920

B. Sonerud et al.: Dielectric Response Measurements Utilizing Semi-Square Voltage Waveforms

Dielectric Response Measurements Utilizing Semi-Square Voltage Waveforms Björn Sonerud, Tord Bengtsson*, Jörgen Blennow and Stanislaw M. Gubanski Department of Materials and Manufacturing Technology Chalmers University of Technology SE 412 96, Göteborg, Sweden *also with ABB Corporate Research, SE 721 78 Västerås, Sweden

ABSTRACT Dielectric response measurements belong to the group of diagnostic tools used for quality evaluation of high voltage insulation systems. A measurement system called Arbitrary Waveform Impedance Spectroscopy (AWIS) is presented here which is capable of utilizing the harmonic content of repetitive voltages with a semi-square waveform and the corresponding current response to perform dielectric response measurements. Extensive circuit modeling and calibrations are required in order to perform accurate measurements. AWIS could offer new possibilities, especially for continuous monitoring of the dielectric properties of high voltage components and systems. The accuracy of the technique in both low and high voltage applications is verified by comparison with results obtained by means of the Frequency Domain Spectroscopy (FDS) technique. Index Terms — Dielectric measurements, discrete Fourier transforms, aging, monitoring, dielectric materials, insulation testing.

1 INTRODUCTION THE long term properties of high voltage insulation systems are mostly studied through investigations of time to breakdown, breakdown strength and dielectric response. The tests of time to breakdown and breakdown strength are simple to perform but result in the destruction of the tested objects. At the same time these tests offer limited information about the progress of involved degradation processes and causes of breakdown. In contrast, dielectric response measurements are non-destructive, relatively simple to perform and provide more information concerning the insulation system studied, although they are difficult to use in on-line applications. Dielectric response measurement techniques can be divided into two categories: time domain and frequency domain. In general, the time domain techniques involve measurements of responses to either impulse or step voltage excitations. To diagnose high voltage insulation system quality, the latter approach is mainly used in the form of polarization and depolarization current (PDC) measurements and recovery voltage measurements (RVM). PDC measurements are used, for instance, in assessing the condition of the insulation in power transformers [1-3]. The frequency response measurements, also known as frequency domain spectroscopy (FDS), utilize sinusoidal signals of Manuscript received on 7 November 2007, in final form 27 February 2008.

variable frequency as the source of excitation. The areas of application in high voltage engineering are, for instance moisture estimation of oil-paper insulation in transformers and characterization of water treeing in XLPE cable insulation [2, 4, 5]. In general, aging and degradation of insulation materials is likely to affect the dielectric response, but the severity of the changes and the frequency range in which effects are discernible will vary between different materials and stresses. The expected dielectric response signature from a deteriorating insulation system is key knowledge for a serious diagnostic measurement. The time and frequency domain techniques have different areas of application due to inherent advantages and drawbacks. The PDC and RVM techniques employ waveforms containing harmonics of numerous frequencies. This makes them sensitive to disturbances from the used dc voltage source. Relay switching times as well as the procedure of connecting and disconnecting the voltage limit the ability of measuring short time responses. The rise time of the voltage source may also influence the results. Memory effects of material polarization must be taken into account and make it necessary to discharge the object sufficiently before commencing new measurements. These limitations make PDC and RVM techniques most useful at low frequencies (< 1 Hz) for materials with high dielectric losses (tan δ > 10-2) or significant dc conduction [3].

1070-9878/08/$25.00 © 2008 IEEE

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 15, No. 4; August 2008

The FDS is a narrow band technique which allows effective filtering in order to improve the accuracy. However, frequency domain measurements become increasingly time consuming at low frequencies [3]. Since it is preferable to record object behavior during at least two periods of the waveform, this makes the technique most suitable at frequencies higher than 0.01 Hz. Another major drawback of the techniques presented above is the inability to use them for on-line measurements. This is a severe limitation, especially in situations where the test object holds an important place in power system, such as a power transformer or cable, for example. Another drawback is that the voltage used for performing dielectric response measurements not always has the same magnitude and waveform as the one actually experienced by the characterized object or system in operation. Measurements of dielectric parameters that utilize waveforms similar to those experienced by the test object in operation have proven useful for characterization of, for instance, non-linear stress grading systems [6]. This indicates that the possibility of using non-sinusoidal waveform is a desirable ability for a dielectric response technique. Techniques for on-line monitoring of dielectric response have previously been developed and some are presented in [7, 8]. Common for these techniques is that the response is measured at power frequency only and the Fourier transform is used for filtering the harmonics rather than obtaining a wider dielectric response spectrum. Measuring at one frequency only gives a limited amount of data, whereas use of the harmonics close to applications such as HVDC or PWM (Pulse Width Modulation) provides a much wider spectrum. To meet the demand of online dielectric response measurements and to utilize the possible richness of harmonics in the power system, a new measuring system, called arbitrary waveform impedance spectroscopy (AWIS) is proposed and described in this paper. As regards the functioning principle, it can be placed in between the both categories, i.e. time and frequency domains, since it utilizes Fourier transforms of repetitive non-sinusoidal waveforms in order to obtain the dielectric response in frequency domain. The system is capable of utilizing the inherent harmonic content of applied voltage waveforms. Similar techniques have been developed in the field of electrochemistry, more specifically in the area of voltammetry, but those measurement systems are adapted for low voltage measurements and apply a controlled voltage waveform consisting of a number of superposed sinusoidal voltages [9] or use square voltages with low fundamental frequency (~ 10 Hz) [10]. The complications associated with the higher voltages and frequencies necessary for material characterization and diagnostics motivates a more high voltage oriented study of the technique. Some aspects related to the principal operation of AWIS have already been presented in [11-13], where comparisons of the accuracy of low voltage measurements with different waveforms are made, high voltage measurements are evaluated and the systems capability of monitoring change in time of dielectric properties is investigated.

921

2 ARBITRARY WAVEFORM IMPEDANCE SPECTROSCOPY (AWIS) The basic operation relies on separately measuring the voltage (V0) over the test object (Zto) and in series connected shunt (Zsh) as well as the voltage over the shunt (V1), as schematically illustrated in Figure 1. The correlations between the voltage over the test object (Vto) and the current flowing through it (Ito) are obtained from equations (1) and (2), whereas the impedance of the test object (Zto) is as in equation (3). Vto = V0 − V1

(1)

I to =

V1 Z sh

(2)

Z to =

Vto I to

(3)

The impedance is found by applying a voltage with an arbitrary waveform and measuring the voltage and the resulting current waveform with a data acquisition card (DAQ). An example of the waveform used for the dielectric response measurements presented here is shown in Figure 2, with the corresponding spectrum up to 1.2 kHz. For clarity, only 50 ms of the measured signal is shown in the figure, but 1.2 s are used for computations. The fast Fourier transform (FFT) is used to compute the harmonic spectrum of both voltage and current waveforms, where the peaks give the complex amplitude at the desired frequencies by fitting to the known line shape. Recording of about twenty periods of the signal is necessary for the curve fitting algorithm to fully resolve the shape of each peak. The basics of this technique can be found in [14] and similar applications were reported in [15-17]. Recording of a large number of samples also reduces the influence of random noise and improves the accuracy of the measurements. This offers advantages over the traditional method of synchronized sampling where it is preferable to analyze full periods and the fundamental frequency must be known. The precise ripple-free sinusoidal or dc voltage waveform required in conventional techniques is no longer

Figure 1. The basic circuit used for response measurements of a test object (Cto). The shunt resistor (Rsh) is used to measure the current passing through the test object.

922

B. Sonerud et al.: Dielectric Response Measurements Utilizing Semi-Square Voltage Waveforms

Figure 2. Example of voltage waveform and the corresponding frequency spectrum used for dielectric response measurements with AWIS.

necessary and it is possible to utilize voltages already present in the system to perform measurements. AWIS is most appropriate for applications at higher frequencies (> 10 Hz). For the same measurement time, the low frequency limit is about 10 times higher than for FDS as at least twenty periods of the repetitive waveform should be recorded compared with the two or four periods used by most FDS instruments. 2.1 LOW VOLTAGE MEASUREMENTS Low voltage dielectric response measurements with AWIS have been performed using a waveform generator (HP 33120A) as voltage source. Data were obtained with a 16-bit, 250 kS/s DAQ (NI PCI 6143). Component properties influencing the high frequency response, such as line inductances and stray capacitances, become important when the AWIS technique is used, especially at frequencies above 100 kHz. In order to limit their influence, the test object, shunt and connections should be shielded to reduce stray capacitances and make them well defined. The wiring should be kept as short as possible to reduce line inductances. The inherent properties of the circuit components should also be taken into consideration, e.g. the relatively high inductance of wire wound resistors. Despite of the efforts to limit stray capacitances and line inductances some will always remain and influence the measurement results. In order to cope with these and the stochastic nature of the parameter values of resistors and capacitors used throughout the circuit, circuit modeling and calibration measurements are performed. The model used is shown in Figure 3. A line inductance (Lto) is added in series with the capacitive test object (Cto). The shunt resistor is modeled as a resistance (Rsh) in parallel with a capacitance (Csh). A line inductance (Lsh) in series is also included. The component values of the model presented in Figure 3 were determined by performing calibration measurements on a standard capacitor (Hotex Standard Capacitor 1409-T, 100.0085 nF, loss factor: 3x10-4 at 1 kHz). The calibration

Figure 3. The low voltage AWIS model. Cto and Rsh are the physical components whereas Lto, Lsh and Csh represent the high frequency properties.

capacitor is assumed to have a capacitance independent of frequency and no losses when the calibration is performed. The shunt resistor (Rsh=10 kΩ ±1%) in the measurement results presented here is of standard electronic wire-wound type. This design will result in high inductance compared to other types, but in the frequency range of interest in this study it is negligible (< 0.1 µH). For measurements at higher frequencies a metal film resistor may be more suitable due to its reduced inductance. The capacitance (Csh) is dominated by the capacitance of the connecting coaxial cable, but also consists of the capacitance of the DAQ and stray capacitances. The results of a capacitance and loss measurement on a 10 nF PET capacitor by means of AWIS are compared with those obtained by two FDS instruments: IDA 200 Insulation Diagnostic System and Hewlett-Packard 4192 LF Impedance Analyzer. IDA 200 has an accuracy of 0.5 % + 1 pF (≈ 51 pF at 1 kHz) for the capacitance and 2 % + 5·10-4 (≈ 6·10-4 at 1 kHz) for the loss factor and a resolution of 10-5. The accuracy for the impedance analyzer is 1.02 % (≈ 102 pF) for the capacitance and 3·10-3 for the loss factor with a maximum resolution of 10-4. As shown in Figure 4a the agreement of the capacitance measurement results is good. The results from AWIS are consistent with the results obtained with the conventional FDS technique. In Figure 4b data from the loss factor measurements of the three different measurement systems are compared. In general the agreement between AWIS and the commercial instruments used here is good. The FDS instruments and the AWIS results seem to agree well although the frequency overlap is not optimal. The results from the impedance analyzer differ slightly from both the results of AWIS and the IDA 200 indicating that comparisons between different commercial instruments also show discrepancies. The deviations observed are in the same order of magnitude as the deviations between the both FDS instruments indicating that the accuracy of AWIS can be similar as the accuracy of the FDS technique.

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 15, No. 4; August 2008

2.2 HIGH VOLTAGE MEASUREMENTS In order to carry out high voltage dielectric response measurements a system consisting of a Spellman High Voltage Supply (RMP60N300/220) modified for positive polarity, a capacitor bank and a Behlke Fast High Voltage Transistor Switch (HTS 301-03-GSM) was used. In high voltage dielectric response measurements with AWIS, circuit modeling and calibration procedures similar to those performed in low voltage measurements are required. The model used is shown in Figure 5. Two main elements are introduced in the high voltage circuit: a voltage divider and a protective device which both have to be carefully designed in order to limit the influence on the measurement results. The high voltage circuit needs to be used when applied voltage exceeds the measurement range of the DAQ. In the present case, this effectively means that all voltages exceeding 5 V are considered as high voltage signals, in the sense that a voltage divider and protective device are needed. The upper voltage range is not limited. By modifying the component values of the voltage divider and the current shunt the same circuit topology can be used for any voltage, provided that the test setup and the components used are sufficiently insulated.

923

A voltage divider was constructed using a standard capacitor as the high impedance component (Cvd1= 95.6 pF) and a polypropylene capacitor as the low impedance component (C= 47 nF ±5%). In Cvd2 the capacitor as well as cable capacitances, stray capacitances and the input capacitance of the DAQ are included. A discharge resistor (R= 7 MΩ ±1%) is placed in parallel with Cvd2 to avoid charge accumulation in the internal impedance of the DAQ. It is modeled together with the input resistance of the DAQ as RDAQ. As a protective device a voltage follower was used. The voltage output from a voltage follower is limited by the voltage supply providing the amplifiers with power. The input to the DAQ is thus kept at safe voltage levels even in the case of test object breakdown. Its frequency response is simple to model with negligible voltage dependence. Examples of the measured response of the voltage follower and corresponding analytical polynomial curve fittings used for compensation of the influence of the protective devices on the measurement results are shown in Figure 6 and Figure 7. The deviation shown in Figure 6 between measured and fitted values can be attributed to the frequency distribution of the measured values and the nature of complex analytical curve fitting. The increased amount of measurement points at higher frequencies causes the fitted function to emphasize the high frequency behavior. This, in conjunction with the KramerKronig relation which the fitted function must obey, causes the discrepancy in measured and fitted values. Dielectric response measurements were performed on a 10 nF PET capacitor. In order to evaluate the accuracy the high voltage dielectric response was compared with a low voltage response from the same capacitor. For calibration the same standard capacitor was used as in the low voltage measurements. The results are shown in Figure 8 and Figure 9. As can be observed there is a difference of less than

a)

b)

Figure 4. a) Capacitance and b) loss factor measured with two commercial instruments (IDA 200 Insulation Diagnostic System and Hewlett-Packard 4192 LF Impedance Analyzer) and AWIS.

Figure 5. High voltage AWIS model. Cvd2 contains the low voltage capacitor of the voltage divider, the capacitance of the cable connecting the divider to the DAQ, the internal capacitance of the DAQ and stray capacitances. RDAQ contains the input resistance of the DAQ and the external discharging resistor.

924

B. Sonerud et al.: Dielectric Response Measurements Utilizing Semi-Square Voltage Waveforms

Figure 6. Absolute values of the relation Vout/Vin of one channel of the voltage follower.

20 pF (2‰) between the capacitance obtained with high voltage and low voltage measurements. The difference in loss factor is less than 10-3. It is difficult to achieve a more accurate calibration. As illustrated in Figures 8 and 9, the deviation between high and low voltage measurements is in the same order of magnitude as the deviation between different voltage levels within the same setup (low or high voltage). This indicates that high voltage AWIS measurements are suitable for obtaining high voltage dielectric response and that the calibration errors are at a sufficiently low level. 2.3 ERROR SOURCES There are several types of errors that can influence measurements with the AWIS system. Random noise errors will influence the results and their magnitude can be estimated by the signal to noise ratio in the spectrum. In order to achieve accuracy in tan δ at a level of 10-3, the ratio must be at least 1000:1. In Figure 2 an example of the spectrum of a repetitive voltage wave is shown where the noise level is at 10-5 V and

Figure 7. The phase difference between Vout and Vin of one channel of the voltage follower.

Figure 8. Comparison between low voltage (2-4 Vpp bipolar) and high voltage (10-200 Vpp unipolar) capacitance measurement.

the fundamental harmonic is at 100 V. In this particular case the accuracy of the extracted amplitude is 10-5. The accuracy can be increased by averaging of the obtained values. In the spectrum in Figure 2 the fundamental and odd harmonics have larger magnitudes than the even harmonics, and the odd harmonics decrease with increasing frequency. For some waveforms the even harmonics can increase in amplitude with increasing frequency due to asymmetry in the waveshape. It is therefore important to only use those harmonics in the analysis which have magnitudes high enough to provide accurate measurement results. Errors can also occur due to insufficient accuracy of the calibration procedure. For instance, the accuracy of the reference capacitor will influence the measurements. There are always some limitations in the model, since stray capacitances and line inductances are distributed, not lumped and together with non-ideal standard capacitors these give rise to errors. Efforts should also be taken to keep non-linear effects of system components, e. g. voltage dependent capacitors, below the required accuracy level.

Figure 9. Comparison between low voltage (2-4 Vpp bipolar) and high voltage (10-200 Vpp unipolar) loss factor measurement.

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 15, No. 4; August 2008

Effects of aliasing, which are known from signal processing theory, must be taken into account as an error source when using FFT [18]. With the DAQ presently used the Nyquist frequency is 125 kHz. Since AWIS could use the waveforms present in a system instead of a controlled waveform there is always a possibility that high order harmonics will influence the results. By monitoring the harmonic spectrum at the initiation of a measurement the influence of aliasing can be assessed. If it is significant a crude anti-aliasing filter can be created by increasing the capacitance Csh or Cvd2 in Figure 5. This is easily done by placing a capacitor in parallel with the current shunt or the low voltage part of the voltage divider, although the component must be included in the circuit modeling and calibration procedures. Thus, even if the voltage waveform has a too rich harmonic content for the required sampling frequency, aliasing can be avoided by an appropriate design of the sensor response. Non-linear effects in the insulation system will generate a harmonic current which will add to the capacitive current. It is therefore in general difficult to separate non-linear and capacitive effects from a single measurement with this method. If measurements at several voltages are available however, the magnitude of the nonlinear harmonic current can be estimated to less than the non-linear part of the current at the fundamental frequency. In the results shown here, the non-linearity at fundamental frequency is low, less than 1 % and the harmonic amplitude is large, thus nonlinear effects may only have a very small influence. If the voltage or the frequency is not under control, it is important to consider possible non-linear effects in the analysis by comparing to laboratory measurements of a similar insulation system. When high accuracy in absolute values is of subordinate interest, such as during continuous monitoring of dielectric properties, measurements containing systematic errors might still be of interest. In those cases focus is on the relative change of the dielectric property, not the absolute value. The most devastating error effect during this kind of measurements is caused by drift during measurement time. This is most notably caused by temperature drift in components and surroundings, although investigations has proved that during normal operating conditions (where the increase of ambient and component temperature is below 3 °C) the influence of temperature remains negligible [13].

3 DISCUSSION As illustrated, the AWIS measurement system provides results consistent with the FDS technique. This indicates that AWIS might be used for characterization and condition monitoring of insulation systems of high voltage equipment. Although presently untested in field applications, the fundamental principle has several advantages compared to other techniques. It can use any waveform, although the range of useful harmonics will be effectively limited by the wave shape. For instance, an applied sinusoidal voltage will only provide results at the fundamental frequency whereas a square

925

voltage will deliver a much wider spectrum. This ability could make it possible to perform on-line monitoring by utilizing voltages already present in the system as a source voltage during on-line conditions. However, it might be difficult to perform precise calibration and focus must then be to monitor changes in the object of interest rather than obtaining absolute values. The harmonic content available in the system during a field measurement will be insufficient for diagnostics of some components and materials. Aging behavior must manifest itself in the available frequency region; otherwise the insulation system cannot be assessed with AWIS. It would be appropriate to measure on components subjected to voltages with a high amount of harmonics, e.g. in close proximity to HVDC or PWM applications. With the DAQ used for the measurement results presented here, AWIS is capable of performing dielectric response measurements on seven test objects simultaneously, provided that the source voltage is the same for all of them. It is possible to obtain a dielectric response every 15 seconds using waveforms with a fundamental frequency in the order of 100 Hz. At frequencies higher than 1 Hz the processing time of the computer is the limiting factor of how many dielectric response spectra it is possible to obtain per time unit. At lower frequencies the period of the waveform determines the measuring time. For instance, twenty periods of a waveform with a fundamental frequency of 0.1 Hz would take 200 s to measure. As indicated in the discussion of error sources there are also several limitations in the accuracy of AWIS. Some of these errors are common for several dielectric response techniques, such as background noise and temperature influence. Others are more specific for AWIS, such as calibration errors and the influence of aliasing. Countermeasures reducing the impact of these errors are available but they cannot be entirely avoided.

4 CONCLUSIONS In all, the AWIS system provides a flexible and accurate system with the possibility of performing continuous measurements with high time resolution. New possibilities opens in the area of dielectric response measurements and in contrast to FDS and PDC, where a significant effort is put into producing the well-defined voltage, it is possible to use any waveform present in the system to perform the measurement with AWIS. By performing continuous response measurements covering a range of harmonics, changes of capacitance and loss factor over time can be monitored. This has uses both in on-line monitoring and diagnostics, where the voltage present in the system can be used as source voltage for the measurements. AWIS is also useful in experimental setups where different aspects related to materials and insulating systems behavior can be investigated. The influence of waveforms, e.g. the rise time and frequency of semi-square voltages and the corresponding development over time of the dielectric properties can be studied. Voltage dependent effects may also be monitored in a similar fashion.

926

B. Sonerud et al.: Dielectric Response Measurements Utilizing Semi-Square Voltage Waveforms

Some new error sources are introduced compared with other dielectric response techniques. Although the influence of aliasing and the need for thorough modeling and calibration procedures are of importance there are ways of dealing with these complications.

ACKNOWLEDGMENT This work has been carried out within ELEKTRA project 3669 financed jointly by the Swedish Energy Agency, Elforsk, ABB, Areva and Banverket.

REFERENCES [1] [2]

[3]

[4] [5] [6]

[7]

[8]

[9]

[10] [11]

[12]

[13]

[14]

W. S. Zaengl, "Dielectric spectroscopy in time and frequency domain for HV power equipment. I. Theoretical considerations", IEEE Electr. Insul. Mag., Vol. 19, No. 5, pp. 5-19, 2003. U. Gafvert, L. Adeen, M. Tapper, P. Ghasemi, and B. Jonsson, "Dielectric spectroscopy in time and frequency domain applied to diagnostics of power transformers", 6th Intern. Conf. Properties and Applications of Dielectric Materials (Cat. No.00CH36347), Xi'an, China, pp. 825-830, 2000. A. Helgeson, Analysis of dielectric response measurement methods and dielectric properties of resin-rich insulation during processing, Ph.D. thesis, Royal Institute of Technology, Department of Electric Power Engineering, 2000. B. Oyegoke, P. Hyvonen, M. Aro, and G. Ning, "Application of dielectric response measurement on power cable systems", IEEE Trans. Dielectr. Electr. Insul., Vol. 10, pp. 862-873, 2003. W. S. Zaengl, "Applications of dielectric spectroscopy in time and frequency domain for HV power equipment", IEEE Electr. Insul. Mag., Vol. 19, No. 6, pp. 9-22, 2003. F. P. Espino-Cortes, Y. Montasser, S. H. Jayaram, and E. A. Cherney, "Study of stress grading systems working under fast rise time pulses", IEEE Intern. Sympos. Electr. Insul. (IEEE Cat. No. 06CH37794), Toronto, Ont., Canada, pp. 380-383, 2006. W. Pei, M. R. Raghuveer, W. McDermid, and J. C. Bromley, "A digital technique for the on-line measurement of dissipation factor and capacitance", IEEE Trans. Dielectr. Electr. Insul., Vol. 8, pp. 228-32, 2001. L. Sang Bin, K. Younsi, and G. B. Kliman, "An online technique for monitoring the insulation condition of AC machine stator windings", IEEE Trans. Energy Conversion, Vol. 20, pp. 737-45, 2005. M. Rosvall and M. Sharp, "A complete system for electrochemical impedance spectroscopy which combines FFT methods and staircase voltammetry", Electrochemistry Communications, Vol. 2, pp. 338-343, 2000. D. J. Gavaghan, D. Elton, K. B. Oldham, and A. M. Bond, "Analysis of ramped square-wave voltammetry in the frequency domain", J. Electroanalytical Chemistry, Vol. 512, pp. 1-15, 2001. B. Sonerud, T. Bengtsson, J. Blennow, and S. M. Gubanski, "Dielectric response measurements utilizing non-sinusoidal waveforms", IEEE Conf. Electr. Insul. Dielectr. Phenomena (CEIDP 2006), Kansas City, Missouri, USA, pp. 43-46, 2006. B. Sonerud, T. Bengtsson, J. Blennow, and S. M. Gubanski, "High and low voltage dielectric response measurements utilizing arbitrarily shaped waveforms", Intern. Sympos. High Voltage Engineering (ISH 2007), Ljubljana, Slovenia, 2007. B. Sonerud, T. Bengtsson, J. Blennow, and S. M. Gubanski, "Utilizing High Voltage Containing High Frequency Components for Continuous Dielectric Response Measurements during Aging", Nordic Insulation Symposium (NORD-IS 07), Lyngby, Denmark, pp. 73-76, 2007. F. J. Harris, "On the use of windows for harmonic analysis with the discrete Fourier transform", IEEE Proc., Vol. 66, pp. 51-83, 1978.

[15] [16]

[17] [18]

J. Hedberg and T. Bengtsson, "Straight Dielectric Response Measurements with High Precision", Nordic Insulation Symposium (NORD-IS 05), Trondheim, Norway, pp. 174-177, 2005. G. Andria, M. Savino, and A. Trotta, "Windows and interpolation algorithms to improve electrical measurement accuracy", IEEE Trans. Instrumentation and Measurement, Vol. 38, pp. 856-863, 1989. V. K. Jain, W. L. Collins, and D. C. Davis, "High-Accuracy Analog Measurements via Interpolated FFT", IEEE Trans. Instrumentation and Measurement, Vol. 28, pp. 113-122, 1979. A. V. Oppenheim, A. S. Willsky, and S. H. Nawab, Signals & Systems, 2 ed. Upper Saddle River: Prentice Hall, 1997. Björn Sonerud was born in Alvesta, Sweden in 1979. He received the M.Sc. degree from Chalmers University of Technology, Gothenburg, Sweden in 2004. He is currently pursuing the Ph.D. degree at Chalmers University of Technology. His research interests include dielectric response measurement techniques, material characterization and aging.

Tord Bengtsson (M’06) was born in Sweden in 1954. He received the Ph.D. degree in mathematical physics from Lund Institute of Technology in 1984 and continued working with theoretical nuclear physics until 1991, when he joined ABB Corporate Research in Västerås, Sweden. Since then, he has been actively engaged in developing diagnostic methods for high voltage apparatuses and systems. Some particular interests are acoustic methods, partial discharges and signal processing. In 2006, he was appointed as a part-time professor in High Voltage Engineering at Chalmers University of Technology in recognition of his efforts to develop research on insulation properties under voltages with fast rise times. Jörgen Blennow (S'94-M'01) was born in Veberöd, Sweden in 1966. He graduated in electrical engineering from Chalmers University of Technology, Gothenburg, Sweden in 1993. In 1996 and 2000 he obtained the degrees of Licentiate of Engineering and Doctor of Philosophy, respectively, in high voltage engineering. In 2001 he became Assistant Professor and since 2006 he holds a position as Senior Lecturer in High Voltage Engineering at the same university. His main research activities have been towards different aspects of insulation exposed to high frequency high voltage, air-solid insulation systems and transformer diagnostics based on dielectric spectroscopy. Stanislaw M. Gubanski (M'89-SM'90-F'01) received the M.Sc. (high voltage engineering) and Ph.D. degrees (material science) from the Technical University of Wroclaw, Poland, in 1973 and 1976, respectively. He was a Research Fellow at the University College of North Wales Bangor, U.K from 1967 to 1977, a senior lecturer at the technical university of Wroclaw, Wroclaw, Poland, from 1977 to 1988. Afterwards he was associate professor at the Royal Institute of Technology, Stockholm, Sweden. Currently, he is Professor in High Voltage Engineering at the Department of Manufacturing and Materials Technology, Chalmers University of Technology, Gothenburg and research leader of High Voltage Valley in Ludvika, Sweden. He is Chair of IEEE-DEIS Nominations Committee. He is also convener of the CIGRE Task Force D1.01.14 “Dielectric Response Methods for Diagnostics of Power Transformers”.

B. Sonerud et al.: Dielectric Response Measurements Utilizing Semi-Square Voltage Waveforms

Dielectric Response Measurements Utilizing Semi-Square Voltage Waveforms Björn Sonerud, Tord Bengtsson*, Jörgen Blennow and Stanislaw M. Gubanski Department of Materials and Manufacturing Technology Chalmers University of Technology SE 412 96, Göteborg, Sweden *also with ABB Corporate Research, SE 721 78 Västerås, Sweden

ABSTRACT Dielectric response measurements belong to the group of diagnostic tools used for quality evaluation of high voltage insulation systems. A measurement system called Arbitrary Waveform Impedance Spectroscopy (AWIS) is presented here which is capable of utilizing the harmonic content of repetitive voltages with a semi-square waveform and the corresponding current response to perform dielectric response measurements. Extensive circuit modeling and calibrations are required in order to perform accurate measurements. AWIS could offer new possibilities, especially for continuous monitoring of the dielectric properties of high voltage components and systems. The accuracy of the technique in both low and high voltage applications is verified by comparison with results obtained by means of the Frequency Domain Spectroscopy (FDS) technique. Index Terms — Dielectric measurements, discrete Fourier transforms, aging, monitoring, dielectric materials, insulation testing.

1 INTRODUCTION THE long term properties of high voltage insulation systems are mostly studied through investigations of time to breakdown, breakdown strength and dielectric response. The tests of time to breakdown and breakdown strength are simple to perform but result in the destruction of the tested objects. At the same time these tests offer limited information about the progress of involved degradation processes and causes of breakdown. In contrast, dielectric response measurements are non-destructive, relatively simple to perform and provide more information concerning the insulation system studied, although they are difficult to use in on-line applications. Dielectric response measurement techniques can be divided into two categories: time domain and frequency domain. In general, the time domain techniques involve measurements of responses to either impulse or step voltage excitations. To diagnose high voltage insulation system quality, the latter approach is mainly used in the form of polarization and depolarization current (PDC) measurements and recovery voltage measurements (RVM). PDC measurements are used, for instance, in assessing the condition of the insulation in power transformers [1-3]. The frequency response measurements, also known as frequency domain spectroscopy (FDS), utilize sinusoidal signals of Manuscript received on 7 November 2007, in final form 27 February 2008.

variable frequency as the source of excitation. The areas of application in high voltage engineering are, for instance moisture estimation of oil-paper insulation in transformers and characterization of water treeing in XLPE cable insulation [2, 4, 5]. In general, aging and degradation of insulation materials is likely to affect the dielectric response, but the severity of the changes and the frequency range in which effects are discernible will vary between different materials and stresses. The expected dielectric response signature from a deteriorating insulation system is key knowledge for a serious diagnostic measurement. The time and frequency domain techniques have different areas of application due to inherent advantages and drawbacks. The PDC and RVM techniques employ waveforms containing harmonics of numerous frequencies. This makes them sensitive to disturbances from the used dc voltage source. Relay switching times as well as the procedure of connecting and disconnecting the voltage limit the ability of measuring short time responses. The rise time of the voltage source may also influence the results. Memory effects of material polarization must be taken into account and make it necessary to discharge the object sufficiently before commencing new measurements. These limitations make PDC and RVM techniques most useful at low frequencies (< 1 Hz) for materials with high dielectric losses (tan δ > 10-2) or significant dc conduction [3].

1070-9878/08/$25.00 © 2008 IEEE

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 15, No. 4; August 2008

The FDS is a narrow band technique which allows effective filtering in order to improve the accuracy. However, frequency domain measurements become increasingly time consuming at low frequencies [3]. Since it is preferable to record object behavior during at least two periods of the waveform, this makes the technique most suitable at frequencies higher than 0.01 Hz. Another major drawback of the techniques presented above is the inability to use them for on-line measurements. This is a severe limitation, especially in situations where the test object holds an important place in power system, such as a power transformer or cable, for example. Another drawback is that the voltage used for performing dielectric response measurements not always has the same magnitude and waveform as the one actually experienced by the characterized object or system in operation. Measurements of dielectric parameters that utilize waveforms similar to those experienced by the test object in operation have proven useful for characterization of, for instance, non-linear stress grading systems [6]. This indicates that the possibility of using non-sinusoidal waveform is a desirable ability for a dielectric response technique. Techniques for on-line monitoring of dielectric response have previously been developed and some are presented in [7, 8]. Common for these techniques is that the response is measured at power frequency only and the Fourier transform is used for filtering the harmonics rather than obtaining a wider dielectric response spectrum. Measuring at one frequency only gives a limited amount of data, whereas use of the harmonics close to applications such as HVDC or PWM (Pulse Width Modulation) provides a much wider spectrum. To meet the demand of online dielectric response measurements and to utilize the possible richness of harmonics in the power system, a new measuring system, called arbitrary waveform impedance spectroscopy (AWIS) is proposed and described in this paper. As regards the functioning principle, it can be placed in between the both categories, i.e. time and frequency domains, since it utilizes Fourier transforms of repetitive non-sinusoidal waveforms in order to obtain the dielectric response in frequency domain. The system is capable of utilizing the inherent harmonic content of applied voltage waveforms. Similar techniques have been developed in the field of electrochemistry, more specifically in the area of voltammetry, but those measurement systems are adapted for low voltage measurements and apply a controlled voltage waveform consisting of a number of superposed sinusoidal voltages [9] or use square voltages with low fundamental frequency (~ 10 Hz) [10]. The complications associated with the higher voltages and frequencies necessary for material characterization and diagnostics motivates a more high voltage oriented study of the technique. Some aspects related to the principal operation of AWIS have already been presented in [11-13], where comparisons of the accuracy of low voltage measurements with different waveforms are made, high voltage measurements are evaluated and the systems capability of monitoring change in time of dielectric properties is investigated.

921

2 ARBITRARY WAVEFORM IMPEDANCE SPECTROSCOPY (AWIS) The basic operation relies on separately measuring the voltage (V0) over the test object (Zto) and in series connected shunt (Zsh) as well as the voltage over the shunt (V1), as schematically illustrated in Figure 1. The correlations between the voltage over the test object (Vto) and the current flowing through it (Ito) are obtained from equations (1) and (2), whereas the impedance of the test object (Zto) is as in equation (3). Vto = V0 − V1

(1)

I to =

V1 Z sh

(2)

Z to =

Vto I to

(3)

The impedance is found by applying a voltage with an arbitrary waveform and measuring the voltage and the resulting current waveform with a data acquisition card (DAQ). An example of the waveform used for the dielectric response measurements presented here is shown in Figure 2, with the corresponding spectrum up to 1.2 kHz. For clarity, only 50 ms of the measured signal is shown in the figure, but 1.2 s are used for computations. The fast Fourier transform (FFT) is used to compute the harmonic spectrum of both voltage and current waveforms, where the peaks give the complex amplitude at the desired frequencies by fitting to the known line shape. Recording of about twenty periods of the signal is necessary for the curve fitting algorithm to fully resolve the shape of each peak. The basics of this technique can be found in [14] and similar applications were reported in [15-17]. Recording of a large number of samples also reduces the influence of random noise and improves the accuracy of the measurements. This offers advantages over the traditional method of synchronized sampling where it is preferable to analyze full periods and the fundamental frequency must be known. The precise ripple-free sinusoidal or dc voltage waveform required in conventional techniques is no longer

Figure 1. The basic circuit used for response measurements of a test object (Cto). The shunt resistor (Rsh) is used to measure the current passing through the test object.

922

B. Sonerud et al.: Dielectric Response Measurements Utilizing Semi-Square Voltage Waveforms

Figure 2. Example of voltage waveform and the corresponding frequency spectrum used for dielectric response measurements with AWIS.

necessary and it is possible to utilize voltages already present in the system to perform measurements. AWIS is most appropriate for applications at higher frequencies (> 10 Hz). For the same measurement time, the low frequency limit is about 10 times higher than for FDS as at least twenty periods of the repetitive waveform should be recorded compared with the two or four periods used by most FDS instruments. 2.1 LOW VOLTAGE MEASUREMENTS Low voltage dielectric response measurements with AWIS have been performed using a waveform generator (HP 33120A) as voltage source. Data were obtained with a 16-bit, 250 kS/s DAQ (NI PCI 6143). Component properties influencing the high frequency response, such as line inductances and stray capacitances, become important when the AWIS technique is used, especially at frequencies above 100 kHz. In order to limit their influence, the test object, shunt and connections should be shielded to reduce stray capacitances and make them well defined. The wiring should be kept as short as possible to reduce line inductances. The inherent properties of the circuit components should also be taken into consideration, e.g. the relatively high inductance of wire wound resistors. Despite of the efforts to limit stray capacitances and line inductances some will always remain and influence the measurement results. In order to cope with these and the stochastic nature of the parameter values of resistors and capacitors used throughout the circuit, circuit modeling and calibration measurements are performed. The model used is shown in Figure 3. A line inductance (Lto) is added in series with the capacitive test object (Cto). The shunt resistor is modeled as a resistance (Rsh) in parallel with a capacitance (Csh). A line inductance (Lsh) in series is also included. The component values of the model presented in Figure 3 were determined by performing calibration measurements on a standard capacitor (Hotex Standard Capacitor 1409-T, 100.0085 nF, loss factor: 3x10-4 at 1 kHz). The calibration

Figure 3. The low voltage AWIS model. Cto and Rsh are the physical components whereas Lto, Lsh and Csh represent the high frequency properties.

capacitor is assumed to have a capacitance independent of frequency and no losses when the calibration is performed. The shunt resistor (Rsh=10 kΩ ±1%) in the measurement results presented here is of standard electronic wire-wound type. This design will result in high inductance compared to other types, but in the frequency range of interest in this study it is negligible (< 0.1 µH). For measurements at higher frequencies a metal film resistor may be more suitable due to its reduced inductance. The capacitance (Csh) is dominated by the capacitance of the connecting coaxial cable, but also consists of the capacitance of the DAQ and stray capacitances. The results of a capacitance and loss measurement on a 10 nF PET capacitor by means of AWIS are compared with those obtained by two FDS instruments: IDA 200 Insulation Diagnostic System and Hewlett-Packard 4192 LF Impedance Analyzer. IDA 200 has an accuracy of 0.5 % + 1 pF (≈ 51 pF at 1 kHz) for the capacitance and 2 % + 5·10-4 (≈ 6·10-4 at 1 kHz) for the loss factor and a resolution of 10-5. The accuracy for the impedance analyzer is 1.02 % (≈ 102 pF) for the capacitance and 3·10-3 for the loss factor with a maximum resolution of 10-4. As shown in Figure 4a the agreement of the capacitance measurement results is good. The results from AWIS are consistent with the results obtained with the conventional FDS technique. In Figure 4b data from the loss factor measurements of the three different measurement systems are compared. In general the agreement between AWIS and the commercial instruments used here is good. The FDS instruments and the AWIS results seem to agree well although the frequency overlap is not optimal. The results from the impedance analyzer differ slightly from both the results of AWIS and the IDA 200 indicating that comparisons between different commercial instruments also show discrepancies. The deviations observed are in the same order of magnitude as the deviations between the both FDS instruments indicating that the accuracy of AWIS can be similar as the accuracy of the FDS technique.

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 15, No. 4; August 2008

2.2 HIGH VOLTAGE MEASUREMENTS In order to carry out high voltage dielectric response measurements a system consisting of a Spellman High Voltage Supply (RMP60N300/220) modified for positive polarity, a capacitor bank and a Behlke Fast High Voltage Transistor Switch (HTS 301-03-GSM) was used. In high voltage dielectric response measurements with AWIS, circuit modeling and calibration procedures similar to those performed in low voltage measurements are required. The model used is shown in Figure 5. Two main elements are introduced in the high voltage circuit: a voltage divider and a protective device which both have to be carefully designed in order to limit the influence on the measurement results. The high voltage circuit needs to be used when applied voltage exceeds the measurement range of the DAQ. In the present case, this effectively means that all voltages exceeding 5 V are considered as high voltage signals, in the sense that a voltage divider and protective device are needed. The upper voltage range is not limited. By modifying the component values of the voltage divider and the current shunt the same circuit topology can be used for any voltage, provided that the test setup and the components used are sufficiently insulated.

923

A voltage divider was constructed using a standard capacitor as the high impedance component (Cvd1= 95.6 pF) and a polypropylene capacitor as the low impedance component (C= 47 nF ±5%). In Cvd2 the capacitor as well as cable capacitances, stray capacitances and the input capacitance of the DAQ are included. A discharge resistor (R= 7 MΩ ±1%) is placed in parallel with Cvd2 to avoid charge accumulation in the internal impedance of the DAQ. It is modeled together with the input resistance of the DAQ as RDAQ. As a protective device a voltage follower was used. The voltage output from a voltage follower is limited by the voltage supply providing the amplifiers with power. The input to the DAQ is thus kept at safe voltage levels even in the case of test object breakdown. Its frequency response is simple to model with negligible voltage dependence. Examples of the measured response of the voltage follower and corresponding analytical polynomial curve fittings used for compensation of the influence of the protective devices on the measurement results are shown in Figure 6 and Figure 7. The deviation shown in Figure 6 between measured and fitted values can be attributed to the frequency distribution of the measured values and the nature of complex analytical curve fitting. The increased amount of measurement points at higher frequencies causes the fitted function to emphasize the high frequency behavior. This, in conjunction with the KramerKronig relation which the fitted function must obey, causes the discrepancy in measured and fitted values. Dielectric response measurements were performed on a 10 nF PET capacitor. In order to evaluate the accuracy the high voltage dielectric response was compared with a low voltage response from the same capacitor. For calibration the same standard capacitor was used as in the low voltage measurements. The results are shown in Figure 8 and Figure 9. As can be observed there is a difference of less than

a)

b)

Figure 4. a) Capacitance and b) loss factor measured with two commercial instruments (IDA 200 Insulation Diagnostic System and Hewlett-Packard 4192 LF Impedance Analyzer) and AWIS.

Figure 5. High voltage AWIS model. Cvd2 contains the low voltage capacitor of the voltage divider, the capacitance of the cable connecting the divider to the DAQ, the internal capacitance of the DAQ and stray capacitances. RDAQ contains the input resistance of the DAQ and the external discharging resistor.

924

B. Sonerud et al.: Dielectric Response Measurements Utilizing Semi-Square Voltage Waveforms

Figure 6. Absolute values of the relation Vout/Vin of one channel of the voltage follower.

20 pF (2‰) between the capacitance obtained with high voltage and low voltage measurements. The difference in loss factor is less than 10-3. It is difficult to achieve a more accurate calibration. As illustrated in Figures 8 and 9, the deviation between high and low voltage measurements is in the same order of magnitude as the deviation between different voltage levels within the same setup (low or high voltage). This indicates that high voltage AWIS measurements are suitable for obtaining high voltage dielectric response and that the calibration errors are at a sufficiently low level. 2.3 ERROR SOURCES There are several types of errors that can influence measurements with the AWIS system. Random noise errors will influence the results and their magnitude can be estimated by the signal to noise ratio in the spectrum. In order to achieve accuracy in tan δ at a level of 10-3, the ratio must be at least 1000:1. In Figure 2 an example of the spectrum of a repetitive voltage wave is shown where the noise level is at 10-5 V and

Figure 7. The phase difference between Vout and Vin of one channel of the voltage follower.

Figure 8. Comparison between low voltage (2-4 Vpp bipolar) and high voltage (10-200 Vpp unipolar) capacitance measurement.

the fundamental harmonic is at 100 V. In this particular case the accuracy of the extracted amplitude is 10-5. The accuracy can be increased by averaging of the obtained values. In the spectrum in Figure 2 the fundamental and odd harmonics have larger magnitudes than the even harmonics, and the odd harmonics decrease with increasing frequency. For some waveforms the even harmonics can increase in amplitude with increasing frequency due to asymmetry in the waveshape. It is therefore important to only use those harmonics in the analysis which have magnitudes high enough to provide accurate measurement results. Errors can also occur due to insufficient accuracy of the calibration procedure. For instance, the accuracy of the reference capacitor will influence the measurements. There are always some limitations in the model, since stray capacitances and line inductances are distributed, not lumped and together with non-ideal standard capacitors these give rise to errors. Efforts should also be taken to keep non-linear effects of system components, e. g. voltage dependent capacitors, below the required accuracy level.

Figure 9. Comparison between low voltage (2-4 Vpp bipolar) and high voltage (10-200 Vpp unipolar) loss factor measurement.

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 15, No. 4; August 2008

Effects of aliasing, which are known from signal processing theory, must be taken into account as an error source when using FFT [18]. With the DAQ presently used the Nyquist frequency is 125 kHz. Since AWIS could use the waveforms present in a system instead of a controlled waveform there is always a possibility that high order harmonics will influence the results. By monitoring the harmonic spectrum at the initiation of a measurement the influence of aliasing can be assessed. If it is significant a crude anti-aliasing filter can be created by increasing the capacitance Csh or Cvd2 in Figure 5. This is easily done by placing a capacitor in parallel with the current shunt or the low voltage part of the voltage divider, although the component must be included in the circuit modeling and calibration procedures. Thus, even if the voltage waveform has a too rich harmonic content for the required sampling frequency, aliasing can be avoided by an appropriate design of the sensor response. Non-linear effects in the insulation system will generate a harmonic current which will add to the capacitive current. It is therefore in general difficult to separate non-linear and capacitive effects from a single measurement with this method. If measurements at several voltages are available however, the magnitude of the nonlinear harmonic current can be estimated to less than the non-linear part of the current at the fundamental frequency. In the results shown here, the non-linearity at fundamental frequency is low, less than 1 % and the harmonic amplitude is large, thus nonlinear effects may only have a very small influence. If the voltage or the frequency is not under control, it is important to consider possible non-linear effects in the analysis by comparing to laboratory measurements of a similar insulation system. When high accuracy in absolute values is of subordinate interest, such as during continuous monitoring of dielectric properties, measurements containing systematic errors might still be of interest. In those cases focus is on the relative change of the dielectric property, not the absolute value. The most devastating error effect during this kind of measurements is caused by drift during measurement time. This is most notably caused by temperature drift in components and surroundings, although investigations has proved that during normal operating conditions (where the increase of ambient and component temperature is below 3 °C) the influence of temperature remains negligible [13].

3 DISCUSSION As illustrated, the AWIS measurement system provides results consistent with the FDS technique. This indicates that AWIS might be used for characterization and condition monitoring of insulation systems of high voltage equipment. Although presently untested in field applications, the fundamental principle has several advantages compared to other techniques. It can use any waveform, although the range of useful harmonics will be effectively limited by the wave shape. For instance, an applied sinusoidal voltage will only provide results at the fundamental frequency whereas a square

925

voltage will deliver a much wider spectrum. This ability could make it possible to perform on-line monitoring by utilizing voltages already present in the system as a source voltage during on-line conditions. However, it might be difficult to perform precise calibration and focus must then be to monitor changes in the object of interest rather than obtaining absolute values. The harmonic content available in the system during a field measurement will be insufficient for diagnostics of some components and materials. Aging behavior must manifest itself in the available frequency region; otherwise the insulation system cannot be assessed with AWIS. It would be appropriate to measure on components subjected to voltages with a high amount of harmonics, e.g. in close proximity to HVDC or PWM applications. With the DAQ used for the measurement results presented here, AWIS is capable of performing dielectric response measurements on seven test objects simultaneously, provided that the source voltage is the same for all of them. It is possible to obtain a dielectric response every 15 seconds using waveforms with a fundamental frequency in the order of 100 Hz. At frequencies higher than 1 Hz the processing time of the computer is the limiting factor of how many dielectric response spectra it is possible to obtain per time unit. At lower frequencies the period of the waveform determines the measuring time. For instance, twenty periods of a waveform with a fundamental frequency of 0.1 Hz would take 200 s to measure. As indicated in the discussion of error sources there are also several limitations in the accuracy of AWIS. Some of these errors are common for several dielectric response techniques, such as background noise and temperature influence. Others are more specific for AWIS, such as calibration errors and the influence of aliasing. Countermeasures reducing the impact of these errors are available but they cannot be entirely avoided.

4 CONCLUSIONS In all, the AWIS system provides a flexible and accurate system with the possibility of performing continuous measurements with high time resolution. New possibilities opens in the area of dielectric response measurements and in contrast to FDS and PDC, where a significant effort is put into producing the well-defined voltage, it is possible to use any waveform present in the system to perform the measurement with AWIS. By performing continuous response measurements covering a range of harmonics, changes of capacitance and loss factor over time can be monitored. This has uses both in on-line monitoring and diagnostics, where the voltage present in the system can be used as source voltage for the measurements. AWIS is also useful in experimental setups where different aspects related to materials and insulating systems behavior can be investigated. The influence of waveforms, e.g. the rise time and frequency of semi-square voltages and the corresponding development over time of the dielectric properties can be studied. Voltage dependent effects may also be monitored in a similar fashion.

926

B. Sonerud et al.: Dielectric Response Measurements Utilizing Semi-Square Voltage Waveforms

Some new error sources are introduced compared with other dielectric response techniques. Although the influence of aliasing and the need for thorough modeling and calibration procedures are of importance there are ways of dealing with these complications.

ACKNOWLEDGMENT This work has been carried out within ELEKTRA project 3669 financed jointly by the Swedish Energy Agency, Elforsk, ABB, Areva and Banverket.

REFERENCES [1] [2]

[3]

[4] [5] [6]

[7]

[8]

[9]

[10] [11]

[12]

[13]

[14]

W. S. Zaengl, "Dielectric spectroscopy in time and frequency domain for HV power equipment. I. Theoretical considerations", IEEE Electr. Insul. Mag., Vol. 19, No. 5, pp. 5-19, 2003. U. Gafvert, L. Adeen, M. Tapper, P. Ghasemi, and B. Jonsson, "Dielectric spectroscopy in time and frequency domain applied to diagnostics of power transformers", 6th Intern. Conf. Properties and Applications of Dielectric Materials (Cat. No.00CH36347), Xi'an, China, pp. 825-830, 2000. A. Helgeson, Analysis of dielectric response measurement methods and dielectric properties of resin-rich insulation during processing, Ph.D. thesis, Royal Institute of Technology, Department of Electric Power Engineering, 2000. B. Oyegoke, P. Hyvonen, M. Aro, and G. Ning, "Application of dielectric response measurement on power cable systems", IEEE Trans. Dielectr. Electr. Insul., Vol. 10, pp. 862-873, 2003. W. S. Zaengl, "Applications of dielectric spectroscopy in time and frequency domain for HV power equipment", IEEE Electr. Insul. Mag., Vol. 19, No. 6, pp. 9-22, 2003. F. P. Espino-Cortes, Y. Montasser, S. H. Jayaram, and E. A. Cherney, "Study of stress grading systems working under fast rise time pulses", IEEE Intern. Sympos. Electr. Insul. (IEEE Cat. No. 06CH37794), Toronto, Ont., Canada, pp. 380-383, 2006. W. Pei, M. R. Raghuveer, W. McDermid, and J. C. Bromley, "A digital technique for the on-line measurement of dissipation factor and capacitance", IEEE Trans. Dielectr. Electr. Insul., Vol. 8, pp. 228-32, 2001. L. Sang Bin, K. Younsi, and G. B. Kliman, "An online technique for monitoring the insulation condition of AC machine stator windings", IEEE Trans. Energy Conversion, Vol. 20, pp. 737-45, 2005. M. Rosvall and M. Sharp, "A complete system for electrochemical impedance spectroscopy which combines FFT methods and staircase voltammetry", Electrochemistry Communications, Vol. 2, pp. 338-343, 2000. D. J. Gavaghan, D. Elton, K. B. Oldham, and A. M. Bond, "Analysis of ramped square-wave voltammetry in the frequency domain", J. Electroanalytical Chemistry, Vol. 512, pp. 1-15, 2001. B. Sonerud, T. Bengtsson, J. Blennow, and S. M. Gubanski, "Dielectric response measurements utilizing non-sinusoidal waveforms", IEEE Conf. Electr. Insul. Dielectr. Phenomena (CEIDP 2006), Kansas City, Missouri, USA, pp. 43-46, 2006. B. Sonerud, T. Bengtsson, J. Blennow, and S. M. Gubanski, "High and low voltage dielectric response measurements utilizing arbitrarily shaped waveforms", Intern. Sympos. High Voltage Engineering (ISH 2007), Ljubljana, Slovenia, 2007. B. Sonerud, T. Bengtsson, J. Blennow, and S. M. Gubanski, "Utilizing High Voltage Containing High Frequency Components for Continuous Dielectric Response Measurements during Aging", Nordic Insulation Symposium (NORD-IS 07), Lyngby, Denmark, pp. 73-76, 2007. F. J. Harris, "On the use of windows for harmonic analysis with the discrete Fourier transform", IEEE Proc., Vol. 66, pp. 51-83, 1978.

[15] [16]

[17] [18]

J. Hedberg and T. Bengtsson, "Straight Dielectric Response Measurements with High Precision", Nordic Insulation Symposium (NORD-IS 05), Trondheim, Norway, pp. 174-177, 2005. G. Andria, M. Savino, and A. Trotta, "Windows and interpolation algorithms to improve electrical measurement accuracy", IEEE Trans. Instrumentation and Measurement, Vol. 38, pp. 856-863, 1989. V. K. Jain, W. L. Collins, and D. C. Davis, "High-Accuracy Analog Measurements via Interpolated FFT", IEEE Trans. Instrumentation and Measurement, Vol. 28, pp. 113-122, 1979. A. V. Oppenheim, A. S. Willsky, and S. H. Nawab, Signals & Systems, 2 ed. Upper Saddle River: Prentice Hall, 1997. Björn Sonerud was born in Alvesta, Sweden in 1979. He received the M.Sc. degree from Chalmers University of Technology, Gothenburg, Sweden in 2004. He is currently pursuing the Ph.D. degree at Chalmers University of Technology. His research interests include dielectric response measurement techniques, material characterization and aging.

Tord Bengtsson (M’06) was born in Sweden in 1954. He received the Ph.D. degree in mathematical physics from Lund Institute of Technology in 1984 and continued working with theoretical nuclear physics until 1991, when he joined ABB Corporate Research in Västerås, Sweden. Since then, he has been actively engaged in developing diagnostic methods for high voltage apparatuses and systems. Some particular interests are acoustic methods, partial discharges and signal processing. In 2006, he was appointed as a part-time professor in High Voltage Engineering at Chalmers University of Technology in recognition of his efforts to develop research on insulation properties under voltages with fast rise times. Jörgen Blennow (S'94-M'01) was born in Veberöd, Sweden in 1966. He graduated in electrical engineering from Chalmers University of Technology, Gothenburg, Sweden in 1993. In 1996 and 2000 he obtained the degrees of Licentiate of Engineering and Doctor of Philosophy, respectively, in high voltage engineering. In 2001 he became Assistant Professor and since 2006 he holds a position as Senior Lecturer in High Voltage Engineering at the same university. His main research activities have been towards different aspects of insulation exposed to high frequency high voltage, air-solid insulation systems and transformer diagnostics based on dielectric spectroscopy. Stanislaw M. Gubanski (M'89-SM'90-F'01) received the M.Sc. (high voltage engineering) and Ph.D. degrees (material science) from the Technical University of Wroclaw, Poland, in 1973 and 1976, respectively. He was a Research Fellow at the University College of North Wales Bangor, U.K from 1967 to 1977, a senior lecturer at the technical university of Wroclaw, Wroclaw, Poland, from 1977 to 1988. Afterwards he was associate professor at the Royal Institute of Technology, Stockholm, Sweden. Currently, he is Professor in High Voltage Engineering at the Department of Manufacturing and Materials Technology, Chalmers University of Technology, Gothenburg and research leader of High Voltage Valley in Ludvika, Sweden. He is Chair of IEEE-DEIS Nominations Committee. He is also convener of the CIGRE Task Force D1.01.14 “Dielectric Response Methods for Diagnostics of Power Transformers”.