Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
HIGH VOLTAGE SQUARE WAVE AND SPWM WAVE GENERATOR
Document Type and Number:
WIPO Patent Application WO/2011/017802
Kind Code:
A1
Abstract:
A high voltage wave generator circuit is provided for generating high voltage square-waves and SPWM-waves. The circuit comprises a plurality of stages, each stage including (a) an input connected to a DC source, (b) a first switch, (c) a second switch, (d) an energy storage device and (e) a wave shaping resistor, wherein the first switch charges the energy storage device and the second switch discharges the energy storage device. Each of the stages is operable to generate waveforms and the stages are connected in cascade, and the stages are configured such that (i) the first switches and the second switches of the plurality of stages are operable to switch substantially at the same time, and (ii) when the first switches are "ON" the second switches are "OFF" and vice versa. The circuit is configured such that by charging the circuit and sequentially switching all of the first switches and then the second switches substantially at the same time, the stages are operable to cumulatively generate as an output high voltage square-waves and SPWM-waves. An insulation test apparatus based on the wave generator is also provided. Also, a novel method for testing rotating machine insulation is provided.

Inventors:
JAYARAM SHESHAKAMAL H (CA)
YU YA TONG (CA)
Application Number:
PCT/CA2010/001233
Publication Date:
February 17, 2011
Filing Date:
August 11, 2010
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
JAYARAM SHESHAKAMAL H (CA)
YU YA TONG (CA)
International Classes:
G01R31/34; A23L3/32; G01R31/00; H03K4/02; H03K7/00
Other References:
YU ET AL.: "High Voltage Square Wave Generator for Motor Coil Insulation Testing", PROCEEDINGS OF THE 2008 IEEE INTERNATIONAL POWER MODULATORS AND HIGH VOLTAGE CONFERENCE, 27 May 2008 (2008-05-27) - 31 May 2008 (2008-05-31), pages 436 - 438, XP031403960
FABIANI ET AL.: "The Effect of Voltage Distortion on Ageing Acceleration of Insulation Systems Under Partial Discharge Activity", IEEE ELECTRICAL INSULATION MAGAZINE, vol. 17, no. 3, May 2001 (2001-05-01) - June 2001 (2001-06-01), pages 24 - 33, XP011091906, DOI: doi:10.1109/57.925300
YIN: "Failure Mechanism of Winding Insulations in Inverter-Fed Motors", IEEE ELECTRICAL INSULATION MAGAZINE, vol. 13, no. 6, November 1997 (1997-11-01) - December 1997 (1997-12-01), pages 18 - 23, XP011085292, DOI: doi:10.1109/57.637150
VAFAKHAH B. ET AL.: "Space-Vector PWM for Inverters with Split-Wound Coupled Inductors", IEEE INTERNATIONAL ELECTRIC MACHINES AND DRIVES CONFERENCE, 3 May 2009 (2009-05-03) - 6 May 2009 (2009-05-06), pages 724 - 731, XP031475853
Attorney, Agent or Firm:
DE FAZEKAS, Anthony (40 King Street West Suite 580, Toronto Ontario M5H 3S1, CA)
Download PDF:
Claims:
CLAIMS

1. A high voltage wave generator circuit for generating high voltage square- waves and SP WM -waves characterized in that the circuit comprises:

(a) a plurality of stages, each stage including (a) an input connected to a DC source, (b) a first switch, (c) a second switch, (d) an energy storage device and (e) a wave shaping resistor, wherein the first switch charges the energy storage device and the second switch discharges the energy storage device;

(b) wherein each of the stages is operable to generate waveforms and the stages are connected in cascade;

(c) wherein the stages are configured such that (i) the first switches and the second switches of the plurality of stages are operable to switch substantially at the same time, and (ii) when the first switches are "ON" the second switches are "OFF" and vice versa;

(d) wherein the circuit is configured such that by charging the circuit and sequentially switching all of the first switches and then the second switches substantially at the same time, the stages are operable to cumulatively generate as an output high voltage square-waves and SPWM-waves.

2. The circuit of claim 1, characterized in that the output is controllable by controlling the first switches of the stages.

3. The circuit of claim 1, characterized in that in charging the circuit, the first switches are "OFF" and the second switches are "ON".

4. The circuit of claim 3, characterized in that in discharging the circuit, the first switches are "ON" and the second switches are "OFF".

5. The circuit of claim 1, characterized in that the circuit is configured such that the energy storage units are charged in parallel by the DC source.

6. The circuit of claim 1 , characterized in that the wave shaping resistors of the stages are connected in series.

7. The circuit of claim 1, characterized in that the energy storage devices are discharged through the wave shaping resistors.

8. The circuit of claim 1, characterized in that the peak value of the output voltage in synchronized operation of the plurality of the stages is equal to the sum total output voltage of the plurality of stages.

9. The circuit of claim 1, characterized in that each stage includes two diodes to define and protect a charging path and discharging path.

10. The circuit of claim 9, characterized in that when the second switches are "ON", the second switches and the diodes form the path for charging the energy storage units in parallel.

11. The circuit of claim 1 , characterized in that the charging circuits for the plurality of stages are provided such that each of the plurality of stages has a charging time that is substantially the same.

12. The circuit of claim 1, characterized in that the charging and discharging is initiated alternatively.

13. The circuit of claim 1, characterized in that a driving circuit is connected to each of the switches, and an isolation transformer (i) provides power to the driving circuits, and (ii) isolates the driving circuits of each of the stages from the other stages.

14. An insulation test apparatus characterized in that it comprises at least one wave generator as claimed in claim 1.

15. An insulation test apparatus as claimed in claim 14 characterized in that it enables testing for effects of electrical stresses and/or thermal stresses caused by varying pulse widths, a fast rise time, high switching frequency and/or impedance of the feeding cable and the motor coil.

6. A method for testing rotating machine insulation characterized in that the method comprises the steps of:

(a) initiating a waveform generator that is operable to generate that high voltage square-waves and SPWM-waves comprises; and

(b) applying pulses generated by the waveform generator to the insulation of the rotating machine and thereby applying transients and enhanced stresses that reflect the conditions of the rotating machine during use.

Description:
HIGH VOLTAGE SQUARE WAVE AND SPWM WAVE GENERATOR

PRIORITY

This application claims priority to U.S. provisional patent application no. 61/272,047 filed on August 11, 2009.

FIELD OF THE INVENTION

The present invention relates generally to square wave and PWM wave generators. The present invention further relates to motor insulation testing and testing systems.

BACKGROUND TO THE INVENTION

Industrial motors are currently the most significant consumers of electricity. The applications of induction motors fed by voltage source converters especially are very common. However, the new pulse width modulated (PWM) voltage source adjustable speed drive motors (ASDs) provide a more complex voltage waveform to the motors than the power frequency sinusoidal voltage waveform; thus, concerns about the adverse effects of such drives on the motor insulation has increased as the applications of drives based on PWM voltage has increased.

The PWM voltage waveform consists of a group of voltage pulses with a fast rise time and varying pulse widths. Several studies have shown that the PWM voltage waveform causes enhanced electrical and thermal stresses on the motor coil that leads to premature stator damage or failure. Unfortunately, the motor manufacturing industry has neither test standards nor test facilities that addresses the issue of enhanced stresses caused by solid state switching drives. Suitable testing equipment is therefore needed for the qualification and acceptance of electrical insulation systems in motors fed by ASDs.

Power electronic-based adjustable speed drives can be classified into two main categories: CSCs (current source converter type speed drive) and VSCs (voltage source converter type speed drive). In CSCs the DC term is a regulated current, and DC inductors are selected to provide the current source. The immediate output is a pulsed current whose fundamental component is controllable in magnitude, phase angle, and frequency through the adjustment of a modulating signal. Output capacitive filters are required in order to eliminate current harmonics from the output current. Thus the voltage and current applied to the motor are quasi-sinusoidal, which is less harmful to the motor insulation system. However, because of the heavy and bulky DC-link choke and the possible occurrence of resonances between the capacitive filter and the motor inductances, current-source converters are less frequently used in industrial applications. The main applications of this type of converter are in active power filters and high-power AC motor drives in which the motors are fed with a long connecting cable.

On the other hand, in VSCs, capacitors are used for the DC-link rather than the heavy and bulky DC-link choke to provide a constant DC regulated voltage. The output voltage of a VSC is a chain of pulsed voltages, and the local average voltage (the average voltage per switching period) of each pulsed voltage in the chain is controlled by the IGBT switches. In a sinusoidal pulse-width modulated or SPWM converter, which is one of the most commonly used in motor drives, the local average voltage of the output voltage is a sinusoidal waveform (also called a fundamental component). Because VSCs are smaller, lower cost, more reliable, and more efficient than CSCs, they have dominated the motor drive market. The main disadvantage of VSCs is the generation of high dv/dt transients and high frequency harmonics, which may cause enhanced electrical and thermal stresses on the motor insulation system and thus results in premature failure of the motors.

Motor Insulation Systems

It is useful to understand the properties of motor insulation systems.

The design considerations for a motor insulation system include both electrical properties and mechanical properties, such as dielectric stress, resistance to partial discharge, corona protection, desired thermal class, and durability. Motor stator insulation is a kind of laminated insulation, which contains multiple layers of insulation made of various materials; conductive layers such as conductive tape for slot corona protection; and stress grading coatings.

Fig. 1 shows both structures of random-wound and form-wound stator insulation systems. The innermost layer is enameled magnet wire coating. In form-wound medium and high voltage motors, several insulation tapes may be needed to cover the magnet wire in order to enhance the turn insulation level. The turn insulation has two purposes: to insulate between turns and to provide good adhesion between the conductor and the impregnating resin. Sometimes the turn insulation may contain conductive additives in order to decrease the partial discharge (PD) activity close to the conductors.

Moving outward from the turn insulation, the next layers are the ground wall and slot insulation and the phase insulation layers. The ground wall and slot insulation provides electrical insulation between the motor winding and the motor core. The phase insulation separates the wire bundles of the different phases in the core slot and coil overhang. Both serve as electrical insulation as well as mechanical protection.

For medium and high voltage motors, the slot portion of the coil is wrapped with layers of conductive armour tape. The main function of this low-resistance layer is to provide a good contact between the coil and the core, eliminating possible partial discharge (PD) and hence the erosion caused by the PD. They also provide protection against vibration damage. The conductivity of the armour tape is constant at about 10 '2 to 10 '5 S/m, thus avoiding short circuiting of core laminations and Eddy current. At the end of the coil, which has a rated voltage above 4.16 kV, layers of voltage stress grading material is used to force the electric field uniform, preventing the occurrence of slot discharges due to sharp changes in voltage or the concentration of the electric field.

PWM converters generally work with switching frequencies of up to 20 kHz. A PWM voltage waveform is a chain of square pulsed voltage with pulse widths varying at a specific mode in one fundamental cycle. The continuous strikes of fast transient voltage clearly increases the electrical and thermal stresses on the insulation system of the motor coil. These problems result from the PWM waveform, which can be characterized according to its varying pulse width, rise time, fall time, peak value, and switching frequency.

The high repetition rate pulse voltage with high dv/dt transients is known to cause increased dielectric loss, space charge injection or charge accumulation in the insulation, and higher partial discharge activity. All these parameters have a negative effect on the insulation system of the motor and result in additional heating within the motor insulation system, causing a general or local temperature rise. Enhanced thermal stress is thus applied to the motor insulation system, which may lead to failure of the motor. Additionally, the enhanced thermal stress on the insulation system of the motor may be caused by the high frequency harmonic content in the PWM voltage waveform. As mentioned before, the motor insulation system is a laminated structure that contains multiple layers of insulation composed of various materials, conductive tape for the elimination of slot discharges, and semi- conductive stress grading coatings. Each of these materials has its own characteristic frequency response; some insulation materials are sensitive to high frequencies, some of them are not so sensitive to high frequencies.

When the multiple sinusoidal voltages of the various frequencies of PWM waveforms are applied simultaneously to the motor insulation system, some of the frequency-sensitive materials in the laminated insulation system generate more heat under the high frequency harmonic stress. Additionally, in a laminated insulation system of motor coils, the diversity of the frequency response characteristics of the materials also changes the voltage distribution along the insulation layers. The voltage distribution under a high frequency harmonic is not the same as the voltage distribution under a power frequency voltage, which is strategically designed. Some layers in the laminated insulation system, such as the conductive armour layers, may be subject to increased voltage drop and therefore generate more heat, which results in a local temperature rise.

Insulation Tests

Insulation tests are designed to evaluate the condition of the electrical insulation in the design, qualification, acceptance, and maintenance of rotating machinery. The current test methods in IEEE and IEC standards include an insulation resistance/polarization index measurement, a high potential test, a voltage-endurance test, a turn insulation test, and a high repetition rate impulse test. These tests are briefly described as follows:

• Insulation resistance/polarization index measurement: Insulation resistance measurement has been used for more than half a century to evaluate the condition of electrical insulation. With the carefully maintained record of periodic measurements, accumulated over months and years of service, this easy method is valuable as a measure of some aspects of the condition of the electrical insulation, such as contamination, absorbed moisture, or severe cracking. However, since the insulation resistance is not related directly to the dielectric strength (unless the defect is concentrated), it is impossible to show clearly that the insulation system of a motor is acceptable. Further, this index measurement is not sensitive to internal insulation defects.

• High potential test: In this test, a voltage higher than the rated voltage is applied to the test object for a specified time (usually 1 minute) for the purpose of determining the dielectric strength of the insulation system. The high potential test can use a power frequency source, a direct current source, or a very low frequency. According to the IEEE standards for stator coils, the power frequency AC test voltage is 1000 V rms plus twice the rated voltage. Alternatively, the DC test voltage is 1.7 times the power frequency AC rms test voltage; the very low frequency AC test voltage is 1.63 times the power frequency AC rms test voltage.

• Voltage-Endurance test: The voltage duration test is designed for the stator coil of large motors and generators. An AC voltage higher than the rated voltage is applied to the stator coil for a long time (250 or 400 hours).

• Turn insulation test: Turn insulation tests are designed for testing the dielectric strength of turn insulation under transient voltage stress. Experience has shown that turn insulation failure can be caused by steep-front surges such as lightning strikes and breaker operations. Thus, in IEEE Std 522-2004, the recommended test voltage is one that has a frequency several orders of magnitude above the power frequency.

• High repetition rate impulse test: In IEC technical specification 60034-18-42, a repetitive impulse or a sinusoidal voltage at a frequency above 1000 Hz is recommended for evaluating the performance of an insulation system in rotating electrical machines that are fed by voltage source converters. This technical specification attempt to address concerns about the enhanced stresses caused by solid state switching devices.

In summary, the insulation system of a motor that is fed by a voltage source converter is not exposed to the traditional power frequency sinusoidal voltage, but instead to a more complex voltage waveform, which may cause increased voltage stress and thus a higher level of thermal stress on the motor insulation, resulting in the premature failure of the motor. The increased stresses caused by PWM-VSCs are a function of the varying pulse width, rise time, fall time, crest value, and switching frequency of the PWM waveform and also of the impedances and frequency response characteristics of the feeding cable and the coil. A variety of techniques have been used to overcome the voltage limitations of power electronic switches and to achieve the high voltages needed by using relatively low voltage switches. The most common approaches are either connecting the semiconductor switches directly to the load, or connecting them to the load through signal or series-connected pulse transformers.

For example, Kim et al., "200KV Pulse Power Supply Implementation", in Power Electronics and Applications, 2007 European Conference, September 2007, pp 1-5" introduced a type of Marx circuit in order to generate a high voltage pulse power supply. This reference discloses a design that is clamed to produce up to 200 kV pulse voltage with a repetition rate of up to 1000 pulses per second (PPS). The rise time of the pulse voltage is less than 1 μs, and the pulse width is up to 4 μs. Although this approach can successfully produce a high voltage pulse at a specific repetition rate, it can not be used to generate high voltage square and PWM waveforms due to the following limitations.

• One limitation is the pulse width. An inductor is used to isolate the DC source from the high voltage output at the moment when the IGBT stacks are turned on. According to the characteristics of the inductor, the inductor can isolate only transient voltage. If the IGBT stacks remained on for a longer period, or in other words, if this prior art Marx circuit produced a wider pulse, the inductor would short-circuit, which might damage the DC source.

• Another concern about this Marx circuit is the risk of an unsynchronized switching operation problem. Because this circuit is actually a series configuration, the simultaneous operation of the switches is critical. If one or more switches in the IGBT stacks fail to operate simultaneously, all the IGBT stacks would then fail. Trigger circuits, protection circuits, and electromagnetic compatibility (EMC) should therefore be designed precisely, which means higher cost and lower reliability.

Redondo et al., in "Analysis of a modular generator for high- voltage, high-frequency pulsed applications, using low voltage semiconductors and series connected step-up transformers" in Review of Scientific Instruments 78, 034 discloses a modular pulsed generator based on low voltage MOSFET switches and step-up pulse transformers. Their generator produces up to 15 kV / 1 A with a 5 μs width at a 10 kHz repetition rate and a pulse rise time of less than 1 μs. A total of three individual stacked modules are used in the generator, each consisting of a modified forward converter that drives a step-up pulse transformer. The secondary side of the transformer in each module is connected in series with the one in the following module.

The advantage of this generator is that it can produce a high pulse voltage by using low voltage MOSFET switches without any risk of switch failure. However, because pulse transformers are used in the generator to boost the voltage, the output waveforms are distorted due to leakage inductance and the residual capacitance of the transformer coils. The pulse width is also limited by the transformer for the same reason. Another disadvantage is that the MOSFET driver circuits are supplied by batteries, which limit the continuous operation time of the generator. Therefore, although this approach provides a safe method of generating a high voltage pulse using low voltage power electronic devices, it can not be utilized to build a test facility that can generate high voltage square wave and PWM waveforms.

A critical consideration in the design, manufacture, maintenance, and evaluation of a rotating machinery insulation is that any insulation test that is used should reflect the actual operating conditions to which the rotating machines are exposed when they are operating. As discussed and reviewed, the motors fed by power electronic-based voltage converters are affected by increased electric and thermal stresses due to varying pulse widths, a fast rise time, a high switching frequency, and the impedance of the feeding cable and motor coil. Until now, most insulation test methods and test facilities are traditional, which have been derived from the initial design of rotating machinery in which the motors are operated at a single power frequency. Thus, the difference between the test conditions of the motors and their actual operating conditions is highly significant. The current test methods and facilities are not directly related to the motor's dielectric strength under a PWM waveform and therefore cannot evaluate the performance of the motor insulation system effectively.

Current laboratory research has utilized fast repetitive exponential decay waves and single high frequency sinusoidal waves and has attempted to analyse the premature failure mechanism of motor insulation fed by a PWM voltage waveform. Although the results have shown that fast repetitive exponential decay or high frequency voltages have a negative impact on the motor insulation system, these factors may be only part of the problem. In fact, the failure mechanism of the motor coil is a result of a combination of a number of factors. The fast repetitive exponential decay type of pulsed voltage or single high frequency sinusoidal waveforms are not the exact voltage waveforms which the motors are exposed to when they are fed by voltage source converters. These test waveforms contain only one or part of the adverse factors, not all of them. The accuracy of the evaluation results obtained by using these voltage waveforms is therefore less than optimal.

There is a need for an IGBT-based pulse voltage generator which can produce high voltage square wave and PWM waveforms. To improve the insulation system of large high voltage form-wound motors, motors should be tested with a voltage waveform similar to the waveform under which the motors are operated. There is a need for a waveform generator design and waveform generator based on such design that is operable to support qualification and acceptance tests of electrical insulation systems.

There is also a need for a square-wave and SPWM-wave generator for more general applications that is operable to generate high voltage pulses that is robust and cost effective to produce.

SUMMARY

A high voltage wave generator circuit is provided for generating high voltage square-waves and SPWM-waves. The circuit comprises a plurality of stages, each stage including (a) an input connected to a DC source, (b) a first switch, (c) a second switch, (d) an energy storage device and (e) a wave shaping resistor, wherein the first switch charges the energy storage device and the second switch discharges the energy storage device. Each of the stages is operable to generate waveforms and the stages are connected in cascade, and the stages are configured such that (i) the first switches and the second switches of the plurality of stages are operable to switch substantially at the same time, and (ii) when the first switches are "ON" the second switches are "OFF" and vice versa. The circuit is configured such that by charging the circuit and sequentially switching all of the first switches and then the second switches substantially at the same time, the stages are operable to cumulatively generate as an output high voltage square- waves and SPWM-waves.

In another aspect of the invention, the output if the circuit is controllable by controlling the first switches of the stages.

In charging the circuit, the first switches are "OFF" and the second switches are "ON". In discharging the circuit, the first switches are "ON" and the second switches are "OFF". In another aspect of the circuit, the circuit is configured such that the energy storage units are charged in parallel by the DC source.

In yet another aspect of the invention, the circuit is configured such that the stages are charged in parallel and discharged in series.

In accordance with the present invention the peak value of the output voltage in synchronized operation of the plurality of the stages is equal to the sum total output voltage of the plurality of stages.

In another aspect of the invention, each stage includes two diodes to define and protect a charging path and discharging path.

When the second switches are "ON", the second switches and the diodes form the path for charging the energy storage units in parallel.

In another aspect of the invention the wave shaping resistors of the stages are connected in series.

The peak value of the output voltage of the circuit in synchronized operation of the plurality of the stages is equal to the sum total output voltage of the plurality of stages.

In another aspect of the invention, the energy storage devices are discharged through the wave shaping resistors.

In yet another aspect of the invention the charging circuits for the plurality of stages are provided such that each of the plurality of stages has a charging time that is substantially the same.

In yet another aspect of the invention, a driving circuit is connected to each of the first switches, and an isolation transformer (i) provides power to the driving circuits, and (ii) isolates the first switch of each of the stages from the other stages.

In another aspect of the invention, an insulation test apparatus based on the wave generator is disclosed.

In yet another aspect of the invention, a method for testing rotating machine insulation characterized in that the method comprises the steps of: initiating a waveform generator that is operable to generate that high voltage square-waves and SPWM-waves comprises; and applying pulses generated by the waveform generator to the insulation of the rotating machine and thereby applying transients and enhanced stresses that reflect the conditions of the rotating machine during use.

In this respect, before explaining at least one embodiment of the invention in detail, it is to be understood that the invention is not limited in its application to the details of construction and to the arrangements of the components set forth in the following description or illustrated in the drawings. The invention is capable of other embodiments and of being practiced and carried out in various ways. Also, it is to be understood that the phraseology and terminology employed herein are for the purpose of description and should not be regarded as limiting.

BRIEF DESCRIPTION OF THE DRAWINGS

Figs. Ia and Ib show representative structures of random-wound (a) and form- wound (b) motor insulation systems.

Fig. 2 illustrates the circuit design (a) and its operation in (b), (c), and (d).

Fig. 3 compares a series connection and a cascade connection, relative to the circuit of the present invention.

Fig. 4 illustrates the circuit design of the present invention in a worst case switching scenario. Fig. 5 illustrates an implementation of the present invention with a cascade connection. Fig. 6 illustrates the charging process in connection with the circuit shown in Fig. 5. Fig. 7 illustrates the discharging process in connection with the circuit shown in Fig. 5.

Fig. 8 illustrates in a block diagram a representative trigger and control system for use in connection with the present invention.

Fig. 9 illustrates a local drive circuit used to drive the switches of the circuit of the present invention.

Fig. 10 illustrates the connection of power supplies for local drive circuit. Fig. 11 illustrates a waveform generating stage.

Fig. 12 illustrates an IGBT based pulse voltage generator.

Fig. 13 is a schematic diagram of the trigger signal generator in accordance with the present invention.

Fig. 14 is a flow chart of the main program in accordance with the software aspect of the invention, in one implementation thereof.

Fig. 15 illustrates in a block diagram and implementation of the TimerO and the prescaler. Fig. 16 is a flow chart that illustrates the TimerO interrupt service routine. Fig. 17 shows a PWM waveform at a CCPx output.

Fig. 18 is a block diagram of a CCP module in accordance with an implementation of the present invention.

Fig. 19 is a block diagram of the Timer2 module in accordance with another aspect of the present invention.

Fig. 20 illustrates square wave output signals generated by the trigger signal generator in accordance with one aspect of the present invention.

Fig. 21 illustrates PWM wave output signals of the trigger signal generator in accordance with one aspect of the present invention.

Fig. 22 is a block diagram of a bipolar square wave and PWM wave generator. Fig. 23 illustrates maximum frequency vs. equivalent capacitance of the test object. Fig. 24 illustrates a simulated circuit for testing an inductive load.

Fig. 25 illustrates test results with a 2mH inductive load, showing a bipolar exponential decay waveform. Fig. 26 illustrates test results with a 2H inductive load, the test voltage waveform being a superimposed waveform consisting of a fundamental sinusoidal waveform and a PWM wave.

Fig. 27 illustrates four types of test voltage waveforms.

DETAILED DESCRIPTION

The generator consists of IGBT switches and other wave shaping components. The circuit design described herein enables the generator to produce the stable high voltage square wave and PWM waveforms.

The basic circuit for high voltage square wave and PWM wave generation is shown in Fig. 2 in one embodiment of the present invention. In Fig. 2(a), R con represents the equivalent resistor of the connecting wire which is very small. Ri is the wave shaping resistor, Ci is the DC link capacitor, and C tr represents equivalent capacitor that is the combination of residual capacitance of the wave shaping resistor and the load capacitance.

First, the capacitor Ci is charged to voltage Vj. When the switch turns on (as shown in Fig 2 (b)), C tr is rapidly charged by the capacitor Ci, and the voltage across C tr increases fast, thus the rising part of a square wave pulse is formed. During the turn-on period, the voltage V OUT can be expressed as:

VOUT ~ V 1 { l^xp[-t X (CiH-Qr)ZRc 0n Ci Qr] } (2.1)

Since C \ » C tr , the rise time of the square pulse is mainly determined by R con and C tr .

When the switch is on, capacitors Ci and C tr discharge through resistor Ri as shown in Fig 2 (c). The crest part of the square pulse is formed in this period. The voltage V OUT in this period can be expressed as:

VOUT = Viexp[-t w /Ri(Ci+C tr )] (2.2) where t w is the on-time of the switch SWi. If t\y is much less than Ri(Ci+C tr ), the pulse width can be mainly decided by the on-time of the switch SWi.

When the switch is off as shown in Fig 2.(d), the capacitor C tr discharges through R 1 . This forms the falling part of the square pulse. The voltage V O uτ can be expressed as the equation (2.3). The equation shows that the fall time of the square pulse is determined by RiQ r . V 0UT = ViexpC-t/RiQ r ) (2.3)

With the properly selected values of the resistors and the capacitors, a square voltage pulse can be generated by turning the switch on and off. If the capacitor is charged from a stiff DC source, a square wave or a PWM wave can be produced by continuously turning on and turning off the switch in a prescribed mode.

It should be understood that MOSFET, GTO, IGBT, and IGCT power transistors may be used to generate high voltage square wave and PWM waveforms due to their high voltage rating. Because the majority of high voltage power electronic switches are also designed with correspondingly high current ratings, the cost of these high voltage switches is very high.

Since high current capacity is not required for insulation testing, it is economical to build the generator based on the components that have relatively low voltage and low current ratings, and consequently, lower in cost. This is made possible by operation of the design of the present invention.

In a representative implementation of the present invention, a low cost discrete IGBT switch (IXMH 16Nl 70) is selected for building the IGBT-based pulse voltage generator of the present invention. The particular IGBT switch selected is manufactured by IXYS Inc. The voltage and current ratings are 1700V and 16A respectively. The typical turn-on time is around 93ns, and the typical turn-off time is 490ns. So theoretically, the maximum switching frequency can reach 1.7 MHz which is far higher than the operating frequency required for the motor insulation testing.

One way to enhance the voltage handling capability of low voltage switches is to connect low voltage switches in series. Prior art applications of this method are known, however the possibility of non-uniform voltage distribution along the chain of switches can not always be avoided. The non-uniform voltage distribution in the series connection could be caused by two reasons.

One reason is that the steady-state and transient characteristic of each individual switch in the chain may not be identical due to variations that come about during manufacture. Therefore, the switches take a different share of the total voltage in the steady-state. For the same reason, the switches in a conventional series chain may not turn-on and turn-off at the same time during switching transients even if the gating signals are applied simultaneously to all the switches. In order to equalize voltage sharing of series switches, in its steady-state as well as during switching transients, external shunt RC circuits are needed on all switches in the chain. However, the use of these circuits may lead to distortion of the generated waveform.

The second reason is that the non-synchronized operation of the switches may be caused by the difference in gate driving signal circuits and the electromagnetic interference on these circuits.

The inventor has discovered that this issue can be addressed with a circuit design that ensures that the switches all turn-on and turn-off simultaneously to avoid damage in the series connection. This comes about when a switch in the chain closes later than the other switches; or if a switch opens earlier than other switches, the total output voltage appears across the out of phase switch, resulting in a damage of the switch. Hence, as soon as the switch fails (permanently short-circuited in most cases), the remaining switches must be able to withstand the total voltage that otherwise would be shared by all switches, which may lead to failure of other switches in the series connection.

In accordance with one aspect of the connection, a cascade connection is used rather than a series connection, specifically as illustrated in Fig. 3, which the inventor has discovered in this application eliminates or significantly reduces the risk of switch failure as described. The cascade connection consists of several isolated stages which are similar to the circuits shown in Fig. 2. Each stage consists of a DC source, a switch, a diode, and a low inductance wave shaping resistor. The outputs of all stages are connected in cascade, and each stage can generate waveforms independently by turning on and turning off the switch. If the switches in all stages turn-on and turn-off simultaneously, the total output voltage is the sum of the voltage of all stages. For N stages, the peak value of the output voltage can be theoretically NxVi, where, Vi is the voltage of the DC source in each stage. The DC sources, in this implementation are isolated from one another.

It should be understood that in a series connection, every switch works simultaneously to share the high DC source voltage V, hence, synchronized operation is critical. Whereas in the cascade connection, the generator is divided into several independent stages and each IGBT switch operates at the stage voltage Vi (=V/N), which is in the range of the rated voltage of the switches. Thus, for the cascade connection, synchronized operation is not as important as the series connection. The unsynchronized operation only affects the value of the output voltage, but do not cause failure of the switches.

In order to illustrate the advantages of the circuit design of the present invention, Fig. 4 shows a worst case scenario for the circuit of the present invention, caused by an unsynchronized operation. As shown in Fig. 4, assuming that one switch is open while other switches are closed. The load current bypasses the wave shaping resistor of "open switch" stage through the freewheeling diode D 3 . In this case, the voltage across the switch can be expressed as: Vsw = V 1 + V D3 , where VQ 3 is the forward voltage of free-wheeling diode D 3 . Since VD 3 is very low, which is less than one volt, the voltage across the switch is still in the range of its rated voltage, So, no switch will be damaged. The only problem caused by the unsynchronized operation is that the total output voltage would decrease from NxVj to (N-I) χ Vi.

As mentioned above, the cascade connection can eliminate or substantially reduce the risk of failure of the switches in generating high voltages. However, in the cascade structure shown in Fig. 3(b), each stage in the generator needs a separate DC source and these DC sources have to be isolated from each other, which increases cost, and the control circuits for all these DC sources are relatively complicated. In order to lower cost and increase reliability, a further implementation of the present invention is shown in Fig. 5, which provides additional advantages especially suitable for certain applications. The implementation is based on the cascade connection shown in Fig. 2. In this new design, the DC sources in the original cascade connection are replaced by capacitors. Instead of a common high voltage DC source to charge the capacitors in all stages, a second IGBT switch SW 2 in all stages is used to control the charging of the capacitors (Ci). The function of switch SWj in the new design is the same as that in original design, Fig. 2.

Therefore the total operation of the circuit of the present invention is operable to initiate two processes; namely, charging and discharging, which are alternatively controlled by two groups of the switches, SWi (also referred to as the first switches) and SW 2 (also referred to as the second switches).

During the charging process, switches SWi in all stages are off, and the switches SW 2 in all stages are on; the capacitors Ci in all stages are charged in parallel by the common DC source as shown in Fig. 6. Because the charging circuits designed for each stage are the same, each stage has the same charging time constant τ c , where τ c = Rc χ C] , and Rc is the internal resistance of the DC source and connecting wires.

During the discharging process, the switches SW 2 in all stages are off, and the switches SWi in all stages are on. The capacitors in all stages are discharged through the wave shaping resistors as shown in Fig. 6. As described above, if the IGBT switches SWi in all stages turn-on and turn- off at the same time, the total output voltage at the load side is the sum of voltage across Ri in each stage.

Both charging and discharging processes are controlled by a microcontroller-based trigger signal generator. In order to avoid short circuiting of the DC source, the charging and discharging operations must run alternatively; for example, when the switches SWi are on, the switches SW 2 should be off, and vice versa. Thus, two trigger signal groups, which are used for driving switches SWi and switches SW 2 respectively, are preferably interlocked from each other.

Trigger and Control System

The implementation of the present invention is further explained by describing a representative trigger and control system for use in connection with the circuit described. Fig. 8 shows a block diagram of an example of a trigger and control system for use in connection with the present invention, used to trigger the power electronic switches used in the circuit described, and wave generators based on such circuit. In order to avoid false triggering from electromagnetic interference (EMI), optical fibre cables may be used to connect the trigger signal generator to the switches which are located in a high electric potential area. The generator of the present invention can produce square wave and PWM waveforms with a peak voltage, for example, of up to 18 kV and with a switching frequency of 600 Hz to 6 kHz. The fundamental frequency of the PWM waveform may be 20 Hz to 1200 Hz, the rise time may be less than 200 ns, and the pulse width may be varied up to several milliseconds.

The trigger and control system of the generator of the present invention may be implemented using a microcontroller-based trigger signal generator, local drive circuits, and trigger signal transmission channels. The trigger signal generated by the trigger signal generator is divided into two interlocked signals; one signal is to control the operation of switches SW] to produce the desired waveform, and another signal is to control the operation of switches SW 2 to charge the capacitors in all stages. Two op-amp buffers are used to amplify the interlocked signals and drive two groups of optical cable communication channels and each group may contain 10 optical channels for example.

The function of the trigger signal generator is to generate square wave or PWM waveform trigger signals, determine the running mode and operation sequence, and provide a control interface for waveform parameter setting. A microcontroller chip is selected for this reason. The trigger signal generator and its auxiliary circuits including the buffer circuit and the interlock circuit is described below.

The local drive circuits amplify the trigger signal to a high level that is required to operate the IGBT switches and drive the IGBT switches successfully into the "on" or "off state, as described above. These local drive circuits are important because the characteristic of the local drive determines the characteristic of the IGBT switches.

For, example an advanced gate driver may be used in connection with the generator of the present invention, such as TD351 which is made by STMicroelectronics. The input of TD351 is compatible with optical-couplers and pulse transformer. The input voltage is internally clamped at about 5 to 7 volts. When connected to an open collector optical-coupler, the resistive pull-up resistor can be connected to either the reference output voltage pin VREF or to the voltage supply.

Local drive circuits may be used in connection the trigger. For example Fig. 9 shows the schematic of the local drive circuit. The signal from the fibre optic receiver is fed into an optical couple, TLP250. Then the output of the optical couple is connected to the input pin of IGBT drive, TD351. The gate on-state voltage value is +12V, and the gate off-state voltage value is - 12V. The gate resistance is 15 Ω. Thus, the approximate maximum drive current is [(12+12)/l 5] = 1.6A, which is within the operating current range of TD351. Each stage of the high voltage generator has two IGBT switches in accordance with the implementation described; hence, two local drive circuits are needed in each stage to drive the two groups of switches SWi and SW 2 .

Significantly, because the electric potential of every stage is not the same, the stages are isolated from each other. In this particular implementation, isolation transformers are used not only to provide power to each drive circuit in each stage, but also to isolate each stage. The connections of the power supplies for each drive circuit are shown in Fig. 10. The isolation transformers are connected in cascade. The low voltage DC sources, which supply the local drive circuit, are therefore isolated from each other. There are two chains of cascade connected isolation transformers in the design, which are used to provide local drives for IGBT switches SWi and SW 2 . It should be understood that isolation between the triggering circuitries is necessary as each switch is at different potential and if we want to trigger them using a single trigger circuit or even using un-isolated trigger circuits, this may damage the circuit and impede its operation.

To decrease size, toroidal transformers may be selected for the isolation transformer in the generator. The parameters of the transformer are shown in Table 1 below.

Table.l: Characteristics of theToroidal Transformer

The trigger signal transmission channels provide a communication connection between trigger signal generator and the local drive circuits of each IGBT switch. These channels also isolate the electromagnetic sensitive trigger signal generator from the local drive circuits which are located at high voltage potential. Thus, the ability to shield against the electromagnetic interference (EMI) is enhanced.

Optical fibres are used as trigger signal transmission medium for example using a communication channel interface connection. The communication channel may consist of an optical transmitter, an optical receiver, and an optical fibre. The transmitter transforms the electrical signal to an optical signal and the receiver then transforms the optical signal to original electrical signal again. The transmitter is connected to the trigger signal generator, and the receiver is a part of a local drive circuit. The signal rate of the channel is one implementation is around 5MBd and the maximum propagation data delay of the optical channel is around 145ns.

Structural design of the generator

In one implementation of the present invention, an IGBT-based pulse voltage generator based on the design described generator consists of ten waveform generating stages connected in cascade. Since each stage is working at a different potential, the stages need to be insulated from each other.

For example, in a representative generator structural design, all components in one stage, which are at same potential, are installed on a fibreglass board, as shown in Fig. 11. The fibreglass board may be made of glass epoxy laminate that has extremely high strength and excellent electrical insulation. The dielectric strength of the board is 550 vpm(Volts/mil) or 21.65kV/mm.

In order to generate a good waveform with a controlled fall time, low inductance resistors are selected for wave shaping. Four resistors are connected in parallel to match the high energy generated in wave shaping. Each resistor is 600W rated, over the operating temperature of -55° to 150°. The inductance of each resistor is less than 8OnH. The heat sink is designed to transfer heat to the ambient to keep the temperature within the specified maximum.

Ten waveform stages are mounted together forming a multi-storey structure, as shown in Fig. 12. Each stage is mounted over another and supported by an aluminum U-channel as shown in Fig. 12. Because the metal supports are mounted on high insulation strength fibreglass board, the stages are insulated from each other. Any electrical or signal related connections between the stages are connected through isolation transformers or optical fibres.

Further Implementation of Trigger Signal Generator

Further detail is provided regarding an example trigger signal implementation, suitable for providing an insulation test apparatus based incorporating or based on the wave generator design of the present invention.

As an insulation test apparatus, the IGBT-based pulse voltage generator requires multiple functions and a flexible control mode to match different insulation test requirements. The requirements and functions of the trigger signal generator can thus be described as follows:

• It can produce square wave and PWM waveform trigger signals according to different settings: the frequency of the square wave and the switching frequency of the PWM waveform needs to be adjustable.

• It provides an interface so that the operator can select the running mode (automatic or manual, continuous or intermittent), waveform type, and running time for the test. An example of a hardware implementation of the trigger signal generator is now described. To realize the requirements and functions described, an 8-bit PIC microcontroller chip PIC16F877A is used as the core control element in the generator. This sophisticated Harvard architecture- based single integrated computer is low cost, easy to program, and multi-functional.

Fig. 13 shows the schematic diagram of the trigger signal generator. The generator consists of a microcontroller and its auxiliary circuit, a signal interlock circuit, a buffer circuit, an input current monitor, and an output for operation of an interlock.

The trigger signal generator may be implemented for example using microcontroller chip PIC16F877A. The main features of the chip are:

• Operating frequency: DC - 20MHz

• Flash program memory (14 bit words): 8K

• Data memory ( bytes ): 368

• Interrupts: 15

• Timers: 3

• I/O ports: Ports A, B, C, D, E

• Capture / Compare / PWM modules: 2

• 10-bit A/D modules: 8 input channels

• Analog comparators: 2

• Instruction set: 35 Instructions

• Packages: 40 pins

Table 2 lists the I/O port and pin assignments of the (micro control unit) MCU in the generator. The pins and their respective names are selected in order to minimize the complexity of the program and the hardware setup.

Table 2 : PIC Microcontroller Chip I/O Port and Pin Assignments.

Name Port Pin number Description

LCD DATA Port B 0 ~ Port B 7 33 - 40 Pins are set to transfer data between MCU and LCD

E Port C 7 26 LCD Enable control line

RW Port C 6 25 LCD Read / Write control line

PORTC is an 8-bit wide, bidirectional port. The 8 pins (port CO to port C.7) of PORTC are multiplexed with an alternate function for the peripheral features on the chip. Setting the Port/Peripheral selection registers permits port Cl and port C2 to be set in normal input / output pin mode, capture module input mode, compare module output mode, or PWM module output mode.

In this design, ports Cl and C.2 are set as the normal output when the square wave is generated. The status of the port pin is controlled by the firmware. To generate a bipolar square waveform, the output of port Cl is set to be opposite to the output of port C.2. When port Cl is high, port C.2 is low, and vice versa. The differential of the two port pins is a bipolar square wave.

The control strategies of the firmware in the microcontroller make the generator flexible.

When the generator is set to generate the PWM waveform, both port Cl and port C.2 are set as the PWM module output. Two independent two-level unipolar PWM waveforms can be generated at port Cl and port C.2. To generate a three-level bipolar PWM waveform, the fundamental components of these two output signals are phase shifted by 180° with respect to each other. The differential of the two outputs is the three-level bipolar PWM waveform.

Significantly, as an operational safety consideration, the discharging switch SWi is designed to be normally closed, whereas the charging switch SW 2 is designed to be normally open. Thus, when no waveform is being generated, the trigger signal of SWi is always high, and the trigger signal of SW 2 is always low.

Port CO is set as the "stop" signal input. When the stop button is pushed, port CO is pulled high, and the generator stops generating the waveform and trips the power supply of the IGBT-based pulse voltage generator.

Port C.3 is set as the manual "start" signal input. In manual mode, when the start button is pushed, port C3 is pulled high, and the generator begins to generate the desired waveform.

A 4 χ 4 keypad ma be used to input the settings. A liquid crystal display (LCD) module may be used for displaying the orders, settings, and status of the generator.

As discussed earlier, each stage in the main circuit in the implementation being described two IGBT switches are used, and the trigger signals of these two switches are interlocked to avoid a short circuit in each stage. Thus, the trigger signal generated from the microcontroller (port Cl or C2) is divided into two interlocked trigger signals in the interlock circuit of the generator; one used to drive SWi and the other to drive SW 2 . As shown in Fig. 13, "and" gate and hex inverter chips are used to realize the interlock function. To eliminate possible signal overlap due to interference, an RC delay circuit is designed in order to insert a "zero gap" between the two signals. Both SWi and SW 2 are in the off state during the period of the "zero gap". An RC delay circuit is used to adjust the duration of the "zero gap".

The waveform signals are then sent to the buffer circuit. The circuit includes 4 buffer units in this implementation. Each buffer unit amplifies the power of the waveform trigger signal and drives ten waveform trigger signal channels. These waveforms are transformed into light signals by the optical transmitters. Optical fibre channels transmit the trigger waveforms to local IGBT switch drives to control the IGBT switches in the improved cascade connection circuit.

The software for the square waveform and PWM waveform trigger signal generator is written in the assembly language of the PIC microcontroller. Fig. 14 shows the flow chart of the program. The program has two subroutines: the square mode cycle subroutine and the PWM mode cycle subroutine. These two subroutines can be selected through the keypad. Square wave signal generation

TimerO module timer/counter is used for square wave generation. TimerO has an interrupt function that can generate an interrupt service routine at a set time to change the status of the waveform output at port Cl and port C.2. Once TimerO is set to timer mode, the TMRO register increments every instruction cycle time multiplied by the prescaler value. TimerO interrupt is generated when the TMRO register overflows from a set value to 0Oh. Fig. 15 is a block diagram of the TimerO module and the prescaler that is shared with the watch dog timer.

Fig. 16 is a flow chart of the TimerO interrupt service routine, which is designed for square wave generation. In the interrupt routine, the interrupt flag is cleared, and the TMRO is set to a given value according to the desired frequency.

When the frequency is set higher than 38 Hz, the product of the data in the TMRO register and the data in the TimerO prescaler register OPTION REG determines the interval time between two interrupts. Because the status of the waveform output pins is changed from high to low or from low to high in each interrupt service routine, this interval time is half the cycle time of the square waveform. Thus, the frequency of the square waveform can be calculated as follows: f Set = f osc /[2x4 χ (TMRO) χ (OPTION_REG )] (3.1) where f set is the frequency setting, and f osc is the frequency of the crystal oscillator.

Accordingly, if the frequency setting and the frequency of the crystal oscillator are known, calculating the initial value of register TMRO and the value of register OPTION_REG is straight forward.

To obtain enough resolution for a variety of square wave frequencies, the data in the prescaler register OPTION_REG is set according to the range of the setting frequency, as shown in Table

2.

Table 3 OPTION_REG setting

The data in the TMRO register is calculated from:

(TMRO) = fosc / [δxfset x(OPTION_REG )] (3.2)

When the frequency setting is lower than 38 Hz, the cycle time of the square wave is longer than the maximum interrupt interval time, which is 256 χ 256 χ 8 χ Tosc (cycle time of the crystal oscillator). Therefore, according to equation (3.2), the value of (TMRO) would be larger than 256, the TMRO register would overflow. Thus, another register (F number) is needed to comprise a 16-bit register with TMRO register in order to store the result of equation (3.2). In the calculation of equation (3.2), the integer quotient is saved in F number, and the remainder is saved in TMRO. In this case, the prescaler rate is fixed at 256, and the interval time between two interrupts is determined by the values in F_number and TMRO registers. The setting frequency is expressed as f osc /[8x (F_number) x 256x256 +8 χ (TMRO) χ 256] (3.3) PWM wave generation

Two Capture/Compare/PWM (CCP) modules are used to generate two independent PWM waveform signals. Each CCP module contains a 16-bit register that can operate as either a 16-bit capture register or a 16-bit compare register or a PWM master/slave duty cycle register.

When in PWM mode, the CPPx pin produces a PWM output with a resolution up to 10 bits. The PWM output, which has a time base (period) and a specific time when the output stays high (duty cycle), is shown in Fig. 17.

Fig. 18 shows a simplified block diagram of the CCP module in PWM mode. The Timer2 module is used as the PWM time base in PWM mode. Timer2 is an 8-bit timer with a prescaler and a postscaler as shown in Fig 19. PR2 is an 8-bit period register in the Timer2 module. The PWM period is a function of the data in the PR2 register and the value of the Timer2 prescaler (T2CON<1 :0>). The PWM period of the PWM wave for the CPPx output can be calculated using the following formula:

PWM period = [(PR2) +1] ><4 χ T 0SC x (Timer2 Prescaler Value) (3.4)

The switching frequency of a PWM waveform is defined as 1/ [PWM period]. For a sinusoidal PWM waveform, the switching frequency of the PWM waveform is the product of the fundamental frequency and the frequency modulation ratio.

The fundamental frequency (f s ) and the frequency modulation ratio (m f ) can be set through the keypad. Thus, the PWM period is determined as following equation.

PWM period = 1 /(switching frequency)

= l/[( fundamental frequency) x (frequency modulation ratio)] (3.5)

If the value of the Timer2 prescaler is preset in the microcontroller chip, the PWM period can be specified the value of (PR2).

The PWM duty cycle is determined by (CCPRxL:CCPxCON<5:4>) and the value of the Timer2 prescaler (T2CON<1 :0>). The maximum resolution of the duty cycle is 10 bits. If the value of the Timer2 prescaler is determined, the PWM duty cycle can be specified by the CCPRlL (or CCPR2L for PWM2) register combining with CCPlCON <5:4> (or CCP2CON for PWM2) bits. The following formula is used to calculate the PWM duty cycle according to the data in these registers:

PWM duty cycle

= (CCPRxL:CCPxCON<5:4>) χ T osc χ (Timer2 Prescaler Value) (3.6)

In the Timer2 module, the value of the PWM time base register TMR2 increases from 0Oh until it matches the value of PR2; once (TMR2) equals (PR2), a PWM interrupt routine is launched. In the interrupt routine, TMR2 is cleared, and the duty cycle of the next PWM cycle is specified through rewriting the data to CCPRxL:CCPxCON<5:4> from a pre-calculated discrete fundamental sinusoidal data form. In the general purpose data memory area (SRAM) of the microcontroller, 50 bytes are specified as the pre-calculated discrete sinusoidal data form area for storing a half cycle of discrete sinusoidal data. Each discrete sinusoidal datum occupies two bytes, which is 16 bits. The size of the sinusoidal data form is flexible and is determined by the frequency modulation ratio (ni f ). A Taylor series is used to calculate each datum in the sinusoidal data form. The Taylor series of the sine function is expressed as sin (x) = x - x 3 /(3!) + x 5 /(5!) - x 7 /(7!) + ... =∑{[(-l) n x (2n+1) ]/[(2n+l)!]} (3.7) where -π < x < π . n is the positive integer = 1 ,2 3, ....

The error of the formula (3.7) to calculate sine function is less than [(-l) n+1 x (2n+3) ]/[(2n+3)!].

If setting x p = (π/ ni f ) x (2p-l), p = 1, 2, 3... ni f /2, where ni f is the frequency modulation ratio and p is the positive integer, the discrete Taylor series expression of the sine function is expressed as: sin (X p ) = x p - X p 3 /(3!) + x p 5 /(5!) - x p 7 /(7!) + ... =∑{[(-l) n x p (2n+1) ]/[(2n+l)!]}

=∑{[(-l) n ((π/ m f ) x (2p-l)) (2n+1) ]/[(2n+l)!]} (3.8)

The discrete fundamental sinusoidal data form can be derived by calculating each sin (x p ) at each discrete x p using equation (3.8). The size of the half-cycle sinusoidal data form in memory area of the microcontroller is 2 χ (rri f /2). Another half-cycle sinusoidal datum is easy to calculate from the half-cycle sinusoidal data form.

The PWM duty cycle in each PWM period is then calculated from

PWM duty cycle (at p cycle) = m a χ sin (x p ) x PWM period (3.9) where, m a is defined as modulation index. In this thesis, the m a is fixed at 0.87.

Once the PWM duty cycle in each PWM period is calculated, the data for CCPRxL:CCPxCON<5:4> in the pre-calculated discrete sinusoidal data form can be then calculated through the equation (3.6).

The signal generator has two PWM output channels; PWMl and PWM2. Each channel can produce a unipolar square wave or a PWM waveform. For the square wave signal, the two channels (1 and 2 in Fig. 20) are set to be opposite to each other. The difference between these two output signals is a bipolar square waveform, in which the peak value is twice the value of each channel (M in Fig. 20). For the PWM wave signal, the fundamental signals of these two output channels (1 and 2 in Fig. 21) are phase shifted 180° with respect to each other. The difference between these two output signals is a three level bipolar PWM waveform (M in Fig. 21).

To generate a high voltage bipolar square waveform or a high voltage three-level bipolar PWM waveform, two improved cascade connection circuits as previously described are needed. The two output channels of the trigger signal generator are used to trigger these two circuits. The differential voltage of these two improved cascade connection circuits is a high voltage bipolar square waveform or a high voltage three level bipolar PWM waveform. The block diagram of high voltage bipolar square wave and PWM waveform generator is shown in Fig. 22.

The IGBT-based pulse voltage generator described can generate both square wave and SPWM waveform with a peak value up to 15 kV peak. The switching frequency range for both waveforms is 600 Hz to 6 kHz, and the fundamental frequency range of SPWM waveform is 20 Hz to 1200 Hz. The rise time is less than 200 ns. The pulse width can be varied from a few microseconds to several milliseconds.

Test Results

Simulations and experimental tests show that the cascade connection enables the generator to overcome the problems related to voltage division and synchronization of switches, which normally exist in a series connected structure.

A generator utilizing the modified cascade circuit design was used, which is based on charging in parallel and discharging in series, is utilized in the generator. The peak value of the output voltage can be boosted to several times of the common DC source input voltage, as a result of the design.

A microcontroller-based trigger signal generator as described was used to generate the trigger signals for IGBT switches. The firmware of the microcontroller enables the generator to produce different waveforms with the same structural topology. In addition, the test mode was set through a keypad in order to satisfy the requirements for both standard and special tests. As described earlier, one of the advantages of the cascade connection used in accordance with the present invention is that the generator in accordance with the present invention can overcome the problems associated with series connected switches that may fail due to the unsynchronized switching operation. These problems are usually caused by electromagnetic interference from both inside and outside the generator. In a test, two generators with different connection structures, namely, series and cascade, where both generators are connected to the same load. Under normal conditions, the switches in both connection structures run simultaneously and the output waveforms and the voltage waveforms at each stage of the two connection structures are almost the same. However, when some of the switches lose synchronism, the conditions of the two connection structures are totally different. In simulation circuits used, the trigger signal of one switch in both generators was connected to ground in order to simulate the worst case operation. One of the switches in both connection structures fails to close and remains open for several cycles, while the other switches operate normally.

When an unsynchronized operation problem occurs in the series connection, the total output voltage drops significantly. The peak value of the output voltage drops from 10.5 kV in normal conditions to only 52 V in an unsynchronized condition. Whereas, when the same unsynchronized problem develops in the cascade connection, the peak value of the output voltage drops from normal 10.5 kV to 9.0 kV. The voltage drop is exactly equal to the voltage of one stage.

The voltage across the switch in the series connection is almost 10.5 kV, which is much higher than the rated operation voltage of the switch. This voltage may damage the switch. As soon as one switch in the series connection is short circuited, all switches in the series connection may be damaged because the remaining switches must withstand the total high voltage which normally would be shared by all the switches. On the contrary, no such problem occurs in the cascade connection. Although the voltage waveform across the switch in the cascade connection is seriously distorted, the peak value of the voltage across the switch is still within the range of the switch's operating voltage. Thus, the switch is safe from damage that might occur. Therefore, the whole generator has zero risk of failure due to unsynchronized operation.

The comparison of the simulation results for the two connection structures confirms that the cascade structure is a more reliable way to generate high voltage waveforms by using low voltage rated IGBT switches. The cascade connection structure avoids the unsynchronized operation problem that may cause damage to the switches in the series connection structure and therefore is safer, more reliable, and less costly.

The equivalent parallel circuit representation of dielectric materials was used where the equivalent resistor of insulation materials is in the range of tens of MΩ to several GΩ, and the range of equivalent capacitance is from tens of pF to several nF.

It is clear that the test object has a significant influence on the wave shape of the pulse voltage produced by the generator, especially when connected to a load with a large capacitance. The wave shaping resistors in the generator were selected to minimize this influence to a certain tolerance. These wave shaping resistors are connected in parallel with the load in the circuit. Based on the waveform generation mechanism described earlier, the rise time, the pulse width, and the fall time of the waveform are predicated by the values of R 00n , Ci, Ri and C tr . The sum of the rise time, pulse width and the fall time determines the cycle of the square wave, so, R con , Ci, Ri and C tr determines the highest switching frequency of the generator.

Considering most insulation materials and commercially available pulse capacitors, the value of Ci may be chosen to be 2 μF. For ten stages, the total capacitance is 0.2 μF, which is much higher than the capacitance of most test objects. So, Ci can be neglected in calculating the rise time. The rise time is determined mainly by the product of R con and C tr . Since the value of R con is very small, the rise time can be limited to less than several hundred nanoseconds.

The fall time is determined by n χ (Ri//Ri oad )C tr , where n is the number of stages in generator. Since Ri is parallel to the load, the smaller the value of Ri, the less the influence caused by the load. However, low resistance causes high power consumption and low efficiency, and the pulse width will decrease with a smaller value Of R 1 . In testing, the value of the resistance used is 5 kΩ.

In the simulated circuit, the equivalent circuit model of a parallel combination of a resistor and a capacitor is used to imitate the any test object. Since nxRi (50 kΩ) is much smaller than the equivalent resistance of test object (MΩ to GΩ), the load resistor can be neglected in the simulation. The test object had an equivalent capacitance of 500 pF. As the frequency increases, the low part of the wave is "chopped" because there is not enough time available to decrease the voltage to zero.

The maximum switching frequency that the generator can produce on different loads, before the waveform chops, is shown in Table 4.1 below. Fig. 23 shows the curve of maximum frequency vs. the equivalent capacitance of the test load.

Table 4.1 Maximum Switching Frequency of Generator with Different Test loads.

As shown in Table 4.1 and Fig. 23 , due to the long decay time, the maximum switching frequency that the generator can produce with a large capacitive load is very low. In order to decrease the decay time and increase the switching frequency for a large capacitive load, an auxiliary external resistor can be connected parallel to the load providing additional path to discharge the load. Table 4.2 shows the maximum switching frequencies for different loads when a 5 kΩ resistor is connected in parallel with the load. The disadvantages of this method are high energy consumption and low output efficiency.

Table 4.2 Maximum Switching Frequency of Generator with Different Test Objects paralleling

with a 5kΩ Resistor.

A simulated circuit for testing with an inductive load is shown in Fig. 24. A 500 Ω resistor is connected in series with the load in order to limit the output current.

Table 4.3. Rise time and fall time of the waveform with no load connected Fig 25 shows the waveforms of the output voltage, the switch voltage, and the output current when the load inductance is 2 mH. Fig. 26 shows the waveforms when the inductance of the load is 2 H.

When testing a low inductance load (2 mH), the output waveform is not a SPWM wave but a bipolar exponential decay waveform (Fig. 25). The exponential voltage pulse occurs at each IGBT switching period. When testing with a high inductive load (2 H), the test voltage waveform is a superimposed waveform consisting of a fundamental sinusoidal waveform and PWM wave (Fig. 26). In both cases, the voltage at the switch is still a PWM waveform with a peak voltage of 1500 V, which is within the operating voltage range of each stage.

Fig. 27 shows the open circuit output waveforms of the generator. The rising and the failing parts of the waveform are also shown in Fig. 27. Table 4.3 lists the rise time and fall time of the waveform at different peak values of the output voltage.

As evident from Table 4.3, the rise time and fall time remain at almost the same level for different peak values of the output voltage.

Three typical test objects were selected for testing: a sample motor coil bar (13.8 kV rated voltage), a cable termination (14 kV rated voltage), and a motor stator coil (4 kV rated voltage). The equivalent capacitance and resistance of the test objects, which are measured at different frequencies, are listed in Table 4.4.

Table 4.4 Equivalent capacitance and resistance of test objects.

Cable termination tests, sample bar tests, and coil tests were completed, in a manner that is known. The upper waveforms show different responses for the various objects tested with the generator. These variations are due to the different parameters and characteristics of the test objects. The parameters of the test objects combined with the parameters of the connecting cable result in different shapes of the voltage and current waveforms, especially in the waveform fronts. Table 4.5 lists the parameters of the waveforms that were obtained from tests with different objects.

Table 4.5 Diversity of voltage waveform with different test objects.

The maximum voltage overshoot amplitude is defined as [V max /( V max - V top )] χ 100%, where V max is the maximum voltage of the waveforms and V top is the top voltage of the waveform (the value of the flat part).

A 4 kV rms form-wound model stator coil was tested under different voltage waveforms; namely, power frequency, exponential decay pulse, square wave, and SPWM wave. The middle part of the coil was clamped with two metal plates to simulate the motor core slot and the test voltages were applied between the coil terminal and the grounded metal plates.

In the test, a FLIR-SC500 infrared camera with an emissivity spectrum between 7.5 and 13 μm was used to measure the surface temperature of the coil. The temperature image was displayed in a 320x240 pixel array. The sensitivity of the detector was 0.07 K at a temperature of 303 K and with an accuracy of 2 K over 273 to 773 K. The emissivity calibration was conducted on the sample at room temperature with a thermocouple reading as a reference. The four types of test voltage waveforms are shown in Fig. 27. The test waveform parameters corresponding to the four test voltage waveforms are given in Table 4.6.

Table 4.6 Parameters corresponding to the test voltage waveforms.

Table 4.7 shows the maximum stable temperature rise of the motor coil under different voltage stresses.

Table.4.7 Coil surface temperature rise

The above test results indicate that the temperature rise on the coil surface varies significantly depending on the type of voltage waveform used, although the peak values of these voltages are the same. From Table 4.7, it can be seen that the highest temperature rise was caused by the SPWM waveform showing a 15 °C rise above the ambient. The second highest temperature rise observed was for the case of the square wave, with a 14 0 C rise. The high repetitive exponential decay pulse causes only a 3.5 0 C temperature rise. Almost no temperature rise was detected in the case of the 60 Hz power frequency test.

However, it is easy to see that the stress due to these waveforms can be significantly severe compared to the other two voltage waveforms. The following paragraphs explain the temperature profiles observed. The peak value of the voltage overshoot produced by transient is a function of rise time, the pulse width, as well as the natural resonant frequency of the circuit. The first two are determined by the voltage source, and the latter one is determined by the impedances of the connecting cable and the coil. In the case where the pulse width matches the natural resonant frequency of the circuit, the overshoot reaches a maximum value that is considerably high in amplitude, and hence, the maximum dv/dt. As a result, the internal partial discharges can be triggered and cause a high temperature rise.

Another factor that contributes to the significant temperature rise is the characteristics of the materials in the laminated insulation system of the motor. Some materials are sensitive to frequency, while others are not as sensitive. The synthetic resin bonded mica tape, which is frequently used in making motor ground wall insulation, is sensitive to frequency. The curve shows that the tape resistance decreases significantly with an increase in frequency. In contrast, the conductive tape, which is used as a corona protector in the slot, is not sensitive to changes in the frequency. The resistance of conductive tape remains the same with the changing frequency.

When high frequency harmonic voltages are applied to the insulation system, the dielectric loss of frequency dependent insulation materials increases; thus, the heat generated due to dielectric loss can increase. When high frequency harmonic voltages are applied to the insulation system, the dielectric loss of frequency dependent insulation materials increases; thus, the heat generated due to dielectric loss can also increase.

Additionally, the frequency response characteristic of different materials also changes the voltage distribution along the insulation layers in a laminated insulation system for motor coils. For some high frequency harmonic components in the applied voltage, a higher voltage is distributed at the conductive layer of the ground wall insulation, which results in an increase of heat generated by the conductive materials.

The invention has been shown with reference to the specific embodiments. However, it should be noted that the invention is in no way limited to the details of the illustrated structures but changes and modifications may be made within the scope of the claims. For example, various alternate circuit implementations are possible that draw on the circuit design principles explained above. With the widely increasing applications of power electronics as well as modern communication signal techniques in power systems, modern electrical apparatuses are no longer operated under the stress of the power frequency sinusoidal voltage, but rather under that of a more complex voltage waveform. Accordingly, test facilities and equipment may be developed that match these changes and guarantee the safety and reliability of power systems. Based on the technique presented in this thesis, a multi-level SPWM voltage source or other voltage waveform related to power electronics can be developed as a next step toward the end goal: the development of a multi-functional, flexible, and power electronic technique-based high voltage test facility.

Future applications of the pulse voltage generator of the present invention can be developed in other areas, which include nero fibre spinning, food treatment, and fault location on transmission lines, high voltage cables, or large motors. In relation to food treatment the technology described may be used for example for cold pasteurization by enabling the injection of high voltage pulses into a volume of liquid (such as milk) without the need for warming. The technology may also be used for filtering air or water, or in the mining industry by using high voltage pulses to aid in metal extraction, with applications for example in the mining industries and other sectors where environmental protection or rehabilitation is required.