Overstressing of High-Voltage Capacitors - IEEE Xplore

4 downloads 0 Views 415KB Size Report
or overcharge, that is, charging to a significantly higher energy density than specified. This paper explores the limits of over- stressing commercially available ...
IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 32, NO. 3, JUNE 2004

1337

Overstressing of High-Voltage Capacitors Anders Heljestrand, Hans Bernhoff, Jan Isberg, and Anders Larsson, Member, IEEE

Abstract—In compact pulsed power conditioning systems, a high electric energy density of its capacitors is of interest. Commercially available capacitors are normally designed for long storage time and long lifetime. The performance of capacitors in terms of energy density capability can be improved by overstress or overcharge, that is, charging to a significantly higher energy density than specified. This paper explores the limits of overstressing commercially available capacitors for short time spans. The selected capacitors have a nominal dc voltage in the range of 2.5–10 kV and a capacitance in the range of 0.015–0.12 F. Two different storage times have been considered: 2 s and 20 ms. A high-voltage test setup was constructed. For the short storage time, it was possible to overstress the energy density as compared to nominal values with a factor of 26 and for the long storage time with a factor of 14 for the best performing capacitor. Other capacitors behaved significantly poorer, especially for the long storage time.

Fig. 1. The basic components of a system with intermediate inductive storage of electrical energy.

Index Terms—Energy density capacitors, overstressing of highvoltage capacitors, pulsed power.

I. INTRODUCTION

T

HE pulsed power conditioning system is a key component in compact systems for generation of pulsed high-power microwaves. In particular, this is pertinent to single-usage systems where the high-current impulse is generated by explosivedriven magnetic flux compression generators [1]. The operation of such a system requires an intermediate energy storage device for electric energy. Using basic electric components, two solutions are possible: inductive storage and capacitive storage [2]. The operating principles for different intermediate storage solutions are shown in Figs. 1 and 2, respectively. The inductive scheme works as follows: the primary energy storage is, by closing the first switch, discharged through the storage inductor. When the second switch is opened, preferably at current maximum, the current is abruptly interrupted and the magnetically stored energy in the inductor is converted to a high voltage across the load. The capacitive scheme works as follows: the primary energy storage, by closing the first switch, charges the intermediate storage capacitor. When this capacitor is sufficiently charged, the second switch closes and discharges the energy into the load. The charging of the storage capacitor can be slow, but the discharge of it must be fast.

Manuscript received April 24, 2003; revised October 7, 2003. This work was supported by a project funded by the Swedish Armed Forces and FMV (the Swedish Defence Material Administration) with the Swedish Defence Research Agency, Tumba. A. Heljestrand, H. Bernhoff, and J. Isberg are with the Department of Electricity and Lightning Research, University of Uppsala, Uppsala S-751 21, Sweden (e-mail: [email protected]). A. Larsson is with the Grindsjön Research Centre, FOI—Swedish Defence Research Agency, Tumba SE-147 25, Sweden, and also with the University of Uppsala, Uppsala S-751 21, Sweden. Digital Object Identifier 10.1109/TPS.2004.828759

Fig. 2. The basic components of a system with intermediate capacitive storage of electrical energy.

The electric energy density in inductive storage reaches about 40 MJ/m [2]. The common capacitors that are available commercially today have an energy density capability of less than 1 MJ/m . Specially made capacitors can reach a higher energy density, but these are often merely for research purposes [3]. Generally it is more attractive to use inductive storage for a compact system. However, capacitive storage has two important advantages. • Storage time can be variable. For the inductive storage, the time for utilizing the intermediately stored energy is determined by the discharge time constant of the primary energy storage–storage inductor system. For the capacitive storage, the storage time is only dependent on the storage performance of the capacitor • Voltage can be multiplied. Several capacitors can be charged in parallel and discharged in series (Marx coupling). Voltage multiplication allows for significantly higher voltage, which is difficult for inductive storage. The performance of capacitors in terms of energy density capability can be improved by overstress or overcharge, that is, charging to a significantly higher energy density than specified. For our primary application of single-usage systems, the capacitor is only overstressed once, and after it has fulfilled its purpose, it is discarded or destroyed. For such a case, lasting effects from previous overstressing is not present. Commercially available capacitors are normally designed for long

0093-3813/04$20.00 © 2004 IEEE

1338

IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 32, NO. 3, JUNE 2004

TABLE I TECHNICAL DATA OF THE CAPACITORS USED

Fig. 3. The capacitors in this study. From the left: Cornell–Dubier, Evox–Rifa, ASC X375, ASC X675, TPM 3 kV, TPM 10 kV.

storage and lifetime. In a normal withstand test, the capacitors is stressed to twice the nominal voltage for 1 min (dc capacitors for voltage less than 20 kV—IEC 61 071-1 Power Electronic Capacitors). This paper explores the limits of overstressing commercially available capacitors for short time spans. The selected capacitors, as introduced in Section II, have a nominal dc voltage in the range of 2.5–10 kV and a capacitance in the range of 0.015–0.12 F. Two different storage times have been considered: 2 s and 20 ms. A pulsed power system capable of charging and discharging the selected capacitors to 50 kV within a few microseconds was constructed and is described in Section III. The results of the overcharging tests are presented in Section IV, and concluding remarks are given in Section V.

Fig. 4. The experimental setup. C1 is the primary capacitor, C2 is the test capacitor, R2 and R3 are the resistor bridges, D1 is the diode pile, and T1 is the thyristor pile. V1 and V2 are surge arresters.

physical volume, and dielectric material in the capacitor. The energy density was calculated by dividing the stored energy with the physical volume of the capacitor (as measured on the outis estimated using side of the package). The stored energy the expression

where is the capacitance of the capacitor and voltage.

is its nominal

III. EXPERIMENTAL ARRANGEMENTS II. CHOICE OF CAPACITORS This study focuses on capacitors with nominal voltage up to 10 kV with a capacitance of about 0.1 F. A market investigation identified only a limited number of capacitors with this combination of properties. Eventually six different capacitors from four different manufacturers were chosen for this study. Four capacitors are of the metallized film type and two capacitors are of the film/foil type. The technical data of the capacitors in this study are shown in Table I and a photograph of them is shown in Fig. 3 [4]–[7]. The table gives the energy density at nominal voltage (nominal energy density), capacitance, design,

A. Setup Overview In order to measure capacitor energy density for different storage times, an experimental setup was constructed, shown in Fig. 4. For safety reasons, the experiments were performed inside a grounded cage. The setup is dimensioned for voltage pulses up to 50 kV with a short rise time. The setup consists of three connected circuits as follows: the primary circuit, the trigger circuit, and the discharge circuit. A general overview is shown in Fig. 5. The primary energy supply C1 charges the test capacitor C2 through a spark gap and a diode pile D1. C2 is discharged, after a preset storage time, by the

HELJESTRAND et al.: OVERSTRESSING OF HIGH-VOLTAGE CAPACITORS

1339

to ground. T1 and D1 are protected by surge arresters V1 and V2 to prevent breakdown of the thyristor pile and of the diode pile. The voltage is measured with an SF -insulated voltage probe (Ross VMP 120-6.2Y-H) and an oscilloscope (Tektronix TDS420A). The rise time of the charge curve depends on the capacitance of the measured capacitor and the resistance in the circuit. Two different resistance setups were used. In the first and ; in the second setup, setup, and . Fig. 5.

General overview of the experimental setup, where C2 is the test object. TABLE II TECHNICAL DATA FOR THE EXPERIMENTAL CIRCUIT

Fig. 6. The primary circuit, to study the energy density in a capacitor.

thyristor pile T1 to earth. The discharge circuit is controlled by the trigger circuit. The parts in the circuits are explained further in Sections III-B, III-C, and III-D. B. The Primary Circuit The primary circuit operates as follows. The experimental setup is dimensioned to withstand high voltage and short rise time pulses. The resistances consist of Arcol NH Power resistors connected in series. The thyristors are of type PCT 5STP 25L5200 and the diodes are of type FRD 5SDF 10H6004, both from ABB Semiconductors, Västerås, Sweden. The arresters are of type EXLIM Q supplied by ABB Switchgear, Västerås, Sweden; see Table II for technical data. A circuit diagram of the primary circuit is shown in Fig. 6. The primary storage capacitor C1 of 3 F is charged through a 10-M resistor R0 from a dc-source G1 with a maximum voltage of 50 kV. A spark gap is closed transferring the energy to the test capacitor C2 through a resistor bridge R2 and a diode pile D1. When the test capacitor is fully charged the diode pile prevents the charge from returning. The remaining energy from the primary storage is discharged through a 9.9-k resistor bridge R1 to ground. A resistor R1.1 of 50 for the trigger pulse is found in series with the R1 resistor bridge (see Section III-D). After the predetermined storage time, the test capacitor is discharged through a resistor bridge R3 and a thyristor pile T1

C. The Discharge Circuit In order to switch the thyristors in the stack of the primary circuit, a control circuit was constructed. Each thyristor must be triggered by a pulse of minimum 2.6 V at floating potential. The discharge circuit operates as follows: the thyristors are switched inductively from a control unit, which contains a capacitor battery as well as dc source to charge the capacitors and a circuit to control the charge/discharge of the capacitor battery. To trigger the control unit a pulse of 5 V is needed (see Section III-D). The circuit diagram is shown in Fig. 7. Then the capacitor battery C is discharged through the resistor R4 and the conductor loop with ten ferrite coils. The pulse is transformed on each ferrite coils to a trigger pulse at potential for each thyristor. Furthermore, two zener diodes protect the thyristor from overvoltage and alternating current. Fig. 8 shows one ferrite coil. To minimize the inductance the conductors in the loop lie close to each other. The trigger pulse rises in 1 s, and the thyristors switches in 3 s when the capacitor battery is charged to 100 V. The conductor loop with the ten ferrite coils and the thyristor pile are shown in Fig. 9. D. The Trigger Circuit A pulse from the primary circuit is needed to trigger the control unit. The trigger circuit works as follows. Resistor R1.1 of 50 is placed at ground of the R1 resistor bridge, seen in Fig. 6. When the primary capacitor C1 is discharged, the R1.1 resistor creates a voltage pulse of a few volts. In order to generate a trigger pulse of 10 V for the HP Function generator, two zener diodes D1 and D2 with a breakdown voltage of 10 V are connected parallel to the resistor R1.1. The current is limited by a resistor R1.2 of 10 . The circuit is protected with an arrester V1. The pulse triggers a HP 33120A Function Generator, which controls the storage time of the test capacitor, which triggers the control unit immediately for the short storage time and after a preset delay for the long storage time. The circuit diagram is shown in Fig. 10. E. Test of the Experimental Setup The experimental setup was tested with two capacitors with a nominal voltage of 100 kV and a capacitance of 0.04 F (connected in parallel) to see if the circuit could withstand the high current. The resistor bridges R1 and R2 were set at 17.6 and 7.6 , respectively. A “max” test was performed where the primary capacitor was charged to 46 kV. The test result is shown in Fig. 11. The inductance of the circuit limits the rise time of the voltage across the capacitor under test. As can be seen in the oscillogram

1340

Fig. 7.

IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 32, NO. 3, JUNE 2004

The discharge circuit with the control unit and the thyristor pile.

Fig. 11. Measured voltage when testing the short storage time with an applied voltage of 46 kV.

Fig. 8.

The ferrite coil with the protecting zener diode.

Fig. 12.

Fig. 9.

Fig. 10.

The thyristor pile with the conductor loop and the ten ferrite-coils.

The trigger circuit.

Measured voltage during capacitor breakdown test.

2.5 s. The delay will depend on which capacitor and resistor bridge that are used. The initial oscillation (close to ) is an artifact originating from the closing of the spark gap. The oscillation after the thyristor pile operation is attributed to LC oscillations in the circuit. The offset voltage after the operation of the thyristor pile (as seen in the oscillogram in Fig. 11) originates from the primary capacitor C1, which has not yet been completely discharged. A breakdown test, i.e., when the test object C2 is destroyed/ short-circuited, was performed with a film/foil capacitor with nominal voltage of 1.25 kV and 0.1 F, to see if the circuit would survive the breakdown of the test capacitor. The measured voltage during the test is shown in Fig. 12. The circuit proved able to withstand a short circuit of the device under test. IV. MEASUREMENT RESULTS A. Short Storage Time Measurements

in Fig. 11, the rise time is slightly above 2 s. Furthermore, delay in the discharge circuit as well as in in the current loop, and the thyristor triggering limits the short storage time to about

The energy density for short storage time was measured. The test object C2 was first subjected to nominal voltage. After the first shot, a number of shots were performed with voltage

HELJESTRAND et al.: OVERSTRESSING OF HIGH-VOLTAGE CAPACITORS

1341

TABLE III RESULTS FROM MEASUREMENTS WITH SHORT STORAGE TIME

Fig. 13. Typical example of voltage oscillograms for tests of metallized film capacitor of type Cornell–Dubier 940 with short storage time.

Fig. 14. Typical example of voltage oscillograms for tests of film/foil capacitor of type Hivoltcapacitor TPM 3 kV with short storage time.

increased in steps (Vstep) until breakdown. The shot before breakdown occured was considered to determine the maximum withstand voltage of the capacitor. Three capacitors of each type were tested accordingly. A fourth capacitor was then charged directly to breakdown voltage in order to probe for fatigue effects. However, in all cases, breakdown occured at this voltage, so no such effects were evident. Examples of voltage traces for metallized film and film/foil capacitors are shown in Figs. 13 and 14. In these examples the nominal voltage, the maximum withstand voltage and an intermediate voltage are shown. Each curve represents a single shot. In the metallized film case, the voltage slowly decreases from the peak voltage down toward zero. In the film/foil case, the voltage drops more rapidly. Self-healing processes in the metallized film capacitor could explain this difference. The energy density at maximum withstand voltage is calculated from the maximum voltage measured. Data from the measurements are shown in Table III, which also presents the technical parameters of each capacitor: the nominal voltage (Vnom) and nominal energy density (Wnom), maximum withstand voltage (Vmax) and maximum withstand energy density (Wmax), voltage factor, and energy density factor. The Vmax, rise time, and storage time are extracted from the measured voltage traces. The different resistor bridge setups are described in Section III-B. B. Long Storage Time Measurements The same procedure was used for measurements with long storage time as for short storage time. The test object C2 was first subjected to nominal voltage. Three capacitors of each type were tested accordingly. Examples of voltage traces for metallized film and film/foil capacitors are shown in Figs. 15 and 16. The voltage levels of these examples were the nominal voltage, the maximum withstand voltage, and an intermediate voltage.

Fig. 15. Typical example of voltage oscillograms for tests of metallized film capacitor of type ASC X675 with long storage time.

Fig. 16. Typical example of voltage oscillograms for tests of film/foil capacitor of type Hivoltcapacitor TPM 10 kV with long storage time.

Each curve represents a single shot. The differences in the two capacitor types are not as obvious for long storage time as for short storage time. However, in the film/foil case, the voltage initially decreases faster than in the metallized film case. The Wmax is calculated from nominal capacitance together with Vmax from the voltage trace. The results are shown in Table IV. The Vmax and the storage time are extracted from the measured voltage traces. No rise time could be extracted from

1342

IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 32, NO. 3, JUNE 2004

TABLE IV RESULTS FROM MEASUREMENTS WITH LONG STORAGE TIME

Fig. 17.

Breakdown in metallized film capacitor ASC X675.

Fig. 18.

Breakdown in film/foil capacitor TPM 10 kV.

the oscillograms due to limitations of the memory depth of the oscilloscope, but it is expected to be similar to that for the short storage time. The different resistor bridge setups are described in Section III-B. C. Inspection of Damage The metallized film capacitor has a self-healing ability: when the dielectric fails at any point, the metal film is vaporized and the short circuit quenches. This leads to a decrease of the capacitance but is normally not detrimental for the operation of the circuit. The final breakdown appears in the dielectric close to the center of the capacitor where the radius is smallest. There appears to be no surface flashover. For the metallized film capacitors the metal vaporizes around the broken dielectric but for film/foil capacitors the dielectric is punctured. Fig. 17 shows breakdown in the metallized film case and Fig. 18 shows breakdown in the film/foil case. The difference in the breakdown behavior is reflected in the voltage curves of a metallized film and a film/foil capacitor for short storage time as seen in Figs. 19 and 20. These oscillograms show that breakdown for a film/foil capacitor is distinct compared to metallized film. In the film/foil case, there is an abrupt breakdown compared with the metallized film case where the breakdown is more controlled. The self-healing processes in the metallized film capacitor explain this difference. The breakdown occurs in a time interval of 1.2–3.4 s from the start of the charging. Similar breakdown phenomena were also observed for long storage time. The difference in voltage traces from breakdown of a metallized film and a film/foil capacitor are shown in Figs. 21 and 22. In the film/foil case, there is an abrupt breakdown compared with the metallized film case, where it appears

Fig. 19. Breakdown in metallized film capacitor of type ASC X675 for short storage time.

to be more controlled. For short storage time, breakdown for a film/foil capacitor is distinct compared to metallized film. The difference is attributed to the self-healing processes in the metallized film capacitor. Breakdown occurs in the same time interval as for short storage time.

HELJESTRAND et al.: OVERSTRESSING OF HIGH-VOLTAGE CAPACITORS

1343

[4] Cornell–Dubilier Electron., Inc. [Online]. Available: http://www.cornelldubilier.com [5] Evox Rifa [Online]. Available: http://www.evox-rifa.com [6] ASC Capacitors [Online]. Available: http://www.ascapacitor.com [7] Hivolt Capacitors Limited [Online]. Available: http://www. hivoltcapacitors.com

Fig. 20. Breakdown in film/foil capacitor of type Hivoltcapacitor TPM 3 kV for short storage time.

Fig. 21. Breakdown in metallized film capacitor of type ASC X675 for long storage time.

Fig. 22. Breakdown in film/foil capacitor of type Hivoltcapacitor TPM 10 kV for long storage time.

V. CONCLUSION Commercially available capacitors of metallized film and film/foil type were overstressed for two different storage times: 2 s and 20 ms. For the short storage time, it was possible to overstress the energy density (as compared to the nominal value) with a factor of 26 and for the long storage time with a factor of 14 for the best performing capacitor. Other capacitors showed significantly less overstress capacity, especially for the long storage time. The results indicate that film/foil capacitors have the best overstressing capability of the studied capacitors. Every capacitor failed when breakdown in the dielectric material occurred. The breakdown occurred in the center of the capacitor. For the metallized film capacitors, the metal vaporizes around the broken dielectric, but for film/foil capacitors, the dielectric is punctured. REFERENCES [1] L. Altgilbers, M. Brown, I. Grishnaev, B. Novac, I. Smith, I. Tkach, and Y. Tkach, Magnetocumulative Generators. New York: SpringerVerlag, 2000. [2] J. Mankowski and M. Kristiansen, “A review of short pulse generator technology,” IEEE Trans. Plasma Sci., vol. 28, pp. 102–108, Feb. 2000. [3] F. Podeyn, H. G. Wisken, and Th. H. G. G. Weise, “High energy density capacitors for pulsed power applications,” presented at the Int. Conf. Pulsed Power Applications, Gelsenkirchen, Germany, 2001.

Anders Heljestrand was born in Östersund, Sweden, in 1975. He received the M.Sc. degree in engineering physics from the University of Uppsala, Uppsala, Sweden, in 2002. Since 2002, he has been with the Ångström Micro Structure Laboratory, Department of Electricity and Lightning Research, University of Uppsala. He works with researchers in material sciences, microfluidics, solar cells, and microelectromechanical systems techniques. He is responsible for development of processes in thin films, diffusion, and ion implanting.

Hans Bernhoff was born in Umeå, Sweden, in 1964. He received the M.Sc. degree in engineering physics and the Ph.D. degree in physics (in high-temperature superconductors) from the Royal Institute of Technology, Stockholm, Sweden, in 1988 and 1992, respectively. He then held a Postdoctoral position with the IBM Research Laboratory, Rueschlikon, Switzerland. In 1993, he joined ABB Corporate Research, Västerås, Sweden, where he was a Project Leader for several innovative projects in the area of electrotechnology, in particular, research on the single-crystal diamond as a wide band-gap semiconductor. In 2001, he became an Associate Professor at University of Uppsala, Uppsala, Sweden, where he teaches in the area of high-performance systems.

Jan Isberg was born in Stockholm, Sweden, in 1964. He received the M.Sc. degree in physics and the Ph.D. degree in theoretical particle physics and quantum field theory (with emphasis mainly on string theory and supersymmetry) from Stockholm University, Stockholm, in 1987 and 1992, respectively. From 1993 to 1994, he held a Postdoctoral position in the Department of Mathematics, King’s College, London, U.K. In 1995, he joined ABB Corporate Research, Västerås, Sweden, where he worked in the area of electrotechnology, in particular, research on the single-crystal diamond as a wide band-gap semiconductor. Since 2001, he has been in the Department for Electricity and Lightning Research, University of Uppsala, Uppsala, Sweden, where he was recently appointed as an Associate Professor. His current research interests include pulsed power and diamond semiconductor physics.

Anders Larsson (M’01) was born in Stockholm, Sweden, in 1963. He received the M.Sc. degree in engineering physics and the Ph.D. degree in electricity from the University of Uppsala, Uppsala, Sweden, in 1989 and 1997, respectively. He was a Development Engineer and Research Project Manager in the Research and Development Department of ABB Transformers AB and at the Swedish Transmission Research Institute (STRI AB). He has postdoctoral experience from the University of Uppsala, the Office National d’Etudes et de Récherches (Onera), and the Lund Institute of Technology. Since 2001, he has been with the Swedish Defence Research Agency, Tumba. Since 2002, he has also been an Associate Professor at the University of Uppsala. He has authored or coauthored more than 20 journal papers and over 30 peer-reviewed conference contributions. His research interests include the physics of electrical discharges and their effects. Dr. Larsson is a member of the Swedish Physical Society, the Society of Automotive Engineers, and Kalmar Nation, Uppsala.