Observations of Acoustic Emission in Power Semiconductors

(1)

Tommi Kärkkäinen

OBSERVATIONS OF ACOUSTIC EMISSION IN POWER SEMICONDUCTORS

Acta Universitatis Lappeenrantaensis 686

Thesis for the degree of Doctor of Science (Technology) to be presented with due permission for public examination and criticism in the Auditorium 1382 at Lappeenranta University of Technology, Lappeenranta, Finland, on the 19^th of December, 2015, at 1 pm.

(2)

Lappeenranta University of Technology Finland

Reviewers Professor Hans-Günter Eckel

Fakultät für Informatik und Elektrotechnik Institut für Elektrische Energietechnik Universität Rostock

Germany

D.Sc. (tech) Panu Maijala

Machinery and Environmental Acoustics VTT

Finland

Opponent D.Sc. (tech) Veli-Matti Leppänen ABB Oy

Finland

ISBN 978-952-265-911-8 ISBN978-952-265-912-5(PDF)

ISSN-L 1456-4491 ISSN 1456-4491

Lappeenrannan teknillinen yliopisto Yliopistopaino 2015

(3)

Abstract

Tommi Kärkkäinen

Observations of Acoustic Emission in Power Semiconductors

Acta Universitatis Lappeenrantaensis 686

Dissertation, Lappeenranta University of Technology 41 p.

Lappeenranta 2015

ISBN 978-952-265-911-8, ISBN 978-952-265-912-5 (PDF) ISSN-L 1456-4491, ISSN 1456-4491

Industrial, electrical power generation, and transportation systems, to name but a few, rely heavily on power electronics to control and convert electrical power. Each of these systems, when encountering an unexpected failure, can cause significant financial losses, or even an emergency. A condition monitoring system would help to alleviate these concerns, but for the time being, there is no generally accepted and widely adopted method for power electronics.

Acoustic emission is used as a failure precursor in many applications, but it has not been studied in power electronics so far.

In this doctoral dissertation, observations of acoustic emission in power semiconductor components are presented. The acoustic emissions are caused by the switching operation and failure of power transistors. Three types of acoustic emission are observed. Furthermore, aspects related to the measurement and detection of acoustic phenomena are discussed. These include sensor performance and mechanical construction of experimental setups.

The results presented in this dissertation are the outset of a research program where it will be determined whether an acoustic-emission-based condition monitoring method can be developed.

Keywords: acoustic emission, power electronics, power semiconductor modules, condition monitoring, reliability

(4)

(5)

Acknowledgments

The research presented in this doctoral dissertation was carried out in the Laboratory of Ap- plied Electronics, LUT School of Energy Systems, Lappeenranta University of Technology between the years 2011 and 2015. The work has been funded by the ABB Research Grant Program and the Academy of Finland.

I would like to thank my supervisor, Professor Pertti Silventoinen for guiding me through the scientific questions and bureaucratic hurdles. I am grateful to the reviewers of this thesis, Professor Hans-Günter Eckel and Dr. Panu Maijala for the valuable time they spent reading and commenting my work, and to my opponent, Dr. Veli-Matti Leppänen for participating in this process.

I would also like to express my gratitude for Dr. Anu Kärkkäinen and Dr. Henry Rimminen from VTT Technical Research Centre of Finland. You pointed us in the direction of acoustic emission in power electronics. Without you, this work would not have come to be. Dr. Elena Mengotti and Dr. Gernot Riedel from ABB Corporate Research Switzerland, you have been keeping up with and contributing to my research from the start. Thank you for that. It has been a pleasure and a privilege to be working with you.

I am very thankful for the support, ideas and aid I have received from my coworkers at LUT, especially Dr. Mikko Kuisma and Dr. Joonas Talvitie. You have donated a lot of your time to solve problems and write papers with me. This work benefited much of your effort. Working with you has been indescribable, in all positive senses of the word.

To all of you from room 6405... I am at a loss for words. What we have in our very special one-office community is amazing. Together we are able to solve any problem inside and outside our field. No matter who is assigned to work at our office, we contaminate them with our twisted sense of humour. Dr Juha-Pekka Ström, Dr. Juho Tyster, Dr. Mikko Qvintus and Dr. Arto Sankala, you have already left to pursue careers outside the university but I still remember what a pleasure it was to work with you. Dr. Juhamatti Korhonen, Mr. Janne Hannonen, Mr. Jani Hiltunen, Mr. Jari Honkanen, Mr. Heikki Järvisalo and Mr. Saku Levikari, you are the current orchestra of coffee break humour. Will we let the world run out of bad jokes?Emme!

I am very grateful to Dr Hanna Niemelä. Your wordsmithery was absolutely invaluable to the

(6)

I would like to thank Walter Ahlström Foundation, the Finnish Foundation for Technology Promotion and Ulla Tuominen Foundation for the personal grants I have received. The true value of these grants is far greater than the combined monetary value. They showed me that people other than working researcher see importance in what I do.

Finally, I want to thank all the people near me, family and friends. It is your support that allowed me to go where I am and become who I am. Mom and Dad, thank you for that.

Ninna and Miika, thank you for being such a great sister and brother. Jussi, thank you for being in my life.

Lappeenranta, December 4^th, 2015

Tommi Kärkkäinen

(7)

List of Symbols and Abbreviations

Roman letters

N number of items

t time

td detection time

tp peak time

ts duration of stoppage u signal, voltage uth threshold voltage Acronyms

DC Direct current DUT Device under test

IGBT Insulated Gate Bipolar Transistor

(10)

(11)

11

Chapter 1

Introduction

Power electronic converters have penetrated into society; low-power converters can be found for instance in computer power supplies, mobile devices, and home lighting applications.

Thus, an average consumer probably uses dozens of power converters daily without being aware of their existence. Higher-power converters can be found in many industrial applications including industrial drives, renewable power generation, and, in the future, also on the power grid. An unexpected fault causes an interruption to the operation of a plant, and in many cases there is a risk of a hazardous situation or an injury. Considerable financial losses may also occur: in addition to the immediate cost of repairs, an interruption to production causes a loss of revenue. According to Eade (1998), the cost of a prolonged unscheduled interruption “could well be in the six-figure bracket.” A stoppage is especially costly in offshore wind turbines, as they are difficult to reach and service (Kärkkäinen and Silventoinen, 2013).

Condition monitoring would significantly mitigate the situation by allowing operators to schedule preventative service of the converters. However, there is currently no widely recog- nized early warning method available for faults.

Acoustic emission has been used as a condition monitoring tool in a variety of applications, the best known ones probably being detection of bearing failures and monitoring of rotational machines (Navarro et al., 2010; Li and He, 2012). Other examples include condition monitoring of pipelines, concrete structures, steel pipes and tools, and monitoring of corrosion (Kaewkongka and Lim, 2007; Kaewkongka et al., 2008; Lamonaca et al., 2012; Lim and Kaewkongka, 2007; Martins et al., 2014).

For electronics, some work has been done on capacitors where acoustic emission has been associated with partial discharge (Bua-Nunez et al., 2014; Smulko et al., 2011). In semiconductor components, acoustic emission has been studied in light-emitting diodes. It has been shown that structural changes in the semiconductor material that are responsible for changes in the spectrum of the emitted light also cause an acoustic emission to occur (Veleshchuk

(12)

et al., 2004; Vlasenko et al., 2008).

In power semiconductor modules, an experiment has been conducted regarding partial discharges occurring in a power module (Berth, 1998). In the paper Berth notes that the acoustic signals in the power module are very small, and that the plastic parts power modules cause significant damping.

As of now, acoustic emission is a generally unknown phenomenon in power semiconductor modules. Indeed, there is no mention of switching- or failure-related acoustic emission in power semiconductor modules in the previous literature aside from the work on partial discharges. Thus, showing that acoustic emission occurs in power semiconductor modules is the main result of this work. The work was conducted within a research project funded by the Academy of Finland and the ABB Research Grant Program.

1.1 Reliability of power electronics

As stated above, power electronics has many applications including consumer electronics, industrial plants, traffic, and the production and transmission of electrical energy. In many of these cases, condition monitoring would allow the plant operator to manage reliability issues and enhance the availability of the plant. Because of statistics readily available, wind energy is used as an example in this section to study the state of reliability in power electronics.

A fault in power electronics was the cause for combined 6219 hours of stoppage in Finnish wind power plants in 2010, constituting 13.1 % of the total stoppage time (Stenberg and Holt- tinen, 2011). In a Swedish survey (Ribrant and Bertling, 2007) the electric system, excluding the generator, was the cause of failure in 14.3 % of cases. Power semiconductor devices, power transistor modules in particular, cause a significant proportion of faults. Statistically, a power semiconductor failure has been reported as the cause of a converter failure in 19 % of power converter faults (Zhu et al., 2011).

A condition monitoring method benefits the user of a system by reducing the downtime caused by a failure (Kärkkäinen and Silventoinen, 2013). Power electronics is a good candi- date for condition monitoring in wind power systems, as the failures related to it tend to cause a relatively long downtime in wind power systems, and faults in power electronics represent the most frequent failure type (Table 1.1).

In solar power systems, plant owners demand 20 years of working life of power electronic converters. Modern power converters are capable of achieving such a life span, and the future development in power semiconductors and packaging will probably extend the lifetime of the converters even further. Even in that case, a condition monitoring system would improve the user experience near the end of the converter life.

1Includes the blades and the blade pitch mechanism

2Reported as the electric system, contains other components beside the power electronics, excludes the generator

(13)

1.2 Condition monitoring in power electronics 13

Table 1.1. Average number of failures per year per system (N), and the average length of a stoppage (hours) caused by a failure per failure type (t_s). Based on data from Stenberg and Holttinen (2011);

Turkia and Holttinen (2013); Ribrant and Bertling (2007)

Finland Sweden

System 2010 2011 2000–2004

N ts N ts N ts

Blade pitch mechanism 65 171 60 141 32.2 91.6¹ Power electronics 48 130 40 89 42.0 106.6²

Gearbox 34 331 22 316 23.6 256.7

Hydraulics 31 164 13 199 32.0 43.2

1.2 Condition monitoring in power electronics

At present, there is no widely accepted condition monitoring method for power electronics.

Most of the proposed methods rely on measuring the collector-emitter voltage of the power transistor (Lehmann et al., 2003). This method is able to detect imminent faults resulting from bond-wire lift-off but is unable to detect failure caused by other mechanisms. Furthermore, the voltage-based condition estimator requires a voltage measurement to be performed on each transistor separately.

In addition to the measurement of the collector-emitter voltage, turn-off time has been sug- gested as a failure indicator (Brown et al., 2012). The gate circuit has also been proposed as a source of condition monitoring information (Zhou et al., 2013). However, both methods require very fast voltage measurements to be performed on each transistor.

1.3 Definition of acoustic emission

Acoustic emission is an established term in the condition monitoring community. It is the experience of the present author that many working professionals in the field of electronics and power electronics initially get a wrong idea of the nature of the phenomena. To prevent confusion, the most important terms are defined here.

In this doctoral dissertation,acoustic emissionis understood as the emission of vibration in a system. The vibration is caused within the system or device being tested, not from an outside source as in acoustic microscopy. It is a common misconception that, given the term ’acoustic emission,’ those vibrations would be at audio frequencies, that is, within the human hearing range. This may be the case, but the present work also considers frequencies below and above the hearing range to be acoustic. In fact, the observations presented extend to hundreds of kilohertz.

(14)

Another common misconception is that the studied acoustic waves would propagate from the device under test to air, where they can be detected by an ear or a microphone. While this would certainly count as acoustic emission, it is not a requirement. In fact, all observations presented in this doctoral dissertation are made by observing the vibrations directly on a solid surface, using an acoustic emission sensor.

1.4 Objectives and research questions

The main objective of this work was to investigate whether acoustic emission exists as a phenomenon associated with power semiconductor modules. This work can then be built up in the future, perhaps ultimately yielding a viable condition monitoring method.

In this doctoral dissertation, the main research question can be formulated asDoes acoustic emission exist in power semiconductor modules? This question is answered for switching- and failure-related acoustic emissions. As a result, three types of acoustic emissions are identified.

How can acoustic emissions be reliably produced and monitoredwas a further key research question. Because of the unknown nature of the phenomenon in the power electronics context, this was not self-evident. This question was answered by developing experimental setups that allowed producing the scientific results of this doctoral dissertation.

1.5 Scientific contributions of the doctoral dissertation

• Discovery of acoustic emission as a phenomenon associated with the operation and failure of power semiconductors

• Development of experimental setups for the investigation of acoustic emissions in power electronics

• Evaluation of the applicability of presently available acoustic emission sensors for investigating power semiconductors

1.6 List of publications

Publication I

Kärkkäinen, T.J., Talvitie, J.P., Kuisma, M., Hannonen, J., Ström, J.-P., Mengotti, E., and Silventoinen, P. (2014), "Acoustic Emission in Power Semiconductor Modules—First Obser- vations,"IEEE Transactions on Power Electronics, vol. 29, no.11, pp. 6081–6086.

(15)

1.6 List of publications 15

In Publication I, it is shown for the first time that acoustic emission is a phenomenon in power semiconductor modules that results from the switching operation of a power module.

The author of this doctoral dissertation is the principal author of the article, and designed and constructed the experimental setup used for producing the results. The analysis of the results was made primarily by the author, in collaboration with the coauthors.

Publication II

Kärkkäinen, T.J., Talvitie, J.P., Ikonen, O., Kuisma, M., Silventoinen, P., and Mengotti, E.

(2014), "Sounds from semiconductors – Acoustic emission experiment with a power module," inProceedings of the 16th European Conference on Power Electronics and Applications (EPE’14-ECCE Europe).

In Publication II, the results of Publication II are confirmed by repeating the experiment using different instrumentation. The same conclusion is drawn: the switching operation of a power module causes acoustic emission to occur.

The present author is the principal author of the paper, and was responsible for conducting the experiment.

Publication III

Kärkkäinen, T.J., Talvitie, J.P., Kuisma, M., Silventoinen, P., and Mengotti, E. (2015),

"Acoustic Emission Caused by the Failure of a Power Transistor," inProceedings of Applied Power Electronics Conference and Exposition (APEC).

In Publication III, observations of acoustic emission caused by the failure of IGBTs are made.

Two distinct types of acoustic emission are detected.

The present author is the principal author of the paper, and designed and constructed the experimental setup used for obtaining the results.

Publication IV

Kärkkäinen, T.J., Talvitie, J.P., Kuisma, M., Silventoinen, P., and Mengotti, E. (2015), "Mea- surement Challenges in Acoustic Emission Research of Semiconductors," inProceedings of the 17th European Conference on Power Electronics and Applications (EPE’15).

In Publication IV, measurements and instrumentation applied to the experiments are discussed. Challenges and solutions involved in the measurement and sensor limitations are presented.

The present author is the principal author of the paper. The paper is based on measurements performed for previous publications. The analysis was made by the author in collaboration with the coauthors.

(16)

(17)

17

Chapter 2

Research methods

The research work was mainly carried out by conducting experiments on power transistors.

Every experiment involved a device under test (DUT) – an IGBT module or a single IGBT – together with other necessary test circuitry, and the instrumentation required for recording the acoustic phenomena being studied, as well as necessary electrical quantities. The circuitry was used to drive the DUT to an operating or failure condition where acoustic emission was expected to occur.

There are also other methods of research: Multi-domain simulation tools capable of analyz- ing electromagnetic and acoustic phenomena simultaneously are available. Because of the absence of prior work on acoustic emission of power semiconductors, it would have been difficult to evaluate the relevance and validity of the simulation results, as there would have been no reference data to compare the simulation results with. For this reason, experiments were considered a better way of investigating these phenomena.

In the course of the research, the measurement of the acoustic phenomena required special attention. For example, attention had to be paid to ensure that the measured acoustic waveforms actually were acoustic and not, for instance, electromagnetic interference. It was also necessary to maintain that the detected acoustic phenomena took place within the power module being investigated and were not of some other origin. A considerable amount of time was spent on ensuring that the data actually represent what they are thought to represent.

2.1 Experimental setups

Much of the work on the experiments was done on designing and constructing experimental setups. This section highlights the questions and problems encountered during this phase of the work.

(18)

2.1.1 Sensors

A wide variety of sensors with diverse characteristics are available in the market. The sensitivity, bandwidth, mechanics, thermal limits, and behavior in a noisy environment were regarded as the most important properties of a sensor for this work. For a practical implemen- tation in a customer installation, other properties such as affordability and durability would be of great interest. In a laboratory setting, however, these matters do not play as significant a role.

For this research, a wide-band sensor was desired. It was anticipated to give the best view into the acoustic phenomena. While it is not uncommon for narrow-band resonance type sensors to be used in acoustic emission projects, it was assumed that such a sensor might not reveal as many interesting characteristics as a wide-band sensor could. It is not immediately obvious which frequencies are found or will be emphasized in the acoustic data. Therefore, selecting a narrow-band sensor with a suitable center frequency would be problematic.

It was also difficult to estimate how sensitive the sensor should be. There was no simple means of estimating the amplitude of the acoustic signal in terms of pressure, displacement, or displacement velocity. There was, however, a simple method to find out whether a particular sensor is sensitive enough for an experiment: to acquire such a sensor and test it.

Two wide-band sensors were selected: Kistler Piezotron and KRN Services KRNBB-PC (Figure 2.1). Both sensors are wide-band, with frequency ranges of 50 to 400 kHz and 100 to 100 kHz, respectively. The KRN sensor is, however, sensitive to much lower frequencies in the 10 kHz range; the datasheet specification is limited to 100 kHz because of the manu- facturer’s test rig limitations. Both sensors are limited to 60^◦C; this restriction required the test in Publication III to be done in an unconventional manner: the temperature limitation of the sensor was circumvented by operating the DUT in a way that does not resemble realistic operating conditions in terms of current, power, and temperature.

Sensors with a high temperature tolerance have been developed, for instance by Noma et al.

(2007). Such sensors were not, however, commercially available at the time of the work described in this dissertation. Since the experiments have been carried out, sensors with temperature ranges reaching 160^◦C have become commercially available and accessible to the author.

Figure 2.1. Examples of wide-band acoustic emission sensors. Two KRN Services KRNBB-PC sensors on the left, and a Kistler Piezotron sensor on the right.

(19)

2.1 Experimental setups 19

Figure 2.2. Examples of methods to attach acoustic emission sensors. The Kistler sensor (left) is held in place by a weight, and the KRN sensor is screwed into a nut that has been glued to the surface. The setup shown here is not used in any of the publications of this doctoral dissertation, but was used in the initial tests where the performance of the sensors was evaluated.

Sensor attachment is an issue that has to be solved in each experimental setup (Figure 2.2).

The Kistler sensor, for example, is designed to be attached with a screw, which requires a threaded hole to be made in the system to be measured. Screwing such a hole for instance into the cooling surface of a power module is not feasible. Instead, the sensor was held in place manually, or by placing a weight on top of it. The KRN sensor, on the other hand, is threaded and intended to be screwed into the system being tested. For some of the experiments, a nut was glued onto the surface where the tests were being made while for some, a larger rig was constructed around the system under test.

The use of a laser doppler vibrometer was also tested. Unfortunately, it was incapable of detecting any signal from the experimental setup. One possible explanation for this is that the surface displacement in the power module is too small for the laser vibrometer to detect.

2.1.2 Mechanical and electrical design

The mechanical design of the experimental setup is crucial for the success of the experiments.

The mechanical construction must allow the acoustic waves to propagate from the intended source to the sensor while also preventing the propagation of other acoustic signals to the sensor.

(20)

The mechanical design is also part of the electromagnetic design of the experimental setup.

The sensors typically produce a relatively low voltage, in the range of tens of millivolts peak- to-peak, which is later amplified to be actually measured. Such signals are susceptible to electromagnetic interference.

For these reasons, the experiments conducted in this work did not use the typical mechanical construction of a power converter where the power semiconductor modules are attached to a heat sink and surrounded by an enclosure. Such an arrangement provides an acceptable level of electrical safety, cooling, and immunity to electromagnetic interference. Constructions of that kind do not, however, usually provide much extra space where an acoustic emission sensor could be installed. Even if room for the sensor could be found, it might be difficult to find multiple measurement locations to gain a better insight into the phenomenon.

Another reason against the typical power converter design was the fact that a purpose-built experimental setup allows better control of the acoustic environment. The number of components affecting the surfaces where the sensors are attached can be reduced, and the acoustic coupling between the parts of the system can be controlled more easily.

Because the mechanical setups used in this work deviate considerably from power converter products, they do not necessarily exhibit the electrical safety built into commercial products.

A choice had to be made. On one hand, one could build an experiment that both maximizes the likelihood of producing and detecting an acoustic emission, and provides electrical safety so that realistic voltages (600 V DC for the modules in this work) can be used. On the other hand, one could use a simplified mechanical construction and reduce the voltage to a safer level. The latter option was chosen mainly in order to expedite the construction of the experimental setups and the delivery of results.

Another instance where a choice had to be made between realistic operating conditions and the ease of experiments is prominent in Publication III. In this case, the main reason for choosing the less realistic option was protecting the sensor from the heat present in the measurement system. Operating the power transistor under realistic conditions would have produced such a large amount of heat that the sensor would have been in danger of destruction.

Here, unrealistic conditions mean that the measured waveforms cannot be expected to resemble those occurring in a power converter.

2.1.3 Experiment on switching-related acoustic emissions

To study whether acoustic emission occurs in connection with the switching operation of a power module, a power module was made to switch and the related acoustic emissions were monitored. A detailed description of the switching circuit is presented in Publication I.

The switching circuit excluding the power module under test is located inside an electronics enclosure (Figure 2.3). The DUT is attached to the lid of the enclosure so that acoustic emissions originating from it can propagate to the lid.

(21)

2.1 Experimental setups 21

Figure 2.3. Experimental setup used in Publications I and II. The device under test is attached to the lid of the electronics enclosure, and the sensor is attached to the outer surface of the lid. The rest of the electronics are located inside the enclosure. From Publication I.

The sensor is attached to the lid, and measurements are made at multiple locations. The propagation delay from the switching instant to the detection of the pulse is measured and compared against the distance of the measurement location from the DUT.

To rule out the possibility of other circuit components affecting the measurement, a test was made where a piece of packaging foam was placed between the lid and the wall of the enclosure. The foam alters the acoustic path between the lid and the rest of the enclosure. If acoustic sources other than the DUT were present in the detected signal, the measurement with the packaging foam should produce a waveform that deviates from the one obtained without it.

However, both the measurements with and without the packaging foam were virtually identical, suggesting that the other components do not emit interfering acoustic emissions to the lid.

2.1.4 Experiment on failure-related acoustic emissions

In order to record acoustic emissions associated with the failure of power transistors, another experimental setup was constructed (Figure 2.4). In this setup, the DUT, a TO220 packaged IGBT, is attached to a piece of sheet metal. The metal allows the acoustic signal to propagate to the sensor while it also absorbs some of the heat generated during the experiment.

The DUT is fed from a constant current supply, in the active region. This causes a substantial amount of heat to be dissipated in the DUT; the thermal power of 100 W causes the failure

(22)

Figure 2.4. Experimental setup used in Publication III. The device under test is attached to a piece of sheet metal, which also conducts the acoutsic signals from the DUT to the acoustic emission sensor. An external power supply is used to heat the DUT, causing it to fail. From Publication III.

of the DUT in about three seconds. Various failure modes took place, including gate-emitter and collector-emitter short-circuits.

2.2 Methods of analysis

Once the experimental setup has been constructed, tested, and functionally verified, it then becomes necessary to perform analyses and calculations on the data to get answers to the questions originally posed. In this work, a typical research question was “Does the event in question cause an acoustic emission under the current circumstances?” Answering the question based on acoustic measurements requires an analysis method that is able to:

• detect acoustic events in acoustic emission sensor data,

• rule out other acoustic sources, and

• rule out signals caused by electromagnetic interference.

Sophisticated wavelet-based methods for the detection of waves have been studied in other applications such as the study of seismology (Jamshidi and Shaabany, 2011; Colak et al., 2009; Suprijanto et al., 2013; Hafez et al., 2010). Other methods that have been applied include empirical mode decomposition (Li and He, 2012), fuzzy inference (Navarro et al., 2010), a coherence- and intensity-based method (Moschioni et al., 2004), a peak analysis (Kaewkongka et al., 2008), and a neural-network-based method (Martins et al., 2014). Most of these methods have been designed for situations where an acoustic emission from a specific

(23)

2.2 Methods of analysis 23

source or caused by a specific event has to be detected from a signal where other acoustic emissions also exist. Some of them, especially the neural-network-based method (Martins et al., 2014), also require prior knowledge of the nature of the phenomenon being observed.

The work presented in this dissertation is done on experimental setups that are designed in a way that acoustically isolates the device under test from the rest of the system. This means that the capability of isolating different sources and mechanisms is not needed in the post- processing of the signals. Therefore, the above-discussed methods do not add value to the analysis in this work.

For this reason, simple and readily applicable methods were proposed in each of the publications. For example, in Publications I and II, acoustic measurement is performed in multiple randomly selected locations around the DUT. The idea is to measure how long it takes for the signal to propagate from the DUT to the measurement location in each case. An analysis method capable of detecting the arrival of an acoustic wave is required. A simple method based on detecting the first peak in the signal was devised: for an acoustic signalu(t)and its derivative function ˙u(t), for each peak in the signal

˙

u(tp) =0 (2.1)

wheretpis the time when the signal is at its peak. It is possible and likely that the signal, containing noise, has peaks before the intended first peak of the acoustic event. For this reason, the analysis should also include a threshold leveluth. The detected time of an acoustic wave can be expressed as the lowestt_dthat satisfies

u(t˙ d) =0

u(td)≥uth . (2.2)

The added threshold allows the method to disregard noise present in the system and low- amplitude interference that couples from the test system. To further reduce interference, the signal was low pass filtered prior to the analysis, as most of the interference was seen to be found at frequencies higher than 1 MHz.

This method was considered to be simple, computationally efficient, and accurate enough for the requirements of this work. Typically, the detection error is one or two signal periods, while the observed differences between measurement locations are much greater.

(24)

(25)

25

Chapter 3

Results

The research work produced several results. First, it was concluded that acoustic emission is a real phenomenon in the context of power semiconductor modules. This result is further validated in Publication II. Acoustic emissions associated with the switching operation of power IGBTs were investigated. In Publication III, observations of acoustic emission related to the failure of an IGBT are presented. Publication III also suggests that different types of acoustic emission can be identified from the acoustic measurements.

In the course of the work, it was discovered that the sensors themselves have an effect on the measurements, and that the limitations of the sensors affect the approach taken to the design and execution of experiments. These aspects are discussed in Publication IV.

3.1 Existence of acoustic emission

To study acoustic emissions associated with switching, a half bridge module was made to switch. The related acoustic events were monitored with an acoustic emission sensor. The experiment was repeated thirty times, placing the sensor each time in a different location.

By plotting the time between the turn-off of the IGBT and the detection of an acoustic event (Figure 3.1), one can see that the further away the sensor was, the longer the time delay was.

This behavior is expected when the power module is the source of the acoustic emission.

The result supports the theory that a switching operation of power transistors causes acoustic emission to occur.

Originally, the experiment was conducted using the Kistler sensor. These findings are presented in Publication I. To enhance the reliability of the conclusion, the experiment was repeated with the KRN sensor. These results are reported in Publication II. Both experiments yielded the same conclusions, although there were some differences in the measurements.

(26)

0 50 100 150 200 0

20 40 60 80 100 120 140 160

x [mm]

y [mm]

30

30 35 35

35

35 40

40

40 45

45

45 50

50

50 50

50

50 60

60 60

70 Lid outline Module outline Semiconductor chip Mounting hole

(a) Propagation delays measured in Publication I

40

60 40 60

60 60

80

80 120

120

120 120

120

160

x (mm)

y (mm)

0 50 100 150 200

0 20 40 60 80 100 120 140 160

Lid outline Module outline Semiconductor chip Mounting hole

(b) Propagation delays measured in Publication II

Figure 3.1. Results of an experiment that showed the existence of acoustic emissions in power semiconductor modules. The contours show how many microseconds it takes for the signal to reach the sensor when measurements are performed in different locations. The measurement locations are indicated by x.

(27)

3.1 Existence of acoustic emission 27

−0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

−0.2

−0.1 0 0.1 0.2

Time (ms)

Sensor output (V)

−0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

−0.5 0 0.5

Sensor output (V)

−0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0 0.51

Figure 3.2. Acoustic event associated with IGBT switching, observed with two different sensors. From top to bottom: the pulse fed to the IGBT, output from the KRN sensor, output form the Kistler sensor.

Perhaps the most important source of these differences can be found by studying the actual waveform captured by each sensor (3.2). It is evident that the KRN sensor captures a strong signal at a much lower frequency than the Kistler sensor. This causes the time resolution of the detection method to be much lower for the KRN sensor than it is for the Kistler sensor.

As such, the contours in Figure 3.1(b) are much coarser than in figure (a).

The time delays shown in Figure 3.1 are longer than those predicted by a calculation based on nominal propagation velocities in aluminum. This can be explained by examining the accuracy of the wave detection method (Figure 5 in Publication I). The detection may take place some periods after the actual arrival of the acoustic wave. These periods correspond to the difference between the calculated and detected delays.

There is also a considerably large difference between the delays presented in Publications I and II. Both measurements use the same wave detection method and are, therefore, subject to similar errors. Because the KRN sensor has a strong low-frequency component, the error caused by a few signal periods is longer than with the Kistler sensor.

Another possible reason for the different time delays is that the sensors may be sensitive to waves with different modes of propagation. For aluminum, the velocity of longitudinal waves is more than double that of the shear wave. The analysis cannot, however, be based solely on the properties of aluminum. Although the enclosure in question is nominally cast aluminum, it contains zinc and is not homogeneous.

In Publication III, acoustic emissions produced by the failure of an IGBT were monitored.

The failure of the TO-220 packaged DUTs was caused by subjecting them to an abnormally

(28)

−10.5 −10 −9.5 −9 −8.5

−0.2 0 0.2

Time (ms)

Output (V)

−7.5 −7 −6.5 −6 −5.5 −5

−1 0 1

Time (ms)

Output (V)

−2.5 −2 −1.5 −1 −0.5

−2 0 2

Time (ms)

Output (V)

(a)

1.5 1.6 1.7 1.8 1.9

−0.5 0 0.5

Time (ms)

Output (V)

58.3 58.4 58.5 58.6 58.7 58.8

−1 0 1

Time (ms)

Output (V)

70.9 71 71.1 71.2 71.3 71.4

−0.5 0 0.5

Time (ms)

Output (V)

(b)

Figure 3.3. Examples of the two observed acoustic emission types associated with the failure of transistors. Immediate emission (a) occurs at the same time as the failure, and post-failure emission (b) typically occurs tens of milliseconds later.

Table 3.1. Number of components where each type of acoustic emission could be observed and the related failure modes; from Publication III.

Failure mode Immediate Post Both None

Gate-Emitter short 1 3

Collector-Emitter 2 7 9 2

Neither 2

high stress of 100 W, which caused a failure in a few seconds. Two different types of acoustic emission were observed: animmediate emission that occurred at the time of the failure, and apost-failure emissionthat occurred tens of milliseconds after the failure (Figure 3.3).

All components tested did not exhibit both emission types (Table 3.1). Some did, others exhibited only one type, and a few components did not emit any acoustic signal at all. It was also observed that there could be multiple post-failure emissions, but only exactly one or zero immediate emissions.

3.2 Acoustic emission types

As discussed above, acoustic emissions related to the switching operation and failure of IG- BTs were observed, and three types of acoustic emission were identified. Comparing the three observed acoustic emission types (Fig. 3.4), it is evident that they deviate considerably from one another. One can see, for example, that the events are of different durations; the

(29)

3.2 Acoustic emission types 29

Figure 3.4. Acoustic emission types associated with power semiconductors. From top to bottom: spectrograms and time domain plots of a switching-related emission, immediate emission, and post-failure emission. The time scales of the spectrograms match those of the time domain plots. Each observation in the figure is measured using the KRN sensor. The post-failure and immediate emissions are from a TO-220 packaged transistor; the switching-related emission from a half bridge module.

(30)

failure-related emissions are significantly shorter than the switching-related acoustic emission. The post-failure emission, in particular, appears to consist of a series of very short

“mini-emissions.” The immediate emission, on the other hand, behaves more like a single decaying oscillation.

The switching-related emission is also a decaying oscillation, but its duration is significantly longer than that of the immediate emission. It is, however, risky to generalize this to be true, as the switching-related emission and failure-related emissions were measured from IGBT modules of different sizes and with different experimental setups.

In terms of waveform and frequency content, the immediate-emission- and switching-related emission have much in common. Both appear to be dominated by a single frequency in the order of 10 kHz. The switching-related emission also seems to contain some other frequencies. The post-failure emission, on the other hand, consists of nonperiodically recurring impulse-like signals.

The spectrograms of the emissions also support these conclusions. The majority of the signal energy in the switching-related and immediate emissions is in the low sub-200 kHz band.

The duration of the immediate emission is shorter than that of the switching-related one. The dissimilarity of the post-failure acoustic emission from the other emission types is also very evident in the spectrograms: the signal is a series of short, wide-band events rather than a single emission.

A reference measurement of the switching-related acoustic was performed at a DC link voltage of 600 V. The purpose of this experiment was to determine whether the phenomenon is different when it is excited by switching occurring at a realistic operating voltage. Comparing the waveforms (Figure 3.5) three differences can be clearly identified. First, the signal amplitude is not noticeably higher in the 600-volt measurement than the 30-volt measurement.

This suggests that the phenomenon is more dependent on current than voltage. Second, the acoustic event is shorter. The reason for this is unknown at this point. Third, the voltage peak caused by the switching of the power module is significantly higher. This is explained by the fact that the sensor is sensitive to capacitively coupled interference, and the increased voltage exacerbates the interference effects.

It is also evident that the measurement at 600 V contains more noise than the 30-volt measurement. Rather than being white noise, it is concentrated on narrow frequency bands spaced about 100 kHz from each other, starting at about 200 kHz. The noise is present in the output even when the sensor is mechanically disconnected from the experimental setup. The voltage supply is perhaps the most likely source for the noise, although this was not confirmed. For- tunately, the acoustic signal is concentrated at frequencies below than 200 kHz so the acoustic signal is not degraded.

In the spectrograms one can see that the signal in the 600-volt measurement is spread over a wider spectrum of frequencies. In both measurements the low frequencies contain most of the signal power. In the 600-volt measurement the acoustic event starts at about 0.4 ms with frequencies between about 30 and 60 kHz. At about 0.5 ms, the low frequencies become

(31)

3.2 Acoustic emission types 31

(a) Measurement at 30 V

(b) Measurement at 600 V

Figure 3.5. Comparison of switching-related acoustic emissions recorded with a DC link voltage of 30 V and 600 V. The time scales of the spectrograms match those of the time domain plots.

(32)

visible. In the 30 V measurement there is significantly less signal at higher frequencies, and the low-frequency signal appears to start more uniformly. Higher frequencies appear to only be present during the peaks of the signal. In the 600-volt measurement, the high frequencies fade out before the low frequencies.

It is unclear why the signals are so different. As stated above, the increased voltage did not increase the observed signal amplitude, which suggests that the main mechanism causing acoustic emission is more related to currents and magnetic fields than voltages and electric fields. It is possible that another voltage-dependent mechanism is causing other acoustic events which causes the differences described above.

One possible voltage-related mechanism is partial discharges occurring in the silicone gel within the power module. In the opinion of the present author this is an unlikely explanation.

In the literature, partial discharge experiments on power modules are typically carried out at voltages that are considerably higher than the 600 volts used here (Breit et al., 2002; Lebey et al., 2004; Berth, 1998). Another possible voltage-related mechanism is the piezoelectric effect. The dielectric in the power module is often made of aluminum oxide or aluminum nitride. Aluminum nitride exhibits piezoelectricity but it is not known which dielectric material is used in the power modules under test.

While it is tempting to also analyze the differences in the amplitudes of the signals, this would be somewhat questionable. The amplitude in the switching emission experiments has been very consistent and repeatable. The failure-related emissions, however, have displayed a significantly varying amplitude, as is evident from Figure 3.3. Further, the significance of such an analysis is unclear. The switching- and failure-related emissions are measured from different experimental setups and with different electrical loads, meaning that the measurement conditions are not comparable. It is, however, worth noting that each acoustic event is strong enough to be picked up by the sensor, meaning that the choice of sensors was a satisfactory one.

3.3 Measurement challenges and sensor performance

Although the primary objective of the work was to investigate the acoustic phenomena in power semiconductors, also measurement issues had to be addressed. Since no previous work on acoustic emissions in the context of power semiconductor modules could be found, there were no reference measurement methods available, neither were there any purpose-built sensors in the market. As such, the experiments provided information not only on the nature of the acoustic phenomena but on how the acoustic emission sensors meet the requirements of performance.

Publications I and II present the same experiment carried out with different sensors. As described in section 3.1, both experiments support the same conclusion. Publication IV further explores the differences and challenges of the measurements.

(33)

3.3 Measurement challenges and sensor performance 33

By comparing an acoustic emission recorded by using two different sensors (Figure 3.2), one can immediately see that the resulting waveforms deviate considerably from each other. The KRN sensor captures much lower frequencies than the Kistler sensor. The higher frequencies present in the output of the Kistler sensor, however, are absent from the KRN sensor signal.

The frequency range of the KRN sensor covers that of the Kistler sensor entirely, and extends to the low frequencies, and up to a megahertz. Therefore, it would be reasonable to expect the KRN sensor to capture all the frequencies that the Kistler sensor does. One explanation for this might be resonances inside the sensors, the frequencies of which would depend on the construction of the sensors. The resonances may be caused by the mechanical elements in the sensors, or by the electrical parts in them.

Another possible source of resonances is the mechanical construction in which the acoustic emissions are measured. This encompasses the device under test, but also other parts of the experimental setup. None of the experiments presented in this work have the sensor directly connected to the DUT, but a metal object acoustically connecting the sensor to the DUT.

This acoustic environment may produce resonances of its own and also attenuate the acoustic signal to some degree.

There is also a distinct difference between the sensors during the IGBT pulse. The KRN sensor captures what looks like an interrupted sine wave during the pulse whereas the Kistler sensor fails to capture anything at that time.

(34)

(35)

35

Chapter 4

Conclusions

This work presents observations of acoustic emission in power semiconductor modules. The main result is that acoustic emission has now been shown to be a real phenomenon associated with power modules. Acoustic events related to the switching and failure of power modules were recorded. Three distinct types of acoustic emission were identified.

Results relevant to the instrumentation and measurement of acoustic phenomena are also discussed. It was found that the wide-band acoustic emission sensors were capable of detecting the acoustic emissions presented in this work. The sensitivity of the commercially available sensors was adequate for the measurements presented.

In Publication I, it was shown that the switching transient in a power semiconductor module causes an acoustic emission. An acoustic wave propagates through the cooling surface of the semiconductor module and is detected by an acoustic emission sensor. In Publication II, this result was confirmed by using a different type of acoustic emission sensor in the same experimental setup. Although significant differences between the outputs of the sensors were observed, the results support the same conclusion: the switching operation of a power semiconductor module causes an acoustic emission. The physical phenomenon that causes the acoustic emission is presently unknown.

In Publication III, IGBTs were made to fail, and the related acoustic emissions were monitored. Two distinct types of acoustic emission were observed: an immediate acoustic emission that occurs at the time of failure, and post-failure acoustic emissions that occur tens of milliseconds after the failure. The fact that there are two distinct emission types suggests that two physical mechanisms may be causing the emissions. The mechanisms themselves are presently unknown. It should also be kept in mind that because of the sensor limitations, this experiment was conducted under unrealistic operating conditions, and therefore, the results cannot be directly generalized to apply to most semiconductor failures.

In Publication IV, questions related to the measurement and instrumentation in this study

(36)

were addressed. Several difficulties in the measurements were identified, the most important being the temperature limitations and the self resonance of the sensors. The publication also discusses the ways in which those limitations have been circumvented in this work.

Although the results presented in the publications are reliable, there are some shortcomings that hinder their generalizability. In the experiments, the modules are operated quite far from their nominal operating point. In the switching experiments, the modules are operated under voltages that are about 5% of the usual working voltage of the modules in question.

Furthermore, the switching experiments were conducted in room temperature, while in a typical converter application the operating temperature would be much higher.

A switching experiment was also conducted at a DC link voltage of 600 V. A similar but not identical waveform was observed in this case. It seems that multiple mechanisms may be causing acoustic emissions; some seem to be excited by current and some by voltage.

In some applications it can be beneficial if different acoustic phenomena can be triggered independently from each other.

Moreover, the mechanical construction used in all experiments was simplified; nevertheless, the experiments were able to demonstrate that acoustic emission originates from the power semiconductor module under test. While this would probably hold true in a typical power converter, such a converter can be expected to be acoustically more complicated. There are several other components besides the power semicondcutors that may contribute to the overall acoustic signal. Even if these other components do not act as an acoustic source, they may still interfere with the acoustic conduction path of the signal, altering the observed signal.

As stated in the introduction, the main application of this work is condition monitoring of power electronics systems. The results presented in the work cannot be directly applied to devise a condition monitoring method, or to deduce whether an acoustic-emission-based condition monitoring method is feasible. It does, however, provide a starting point for future work, which may ultimately lead to a condition monitoring method. This is the scientific impact of this work.

The work presented in this doctoral dissertation does not have a direct societal impact. Future work, on the other hand, may have a considerable societal impact in the form of enhanced availability of power electronics. If a condition monitoring method can be developed, it has the potential to promote the availability, quality of service, and safety in industrial plants, energy production, distribution of electricity, and transportation, to name but a few.

4.1 Considerations on future work

Based on the work presented in this doctoral dissertation, multiple paths are available for future research (Fig. 4.1). The two main paths are related to the physics of the acoustic phenomena and the condition monitoring application. The study of physics could focus on grasping the mechanisms behind the acoustic emissions and phenomena associated with the

(37)

4.1 Considerations on future work 37

propagation of the acoustic wave. Comprehension of these aspects would facilitate the development of practical applications. It would also be the key scientific impact of the research area. The effects of aging on the acoustic phenomena of the power modules should also be studied in order to support the development of a condition monitoring method.

In order to fully exploit the acoustic phenomena for condition monitoring, there is more to study than the mere effects of aging. Unless a very simple solution to the condition monitoring proves to be feasible, considerable research efforts are required on detection methods, for instance algorithms that can discriminate aging information from other acoustic signals and interference. At this point it is, however, difficult to say what questions would have to be answered.

In addition to the physics and methods questions of condition monitoring, there are some very practical multidisciplinary problems that have to be solved before an end-user can ben- efit from any condition monitoring: would the condition monitoring system be permanently installed in the power converters, or would a portable instrument be used to periodically check the converter? Would the end-users of the converters own the condition monitoring system, or would they prefer to purchase condition monitoring as a service? How should the instrumentation be implemented? What kind of infrastructure such as data centers and data transmission would be needed?

The measurement-related aspects of acoustic emission are also an important, supporting research task. As has already been discussed in this work, the current measurement tools are not perfectly suited for this type of research. The measurement requirements of a practical condition monitoring method may significantly differ from those of the research presented in this work, meaning that it may be possible to sacrifice bandwidth in order to extend the temperature range of the instrumentation. Concerning a practical instrumentation, affordability is also an important factor.

One potential path is the study of nuisance noise caused by the power semiconductors. The results from this field could help improve the comfort of users in household systems, electric vehicles, and other applications where a regular user is in the vicinity of a power converter.

(38)

Figure 4.1. Future work on the subject can take several paths, some of the main ones being depicted here.

(39)

REFERENCES 39

References

Berth, M. (1998), “Partial discharge behaviour of power electronic packaging insulation,” in Electrical Insulating Materials, 1998. Proceedings of 1998 International Symposium on, pp. 565–568.

Breit, F., Dutarde, E., Saiz, J., Lebey, T., Malec, D., and Dinculescu, S. (2002), “Partial discharge detection in power modules,” inPower Electronics Specialists Conference, 2002.

pesc 02. 2002 IEEE 33rd Annual, vol. 2, pp. 748–752 vol.2.

Brown, D., Abbas, M., Ginart, A., Ali, I., Kalgren, P., and Vachtsevanos, G. (2012), “Turn-off time as an early indicator of insulated gate bipolar transistor latch-up,”IEEE Transactions on Power Electronics, vol. 27, no. 2, pp. 479–489.

Bua-Nunez, I., Posada-Roman, J.E., Rubio-Serrano, J., and Garcia-Souto, J.A. (2014), “In- strumentation system for location of partial discharges using acoustic detection with piezoelectric transducers and optical fiber sensors,”IEEE Transactions on Instrumentation and Measurement, vol. 63, no. 5, pp. 1002–1013.

Colak, Ömer.H., Destici, T.C., Özen, ¸S., Arman, H., and Çerezci, O. (2009), “Detection of p- and s-wave arrival times using the discrete wavelet transform in real seismograms,”The Arabian Journal for Science and Engineering, vol. 34, no. 1A, pp. 79–89.

Eade, G. (1998), “Financial implications and cost justification,” in ed. A. Davies (ed.),Hand- book of Condition Monitoring: Techniques and Methodology, Chapman & Hall, Padstow, Cornwall, Great Britain.

Hafez, A.G., Khan, M.T.A., and Kohda, T. (2010), “Clear p-wave arrival of weak events and automatic onset determination using wavelet filter banks,”Digital Signal Processing, vol. 20, no. 3, pp. 715–723.

Jamshidi, F. and Shaabany, A. (2011), “Neural network and wavelet transform for classifica- tion and object detection,”Journal of American Science, vol. 7, no. 5, pp. 20–25.

Kaewkongka, T. and Lim, J. (2007), “Statistical estimated parameter for pipeline condition monitoring using acoustic emission,” inIEEE Instrumentation and Measurement Technol- ogy Conference Proceedings, IMTC 2007, pp. 1–3.

Kaewkongka, T., Lim, J., Kanjanaprayut, N., and Daopiset, S. (2008), “Corrosion monitoring using acoustic emission and potentiodynamic method,” inProceedings of IEEE Instrumen- tation and Measurement Technology Conference, IMTC 2008, pp. 717–719.

(40)

Kärkkäinen, T. and Silventoinen, P. (2013), “Considerations for active condition monitoring in power electronic converters,” in15th European Conference on Power Electronics and Applications (EPE), pp. 1–5.

Lamonaca, F., Carrozzini, A., Grimaldi, D., and Olivito, R.S. (2012), “Acoustic emission monitoring of damage concrete structures by multi-triggered acquisition system,” inIEEE International Instrumentation and Measurement Technology Conference (I2MTC), pp.

1630–1634.

Lebey, T., Dinculescu, S., and Malec, D. (2004), “Partial discharges testing of power modules,” inSolid Dielectrics, 2004. ICSD 2004. Proceedings of the 2004 IEEE International Conference on, vol. 2, pp. 896–899 Vol.2.

Lehmann, J., Netzel, M., Herzer, R., and Pawel, S. (2003), “Method for electrical detection of bond wire lift-off for power semiconductors,” inProceedings of IEEE 15th International Symposium on Power Semiconductor Devices and ICs, ISPSD ’03, pp. 333–336.

Li, R. and He, D. (2012), “Rotational machine health monitoring and fault detection using emd-based acoustic emission feature quantification,”IEEE Transactions on Instrumenta- tion and Measurement, vol. 61, no. 4, pp. 990–1001.

Lim, J. and Kaewkongka, T. (2007), “Micro cracking in stainless steel pipe detection by using acoustic emission and crest factor technique,” inProceedings of IEEE Instrumentation and Measurement Technology Conference Proceedings, IMTC, pp. 1–3.

Martins, C.H.R., Aguiar, P.R., Frech, A., and Bianchi, E.C. (2014), “Tool condition monitoring of single-point dresser using acoustic emission and neural networks models,”IEEE Transactions on Instrumentation and Measurement, vol. 63, no. 3, pp. 667–679.

Moschioni, G., Saggin, B., and Tarabini, M. (2004), “Sound source identification using coherence and intensity based methods,” inProceedings of the 21st IEEE Instrumentation and Measurement Technology Conference, IMTC 04., vol. 3, pp. 1959–1964 Vol.3.

Navarro, L., Delgado, M., Urresty, J., Cusido, J., and Romeral, L. (2010), “Condition monitoring system for characterization of electric motor ball bearings with distributed fault using fuzzy inference tools,” inProceedings of the IEEE Instrumentation and Measure- ment Technology Conference (I2MTC), pp. 1159–1163.

Noma, H., Tabaru, T., Akiyama, M., Miyoshi, N., Hayano, T., and Cho, H. (2007), “High- temperature acoustic emission sensing using aluminum nitride sensor,”Journal of Acoustic Emission, vol. 25, pp. 107–114.

Ribrant, J. and Bertling, L. (2007), “Survey of failures in wind power systems with focus on Swedish wind power plants during 1997 – 2005,”IEEE Transactions on Energy Conver- sion, vol. 22, no. 1, pp. 167–173.

Smulko, J., Józwiak, K., Olesz, M., and Hasse, L. (2011), “Acous- tic emission for detecting deterioration of capacitors under aging,”

Microelectronics Reliability, vol. 51, no. 3, pp. 621 –627, URL http://www.sciencedirect.com/science/article/pii/S0026271410005391.

(41)

REFERENCES 41

Stenberg, A. and Holttinen, H. (2011), “Wind energy statistics of Finland – yearly report 2010,”VTT Working papers, vol. 178.

Suprijanto, Istiana, T., Hariyanto, and Dian, N.A. (2013), “Detection of p-wave on broadband seismometer using discrete wavelet denoising,” inProceedings of the 3rd International Conference on Instrumentation Control and Automation (ICA), pp. 110–114, iD: 1.

Turkia, V. and Holttinen, H. (2013), “Wind energy statistics of Finland – yearly report 2011,”

VTT Technology, vol. 74.

Veleshchuk, V., Lyashenko, O., Myagchenko, Y., and Chuprina, R. (2004), “Evolution of the electroluminescence spectra and the acoustic emission of the epitaxial structures of GaAsP,”Journal of Applied Spectroscopy, vol. 71, no. 4, pp. 553–557.

Vlasenko, A., Lyashenko, O., Oleksenko, P., and Veleschuk, V. (2008), “Fluctuations of current, electroluminescence and acoustic emission in light-emitting A³B⁵heterostructures,”

Semiconductor Physics, Quantum Electronics & Optoelectronics, vol. 11, no. 32, pp. 230–

235.

Zhou, S., Zhou, L., and Sun, P. (2013), “Monitoring potential defects in an igbt module based on dynamic changes of the gate current,”IEEE Transactions on Power Electronics, vol. 28, no. 3, pp. 1479–1487.

Zhu, P., Liu, Y., Robert, R., and Hao, X. (2011), “Offshore wind converter reliability evaluation,” inIEEE 8th International Conference on Power Electronics and ECCE Asia (ICPE ECCE), pp. 966–971.

(42)

Observations of Acoustic Emission in Power Semiconductors