You are on page 1of 40

IEEE PC57.127/D3.

0, February 21, 2004

Draft Guide for the Detection and Location of Acoustic Emissions from Partial Discharges in Oil-Immersed Power Transformers and Reactors
Sponsored by the Transformers Committee of the IEEE Power Engineering Society

Copyright 2004 by the Institute of Electrical and Electronics Engineers, Inc. Three Park Avenue New York, New York 10016-5997, USA All rights reserved. This document is an unapproved draft of a proposed IEEE Standard. As such, this document is subject to change. USE AT YOUR OWN RISK! Because this is an unapproved draft, this document must not be utilized for any conformance/compliance purposes. Permission is hereby granted for IEEE Standards Committee participants to reproduce this document for purposes of IEEE standardization activities only. Prior to submitting this document to another standards development organization for standardization activities, permission must first be obtained from the Manager, Standards Licensing and Contracts, IEEE Standards Activities Department. Other entities seeking permission to reproduce this document, in whole or in part, must obtain permission from the Manager, Standards Licensing and Contracts, IEEE Standards Activities Department. IEEE Standards Activities Department Standards Licensing and Contracts 445 Hoes Lane, P.O. Box 1331 Piscataway, NJ 08855-1331, USA

Abstract: This guide is applicable to the detection and location of acoustic emissions from partial discharges and other sources in oil immersed power transformers and reactors. It is intended to provide a means of associating the relative magnitude and position of partial discharges and other sources with the acoustic signals obtained by strategically located transducers. Keywords: acoustic emission (AE), attenuation, burst, gas-in-oil analysis, low-amplitude discharges, partial discharge (PD), power transformers, reactors.

Copyright 2004 IEEE. All rights reserved. This is an unapproved IEEE Standards Draft, subject to change.

IEEE PC57.127/D3.0, February 21, 2004

This introduction is not part of IEEE xxxxxx, Guide for the Detection and Location of Acoustic Emissions from Partial Discharges in Oil - Immersed Power Transformers and Reactors. The Guide is an expansion of PC57.127 Trial Use Guide for the Detection of Acoustic Emissions from Partial Discharges in Oil-Immersed Power Transformers. It has been expanded to include more theory and signal interpretation information, newer techniques for detection and the concepts for location. Active workers in the field are constantly trying to improve their methods. More effective methods may appear in the future.

At the time this guide was completed, the Working Group on Partial Discharge Tests in Transformers had the following membership: John W. Harley, Chair Donald Ayers Ron Barker Barry Beaster Jeff Benach Tord Bengtsson Paul Boman John Bosiger Pierre Boss Carl Bush Alvaro Cancino William Carter Yunxiang Chen Bill Chiu Roy Colquitt Jerry Corkran Ed Cromer John Crouse Alan Darwin Ron Daubert Fred Elliott Don Fallon Norman Field Michael Franchek Jim Fyvie Robert Ganser Andreas Garnitschnig Richard Graham William Griesacker Sergio Guerrero Ernst Hanique Tom Harbaugh Peter Heinzig Keith Highton Thang Hochanh John Holland Anthony Jonatti Steve Jordan Samer Khaled Vladimir Khalin Emil Kowal John Lackey Robert Langan Mike Lau Eberhard Lemke Raymond Lortie Richard Lowe Andre Lux Tamyres Luiz Machado Jim McIver Martin Navarro Van Nhi Nguyen Arturo Nunez Mark Perkins Paul Pillitteri Bertrand Poulin Gusftav Preininger George Reitter John Runski Dirk Russwurm Ewald Schweiger Hemchandra Shertukde James Smith Brian Sparling Ed Tenyenhuis Subhash Tuli Albert Walls Barry Ward Eduardo Garcia Wild

Others who were active in the writing of this guide are:

The following persons were on the balloting committee: (To be supplied by IEEE Standards Project Editor at time of publication)

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change

IEEE PC57.127/D3.0, February 21, 2004

Introduction ... ii 1. Overview .. 1.1 1.2 1.3 Scope . Purpose . Safety Warning .. 5 5 5 5

2. 3.

Definitions Introduction to Acoustic Partial Discharge Systems ... 3.1 3.2 3.3 The all-acoustic system . The acoustic system with an electrical PD trigger The on-line (continuous) acoustic system .

6 8 8 8 9 9


Acoustic Signal Transmission Characteristics . AE Systems: Equipment Specifications .. 5.1 5.2 5.3 5.4 5.5 5.6 5.7 Introduction ... External sensor .. Internal sensor ... Band-pass filter (optional) . Acoustic systems that record single events ... Acoustic systems that record data over extended periods . On-line (continuous) acoustic partial discharge systems .


11 11 11 12 12 12 13 14


Acoustic Emission Testing: Field vs. Factory Test Differences .. 14 6.1 Field vs. factory differences.. 14 6.2 General considerations 15 Acoustic Emission Field Test Procedure . 7.1 7.2 7.3 7.4 7.5 7.6 Introduction ... Test setup .. Sensor placement and initial scan . Monitoring techniques ... Locating the source of the signal ... Reporting and follow-up 16 16 16 16 17 18 18 18 18 19 21



Factory Test Procedure with an Electrical Trigger .. 8.1 8.2 8.3 Introduction to factory PD testing . Initial sensor placement . Measurements and changing of sensor placement


Characterization of Acoustic Emission Signals .. Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change

22 3

IEEE PC57.127/D3.0, February 21, 2004 9.1 9.2 9.3 9.4 9.5 9.6 9.7 General alternating current systems... Acoustic systems that record single events .. On-line acoustic systems ...... General direct current systems .. Characteristics of partial discharge from static electrification ..... Acoustic activity from thermal faults, the core, mechanical noises and other sources.............................................................................................. Comparison between electrical and acoustic signals 22 23 25 27 27 28 28

10. 11. Annex A Annex B Annex C Annex D Annex E

Integrating AE Results with Data from Oil Analysis .. Acoustic Activity Interpretation .. Bibliography Signal Processing . Wavelet Signal Processing Theory .. Instrumentation calibration .. Calibration of transducer and preamplifier ..

29 30 32 35 36 38 39

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change

IEEE PC57.127/D3.0, February 21, 2004

IEEE Guide for the Detection and Location of Acoustic Emissions from Partial Discharges in Oil-Immersed Power Transformers and Reactors
1. Overview
1.1 Scope
This guide is applicable to the detection and location of acoustic emissions from partial discharges and other sources in oil immersed power transformers and reactors. Both electrical (partial discharge) and mechanical sources (such as loose clamping, bolts or insulation parts) generate these emissions. There are descriptions of acoustic instrumentation, test procedures, and interpretation of results. When this guide is used with oil-immersed reactors, it must be understood that interpretation of signals may be different because of the construction of the reactor. Accuracy of location depends on the type of fault, configuration of tank, type of instrumentation and experience.

1.2 Purpose
This guide is intended to provide a means of associating the relative magnitude and position of partial discharges and other sources with the acoustic signals obtained by strategically located transducers. Users are intended to be persons knowledgeable in this field such as utility engineers, consultants, academics and manufacturers.

1.3 Safety warnings

The safety warnings in this subclause apply only to work done on transformers installed in the field, not to factory testing. Refer to factory test codes for safety warnings for these situations. PD location should only be attempted by those technicians and engineers trained in working on high-voltage transformers. WARNINGS

1. The transformer tank must be connected to a low resistance ground to limit the extremely high voltages being induced into the ground circuit and the tank if a high voltage to ground failure occurs. The personnel risk is very high if the transformer fails to ground. Even when grounded properly, the voltage on the tank to a different ground source may be LETHAL at the instant the failure occurs. 2. If the transformer is being energized or de-energized, or there is another type of power system voltage, all personnel should maintain a reasonable distance from the transformer and equipment electrically connected to the tank due to the possibility of a failure. It is recommended that acoustic measurement equipment connected to the tank be electrically isolated from the transformer tank, e.g., by optical means or by high-voltage electrical insulation, when measuring during transient events to eliminate the danger to the equipment or operators. 3. It is preferable to make all connections to the tank with the transformer de-energized, but in no case should the transformer voltage be above normal voltage while the sonic measuring devices are
Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 5

IEEE PC57.127/D3.0, February 21, 2004

installed. Personnel must not access areas where high voltages are within striking distance, such as on top of energized transformers or in bushing compartments. 4. The transformer ground circuit must never be changed (connected or disconnected) while the transformer is energized. Even with the transformer de-energized, it is possible to have circulating currents in substation ground circuits; therefore, appropriate care should be exercised when connecting or disconnecting ground circuits.

2. Definitions
2.1 Acoustic couplant: a material used at the structure-to-sensor interface to improve the transmission of acoustic energy across the interface during acoustic emission monitoring. All liquids and many gels meet this criterion. Couplants produced for ultrasonic non-destructive testing purposes are generally suitable. However, gelled glycerin and silicone grease are particularly efficient and are recommended. Acoustic emission (AE) oscillation: An oscillation produced by a resonant piezo-electric crystal when perturbed by a shock wave, which could be caused by a partial discharge. Acoustic emission oscillation rate or count rate: The number of AE oscillations that exceed the counter threshold level in a time interval, often one-second, or a number of cycles, depending on the instrument being used. For example there are 13 oscillations in the detail of Figure 1. The time interval in the detail is not defined.



Bursts 1 2

5 6

9 Threshold

13 Oscillations

Figure 1 - Typical Acoustic Emissions Oscillations


Acoustic impedance: The decisive factor for determining reflection and transmission properties when passing from one acoustic medium to another. The acoustic impedance is denoted Z and defined as (density) x (propagation velocity), Z = r v. Attenuation: The decrease in AE amplitude per unit distance, normally expressed in dB per unit length. Attenuation is due to a combination of wave diffusion and losses due to molecular collisions, as well as reflection / transmission at media interfaces.


Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change

IEEE PC57.127/D3.0, February 21, 2004 2.6 Barkhausen effect or magnetostriction noise: noise associated with the deformation of magnetic domains in the core of a transformer. The primary frequency occurs at twice the power frequency; the amplitude of this signal in both halves is about the same. Over fluxing may create considerable noise, which may have high frequency components that reach to 30-40 kHz or above. Critical angle: The largest incidence angle (from normal) for which a wave can enter a medium that has a higher propagation velocity. This angle is given by sin ()=Vin/Vout; for an oil-steel interface it is about 25 degrees for transverse waves in steel. Decibel: The decibel (dB) scale is a logarithmic scale commonly used to express sound pressure or energy levels. The conversion between decibel level and linear scale is: Sound Pressure Level (dB) = 20 log (P/Pref), where: the reference pressure, Pref, is typically 1 Pascal. Sound energy, which is proportional to the square of the pressure, increases 26% for each decibel increase. This means that the sound energy doubles for each 3-dB increase. 2.9 Diffraction: The distortion of an acoustic wave front by an object in the sound field. It is characterized by the "bending" of sound waves around the obstruction.



2.10 Direct acoustic (oil-borne) path: Propagation of the PD acoustic signal through the oil directly to the sensor location on the tank wall. 2.11 Harmonics: Higher frequency components of various noises are sometimes difficult to distinguish from partial discharge. They can be repetitive, but are usually at lower energy levels than significant partial discharge

2.12 Longitudinal waves or pressure waves: Consist of alternate compression and rarefaction of oil in the direction of propagation. The molecules vibrate back and forth about their rest positions parallel to the wave axis. Normally this is the only kind of wave that propagates in liquids and gases. They are easily propagated by molecular collision. 2.13 Out-of-plane waves: In a flat plate or shell, an out-of-plane wave yields motion that is perpendicular to the plane of the plate. A pressure wave in the oil impinging at normal incidence on the tank will generate out-ofplane motion in the tank wall. These waves are also called bending or flexural waves. Out-of-plane waves have a propagation velocity that is dependent on the frequency and are thus difficult to use for localization. 2.14 Partial discharge (PD): An electrical discharge that only partially bridges the insulation between conductors. 2.15 Reflection: Scattering of propagating waves back in the direction of origin by an obstruction or discontinuity in the propagation path. 2.16 Sensors: For the applications discussed here, damped acoustic emission piezo-electric sensors are generally used to transform the particle motion produced by an elastic wave into an electrical signal. They are typically mounted externally on the transformer tank or mounted on a wave-guide that is submerged in the oil inside the transformer tank. Different investigators are using sensors with a sensitive region ranging from about 20 kHz to 500 kHz. As sensor output is usually inversely proportional to bandwidth, often a sensor with a narrower bandwidth centered at either 60 kHz or 150 kHz is used in PD detection. Note that the sensor is sensitive to pressure waves in its frequency range that may not be from a PD source. 2.17 Speed of sound in oil medium: 1413m/s at 20 C typical. Corrections to speed of sound for temperature and moisture content are not generally made to increase accuracy because the uncertainties due to material propagation are usually much larger. The values of the speed of sound in oil may change depending on the properties of the oil. 2.18 Speed of sound in steel: pressure wave (bulk) about 5800 m/s, (plate) about 5200 m/s; shear/transverse wave Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 7

IEEE PC57.127/D3.0, February 21, 2004 (bulk) about 3250 m/s, (plate) about 3200 m/s 2.19 Structure-borne path: Propagation of the PD acoustic signal through the transformer structure. 2.20 Transverse or shear waves: The particle motion is perpendicular to propagation direction similar to a vibrating string. A necessary condition is that there is sufficient attraction force between particles, so that as each particle moves back and forth, it pulls its neighbor along with it. The reason they do not occur in liquid and gases is that there is only a small attractive force between molecules. Typically shear waves travel at about half the velocity of longitudinal waves. At wavelengths comparable to or larger than the thickness of the material, the transversal waves are gradually transformed to out-of-plane waves (definition 2.13).

3. Introduction to Acoustic Partial Discharge Systems

Partial Discharge (PD) detection and location work is carried out in the factory and in the field, the latter being done with the transformer either connected to the grid or supplied by a separate power source. There are two general categories of acoustic location systems: all-acoustic systems and acoustic systems with an electrical PD trigger. In addition, on-line (continuous) acoustic monitoring systems are being used primarily to detect and trend emissions. Typical implementations of each of these types of systems are given below along with the respective advantages and limitations.

3.1 The all-acoustic system

The first category, the all-acoustic system, consists of one or more ultrasonic transducers that are sensitive to the acoustic emissions generated by a PD event. The detection and coarse location of one or more sources can be accomplished by moving one or more externally mounted sensors to different locations on the transformer tank. A more precise location of a PD source may be determined by the relative arrival times of the acoustic signals at each of the sensors. No voltage or current readings are required on the transformer. This makes the all-acoustic system a suitable tool for source location on operating transformers in the field. The acoustic transducers can be mounted on the exterior of the transformer tank to detect the acoustic signal as it hits the tank wall, or inside the transformer to detect the signal in the oil. Benefits of the externally mounted sensors include the ability to reconfigure the sensors as necessary to obtain a clearer acoustic signal, the flexibility to move the system to another transformer, and the ease of retrofitting existing transformers. The disadvantage of the external configuration is that it is more sensitive to external noise sources. Benefits of the internally mounted sensors may include a clearer, louder measurement with a better signal-to-noise ratio. Disadvantages are that installation of the sensors is invasive to the transformer and, once installed, the sensors can not be moved around to achieve a clearer line of sight to the source, nor can they be easily removed and installed on another transformer.

3.2 The acoustic system with an electrical PD trigger

The second category, the acoustic system with an electrical PD trigger, pairs the array of acoustic sensors described above with a current or voltage measurement device that detects the PD signal electrically. The electric signal is usually considered as detected instantaneously. When using this assumption, the arrival time of the electric signal is used as time zero for the PD event. The difference in arrival times of the electric signal and an acoustic signal is the propagation time between the PD source and that sensor location. PD location is based on the absolute arrival time at each sensor, as opposed to the all-acoustic system described above which uses the difference in arrival time between sensors. The assumption of instantaneous detection of the electric signal is correct when it appears e.g. at the bushing tap of the transformer. At the most this delay is in the range of a few micro-seconds and therefore can be neglected. However, an acoustic system with an electrical PD trigger is processing the electrical signal first. Here the

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change

IEEE PC57.127/D3.0, February 21, 2004 processing time may be taken into account in order to decrease location error. The following time lags due to the analog processing time were measured in three different electrical PD detector types.

Type of detector Wide-band with active Integrator (100-400 kHz) Narrow-band at 1 MHz With 9 kHz bandwidth Narrow-band at 1 MHz With 4.5 kHz bandwidth

Time-lag to t0 0.2 s 16 s 16 s

Signal rise time (10-90%) 1.2 s 35 s 50 s

Approximate location error in oil Negligible (< 2 mm) 22-70 mm (1~3 in.) 22-95 mm (1~4 in.)

The numbers are provided only as an example. Other detectors from different manufacturers may have different time lags since this property is not specified in the standards. It can be seen that a narrow-band detector may introduce a time lag that should be taken into account if very accurate location measurements are desired. For the calculation of distance of PD source from the sensor, following are estimates of the velocity of sound in oil. The velocity will be different if there are other materials in the wave path (see also definition 2.17). Temperature of oil - C 50 80 110 Velocity m/s 1300 1200 1100

One advantage of the combined system is that the electrical measurement provides confirmation that the acoustic sensors are locating a PD event as opposed to another acoustic noise source. Furthermore, the electrical signal is a convenient trigger that can be used to start the data acquisition at the acoustic sensors. The major disadvantage of the combined system is that it may be difficult to obtain a clean electric PD measurement due to electrical noise in the field. Hence, the combined acoustic-electrical PD locator system is more suitable for use in the factory or plant than in the field.

3.3 The on-line (continuous) acoustic monitoring system

The main purpose of permanently installed on-line acoustic monitoring systems is to provide an early indication of an incipient fault to a remote location. These systems usually consist of multiple sensors, amplifier and data acquisition/processing modules. The sensors are placed at locations where faults may be anticipated based on past experience or highest probability of problems occurrence. The data acquisition/processing systems that are able to transmit collected data and/or warning alerts to locations outside the substation. The data is often limited to activity levels rather than specific waveforms. Location information is often limited to knowledge of which sensor is most active. Warning alerts caused by level of activity above a baseline or trend of activity are usually reason to take gas samples and perhaps perform more extensive acoustic or other testing.

4. Acoustic Signal Transmission Characteristics

Active PD sources in oil-filled transformers produce acoustic emission signals that propagate away from the source in all directions. The acoustic signals travel through the intervening material to eventually arrive at the transformer tank wall. The distance traveled in a particular medium is dependent on the time that it takes for an acoustic signal to complete this journey as shown below: distance traveled = acoustic wave speed x time Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 9

IEEE PC57.127/D3.0, February 21, 2004 Consequently, sensors placed at different locations on the tank wall, i.e. at different distances from the source, will experience different signal arrival times. For the case of the all-acoustic system, it is possible to detect the difference in signal arrival time at one sensor relative to another. This in turn allows for the estimation of the difference in the two propagation path distances. For the case of the combined acoustic-electric system, the time delay between the source and the sensor can be detected, and the absolute path length between the source and a given sensor can be determined. Each of these system types requires a slightly different procedure to locate the source, as described in the sections on Operating Procedures, Sections 7 and 8. The fundamental principle of acoustic PD source location assumes that the acoustic signal travels a direct, straightline route from the source to the sensor. Unfortunately, this is not always the case as the acoustic field inside the tank is very complex due to wave reflection, diffraction in different materials and other parameters. For example, if there is an obstruction blocking the line of sight between the source and the sensor location, the sound may travel around the obstruction. This results in a longer propagation time that would imply a greater distance between the source and the sensor than actually exists. Alternately, the sound may travel directly through the obstruction at a wave speed that is greater than in the oil. The resulting arrival time would be earlier, which would imply a shorter distance between the source and the sensor than actually exists. To help avoid these misinterpretations, it is important to confirm the estimated source location by repeating the distance calculations with several sensor locations. Structure-borne propagation paths within the tank wall present a further technical challenge. As the acoustic wave hits the tank wall, its frequency characteristics remain the same, but its mode of propagation and propagation speed change. Take the example in Figure 2 in which the sensor is located on the far side of the tank, away from the source.

Structure-borne path

Direct acoustic path PD Source

Acoustic sensor Primary reflection wave Transformer tank

Figure 2 - Illustration of typical propagation paths for the acoustic PD signal. Primary reflection wave propagation can take place within the confines of the tank as shown. The speed of sound for this wave depends on the media encountered by the sound wave. Acoustic waves hitting the nearby tank wall will create an alternate propagation path via the tank wall to the sensor on the other side. The wave speed in metal is greater than in oil. Therefore the wave traveling this structure-borne path may arrive at the sensor earlier than the wave traveling the direct acoustic path. If the distance calculations were based on the arrival time of the structure-borne wave using the wave speed in oil, this would imply an incorrect distance between the source and the sensor. It is crucial that distance calculations are based on the direct acoustic path. It is critical to confirm the estimated source location by using a variety of sensor locations. Another way to distinguish the structure-borne waves from the oil-borne waves is to analyze the mode of vibration. Fluids, such as oil, will only support pressure waves. Solids, such as steel, can support many types of wave motion. Waves in the oil give rise to both pressure waves and shear waves in the tank wall.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

Source Critica l a ngle for: pressure wa ve she ar wave Dire ct wave

Tank wall

Struc ture -bo rne waves Dete ctor

Pressure (Longitudinal) Waves Shear (Transversal) Waves

Figure 3 - Illustration of the longitudinal and transversal waves in the enclosure and how they are created from direct waves. Note: the wavelengths are not to scale. Two wave fronts will be seen in these cases, as illustrated in Figure 3. The transversal (also called shear or slow) wave has the largest amplitude and can be identified in this way. Using sensors that are more sensitive to out-ofplane waves than in-plane waves can help to minimize structure-borne interference. The problem of structure-borne waves is reduced significantly if the acoustic sensor is located inside the tank.

5. AE Systems: Equipment Specifications

5.1 Introduction
Many different types of instrumentation are available for location of acoustic emissions. Several typical systems are described here which have been shown effective in certain transformer arrangements; however, other systems may be equally or more effective, depending on the transformer physical parameters and the location of the partial discharge.

5.2 External sensor

The sensor mounted on the external surface of the transformer tank is a piezo-electric displacement transducer operating in its compression mode. Users typically choose sensors with resonant frequency (for longitudinal waves) in the 60 or 150 kHz range. It has been shown [B5] that the main frequency of a partial discharge about 150 pC magnitude is 100 kHz. For larger discharges, frequencies should decrease. Also, attenuation affects high frequencies more than low. These factors favor the sensor with the lower range. In the field, however, numerous noises or harmonics of noises are encountered in the lower frequency range that may lead to false readings. A number of users favor the higher range sensor for this reason. Being a piezo-electric device, the sensor will also respond to varying electromagnetic fields such as those found in substations. To minimize this effect, the transducer can be either a "differential" type utilizing two crystals (mounted out of phase for noise reduction) or a shielded single crystal transducer with an integral pre-amplifier circuit. The latter is the preferred and most common configuration because its comparatively high amplitude, low impedance output is less susceptible to degradation due to noise pick-up in the connecting cables. The acoustic impedance of a sensing crystal differs from that of the steel transformer wall. Therefore, for efficient transfer of the signal from the steel to the crystal, some users interpose a "matching piece". Although several materials may be used for this purpose, a hard epoxy resin is convenient because it also provides some thermal and Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 11

IEEE PC57.127/D3.0, February 21, 2004 electrical isolation. However, care should be taken to select a resin which exhibits low acoustic attenuation (usually a function of the fillers used) so that it does not adversely affect the amplitude of the transmitted signal. Furthermore, as the acoustic impedance of epoxy resin does not numerically fall between that of steel and crystal the thickness of the matching piece should be equivalent to half the wavelength of the signal propagating in it (in this case, 60 or 150 kHz longitudinal waves). The acoustic couplant gel or grease defined in 2.1 should be applied to the face of the transducer or matching piece just prior to test.

5.3 Internal sensor

The internal acoustic emission sensor or waveguide is a device immersed in the oil that couples acoustic energy from the oil to an externally mounted sensor. An example of a waveguide would be a solid fiberglass rod inserted through a seal into the transformer tank with a sensor mounted on the external end of the rod. Careful attention must be given to maintaining dielectric properties and dimensions. The waveguide is (1) less directional than external tank wall-mounted sensors because the impedance of the steel tank wall is not a factor and (2) is less sensitive to external noise such as storms or loose fittings. A new technique under development uses an optical interferometric technique. A laser light source is coupled to the sensing head through an optical fiber. The sensing head consists of an air gap and a thin silica glass diaphragm. The optical signal is a function of the air gap length. Variations in the length, caused by an acoustical pressure wave, are proportional to the intensity of the acoustic signal.

5.4 Band-pass filter (optional)

The filter is a band-pass type with lower and upper cutoff frequencies of FL and FH. These are frequencies at which the response to a constant sinusoidal input voltage has fallen by 3 dB from the maximum value. When used with a 150 kHz sensor, FL should be about 100 kHz and FH should be about 300 kHz. The roll-off characteristics of the filter shall be a minimum of 48 dB/octave (240 dB/decade) for the high-pass section. This means that, relative to the signal of interest (150 kHz), a 50 kHz signal would be attenuated by 48 dB. The low-pass filter should roll off at not less than 24 dB/octave (120 dB/decade) so that a 600 kHz signal would be attenuated by 24 dB. The purpose of the filter is to negate as many of the effects as possible of signals that are not associated with PDs. These include vibrations caused by the magnetostrictive action of the core (Barkhausen noise), pumps and fans. Most of these fall below 30 kHz; however, the Barkhausen noise emanating from the core has sometimes been found to be in the 50 kHz range. Hence, a 100 kHz high-pass section with a rapid, roll-off response characteristic is needed. The reasonably generous band-pass (200 kHz) allows for variations between different transducers, in so far as their resonant frequencies are concerned.

5.5 Acoustic systems that record single events

A typical system consists of the following components. a) Sensors: As discussed in 5.2 and 5.3

b) Sensor holders - these must be practical even when reaching far while standing high up on a ladder. c) Cabling and power supply units for the sensors. d) Measuring device for measuring sensor positions, etc. e) A digital detection system. This may consist of one or more standard 4-channel digital oscilloscopes with sampling rate for each channel greater than 1 mega-sample/second and memory depth greater than 5000 samples. Data acquisition units available for computers could be used if they fulfill these requirements. Features such as averaging, peak detection, zoom, measurements and storage are very useful.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

5.5.1 Variation of acoustic systems that record single events

Optimization of sensitivity and rejection of non-PD signals are the two basic requirements in the design of this system. Further, it should be applicable in a factory as well as in the field. If available, it should be able to take advantage of electric PD signals for noise rejection, signal enhancement and localization. It should also be as easy to work with as possible. These requirements result in a system that consists of the basic parts: a) An acoustic detector frame that holds three piezo-electric transducers in fixed relative positions. Thus enabling immediate determination of the direction to a source relative to the position of the frame.

b) One or more 4-channel digital oscilloscopes for signal acquisition, one oscilloscope for each detector frame. With digital oscilloscopes, a great flexibility in choosing the trigger method, channel and level is obtained. Further, the fast refreshment rate of the screen is invaluable as a help to discriminate against spurious signals. Other facilities that have proven to be very important are the averaging and persistency functions normally available. c) A computer for storage of the signals acquired by the oscilloscopes.

This system is intended to be operated by an experienced person. In searching for possible PD signals and their origin, the detector frame is moved to selected positions on the transformer tank. At each position, a set of different trigger and time frame settings are studied on the oscilloscope, which is thus the main instrument. For reference, a few signals can be stored in the computer. When a signal to be localized is found, the computer aids in calculating the estimated position of the source.

5.6 Acoustic systems that record data over extended periods

The following is representative of systems that have been implemented by commercial suppliers for higher precision of PD location: a) Data acquisition system with general DSP (digital signal processor) controller with DSP cards. The system should come with suitable CPU, RAM, hard drive, fast video, monitor, keyboard, mouse and cache. b) Suitable software with provision for data collection, storage, display and replay capability for linear and/or three-dimensional location, FFT (fast front transient) analysis or advanced Wavelet Transform analysis, waveform recording, real time feature extraction and suitable GUI (graphic user interface) for PC based platforms.

c) DSP two channel digital AE card with minimum 512kB RAM (expandable), two analog and one digital parametric and hardware and parallel capture of transient and classical AE feature extraction data, I/O ports, PC interface, sensor preamp, BNC auto sensor calibration, and software selectable filters. An alternate digital hardware that can increase the number of channels that can be handled by the DSP board with suitable interface box can be used. This should have a minimum of 1 MHz as a sampling rate per channel per sensor from which data is being acquired. d) An acoustic sensor as described in 2.16, 5.2 and 5.3 above e) Hold-downs for the acoustic sensors f) Suitable User's Manual with proper instructions and suitable examples to illustrate working of the device and interpretation of results g) Transportation cases for data acquisition system and accessories are suggested

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

5.6.1 Optional equipment

a) Preamplifier with 40/60 dB Gain w/100-300 kHz band pass filters. Filters are described in 6.4 above. b) RG58 15 meters minimum of signal cable (BNC/BNC or SMB/BNC terminations) c) Inkjet or suitable printer, associated cable and software drivers for above products d) Industrial computer chassis, rack mounted capable of housing sixteen DSP PC cards e) Acoustic verification rod about 1 m length

5.7 On-line (continuous) acoustic partial discharge monitoring systems

There are two types of systems that utilize continuous monitoring of PD. a) b) Portable systems to identify, qualify and locate PD sources Permanently installed on-line monitoring systems to detect and trend PD signals

These systems usually consist of (1) multiple sensors, which are placed at locations where faults may be anticipated based on past experience or highest probability of problems occurring, and (2) amplifier and data acquisition/processing systems that are able to transmit collected data and/or warning alerts to locations outside the substation. Supply of power from the station DC source or another uninterruptible supply may be specified. For long-term robustness, the systems should require protection of inputs, outputs and grounds similar to electronic relays. A test to establish the baseline data of acoustic activity on the transformer is performed during the installation process. This test provides information to establish the best hardware and software settings for permanent monitoring. It is helpful for gain settings to be the same as adjacent transformers in order to compare readings if signals are thought to be the result of network disturbances. Permanent mounting of sensors may be desirable for long term installations. This may be accomplished with a suitable epoxy and/or a mechanical holder.

6. Acoustic Emission Testing

6.1 Field vs. factory test differences
Acoustic emission testing in factories or field situations has many differences. Some of the specific ones are the following: Field Limited control; e.g. noise from rain or hail. Signal processing or post test analysis can reduce or eliminate noise in some cases. May be required by system operating requirements. Limited control, e.g. noise from rain. Signal processing or post test analysis can reduce or eliminate noise in some cases. Isolation, Filtering, and/or UPS required Single point grounding for test object Factory Controllable

Ambient Noise sources

Cooling Pumps Weather

Can be shut off Testing done indoors

Power Supply Grounding

Clean power available Appropriate grounding provisions

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004 and equipment needs to be established. Precautions needed for working near energized facilities. Generally limited knowledge of what is inside the tank. Limited adjustment range unless a test power supply is taken to the transformer location Use care that sensors are mounted below oil level. Also, bubbles may be present during certain operating conditions. built into test bay.

Equipment Access Transformer Design Test Voltage

Full design information available. Full control of voltage levels

Gas blanket (air or nitrogen)

More information appropriate to the two environments is included in Sections 8 and 9.

6.2 General considerations on the application of sensors

Transducer placement and mounting is critical to getting good results [B4]. Some things to consider include the following: a) The contact between the transducer and the transformer tank is critical. Simply placing a transducer on the transformer tank surface often produces a very weak signal. It is advisable to wipe the area free of dirt, oil, bugs, etc. and polish it with a mild abrasive or abrasive cloth before placing the transducer. b) An acoustic couplant is essential for enhancing the mechanical and acoustical coupling between the transducer and the tank surface. It should be evenly applied to the clean mounting surface of the transducer before placement. c) A sound transmitting epoxy may have to be used if the mounting location is non-magnetic. d) Transducers mounted on the tank walls may detect both direct and wall-borne signals. Those mounted on a bolted cover or other gasketed surface may receive the direct signal more clearly, but the wall-borne signals may be distorted or dissipated because of the gasket. e) Avoid locations where there is magnetic or non-magnetic tank shielding, which will cause extra signal attenuation f) On transformers built with double wall construction, transducers should be located on the welded ribs that span between the two tank walls to provide a strong signal. The air in the cavity between the walls attenuates acoustic signals. g) Avoid locations above nitrogen blankets. This adds extra impedance in the transmission path, which produces additional attenuation h) Avoid tank stiffeners i) For safety reasons, avoid locating sensors in areas of high voltage j) Provide sufficient spacing between sensors to insure independent signals. Distance depends on sensor model k) Verify sensor operation, either the entire system or each sensor. Amplifiers should be set for similar sensitivity for all sensors. l) An acoustic verification rod may be helpful in determining an initial location for a sensor m) Avoid the ends of tanks with three-limb core single phase and five-limb core three phase Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 15

IEEE PC57.127/D3.0, February 21, 2004 n) Aluminum or stainless steel walls are non-magnetic. Sensors may have to be placed using epoxy. o) Do not mount sensors on control boxes p) LTC operation contains a high electromechanical energy that usually propagates through the entire transformer. Actions should be taken in order to identify operations during the test and distinguish them in post-test analysis to identify the source of acoustic emissions. q) Shell-form transformers: locate the sensors on top (above the core) or bottom (below the core). r) Transient voltage protection must be applied to the input of test instruments/devices. s) Extraneous electrostatic and magnetic signals may cause false indications and damage equipment.

7. Acoustic Emission Field Test Procedure (Typical)

7.1 Introduction
Acoustic emission testing is usually done in response to gas-in-oil test results or noises that may indicate partial discharge activity. In most cases, the transformers are monitored while carrying load. This requires safety precautions when placing the transducers, assuring that minimum approach distances are not compromised.

7.2 Test setup

Approximately two to three hours are usually required to prepare for monitoring after arrival at a site. This will depend on number of sensors to be installed and accessibility to the unit, power source, etc. If a test power supply is being used, extra time will be needed to setup and check out the power supply. Set up the test equipment at a convenient location for access to the transformer and the station service source. Establish a common point ground for all of the test equipment and the transformer. The transducers are connected to the test equipment with coaxial cables. An operational test should be followed in order to assure the same sensitivity of all sensors. A common method is to break the lead on a mechanical pencil near the sensor. Cable routings should be chosen to minimize interference pickup from bus work and grounding connections in the test area. Verify proper operation of the monitoring system. Tapping on the tank is the usual method of causing a test event. The signals recorded by the various transducers can be compared to see if they are reasonable and consistent.

7.3 Sensor placement and initial scan

Unless on-line PD monitoring is being used, there is usually no indication of where to look for the noise source. Therefore, a scan covering the complete transformer is the first step. a) For a three-phase core form transformer, install one sensor in the general area of each bushing bottom connection. Place additional sensors at approximately the center of each winding limb on the high and low voltage sides of the tank. If extra sensors are available, place them on the tank ends. b) Similarly, for a single-phase core form transformer, install one sensor in the general area of each bushing bottom connection. Place additional sensors at approximately the center of each limb of the windings on Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 16

IEEE PC57.127/D3.0, February 21, 2004 each of the 4 sides of the transformer tank. c) If the transformer has a preventive autotransformer inside, consideration should be given to placing one or two transducers close to its location. d) Using the same sensor locations and software settings when testing identical transformers permits establishing ambient baseline data for that particular design. The sensors may have to be moved to improve reception of signals from a specific source. e) Sufficient couplant gel is to be applied to the face of the transducer to ensure efficient transmission of the AE signal from the tank wall to the sensing crystal. Too much couplant will not be harmful (though wasteful), whereas too little couplant can seriously inhibit the transducers sensing capabilities. f) Magnetic shielding blocks the signal from the sensors. The location where shielding is installed is usually obtained by referring to pictures or notes from past experience with the transformer or from the manufacturer. g) In the case of transformers indicating arcing, a high frequency split-core CT or ammeter (Rogowsky coil type) around the transformer case ground can be used as a direct signal to the digitizer. This can be used as the trigger transducer to indicate time zero. The time period for the initial scan is dependent on the acoustic activity rate. Some users monitor for about 4 hours before changing transducer placement. This basically assures the emission patterns are consistent and repeatable. If activity rates are low or erratic overnight monitoring may be required. The face of the transducer with its film of couplant should be brought into contact with the transformer tank wall with only sufficient pressure applied in order to get a good signal and hold it in position. It is only necessary to hold the sensor steady so that no signals are generated due to relative movement between the sensor and tank wall. This can be achieved by means of a magnetic clamp, adhesive tape or epoxy.

7.4 Monitoring techniques

When the transformer is energized from the power system, voltage levels are fairly constant making inception and extinction levels of acoustic activity very difficult to determine. Techniques for varying system conditions to help characterize the acoustic emission source include: a) Varying transformer loading to see if the acoustic activity is load related b) Raising or lowering the bus voltage several kV by tap-changing and/or capacitor/reactor switching c) Moving the transformer on-load tap-changer(s) up or down a step at a time to see the impact on the acoustic activity If the transformer is energized from a test power supply, the test voltage can be raised and lowered at will for determination of acoustic activity inception and extinction levels. Signal sensitivity is unique to each particular acoustic emission monitoring system. Trigger level settings depend on the normal operating noise on the transformer to be tested (this depends on core type, pump/fan operation noise, etc.). The best initial settings are the least sensitive settings that give a clear AE pattern (avoiding the noise coming from the core, pumps/fans or other sources). The settings can then be optimized during the signal location process. A three dimensional layout can be constructed, for instance, using a convenient corner of the tank for zero coordinates. For tanks with rounded corners, consider squaring them off for reference. After sensor locations are plotted, the time delay in each transducer signal can be used to estimate the source location. Using the maximum number of sensors possible improves the accuracy and usability of this approach. Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 17

IEEE PC57.127/D3.0, February 21, 2004

7.5 Locating the source of the signal

An approximate source location can be estimated based on relative timing of signal arrivals at the various transducers and assumptions about the speed of sound in oil. This is not a foolproof method because of the different types of material (insulation, steel, copper, oil, etc.) in the paths between the signal source and the various transducers. Also, there may be more than one signal source in the transformer. Reposition transducers to focus on the area of the transformer tank where the suspected source is located and continue the monitoring process. The sensors can be arbitrarily close as long as a time difference between them can be measured. Sometimes, even when transducers have been relocated to a suspected area, only one sensor detects the acoustic activity. This can happen either because the source detected is very low in intensity or highly attenuated or because there is a problem with the instrumentation. Care should be taken not to automatically disregard this activity as noise. The signal source location can be confirmed by a set of consistent signal timings. A clear consistent picture of the signal characteristic helps to identify the source type.

7.6 Reporting and follow-up

Comparison of actual problems and the predictions from acoustic monitoring are important for building the knowledge base on the response of a particular monitoring system. This greatly enhances the usefulness of acoustic monitoring as a diagnostic tool. The investigation report should include information on the test setup, test procedures, test waveforms (if any), and predicted signal sources based on experience, internal assembly drawings and/or photographs of the transformer under test. Follow-up reporting should include comparison of the actual problem and the predictions from the investigation. Useful information includes: a) b) c) d) What was found Where was it located Differences between predicted and real locations How the problem was corrected

Acoustic monitoring after repair allows comparison of before and after acoustic signatures and allows establishing the new baseline data of the transformer.

8. Factory Test Procedure using an Electrical Trigger

8.1 Introduction
An important difference between the factory test and the in-field test is the advantage of being able to use an electrical partial discharge signal to trigger the acoustic data acquisition. This electrical trigger, occurring virtually instantaneously as the partial discharge emissions leave their source, provides a convenient "time-zero" mark to use to measure the time differential between this signal and the acoustic signals, and hence calculate the distance from the acoustic sensors to the PD source. The input of the electrical PD detector is typically connected to the bushing capacitance tap. The output is connected to one of the channels of the acoustic location system. It is recommended to use an electrical apparent charge detector rather than a radio-influence voltage meter since the frequency range of the former is much closer to that of the acoustic emissions than the latter. However, either can be used for the trigger. The acoustic emission Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 18

IEEE PC57.127/D3.0, February 21, 2004 sensors are connected to the transformer tank wall. The outputs of the sensors are connected to the acoustic location system for viewing the acoustic signal and its time delay from the electrical signal.

8.2 Initial sensor placement

The initial placement of the acoustic emission sensors onto the tank wall can make the difference between an efficient location of the PD source and a more time consuming one. Guidance in making an informed estimate of where to initially place the sensors can come from the Induced Voltage Test. This test, in most cases, will indicate which phase of the transformer contains the PD source. If so, the sensors should then of course be placed in the area of the problem phase. Caution should be exercised because there has been at least one experience [46] where the induced voltage test indicated the PD to be located in one phase and the acoustic test located it on a different phase, the latter being verified through internal inspection. A typical example would be a delta winding where each bushing is connected to two phases. Another source of information is to examine the transformer internal assembly drawings and photographs if available. They may show likely areas of PD sources, and indicate whether the problem may be on the LV or HV side, or an end wall, and whether it is towards the top, middle, or bottom of the tank. An example of a simplified HV-Side transformer assembly is shown in Figure 4. A search for the problem PD source within such an assembly can include placing sensors along the windings, over each de-energized tap-changer, and/or along the high-voltage leads connecting each HV winding to its respective bushing.

Figure 4 - Example of HV-Side of Three-phase Transformer Assembly Ideally, initial placement of sensors would resemble one of the three arrangements shown in Figures 5a, 5b, and 5c. Figures 5a and 5b are for the case when the problem phase is known; Figure 5c for when it is not known.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

Figure 5a - Sensor Locations with PD Source in Center Phase. Typically sensors are located on each side wall.

Figure 5b - Sensor Locations with PD Source in Outer Phase. Typically sensors are located on each side wall and on the end wall.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

Figure 5c - Sensor locations when phase of PD source is not known. Sensors are located initially on one side wall.

In the above figures, the top, middle, and bottom of the suspect phases are tested, in an effort to obtain general overall coverage of each phase. In actual practice, however, often the most convenient placement of the sensors is not possible due to obstructions such as radiators, tank stiffeners, on-load tap changers, and control cabinets. Figure 6 shows the plan view of an actual transformer tank having all of the aforementioned obstructions.

Figure 6 - Transformer Tank Plan View Some creativity is therefore required, and the actual sensor placement may be significantly different from the arrangements shown in the previous figures. For example, due to the presence of radiators or internal wiring trays, the sensors for one phase might need to be placed on the opposite side-wall than the other two phases. Mounting the sensors on manhole or other covers may provide an AE signal path that is clear of tank wall - carried signal if the cover is isolated from the tank by flexible gaskets. The transformer oil must be in contact with the inside surface of the cover for the sensor to be effective. Shell form type transformer arrangement requires special mention owing to the shielding used inside the tank walls. Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 21

IEEE PC57.127/D3.0, February 21, 2004 It may only be possible to place sensors on the top of the tank or in areas where there is no shielding.

8.3 Measurements and changing of sensor placement

With the sensors in place, a simple check should be performed to ensure that they are indeed operational. Using low signal amplification setting, this check can be accomplished by gently tapping your fingertip on the tank wall immediately beside one of the sensors. This will create an acoustic signal, which should be detected by the sensor and displayed on the oscilloscope. Use this procedure to check each sensor in turn. Now, the first set of acoustic emission measurements is ready to be taken. To do so, energize the transformer in the same manner as during the routine Induced Voltage Test. Increase the voltage slowly until the electrical apparent charge detector measures sustained partial discharge activity, or until the full test voltage level is reached that can be safely sustained continuously. (Do not raise the voltage level beyond this amount, as a higher voltage level, though possibly producing greater PD activity, unduly stresses the transformer insulation). At this voltage level, and with a medium acoustic signal amplification setting, the oscilloscope trigger sensitivity level must be calibrated to the corresponding electrical PD detector signal. Now, for every half-cycle of the voltage wave, the oscilloscope will be triggered to begin data acquisition from both the electrical PD detector and the acoustic emission PD sensors. "Timezero" will be the instant in time that the data acquisition is triggered to begin. As many sensors as the oscilloscope will allow should be displayed at once. Data that invites analysis can be "frozen" on-screen and stored in memory using the digital oscilloscope. The interpretation of the acoustic signals displayed on the oscilloscope is done according to Section 10.1 of this guide. If none of the sensors indicate PD activity, increase their signal amplification slightly and repeat the test. If there are still no clear signs of activity, the sensors will have to be moved manually, perhaps by a displacement of half a meter either vertically or horizontally, and retested again. Another suggestion is to move some of the sensors to the opposite side of the tank. This is a trial-and-error process that continues until at least one of the sensors indicates the presence of PD activity. Again, this process can be expedited with the examination of the transformer internal assembly drawings and photographs. When one or more of the sensors shows activity, assuming in the simplest case that only one problem PD source is present; the rest of the sensors shall be manually moved into the vicinity of the active area. The objective is to obtain acoustic emission signals from all sensors, and to position them so as to minimize the time differential between "time-zero" and the beginning of the acoustically detected PD signal. Ideally, this minimum time difference will correspond to a maximum number of individual oscillations in the signal as measured by the counting circuit, however this may not necessarily be so (refer to Sections 3.2 and 4 for calculation of the distance from the sensors to the PD source). Using this procedure for sensor placement and acoustic emission measurement, the PD source can often be located to within 25 cm or less.

9 Characterization of Acoustic Emission Signals

9.1 General alternating current systems
In general, two types of results (other than zero) may be encountered: one in which "continuous" readings are obtained, while the other produces "sporadic" readings. By continuous, it is meant that activity is present all the time, though it may have a varying AE amplitude. The repetition rate is on the order of the excitation frequency. This type of signal is typical of that produced by an energetic PD source or one not attenuated by structure. Sporadic activity can be further subdivided into two types. a. Sporadic AE from a continuous PD source is characterized by activity that is present most of the time, but short quiescent periods are also encountered. This type of signal is usually produced by exposed sources on conductors, connectors, worn tap changers, foreign objects or ungrounded hardware.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004 b. Sporadic PD is typified by lengthy quiescent periods (perhaps minutes) followed by short periods of very high activity. This type of signal usually has been found to be associated with floating static shields. Often short-lived arcs are associated with this type of fault, and these produce very energetic AE signals during active periods.

As previously described, it is often possible to determine the position on the tank wall where the transducer is closest to the PD source. This does not give information as to the distance into the tank (from that location) to the source. However, observing the signal on an oscilloscope (a digital transient recorder is recommended), it is possible to form an opinion regarding this. For example, the pulse shown in Figure 1 has suffered very little attenuation. This is evidenced by the high rise rate of the leading edge of the burst envelope, resulting in the characteristic "arrow head" shape. To achieve this, the propagation path is almost entirely in oil with little solid insulation involved. If the same signal had propagated through layers of insulating materials, the resulting attenuation would not only have affected the overall amplitude, but also modified the burst envelope by "rounding off" the leading edge. In the extreme, the burst envelope becomes "egg-shaped" or "ellipsoid-shaped" as shown in Figure 7. By utilizing this phenomenon, it may be possible to estimate whether the source lies close to the surface or is buried well within the insulation system.

Figure 7 AE burst showing effect of attenuation

The attenuation shown in the Figure 7 AE burst can also be caused by the location of the sensor on the wall of the transformer in a position that is beyond the critical angle. If the sensor is placed in a position beyond the critical angle, the signal goes into the plane of the tank wall as a shear or transversal wave instead of directly through the tank wall. The combination of some transversal wave reaching the sensor first (the velocity of the signal through steel is faster than through the oil) and then followed shortly by the pressure wave looks like the attenuated AE burst in Figure 7. The user must always be aware that the responding characteristics of the sensor may at times be more in evidence than the forcing characteristics of the PD signal. In addition to the characteristics noted above, PD usually has a correlation to the excitation frequency waveform. In some cases there is a slight "jitter" back and forth from a constant position on the waveform.

9.2 Acoustic systems that record single events

The frequency range of a signal is between 50 kHz and 350 kHz. In [B7] the main frequency for an approximate 150pC discharge is about 100 kHz with the expectation that for larger discharges, frequencies will decrease. The amplitude and time duration depend on the physical size of the transformer tank and the location of the sensors and their actual separation from the estimated PD source. The subsequent attenuation of the signals is due to the heterogeneous structure of the transformer core/coil assembly. Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 23

IEEE PC57.127/D3.0, February 21, 2004 Attenuation of the signal inside the tank affects high frequencies more than low. The attenuation of the signal also creates a lack of correlation of scaling factor between the acoustic signal and electrically measured apparent charge. Some examples of signals are shown in the figures below.

0.3 0. 0.2 0.1 0.2E+06 0.3E+06 0.4E+06

FFT:direct PD signal 0.3 0.2 0. 0.1E-03 0.1 0. -0.1 -0.2 direct PD signal 0.3E-03 0.4E-03 0.5E-03 0.6E-03 0.7E-03 0.8E-03

Figure 8 - A laboratory recorded direct PD signal (bottom) and its power spectrum (top) For the signal in the time frame of Figure 8, the x-scale is in seconds, while the frequency frame x-scale is Hz. The sensor used has its main sensitivity in the range 20 - 120 kHz, which is clearly visible in the spectrum. Note the sharp initial rise in the time frame; maximum amplitude is reached in the first oscillation. The second, more dilute, burst at about 0.7 ms is due to reflection.
1. 0. 0.5 0.2E+06 0.3E+06

FFT:ttd1 0.5 0.7E-03 -0.6E-03 0. -0.4E-03 -0.3E-03 -0.2E-03 -0.1E-03 0. 0.1E-03 0.2E-0


Figure 9 - A laboratory recorded signal with clear signs of propagation in the tank wall, together with its power spectrum The time frame signal in Figure 9 shows a typical two-step behavior. The longitudinal wave arrives first with lower amplitude than the transversal wave that comes about 0.04 ms later. The frequency content will be slightly dependent on the wall thickness, length of wall path etc., but is mainly unchanged from a direct signal.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004


0.1 0.05



FFT noise 0. 0.5E+05 0.1E+06

0.1 0.05

FFT entire signal 0.05 -0.001 0. -0.5E-03 0. 0.5E-03

time domain

Figure 10 - A weak time domain PD signal together with power spectra for entire signal (middle) and for the noise part (top) Comparison of the two spectra in Figure 10 reveals that the PD signal main amplitude is around 50 kHz. This is because the source is under paper insulation, which attenuates higher frequencies. The signal is just slightly larger than acquisition system noise.

0.01 -0.2E-03 0. -0.1E-03 0. 0.1E-03

Signal 256 averages

0.01 -0.4E-03 0. -0.2E-03 -0.1E-03 0. 0.1E-03 0.2E-0


Figure 11 - A clear signal with indications of wall propagation (bottom). These indications are confirmed by averaging (top) Averaging can be a very effective method to reduce noise, as illustrated in Figure 11. The trigger was in this case set on the high amplitude oscillation in the acoustic signal. A stable trigger and a high sample rate are required for averaging. Several independent averages should always be performed to avoid chance coincidences. If applicable, averaging is often the most effective method to reduce noise. Note, for example, the very weak indications of a signal appearing around -0.3 ms in the upper trace. Such observations are often the key to a successful localization.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004











Figure 12 - A PD signal that has passed from the high-voltage side to the low-voltage side of a transformer The wave front in Figure 12 is not sharply rising; it grows slowly (note time scale) and starts roughly at -1.5 ms in the figure. The power spectrum shows a strong peak around 35 kHz, indicating the passage of the signal through considerable amounts of attenuating material.

9.3 On-line acoustic systems

Partial discharges obtained by conventional electrical methods use a threshold to predict severe activity. This level is generally about 300-500 pC. Because of variations in the acoustic signal caused by distance and interfering materials, there is no similar threshold for acoustic systems. A strong signal buried deep within a winding may be very weak by the time it reaches the acoustic sensor. Also, differences in amplifier gain settings cause differences in magnitude. Some commercial systems use signal length of 40 - 220 microseconds as an identification criterion for PD. For on-line acoustic systems, the two most important waveforms obtained are the signal characteristics representing the amplitude vs. time and the power spectral density. In addition, the other attributes that can be obtained are: a) Signal energy counts b) PD hits per unit time c) PD hits vs. amplitude (dB) d) Signal duration (micro seconds) vs. amplitude (dB) e) PD hits vs. power phase where phase information is available Software in these systems may take advantage of characteristics such as the PD signal always being asynchronous. The signals from sensors 1, 4 and 6 in Figure 13 do not repeat with a predictable average frequency. Such a signal is called "asynchronous." The signal for channel 1 (sensor 1) in Figure 13 shows an ellipsoid-shaped signal, which is an attenuation of a PD arrowhead signal. The attenuation indicates the signal is passing through different media before being sensed by this sensor or it indicates the location of the sensor is not optimum. The signal from sensor 4 resembles the arrowhead signature of a direct PD signal (also see Figure 8).

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

Figure 13 - PD waveforms for signal on different sensors placed on a transformer tank. The signal at sensor 6 is a developing ellipsoid-shaped signal. It is delayed, indicating that it is far away from the PD source compared to sensor 4. An initial trigger of channel 6 is caused by the signal crossing a known amplitude threshold at zero micro-seconds. The pre-trigger time is generally 20 microseconds for a sampling frequency of the A/D of 1 MHz. The maximum value of this signal has not been attained yet in the time window set for this snapshot of a total of 120 microseconds. This indicates that the travel of the signal is possibly through several media or the sensor location has caused the propagation time to last for such a long time before reaching the sensor. If the signal is synchronous, it is usually noise. The waveforms in Figure 13 from sensors 2, 3 & 5 show that the signature of these signals is repeating (periodic) with an average period that can be estimated in relation to some frequency in the bandwidth of the sensors. Such a signal with a predictable frequency is typically called synchronous. Software with a PD locator algorithm uses data collected by the system and three dimensional details of the transformer tank to generate the best estimate of the PD location. This software may use wavelet transform methods to estimate the PD location (Annex C).

9.4 General direct current systems

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

9.5 Characteristics of PD from static electrification

The failures of a number of large power transformers have been attributed to accumulation of charge and subsequent flashover discharge. Field tests have indicated a correlation of the charge buildup, accompanied by an increasing level of PD activity, with oil temperatures below approximately 40 C and relatively high oil velocity [B12, B37]. This phenomenon is often called static electrification. DGAs taken during the field tests [B37] did not indicate any increase in levels of hydrogen or other gases. This was probably because of the relatively short period of time and sporadic occurrences of the static electrification. External sensor placement is mostly toward the top of the transformer using care to avoid internal shielding. However, the high energy discharges of advanced static electrification can be detected by sensors located at almost any location clear of shielding. And these discharges can sometimes be heard as loud bangs by persons standing near the transformer. Static electrification discharges can be detected by acoustic methods well before they become damaging [B16]. Field test experiences [B16, B42] have shown that the most critical situations where static electrification can occur is during start-up or lightly loaded operation in cold weather. Acoustic activity due to static electrification is gradually reduced as oil temperature is increased. It practically stops after reaching 50 C.

9.6 Acoustic activity from thermal faults, the core, mechanical noises and other sources
Acoustic emission techniques have typically been used only for the detection and location of PD/arcing in power transformers when an indication of a problem is pointed out by other techniques (usually gas-in-oil). However, it has been observed through the years that transformers without gassing history and/or no indication of an electrical problem produce acoustic activity when monitored over a fixed period of time [B42]. These signals must be differentiated from PD signals for the results to be effective. Some examples and their characteristics follow. a) Core magnetostriction noise (Barkhausen effect): harmonics of this noise are often in the 30-40 kHz frequency range. b) Pumped liquid noise: the discharge from oil circulating pumps may interfere with PD signals with transducers placed low on the transformer tank. They usually have no correlation with the 50 or 60 Hz waveform. c) Loose nameplates, pipes hitting each other, fan noise, etc. may sometimes have repeatable waveforms close to PD signals. The length of the signal is often much longer than the PD signal, which is characteristically under 150s. d) Loose shielding connections in the transformer tank may cause large PD indications, but not be detrimental to transformer operation. The location of the signals must be considered and correlated with transformer design drawings. e) Wiring from the sensors to the amplifiers, if not properly shielded, will pick up spurious PD signals. Look at Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 28

IEEE PC57.127/D3.0, February 21, 2004 every possible source before concluding there are major problems inside the transformer. f) Trucks, environmental noises (thunderstorms, rain, snow, hail, wind, etc.), and miscellaneous events near the transformer tank may activate the acoustic sensors. These are random signals lasting longer than one ms. g) Switching and load tap changer movement are random signals. h) Thermal faults may cause random signals [B15, B36] The correlation of additional parameters of the transformer during the test (current, voltage, pump/fan current, oil temperature, winding temperature, phase angle, etc) with the acoustic emission data may assist in identification and separation of the different acoustic sources detected during a test [B35]. If possible, a correlation with electrical PD detection can aid in distinguishing between PD and noise signals.

9.7 Comparison between electrical and acoustic signals

A partial discharge exhibits, besides other phenomena, a fast transient electrical pulse and an acoustic "bang". Depending of the location of the PD and the coupling path between the event and the detector, the electric or acoustic signal can be used to detect the PD. Both methods have different detection ways and sensitivities for unwanted signals (noise). The acoustic PD detection is restricted to events within the line-of-sight of the acoustic transducers. This limits the detection range, but also the amount of noise. The electric PD detection covers a wider area, including e.g. bushing and tap changer. External noise will also be detected and is difficult to remove. The correlation between instrument reading and actual discharge magnitude is better than with the acoustic method. Several international standards exist that define the instrument response, which is the readout in pico-Coulomb or micro-Volt, allowing a better comparison between manufacturer and in-field measurements. The following table gives a rough overview. Electrical Acoustic Remark Detection Detection PD generated within the transformer PD on the outside of the winding Yes Yes Best use for acoustic detector, location (triangulation) PD within the winding Yes unlikely Strong acoustic attenuation inside the winding. RIV-detector signal affected by circuit L-C resonance PD between winding and core Yes difficult acoustic signal reflection at the core required Arcing / tracking of the oil surface Yes Yes Arcing / tracking of the bushing Yes Yes surface in the oil PD in the bushing Yes possible )3 PD in the no-load tap changer Yes Yes PD in the on-load tap changer Yes Yes )4 sensor placement on LTC compartment Noise sources PD outside the transformer, (e.g. Yes No )1 corona on busbar, PD in switchgear) Strong tracking / leader discharge Yes possible )1, )2 Nearby lightning strokes Yes possible )1, )2 unlikely single events Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 29 Source

IEEE PC57.127/D3.0, February 21, 2004 Nearby car ignition Switched electronic power supplies Radio stations, transmitter Weather (rain, sleet, snow, hail) Moisture / Degradation in the oil Yes Yes Yes No No No No )1 no relation to phase angle )1 Electrical noise filter (band-stop) or shift in RIV frequency required

Yes Other events No No

Other diagnostic methods have to supplement the PD detection (PF, DGA etc.)

)1 Successful electrical PD detection requires noise suppression (e.g. gating together with an external noise antenna) and observation of the PD - phase pattern (fingerprint) )2 Very strong electrical signals can couple into the transducer or amplifier if the detector is not sufficiently screened )3 Field tests have proven (B32) that acoustic detection of PD in the bushing can be achieved if acoustic sensors are placed on the lower part of the bushing, close to the capacitance tap. See SAFETY WARNING Subclause 1.3. )4 Several cases have been reported (B31, B41) where PD/Arcing and/or tracking has been detected inside LTC compartments, either on the insulating support bars or on the diverter switch cylinder.

10. Integrating AE Results with Data from Oil Analysis

Although AE data is useful in its own right, it becomes even more so when used in conjunction with dissolved gasin-oil data. For example, if the problem has been present for some time, as is typical of situations that develop in the field, good correlation should be expected between gas analysis and AE data. As with any diagnostic tool, a baseline reading should be established before or at the start of of monitoring AE signals. The increase and rate of increase of combustible gases from the baseline dissolved gas analysis (DGA) reading, in combination with events that occur on the transformer, then become relevant diagnostic data. Partial discharge PD and/or arcing in the presence of oil produce hydrogen (H2) and other gases. PD and the generation of H2 due to PD, are the earliest (first) warning signals of most incipient faults known to the industry. If DGAs show a continuing increase of H2 in combination with AE signals emanating from a source determined to be in a specific area, there is a good possibility that the partial discharge is taking place in the specific area. Refer to IEEE Standard C57.104 Guide for the Interpretation of Gasses Generated in Oil Immersed Transformers for more information. The two sources of data can often supplement each other in yet other ways. For instance, sometimes the breakdown of constituents in the gas analysis is so complex that, although it is obvious that a significant problem is involved, it is not possible to determine whether the cause is due to partial discharges or is thermal in origin. The AE system responds to internal signals produced by partial discharges, arcs or gas bubbles. Thermal phenomena that are below the threshold of causing gas bubbles can produce acoustic signals [B15], but with different characteristics than the signals produced by partial discharge. Therefore, the existence of any AE signal having the characteristics associated with partial discharge together with the continuing increase or acceleration of hydrogen gas, verify the existence of PD in oil. The absence of acoustic emission activity in this case may indicate that the problem is located in an area where the signal is highly attenuated. If AE signals continue and/or increase, with no further increase or acceleration of H2 or other combustible gases, the AE signals may be due to components external to the transformer tank. As the combination of information produced by these two techniques is so advantageous, it is particularly recommended that gas analysis results be taken into account when interpreting AE data. When evaluating transformers on the shop floor, the same good correlation between dissolved gas and partial discharge is normally obtained after HV dielectric tests and temperature test. Acoustic emission technique provides essentially real time data relative to activity occurring at that instant. Oil analysis on the other hand is to some extent historical in nature. Depending on the location of the source and the activity level of PD, there may be a delay of seconds to much longer time periods before gassing is observed. It cannot be assumed that because there is gas there is PD and vice versa. Normally, "RIV" or "Apparent Charge" detection is carried out in the factory and this provides Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004 an alternate database for correlation with acoustic emission data in the signal interpretation process.

11. Acoustic Activity Interpretation

In the laboratory it may be possible to correlate AE characteristics with electrical partial discharge (pico-Coulomb) levels. However, this correlation can not be used in the field due to the external disturbances and effects of attenuation in both the acoustic and electrical signals. This means that no absolute value of partial discharge activity can be determined from AE measurements made in the field. It is important to verify whether the acoustic signal is due to internal partial discharges or if it is caused by some other source such as mechanical noise. To make this determination requires expertise of the investigator and other evidences such as the presence of indicating combustible gasses or electrical PD. Data processing technology may also be helpful. The classification is performed by separating the acoustic signals based on the particular characteristics of each individual source (amplitude, duration, energy, counts, etc.). In general, a more intense PD source will typically produce a higher AE magnitude and count rate than a weak source. This is because at the site of an intense discharge there may be multiple locations or perturbations that are producing higher energy PD and AE. It is necessary to consider the transformer design when making acoustic measurements on transformers. a) The locations of the partial discharges that are likely to lead to failure may be in areas where the attenuation of the PD signal is great. These would include areas within the windings and in the high-low spaces. The attenuation is caused by the acoustic signal not having a clear path out of the winding, thicker tank walls, tank wall shielding, the presence of more insulation barriers, directed oil wraps, etc. A source that is likely to lead to failure may be attenuated to the point that the PD count rate is low or nonexistent. Since the value of many power transformers is high and the cost of a catastrophic failure is great, the detection of any internal partial discharges in such transformers should be a cause for further investigation. This might include close monitoring of the behavior of the discharge with time, more frequent samples of oil for combustible gas measurements, and other advanced diagnostic measurements. b) Some transformers do not have tank wall shielding or directed flow barriers and have tank wall thickness 6 mm or less. A given partial discharge will probably produce more acoustic energy at the transducer location than a transformer with 10 mm tank wall thickness and tank wall shielding. c) The acoustic energy at the sensor reflects a combination of magnitude and repetition rate of the PD, the distance it is away from the transducer and the effect of intervening materials. The AE characteristic that warrants the need for further investigation varies depending on the size of transformer as well as the magnitude and type of PD. The data obtained from acoustic tests is not sufficiently definitive to warrant its use as either acceptance or go/no-go criteria and should not be used for such. By taking into account the type of signal obtained, the approximate location of the emitting source and an estimate of the level of the activity involved, acoustic measurements can be used as a means of identifying potential partial discharge problems. The acoustic measurement alone does not provide an estimate of the severity of the problem. A database and grading system based on case histories of similar problems may help with this analysis. However, it can indicate the need for other diagnostic measures which, when combined with the acoustic data, will often provide the means for identifying the cause and severity of the problem.

Annex A (Informative)
Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change 31

IEEE PC57.127/D3.0, February 21, 2004

Since the early 1950s, there has been much activity in the area of ultrasonic/acoustic emission detection of partial discharges. The following bibliography is intended to give a broad overview of the subject and provide references for further study. [B1] "Acoustic Emission Detection of Partial Discharges in Power Transformers," EPRI final report no. EL-4009, 1985. [B2] Anderson, J.G., "Ultrasonic Detection and Location in Insulating Structures," AIEE Transactions No. 75, 1956. [B3] Application Note AN23, "Table of Ultrasonic Properties," Metrotek Inc., Richland WA. [B4] ASTM Standard E 650-85 "Standard guide for mounting piezoelectric acoustic emission sensors." [B5] Bartnikas and McMahon, "Engineering Dielectrics," Vol. 1, Chapter 10, by R.T. Harrold, ASTM Publication, STP669-1979. [B6] Bengtsson, T. and Jnsson, B., "Transformer PD Diagnosis using Acoustic Emission Technique," ISH-97, paper no. 115. [B7] Bengtsson, T., Leijon, M. and Ming, L., "Acoustic Frequencies Emitted by Partial Discharges in Oil," ISH-93, paper no. 63.10 [B8] Berent, D., "Acoustic Monitoring and Gas-In-Oil Analysis for Transformers," Report #62PAIC95, pp 83.183.6, Doble Engineering Company. [B9] Boczar, T., "Identification of a Specific Type of PD from Acoustic Emission Frequency Spectra," IEEE Transactions on Dielectrics and Electrical Insulation, pp 598-606, August 2001, New York, NY. [B10] Boguslaw, J. "Utilization of Acoustic Emission for Detection, Measurement, and Location of Partial Discharges," AEWG-Second International Conference on Acoustic Emission, Lake Tahoe, 1985. [B11] Carpenter, J.H., Kresge, J. S. and Musick, C. B., "Ultrasonic Corona Detection," IEEE Transactions PAS-86, No. 8, 1965. [B12] Dahlgren, D. and Harley, J.W., "Implementing a Static Electrification Diagnostic Model," EPRI Symposium on Static Electrification, May 19-21, 1999, Monterey, CA. [B13] Edmonds, P.D., "Methods of Experimental Physics," Vol. 19 (Ultrasonics), pages 18-19, Academic Press, New York, 1981. [B14] Eleftherion, P.M., "Partial Discharge XXI: Acoustic Emission-Based PD Source Location in Transformers," Proceedings IEEE Electrical Insulation Magazine, 08837554/95,1995. [B15] EPRI Report "Development of a New Acoustic Emissions Technique for the Detection and Location of Gassing Sources in Power Transformers: Phase 1 Results," EPRI, Palo Alto, CA, Allegheny Power, American Electric Power, Bonneville Power Administration, Consolidated Edison Company of New York, Inc., Electricite de France, Exelon Corporation, and Public Service Electric and Gas Company: 2002, Product ID 1007176. [B16] EPRI Report TR-111386 "RAMAPO static electrification tests: EPRI and Consolidated Edison Company of New York." [B17] Fredrick, J.R., "Ultrasonic Engineering," John Wiley & Sons, New York, 1965, pp. 13-21. [B18] Harold, R.T., "The Relationship between Ultrasonic and Electrical Measurements of Under-Oil Corona Sources," IEEE Transactions, Vol. EI-11, pages 8-11, March 1976. [B19] Harrold, R.T., "Acoustic Waveguides for Sensing and Locating Electrical Discharges in High Voltage Power Transformers and Other Apparatus," IEEE Trans. On PAS, Vol. PAS-98 No. 2, April 1979, pp. 449-457. [B20] Harrold, R.T., "Acoustical Properties of Insulating Liquids and Gases," IEEE International Symposium of Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004 Electrical Insulation, Philadelphia, 1978. [B21] Harrold, R.T., IEEE transactions on Electrical Insulation, Vol. E.I.10. No. 4, December 1975. [B22] Heroux, P., "A Statistical Study of Electrical and Acoustical Characteristics of Pulsative Corona," IEEE paper AT6 122-2, Winter Power Meeting 1976. [B23] Herzfeld, K. F. and Litovitz, T.A., "Absorption and Dispersion of Ultrasonic Waves," Academic Press, New York, pg. 454, 1959. [B24] Howells, E. and Norton, E. T., "Parameters Affecting the Velocity of Sound in Transformer Oil," IEEE Transactions, Vol. PAS-103, pages 1111-1115, May 1984. [B25] Howells, E. and Norton, E., "Detection of Partial Discharges in Transformers Using Acoustic Emission Techniques," IEEE Transactions, PAS 97, No. 5, 1978. [B26] Howells, E., "Acoustic Emission from Stressed Dielectric Liquids," ISA paper no. 74219 Instrumentation Symposium, Albuquerque, 1974. [B27] IEEE C57.113-1991, IEEE Guide for Partial Discharge Measurement in Liquid Filled Power Transformers and Shunt Reactors. [B28] Kawanda, H., Honda, M., Inoe, T. and Amemya, T., "Partial Discharge Automatic Monitor for Oil-Filled Power Transformers," IEEE paper no. 833M424-8, PES Summer Meeting 1983. [B29] Kim, T. Y., Suh, K. S., Nam, J. H. and Takada, T., "Acoustic Monitoring of HV Equipment with Optical Fiber Sensors," IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 10, No. 2, pp. 266-270, 2003. [B30] Krautkramer, "Ultrasonic Testing of Materials," Springer-Verlag, New York Inc., 1969, pp. 20-23. [B31] Lin, R., Nez, A. and Jimnez L., "On-line ultrasonic diagnosis to prevent catastrophic failures in power transformers," EPRI 2000 Substation Equipment Diagnosis Conference VIII, February 20-23, New Orleans, Louisiana. [B32] Lopez, A., Equihua, H., Lin, R. Alvarez, R. and Nez, A., "Detection of incipient failures on power transformers using on-line non-intrusive diagnosis techniques," IEEE XI Power summer meeting 1998, Acapulco, Mexico. [B33] Lundgaard, L. E., "Partial Discharge-Part XIV: Acoustic Partial Discharge Detection-Practical Application," IEEE Electrical Insulation Magazine, Vol. 8, No. 5, pp. 34-43, 1992. [B34] Meunier, R. and Vaillancourt, G.H., "Propagation behaviour of acoustic partial discharge signals in oil-filled transformers," Conference Record of the International Conference on Conduction and Breakdown in Dielectric Liquids, July 15-19, 1996, Roma, Italy. [B35] Miller, R.K. and Nez, A., "Detect and locate sources of Power Transformers deterioration using high speed Acoustic Emission waveform acquisition with location and pattern recognition," EPRI 2002 Substation Equipment Diagnosis Conference X, February 17-20, San Antonio, Texas. [B36] Miller, R.K., Shu, F., Nez, A. and Ternowcheck, S., "Advances in Acoustic Emission Testing for Detecting, Locating and Assessing Electrical and Thermal Faults," EPRI 2003 Substation Equipment Diagnosis Conference X, February 23-26, New Orleans, Louisiana. [B37] Moore, H.R, Savio, L.J., Chu, D. and Sandstrom, J.E., "Report on Studies of Static Electrification performed on a Service-Aged Transformer," Sixty-Third International Conference of Doble Clients, March 25-29, 1996, Boston MA. [B38] Morel, J.F. and Fallow, B., "Detection and Location of Partial Discharges by Ultrasonics," ERA Transactions 2947, Rev. General Electric, Vol. 80, pages 225-228, 1971. [B39] Noro, A., Nakamura, K., Horii, K., Katsukawa, H. and Matsuoka, R., "Acoustic-Based Real-Time Fault Location in Power Substation," Third International Symposium on Electricity Distribution and Energy Management, ISEDEM Proceedings, vol., pp. 106111, 1993. [B40] Noro, A., Nakamura, K., Watanabe, T. and Morita, T., "Acoustic-Based Real-Time Partial Discharge Location in Model Transformer," Proceedings ICSPAT 1994, pp. 10771082.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004 [B41] Nez, A., "Acoustic Technique for the detection and location of partial discharge in power transformers," Thesis, Instituto Politecnico Nacional-Instituto de Investigaciones Electricas, Mexico, 1999. [B42] Nez, A., Miller, R. K. and Ward, B., "Reduction of failure risk in power transformers through the detection and location of incipient faults using acoustic emission," TechCon 2003 North America, February 3-4, 2003, St. Petersburg, Florida, USA. [B43] "Partial Discharge Detection Using Acoustic Emission Report," Physical Acoustics Corporation, P.O. Box 3135, Princeton, NJ 08543. [B44] PC57-127-2000, Trial Use Guide for the Detection of Acoustic Emissions from Partial Discharges in Oil Immersed Power Transformers. [B45] Phung, B. T., James, R. E., Blackburn, T. R. and Su, Q, "Partial Discharge Ultrasonic Wave Propagation in Steel Transformer Tanks," Proceedings of 7th ISH, Dresden, Germany, Vol. 7, pp. 131-134, 1991. [B46] Reason, John, "Online Transformer Monitoring," Special Report, Electrical World, pp 19-26, October 1995. [B47] Rossi, M., "Acoustics and Electroacoustics," translated by Patrick Rupert, Windsor Roe, Artech House, 1988. [B48] Shertukde, H. M. and Bar-Shalom, Y., "Target Parameter Estimation in the near field with two sensors," IEEE Transactions on Acoustic Speech and Signal Processing, Aug., 1988. [B49] Shertukde, H.M. and Shertukde, R.H., "Monitoring vs. Diagnostics- FAN-EPT: A Novel Tool for On-Line Monitoring and Diagnostics of Large Power Transformers," Proceedings IEEE T&D Conference, Japan, October 7, 2002. [B50] Shertukde, H.M., Alnajjar, H. and Prabhune, U., "Fault Detection Device For Electrical Power Transformers Using Novel DSP Scheme," Proceedings ICPAT '96, pp 1219-1223, October 7-10, 1996, Boston, MA. [B51] Shertukde, R.H. and Shertukde, H.M., "Manufacture of Fault Diagnostic Device for Electrical Power Transformers (FD2EPT)," Proceedings ICSPAT'96, pp 1229-1233, October 7-10, 1996, Boston, MA. [B52] Titto, S., Olata, M. and Saynajakangas, S., "Non-Destructive Magnetic Measurements of Steel Grain Size," Non-Destructive Testing Journal, England, 1976. [B53] Train, D., Mercier, A. and Thorne, D., "The Detection of Partial Discharges in High Voltage Potential Transformers in Service," IEEE Transactions, Vol. PAS-93, pages 1909-1916, November 1974. [B54] Van Brunt, R. J., "Stochastic Properties of Partial-discharge Phenomena," IEEE Transactions on Electrical Insulation, Vol. 26, No. 5, October 1991. [B55] Vohl, P. E., Gervais, Y. and Mukhedkar, D., "Model Analysis of Pulses Generated by Partial Discharges," IEEE paper no. A76-4168, PES Summer Meeting, 1976. [B56] Von Glahn, P., Stricklett, K.L,, Van Brunt, R. J. and Cheim, L. A. V., "Correlation Between Electrical and Acoustic Detection of Partial Discharge in Liquids and Implications for Continuous Data Recording," IEEE ISEI, Vol. 1, pp. 69-74, 1996. [B57] Wood, J.W., Hickling, G.H., Hindmarch, R.T. and Raju, B.P., "Electrical and Ultrasonic Characteristics of Partial Discharges in Oil Immersed Insulation," IEEE Conference on Dielectric Materials, Measurements and Applications, Cambridge, England, 1975. [B58] Young, R.K., "Wavelet Theory and Its Applications," Kluwer Academic Publishers, MA 1994.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

Annex B (Informative)
Signal Processing
To evaluate the source of a PD signal, it is necessary to establish its physical disposition or location within the confines of the transformer. A technique [B43] that is sometimes utilized is based on the presumption that the sensor location resulting in the highest AE oscillation or count rate would be closest to the source. This is not always true because of the attenuation that can be caused due to passage of the sound wave through different insulation materials and copper. Even under most favorable circumstances this technique can only indicate the general zone wherein the source may probably lie. For more accurate location more precise solutions need to be used. Some of the techniques are: a) Time domain: finding the start of a candidate signal in the time domain is the simplest and most reliable approach. b) Cross-Correlation: this requires two channels of data from two acoustic sensor locations, with the data in one channel artificially delayed with respect to the other channel of data. One then attempts to evaluate the Cross-Correlation of these data by multiplying the corresponding data points. When the artificial time delay is equivalent to the real time delay between the two data channels a maximum occurs which indicates the possibility of source location after corresponding mathematical manipulations are conducted. This method is hampered by the necessarily different wave forms recorded at different locations. c) Signal Enhancement: averaging is a very powerful technique for recovering repetitive signals from random noise when there is a stable trigger. In this method, randomness in the signal is reduced to almost zero amplitude while the true repetitive portion of the signal remains unaffected. However, as AE signals vary in amplitude, the degree of enhancement obtained is usually less than that predicted by theory. d) Fast Fourier Transform: the signal can be analyzed in the frequency domain by using suitable Fourier techniques [B39, B48]. This can result in corresponding power spectral densities which can be manipulated mathematically to get the estimate of the time delay and then correspondingly the source location. e) Wavelet Transform: the signal can be analyzed in the frequency as well as the time domain using suitable wavelet transform techniques [B50, B51, B57]. This can result in corresponding wide band cross ambiguity functions with better estimation of the time delay and correspondingly better source location accuracy. These estimated time delays in conjunction with proper signal characteristics (amplitude, signalto-noise ratio and shape) yield better PD source location accuracy. Use of suitable location algorithms is illustrated in Annex C.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

Annex C (Informative)
Wavelet Signal Processing Theory
Incipient faults generally occur due to the breakdown of a small part of the transformer insulation, generally caused by the inclusion of moisture or a cavity. This activity results in an instantaneous shunting of a small partial capacitance of the entire insulation. The discharge of this partial capacitance reconfigures the distribution of charges within the insulation instantaneously causing a short current pulse. This activity can continue in a sequence for a while setting up a train of pulses. These pulses can have a particular shape, size and frequency depending upon the spatial arrangement of the inclusion. In the following analysis, wavelet transform techniques are used with data from external sensors to estimate the spatial location of the incipient fault.
sensor 2 r2 (t) sensor 1 r1 (t)

sensor3 r3 (t)

3, s3

1, s1


0 , s0
sensor 0 r0 (t)

sensor i ri(t)

n, sn
sensor n rn (t)

Figure 14 PD source with sensors located on the external surface of the transformer tank Let the PD source be located as shown in Figure 14. The different sensors are located at different positions on the transformer tank. Let ri(t) be the measurements received at the n sensors consisting of the signal corrupted by Gaussian noise, independent and identically distributed, with mean zero and variance

ri (t ) =

t i + ni (t ) x si si = xi (t ) + ni (t )



x (t ) = e t e



Then the wavelet transform of the reference measurement received at sensor 0 and the measurement at sensor i is given by

Wr0 ri ( s, ) = Wx0 xi ( s, ) =

1 t ri (t )r0 dt s s 1 s


x (t ) x
i 0

t dt s


Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

Wx0 xi ( s, ) =

0 i

s + s 0 i Wx x s 0 , si si s0 si


Wx x will have a maximum value at (s,) = (1, 0), equation (5) will have a peak when
s= si si and = i s0 s0 0

t 0 t i and x 0 ( t ) = x xi ( t ) = x si s0
If then

(7) (8) (9)

t = t 0
t i = t ( i 0 )


t ( i 0 ) xi ( t ) = xi ( t + 0 ) = x si
and x 0 ( t ) = x 0 t + 0 = x


t s0


Now x 0 ( t ) and xi ( t ) are scaled and delayed versions of x( t ) . In the new time domain, using the scale/delay property of WT, equation 5 will have a peak when


si s0


= i 0


sensor 1 (x1,y1,z1)

sensor 2 (x2,y2,z2) z y

(x,y,z) PD

sensor 0 (x0,y0,z0)

sensor n (xn,yn,zn) x


Figure 15 Representation of 3-D transformer tank with sensor locations for calculation of PD source location coordinates

If the speed of sound in oil is C feet / sec, then, in Figure 15,

( x x i ) 2 + ( y y i ) 2 + ( z zi ) 2 = C 2 . i 2
From the above equation the location of the PD source (x, y, z) can be easily estimated.


Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

Annex D (informative)
Calibration of equipment external to transducer and pre-amplifier
The circuitry in the instrument external to the sensor is required to perform the following functions: a) Provide 60 dB of gain to the incoming signal b) Severely attenuate signals outside the frequency range of 100300 kHz c) Produce a 3:1 signal-to-noise ratio for actuation of the pulse counting process The procedure below is designed to verify satisfactory operation.

Equipment required
The following equipment is required: a) Signal generator (with low-output voltage capability) b) Frequency counter c) Precision voltmeter with high-frequency (500 kHz) response capability and preferably peak voltage indication

Connect the signal generator to the input of the PD detector. Connect the voltmeter across this input so that it monitors the voltage level of the applied signal. Connect the counter to the output of the signal generator to monitor the frequency of its output. Completion of these checks ensures satisfactory performance of the power amplifier, filter, and signal-to-noise ratio discriminator and counter circuits. a) Adjust the signal generator to supply a 150 kHz sinusoidal signal with peak amplitude of 1 mV b) Set the power amplifier gain to 60 dB 1 dB c) Set the pulse counter circuitry trigger level to 1 V. The instrument display should indicate an oscillation count rate of 150,000 pulses/s d) Raise the instrument counter trigger level to 3.00 V. The counter should now indicate 0. e) Slowly increase the amplitude of the applied signal (still at 150 kHz), noting the voltage necessary to activate the counter circuit and result in a 150,000 pulses/s display. This amplitude should be no less than 3.00 mV and no more than 3.20 mV. f) With the system setup as previously described (3.00 V counting level; 60 dB power amplifier gain; and 3 mV, 150 kHz input), reduce the frequency of the input signal to 100 kHz. The counter should now indicate 0. g) With the same setup, increase the frequency of the input signal to 300 kHz. The counter should again indicate 0. h) Without changing the instrumentation configuration, slowly vary the input signal (at constant 3 mV peak amplitude) over the range of 100300 kHz. Correct indication should only be obtained when the frequency of the input signal is between 120 kHz and 280 kHz.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004

Annex E (informative)
Calibration of transducer and preamplifier
The sensor utilizes a piezo-electric crystal with a nominal resonant frequency of 150 kHz when excited by a longitudinal waveform. The output from the crystal is a very low-amplitude, high-impedance signal that requires processing in a preamplifier before it is useful. The transducer element and preamplifier is to be considered as a complete system, whether they are contained in the same package (as in the preferred shielded single crystal sensor) or as separate units.

Instrumentation required
a) Heavily damped, 5 MHz, ultrasonic, non-destructive testing immersion transducer b) Transducer excitation pulse circuitThis is required to provide a positive going pulse, achieving a peak amplitude of 300 V in 500 ns, and decaying to 0 amplitude in 3 s. The circuit is required to work into a high-impedance load and have a pulse repetition rate of approximately 1 kHz. c) Transient recorderA digital oscilloscope with a 500 ns sampling rate is recommended. d) Spectrum analyzerThis should be capable of analyzing transients and processing signals with a frequency content up to at least 300 kHz. e) Ultrasonic immersion test tank f) Appropriate preamplifier power supply

The intent of this procedure is to determine the output of the PD sensor/preamplifier combination to a well-defined longitudinal mechanical pulse. To achieve this, the pulsing circuit excites the ultrasonic (driving) transducer so that it outputs a well-defined mechanical pulse. Having been submerged in water, this pulse propagates exclusively in the longitudinal mode and subsequently excites the sensor being evaluated. The output of the sensor/preamplifier combination is then supplied to a transient recorder where the time-domain record is obtained. At the same time, the signal is supplied to a spectrum analyzer for frequency analysis. a) Connect pulser circuit to ultrasonic (transmitting) transducer b) Connect appropriate power supply to sensor/preamplifier combination c) Connect sensor/preamplifier output to transient recorder and in parallel to the spectrum analyzer d) Set transient recorder sampling period to approximately 500 ns and transient capture trigger level to approximately 2 V e) Set spectrum analyzer in the transient analysis mode and select a frequency range that embraces at least 0250 kHz f) Immerse both driving transducer and discharge detector sensor in the water-filled immersion tank. Ensure that they face each other squarely and are 1617 cm apart. The transducers should be located at least 8 cm away from any reflecting objects such as the tank walls. It is also important to ensure that no bubbles adhere to the face of either the transmitting transducer or sensor.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change


IEEE PC57.127/D3.0, February 21, 2004 g) Energize transducer excitation pulse circuit and preamplifier power supply. The time-domain signal displayed by the transient recorder should be that of a "burst" made up of many oscillations. The leading oscillations should be high in amplitude with the remainder decaying to zero, similar to that shown in Figure 1. The requirements are that the maximum peak-to-peak voltage be no less than 5.80 V and no higher than 6.20 V. The duration of the burst should be no less than 80 s and no longer than 150 s. (Note: if the duration is longer than 150 s, then it is likely to be mechanical noise rather than partial discharge noise.) To avoid the confusing effects of random noise, it is recommended that the spectrum be enhanced by averaging at least eight separate spectra. The resulting spectrum should show a dominant peak between 120 kHz and 160 kHz. The resonant characteristic of the crystal should be evident by the amplitude of this peak being at least 40 dB and no more than 43 dB above the spectrum reference level. In meeting these criteria, the sensor is shown to have a lightly damped crystal of the correct resonant frequency and the preamplifier is producing the required 40 dB of gain.
Notethe foregoing procedure requires that the PD detection sensor be completely immersed in water. If a sensor that is not suitable for total immersion is used, the same result can be obtained by utilizing a vertical water column. In this case, the driving transducer is located about 8 cm from the bottom of the column, while only the face of the sensing transducer is required to enter the water surface. If this approach is used, the same precautions relative to reflections from the tank sides, avoidance of bubbles, and separation distance between the sensors are still appropriate.

Copyright 2004 IEEE. All rights reserved This is an unapproved IEEE Standard subject to change