 of the Set-Up Routine and has not yet reclosed, both resistors 2~
and 29 are now effectively in series in this parallel branch. The magnitude of armature current is initially very high, limited only by the internal resistance of the battery, the negligible resistance of the armature windings, and the total resistance of the two resistors 28 and 29. At the time TX and TY
start conducting load current, the field winding 18 is being excited by the residual of the current that had previously built up therein during the interval of time between the execution of step 114 (Figure 5) and the execution of step 125, 135 or 136 (Figure 7) in the Set-Up Routine. The magnetic fields generated by current in the armature windings now intereact with 3~ the excitation current in the field winding to produce in the rotor 15 a torque (proportional to the product of the magnitudes of these currents) that tends to turn the crankshaft of the engine 16 in the desired direction.

~7~

In one practical embodiment, currents in the armature and field windings were high enough, with the battery not fully charged, to produce a "breakaway"
torque of at least 3,600 foot-pounds which is S sufficient to turn the crankshaft of a 4,000 horsepower diesel engine. As the crankshaft and rotor start xotating, current (and torque) tends to decrease in magnitude due to the rising amplitude of the back emf that is induced in the armature windings and that opposes the battery voltage. The instantaneous magnitude of the back emf in each of the three phases of the synchronous machine will alternate sirlusoidally between relatively positive and negative peaks as the rotor accelerates Erom rest and its angular position advances. In due course the rotor will pass through a location where the increasing voltage magnitude of the oncoming or relieving phase (e.g., B) just equals the decreasing voltage magnitude of the offgoing or relieved phase (e.g., A), whereupon one of the three zero-crossing detecting signals (e.g., XC) will change up or down to mark the transition to the next state.
As is shown in Figure 8, after generating the firing signals for valves TX and TY (step 141), the Normal Cranking Routine proceeds to a step 142 in which a K2 timer is started. This timer will run for a predetermined length of time (e~g.~ approximately three seconds) after being started. Step 142 is followed by a step 1~3 in which the microcomputer:
(1) reads data representing the hi~h or low conditions of the signals XA, XB, and ~C, (2~ uses a logical combination of this data to find, in an appropriate encoded look-up table, whichever one of the six different states 1 through 6 is extant, and (3) places the number of the extant or actual state in its memory where this number is saved as "new state." From 20-TR-1~20 step 1~3 the control proceeds to an in~uiry point 14 which determines whether or not the new state is the same as oldstate. So long as the answer is affirmative, the control next inquires, at a point 1~5, as to wllether or not a cranking "finish"
flay is on, and if not it then proceeds to an inquiry point 146 where the open or closed status of contactor K2 is tested. If K2 is closed, the control returns to step 143, if not, the status of the K2 timer is tested at point 147. Assuming tha~ the K2 timer is still runninc~, the control immediately returns to step 143. Otherwise, the control proceeds ~rom the in~uiry point 147 to a step 148 in which the contactor K2 is commanded to close, whereupon the contxol returns to stap 143. It will now be apparent that so long as the inquiry step 144 determines that the new state is the same as oldstate, the control steps repetitively around a loop comprising the step 143 and inquiry points 144, 145, 146, and 147 while the K2 timer is running, whereas, it steps repetitively around a subloop comprising step 143 and inquiry points 144, 145 and 146 after the lenyth o~
time programmed in the K2 timer is overO In response to the expiration of this predetermined length o~
time, step 148 is implemented to close the contactor K2 which then short circuits the resistor 29 (Figure 1), thereby reducing the ohmic value o~ the resistance in parallel with the field winding 180 As a result, more current can flow in the branch of the load current path between conductors 26 and 27, the field excitation is wea~ened, and higher cranking speeds can be achieved.
As soon as the inquiry point 14~ determines that the new state is not the same as the oldstate (i.e., in response to a state change of one of the 20-T~-1420 phase-to-phase stator voltage zero-crossing detecting signals XA, X~, and XC), the control transfers from the above-described loop to an alternative loop comprising the step 143, the inquiry p~int 1~4, and, in the following order, a "Commutation Subroutine" 1~1, a "Next State Subroutine" 152, an inquiry point 153, and a "System Synch ~ubroutine" 154. The three subroutines 151, 152, and 154 are therefore executed each time the inquiry point 144 detects a ~tate change.
The Commutation Subroutine 151 of the ~ormal Cranking Routine 84 is shown in Figure 9. It has two functions: (1) to initiate commutation of the offgoing main valve by ordering the production o~ a firing signal for the appropriate one o~ the two auxiliary valves Tp and Tn, and (2) to delay the execution of the Next State Subroutine 152 until a programmed interval of time has elapsed following each resulting zero crossing of the voltage on the commutation capacitor ~5. The Commutation Subroutine is entered at a point la~eled "Commutation" and then proceeds to a decision point 156 which determines whether or not oldstate was an odd number. If oldstate was odd (i.e., 1, 3 or 5), the control proceeds to a step 157 in which the controller is instructed to produce a Eiring signal 69 for turning on the auxiliary valve Tn. If oldstate was even (2, or 6), the control proceeds from point 156 to a step 158 in which the controller is instructed to produce a firing signal 68 for turning on the auxiliary valve Tp. Consequently one or the other of the auxiliary valves is turned on to connect the commutation capacitor across the offyoing main valve.
The capacitor voltage will now have the correct polarity to force load current to transfer to the conducting auxiliary valve, whereupon the offgoing 6q~

main valve stops con~ucting. The load current in the commutation circuit first discharges the capacitor and then recharges it with reverse polarity, as i8 illustrated in the previously described Figure 3.
More specifically, if oldstate was odd, Tn is fired, and at time t3 the capacitor voltage will change from positive to negative as indicated by a high-to-low state change of the output signal VEl of the first bistable capacitor voltage sensor in the voltage processor 48. On the other hand, if oldstate was even, Tp is fired, and at time t~ the capacitor voltage will change from negative to positive as indicated by a low-to-high state change o-f VE1.
As is shown in Figure 9, step 157 (or 158) of the Commutation Subroutine 151 is immediately followed by an inquiry, at point 160 (ox 161), as to the high or low state of the capacitor voltage polarity indicating signal VEl. As soon as VEl changes state, the control proceeds from point 160 (or 161) to another inquiry point 162 where the status of the K2 timer is measured. For a predetermined initial period of time, which starts when step 142 of the Normal Cranking Routine 84 starts the K2 timer, the inquiry 162 yields a negative answer, and thereafter the answer will be affirma-tive. The initial period is preferably approximately 1.5 seconds, or approximately half of the length of time that the K2 timer is programmed to run.
In response to a negative answer at the inquiry point 162, the next step 163 in the program is to load a predetermined maximum capacitor recharging time into a fourth timer. Step 163 is followed immediately by a step 164 that starts tlle fourth timer and then by an inquiry, at a point 165, as to the state o~ the second capaci-tor voltaye sensor in the S~

voltage processor ~, as indicated by the high or low state of the signal VE~. So long as the capacitor voltage is lower than the predetermined maximum level (e.g., ~00 volts), the answer to inquiry 165 is negative, and the control proceeds to an inquiry point 166 where the status of the fourth timer is tested. So long as the timer #4 is still running, the answer to inquiry 166 is negative, and the control r~turns to the preceding inquiry point 165. But whenever an affirmative answer is obtained at either inquiry point 165 (revealing that the magnitude o~
capacitor voltage has attained the aforesaid maximum) or inquiry point 166 (revealing that the time delay interval of the timer #4 is over), whichever is ~irst to occur, the Commutation Subroutine 151 is exited, and the control is transferred to the Next State Subroutine 152 of the Normal Cranking Routine.
Preferably the maximum recharge time that is loaded into timer #4 in step 163 is selected to have a relatively long, fixed duration (e.g., approximately 1.2 milliseconds) so that an affirmative answer will ordinarily be obtained from inquiry 165 earlier than from inquiry 166 throughout the aforesaid initial pariod (i.e., before the inquiry point 162 yields an affirmative answer). In other words, whenever the Commutation Subroutine is executed during the initial period of time, it will be completed as soon as the voltage on the commutation capacitor reverses polarity and rises in magnitude to the aforesaid maximum level (but no later than the expiration of the maximum recharging interval that was loaded in timer #4 at step 163~. ~igh capacitor voltage is required for successful commutation during this period when load currant is relatively high. At the same time, the relatively long maximum capacitor recharging time is 20-TR-1~20 permissible because the rotor speed (and hence the frequency of state changes) is now relatively low.
In response to an affirmative answer at the inquiry point 162 (which will be true anytime the Commutation Subroutine 151 is executed after expiration of the aforesaid initial period of time), the program shown in Figure 9 proceeds from point 162 to a decision point 1~7 which is a cluplicate of 156.
If oldstate was odd, the next step 168 is to load a "negative" capacitor recharging time into timer ~4.
Alternatively, if oldstate was even, the next step 169 is to load a "positive" capacitor rP~harging time into the same timer. In either case, the control then proceeds, as before, to start timer ~4 at step 164 and then repetitively to check for maximum capacitor voltaye at point 165 and to test the status of the timer at point 166. In accordance with the present invention, the delay intervals that are loadsd into timer #4 at steps 168 and 169 are shorter than ~he maximum capacitor recharging intexval that is loaded at step 163. While both intervals could be equal to each other if desired, in the illustrated embodiment the negative recharging interval is shorter than the positive recharging interval. By way of example, the negative recharging interval is approximately 300 microseconds, and the positive recharging interval is approximately 500 microseconds. These intervals, which in Figure 3 are respectively represente~ by the delta t- and t+ symbols, are sufficiently short so that, after the aforesaid initial period expires (i.e., when the inquiry point 162 yields an affirmative answer), an affirmative answer will ordinarily be obtained from inquiry 166 earlier than from inquiry 165. In other words, whenever the Commutation Subroutine is executed after the initial ~.~7~ ~

20-T~-1420 period of ki~e, it is completed as soon as the delay interval that was loaded in timer #4 at step 168 or 169 i5 over, and this does not provide enough time for the commutation capacitor to recharge to the aforesaid maximum level of voltage. Consequently, as was explained hereinbefore in conne,ction with the description of ~igure 3, the actual magnitude of capacitor voltage at the conclusion of ~he Commutation Subroutine i~ a function of the magnitude of load current. It decreases with current, and as a result the length of the co~nmutation interval is desirahly reduced as the rotor speed (and frequency) increases after the aforesaid initial period of time.
The Next State Subroutine 152, which is executed immediately after the Commutation Subroutine 151, is shown in Figure 10. Its purposes are (1) to calculate and save the "next state" and (2) to complete the third harmonic commutation process by ordering the production of a firing signal for the oncomin~ or relieving main valve. rrhis subroutine is entered at a point labeled "Next State" and then proceeds to a step 171 in which the microcomputer (1) subtracts the previously saved oldstate (step 120 in Figure 7) from the previously saved new state (step 143 in Figure 8) to find the difference therebetween, the difference being ~1 for the assumed phase rotation A B-C of stator voltages but -1 if ~he phase rotation were C-B-C, (2) increments the oldstate by one if the difference is +l (or decrements it if the difference were -1) to give the number of the "next state", i.e., the state which comes after oldstate and which therefore should coincide with the new state, and (3) places the number of the "next state" in the selected register of its memory where this number replaces the ~reviously saved oldstate and ~,,"~'~d JJ ~

~ 46 -is saved as a calculated new oldstate. Step 171 is followed by a step 172 in which the new oldstate is used to find, in an appropriately encoded look-up table, the identity of the pair of main valves (TX, TY) that normally should be conducting load current during such state. For example, if new oldstate = 3, TX is T2, and TY is T3. r~y is the oncoming or relieving valve of the pair. Step 172 is followed immediately by a step 173 in which the controller is instructed to generate the proper pair of firing signals (e.g., 72 and 73) to turn on the identified valves TX and TY. It will be apparen-~ that the firing signal for valve TX is redundant, as rrx in this subroutine is the same valve as TY which was turned on earlier in the program. Once the oncoming valve TY is turned on by a firing signal produced at step 173, load current can transfer to it from the auxiliary valve that was turned on by step 157 or 158 of the Commutation Subroutine 151 (Figure 9), thereby completing the commutation process. While load current is decaying to zero in the auxiliary valve, the commutation capacitor continues recharging to a peak magnitude somewhat beyond its level of voltage at the conclusion of the Commutation Subroutine.
The firing signal that is produced by step 173 of the Next State Subroutine 152 for the oncoming main valve TY is represented in Figure 3 by the pointer 7X. It is produced at time t4 if the calculated new oldstate is an even number and at time t9 if odd. Both of these times are delayed with respect to the preceding state change of the capacitor voltage polarity indicating signal VEl (as detected by the inquiry point 160 or 161 of the Commutation Subroutine 151). Each time the Next State Subroutine is executed during the aforesaid initial period (which - ~7 -is determined by step 162 of the Commutation Subroutine), this delay will depend on how long the commutation capacitor takes to recharge to the vol-tage level at which an affirmative answer is obtained at the inquiry point 165 in the Commutation Subroutine.
But each time the Next State Subroutine is executed after the initial period, the firing signal 7X is delayed until a programmed interval of time has elapsed. If the new oldstate is an even number, the proyrammed delay interval is determined by the negative recharging time loaded into timer ~4 at step 168 of the Commutation Subroutine, and otherwise it is determined by the positive recharging time loaded into the same timer at step 169.
Having generated the firing signal (7X) for the main valve TY at step 173, the Next State Subroutine 152 returns to the Normal Cranking ~outine (Figure 8) where the calculated new oldstate (step 171 in Figure 10) is checked, at the inquiry point 153, to be sure that it is in fact the same as the previously saved new state (step 143 in Figure 8). If not, after waiting for a fixed period of approximately 50 milliseconds (step 174), the control is retransferred to the Commutation Subroutine 151, and the two ~5 subroutines 151 and 152 are executed again. This process is repeated, if necessary, until the calculated new oldstate coincides with the new state, and then the control is trans~erred to the System Synch Subroutine 153.
The presently preferred embodiment of the System Synch Subroutine is shown in Figure 11. It i5 entered at a point labeled "System Synch" and proceeds to an inquiry point 175 where the open or closed status of the contactor K2 is checked. Initially, 35 while the K2 timer is running lsee steps 142 and 147 - 4~ -in ~igure ~) and therefore prior to impleMentation of the step 148 that commands the contac~or K2 to close, this inquiry will reveal that K2 is open.
Consequently, after waiting for a certain delay interval (step 176) the control is returned directly to the step 143 of the Normal Cranki}lg Routine 84.
The delay introduced by step 176 will allow time, after implementing step 173 of the Next State Subroutine (~igure 10) and before returning to the step 143, for the above-describead commutati~n from tlle auxiliary valve (Tp or Tn) to the oncominy main valve TY to be comple~ed and fox the resulting electrical transients ("noise") to subside in the voltage processor 48. The delayed return to step 143 is desirable when load current is relatively high (as is true initially), because the aforesaid transients might then be severe enough to cause false data to be supplied on lines 51, 52 and 53 to the controller 44.
Once the control returns from step 176 (Figure 11) to step 143 (Figure 8), the loop comprising step 143 and in~uiry points 144, 145, 146, and 147 is repeatedly executed until the next state change is indicated by a negative answer to inquiry 144 (i.e., until the new state, as determined by step 143, is no longer the same as the new oldstate that was calculated by step 171 of the Next State Subroutine 152), whereupon the control again transfers to the Commutation Subroutine 151.
As the angular position of the rotor of the machine 14 advances wi-th increasing speed during the cranking mode of operation, the above-described execution of the steps 143-147 in the Normal Cranking Routine 84 and of the subroutines 151-154 are automatically repeated until the K2 timer stops running. The Commutation Subroutine 151 is initiated 56'~'~
20~TR-1~20 each time the inquiry 144 indicates a state change, and such changes occur with increasing fre~uency as the rotor accelerates. In practice it may take approximately two seconds for the rotor to complete its first revolution and another second for the second revolution. By the end of two revolutaions the rotor may have attained a speed on the order of 100 rpm, and it is around this time that the K2 timer stops runnin~
and the contactor K2 is closed to further weaken the field and permit higher speed cranking.
Each time the System Synch Subroutine 154 is executed after contactor K2 is closed, the control will proceed from inquiry point 175 to an inquiry point 177 in which the status of the first timer is tested. As shown in Figure 11, if timer #1 has stopped running the next step 178 of this subroutine will cause the controller 44 to command all contactors to open, and the Normal Cranking Routine is then aborted at a stop point 179. Assuming, however, that timer #1 is still running, the control proceeds from point 177 to a speed check step 181 where the rotor speed is measured. Any suitable m0ans can be used for this purpose. One simple yet effective means for measuring rotor speed is to count the number of times the Commutation Subroutine 151 is executed over a known period of time. It can be shown that this count is proportional to speed. There are six state changes and hence six commutation intervals per cycle of the fundamental component of alternating voltages on the stator windings of the machine 14, and one complete revolution of the rotor corresponds to five such cycles in a 10-pole machine. Thus the predetermined threshold speed of 240 rpm corresponds to a fundamental frequency of 20 ~ertz which is indicated 3~

if 12 commutations are counted in a period of Ool second.
From step 181 the System Synch Subroutine proceeds to a point 182 that inquires as to whether or not the rotor speed has attained a predetermined rate (i.e., the aforesaid threshold speed of 240 rpm). If not, the control is then returned directly to the step 143 of the Normal Cranking Routine 84 (Figure 8). Now the subloop comprising step 143 and inquiry points 144, 145, and 146 will be repeatedly e~ecuted until the next state change taXes place, whereupon the control once again transfers to the Commutation Subroutine 151.
The above-described execution of the steps 143-146 and of the subroutines 151-154 are automatically repeated until the rotor is ~otating faster than the aforesaid predetermined rate~ Once this threshold speed is exceeded, the inquiring point 182 of the System Synch Subroutine (Figure 11) yields an affirmative answer. In this event the control proceeds from point 182 to a step 183 that causes the controller to issue an appropriate signal that the engine is running. Step 1~33 is followed immediately by a step 184 that sets the cranking finish flag in an "on" state, and later by a step 185 that will cause the controller to command the opening of contactors Klp, Kln, K2 and K3. From step 185 the control returns to step 143 of the Normal ~ranking Routine (Figure 8). Now ths step 143 and the inquiry points 144 and 145 are passed through again, and from the inquiry point 145 the control can proceed to a finish point 186 which marks the conclusion of the cranking mode of operation.

5~

While a preferred embodiment of the invention has been shown and described by way of example, many modifications will undoubtedly occur to persons skilled in the art. The concluding claims are S therefore intended to cover all such modifications as fall within the true spirit and scope of the invention.

Claims (15)

1. An improved third harmonic auxiliary impulse commutated electric power inverter that is operative to supply variable frequency alternating current to the three different phases of a 3-phase inductive load circuit from a source of relatively smooth direct current, said inverter including at least three pairs of alternately conducting main controllable electric valves arranged in a 3-phase, double-way bridge configuration for interconnecting said source and said load circuit, and a commutation circuit including at least first and second auxiliary controllable electric valves interconnected in series aiding fashion across said source and connected via a precharged commutation capacitor to said load circuit, wherein the improvement comprises:-a. bistable first means coupled to said capacitor for sensing the electrical potential difference across said capacitor, said first means being in one state whenever the potential on one side of the capacitor is measurably positive with respect to the other side and being in a different state whenever said potential is measurably negative;
b. second means coupled to said 3-phase load circuit for detecting all zero crossings of the fundamental phase-to-phase alternating voltages that are developed at line terminals of the respective phases of said load circuit;
c. control means coupled to both said first and second means and having a third harmonic commutation mode of operation in which it cylically produces a family of periodic firing signals that cause said valves to turn on selectively, said family including a first series of firing signals respectively produced in response to said second means detecting consecutive zero crossings of said phase-to-phase voltages for alternately turning on said auxiliary valves, whereupon load current can immediately transfer from an offgoing main valve to a parallel path including the turned-on auxiliary valve and said capacitor which is first discharged and then recharged with reverse polarity by such current, and said family also including a second series of firing signals respectively produced in delayed response to successive state changes of said first means for turning on said main valves in a predetermined sequence, whereupon load current can then transfer to the oncoming main valve from the turned-on auxiliary valve; and d. said control means including time delay means effective at least after a predetermined initial period of time, which starts when said third harmonic commutation mode of operation commences, for delaying the production of each firing signal in said second series until a programmed interval of time has elapsed following each state change of said first means.

2. The inverter as in claim 1, in which said load circuit comprises the stator of windings of a rotatable synchronous machine.

3. The inverter as in claim 2, in which said stator windings are star connected and said capacitor is connected between the neutral of said windings and the juncture of said auxiliary valves.

4. The inverter as in claim 2, in which said source comprises an electric storage battery.

5. The inverter as in claim 2, in which load current flows between said source and said inverter through an impedance that has appreciable electrical inductance.

6. The inverter as in claim 5, in which said impedance comprises the field winding of said machine.

7. The inverter as in claim 6, in which said source comprises an electric storage battery.

8. An inverter as in claim 1, in which voltage sensing means is coupled to said capacitor for detecting whether or not the capacitor voltage has a magnitude exceeding a predetermined level, and in which said control means is so arranged that each firing signal in said second series is produced, after said first mean changes state, in response to either (i) the elapse of the programmed delay interval or (ii) the voltage sensing means detecting that said capacitor has been recharged to a voltage magnitude in excess of said predetermined level, whichever is first to occur.

9. An inverter as in claim 8, in which said time delay means is programmed so that throughout said predetermined initial period of time it will delay each firing signal in said second series for an interval of preselected fixed duration, and thereafter the delay intervals are shorter than said preselected fixed duration.

10. An inverter as in claim 9, in which said initial period of time is approximately 1.5 seconds, said preselected fixed duration is approximately 1.2 milliseconds, and said shorter delay intervals do not exceed approximately 0.5 milliseconds.

11. An inverter as in claim 9, in which said time delay means is additionally programmed so that after said initial period of time the duration of alternate delay intervals differs from the duration of intermediate delay intervals.

12. An inverter as in claim 11, in which said alternate delay intervals have a shorter duration than said intermediate delay intervals.

13. An inverter as in claim 1, in which said second means has six different states for indicating, during each cycle of operation, the six different combinations of relative polarities of said phase-to-phase voltages, said second means experiencing a state change each time the magnitude of any of said phase-to-phase voltages crosses zero, and in which said control means is effective to produce firing signals for turning on the second auxiliary valve in immediate response to said second means changing from a first state to a second state and from a third state to a fourth state and from a fifth state to a sixth state, said control means being similarly effective to produce firing signals for turning on the first auxiliary valve in immediate response to said second means changing from said sixth to said first state and from said second to said third state and from said fourth to said fifth state.

14. An inverter as in claim 13, in which said time delay means is programmed so that throughout said predetermined initial period of time it is effective to delay each firing signal in said second series for an interval of preselected fixed duration, and thereafter the delay intervals are shorter than said preselected fixed duration.

15. An inverter as in claim 14, in which said time delay means is additionally programmed so that after said initial period of time the duration of alternate delay intervals is shorter than the duration of intermediate delay intervals.