861 Improvements To PD Measurements For Factory and Site Acceptance Tests of Power Transformers

A2/D1
Technical Brochure
Improvements to PD measurements
for factory and site acceptance tests
of power transformers
Reference: 861
February 2022
TECHNICAL BROCHURE
Improvements to PD measurements
for factory and site acceptance
tests of power transformers
JWG A2/D1.51
Members
S. COENEN, Convenor DE T. LINN CH

S. MARKALOUS, Secretary DE P. MRAZ CH
M. BELTLE DE A. NADERIAN CA
R. CSELKÓ HU V. SCHMIDT DE
P. FEHLMANN CH R. SCHWARZ AT
J. FUHR † CH M. SIEGEL DE
R. GATECHOMPOL TH M. SÖLLER DE
M. HÄSSIG CH J. SZCZECHOWSKI US
S. HOEK DE D. TABAKOVIC US
M. JUDD GB S. TENBOHLEN DE
U. KEMPF DE M. WEBER DE
Copyright © 2022
“All rights to this Technical Brochure are retained by CIGRE. It is strictly prohibited to reproduce or provide this publication in any
form or by any means to any third party. Only CIGRE Collective Members companies are allowed to store their copy on their
internal intranet or other company network provided access is restricted to their own employees. No part of this publication may
be reproduced or utilized without permission from CIGRE”.
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to the
accuracy or exhaustiveness of the information. All implied warranties and conditions are excluded to the maximum extent permitted
by law”.
WG XX.XXpany network provided access is restricted to their own employees. No part of this publication may be
reproduced or utilized without permission from CIGRE”.
ISBN : 978-2-85873-566-2
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any
TB 861 - Improvements to PD measurements for factory and site acceptance tests of power transformers
Executive summary
This Technical Brochure (TB) deals with possible improvements to partial discharge (PD) measurements
for factory acceptance test (FAT) and site acceptance tests (SAT) of power transformers. In the joint
working group JWG A2/D1.51, it was generally agreed, that a sensitive PD measurement is in the
common interest of the transformer manufacturer as well as the buyer and end user of the transformer.
Hence, the main task of PD measurements during FAT or SAT is to evaluate the quality of the insulation
of a transformer and to minimize the risk of failures during operation. In this regard, it is neither the
intention of this brochure to adopt acceptance criteria to allow easier passing of FAT/SAT, nor to tighten
the acceptance criteria by introducing an additional measurement method without a clear indication of
critical PD activity.
The main improvement for FAT and SAT acceptance tests provided by this brochure is the possibility to
provide repeatability and comparability for UHF PD measurements using systems from different
manufacturers. As with conventional electrical measurements to IEC 60270, these requirements must
be met for the test method to be standardized. It must be mentioned that the findings in this brochure
do not aim to establish any correlation between the PD level of electrical measurements according to
IEC 60270 (apparent charge pC) and UHF PD measurements (electric field V/m).
To achieve comparability among UHF PD measurements performed using different sensors and
measurement instruments, the brochure sets out in detail a general UHF calibration procedure with
several steps that should be performed. By using this defined calibration procedure, UHF PD
measurements can be applied more extensively in the future since reproducibility and comparability will
be ensured.
The brochure contains clear guidelines about:
- Usable types of UHF sensors and their general advantages, disadvantages, and characterization
- Number and location used for UHF sensors at transformer tanks
- Frequency range used for UHF PD measurements
- Calibration of the UHF PD measuring system and necessary documentation
- Improving signal to noise ration through knowledge of the most common noise sources
- Performance check procedure of the UHF measurement setup
Outcomes from collected case studies and documented measurement campaigns are included to
demonstrate the traceability of the findings and help readers to perform effective UHF PD measurements
themselves.
The brochure documents development steps for measuring instruments that will ensure effective
implementation of the proposed calibration procedure.
The brochure ends with a collection of proposals for further research and a discussion concerning the
definition of UHF thresholds that might be used as acceptance criteria for FAT and SAT in future once
more experience has been gained.
3
In Memoriam
Dr.-Ing. Jitka Fuhr
Dr.-Ing. Jitka Fuhr Aschwanden passed away
suddenly in a tragic car accident on July 26,
2020. She began to study electrical engineering
in Prague in the 1967–1968 academic year. In
1968, during the Prague Spring, Jitka was on a
leave in England, when the Russians occupied
CSSR. On her way back home from England to
CSSR, she decided to stay in Germany and to
continue her studies at the Technical University
Aachen. In December 1973 she graduated from
the Technical University ‘Fridericana’ in
Karlsruhe with a master's degree (Dipl-Ing.) in
electrical engineering, specializing in high
voltage technology. From 1974 to 1980, she
worked as a development and design engineer
at Brown Boveri Company (BBC) in the
transformer factory in Mannheim. In this first
job at BBC she became familiar with the design
and testing of large power transformers. From
1980 to 1985, Jitka was a scientific staff
member at Hahn-Meitner-Institute for Nuclear
Research (which became the Helmholtz-
Zentrum in 2009) in Berlin-West, where she
investigated the electrical breakdown
mechanism in dielectric liquids. In December
1985 she was awarded the title Dr.-Ing. (PhD) from the Technical University of Darmstadt with her
thesis “Experimental investigation on the time-dependency of electrical breakdown in liquid
hydrocarbons”.
From 1985 to 1988, Dr. Jitka Fuhr was a research associate at the Polytechnic University of New York,
where she taught graduate students in the fields of electro-physics and high voltage engineering and
was also an advisor to PhD students in the high voltage laboratory. Returning to Europe in 1988, Jitka
joined Tettex Instruments in Zürich, Switzerland, a manufacturer of high precision measuring
instruments for high voltage laboratories and dielectric testing of insulating materials. Here, she was
responsible for the development of new applications of partial discharge (PD) measuring systems. In
1989 Jitka Fuhr entered the ABB Corporate Research Center in Baden-Dättwil (Switzerland), where she
was a leader of the project “PD-diagnosis of electrical power equipment.” In 1995 she moved to the
ABB transformer factory in Geneva (ABB Sécheron), where she was responsible for the development
and application of advanced diagnostics methods (“fingerprints”) for condition-based maintenance of
power transformers. In 2001 she began working as an expert in the ABB Business Area “Power
Transformers” for solving PD-related problems.
In 2007 Dr. Fuhr began to work as a senior consultant for the Swiss utility BKW Energy Ltd. in the field
of condition assessment of HV components (generators, transformers, and GIS). In 2012 she founded
her own company, AF Engineers +Consultants GmbH, in Iseltwald, (Switzerland), together with her
husband Dr. Thomas Aschwanden. Her activities as an independent consultant were focused on design
review and factory acceptance tests, condition assessment, and root cause analysis of failures in all
types of power generation and transmission equipment, in particular in power transformers and power
generators.
Throughout her career, Jitka made many important contributions to the understanding of electrical
breakdown processes in dielectric liquids and in the diagnostic methods for condition assessment of
large power transformers, with particular emphasis on the detection and analysis of partial discharges
4
and the subsequent degradation of insulation materials. Jitka Fuhr pioneered the application of
advanced PD-measuring techniques with on-site testing of large power transformers. She is a coauthor
of the ABB book Testing of Power Transformers and Shunt Reactors, and her broad experience gained
over decades working in the fields of high voltage testing and partial discharge diagnostics are
documented in more than 50 technical papers presented at international conferences and published in
peer-reviewed technical journals. In addition to her daily work, Jitka was a member of IEEE (M’04),
including DEIS, the IEEE Transformer Committee, Electrosuisse, and the mirror committees in
Switzerland of IEC TC14 and TC 42. She was active in CIGRE Study Committees A2 and D1 for many
years, being the convenor of Working Group D1.29 “Partial discharges in power transformers,” which
published the CIGRE Technical Brochure 676 (2017).
The death of Jitka is a great loss which came suddenly. It is the loss of an expert in the preparation of
CIGRE brochures and reports for the coming generations. Unforgettable for everyone who worked with
her, her humor and positive charisma, which always accompanied her.
On behalf of all members of CIGRE working bodies, we would like to present our deepest condolences
to the family of this great colleague and friend of us, in particular to her husband Thomas.
Jitka’s spirit will stay with us forever.
5
Content
Executive summary ............................................................................................................. 3
1. Introduction.............................................................................................................. 12
1.1 Objective of Brochure ............................................................................................................................. 12
1.2 Structure of Brochure ............................................................................................................................. 13
2. Definitions And Initial Considerations ................................................................... 14

2.1 Definition of PD ........................................................................................................................................ 14
2.2 Definition of Calibration .......................................................................................................................... 14
2.3 Definition of Factory Acceptance Test (FAT) ........................................................................................ 15
2.4 Definition of Site Acceptance Test (SAT) .............................................................................................. 16
2.5 Motivation for improvement of PD measurements during FAT and SAT ........................................... 16
2.6 Relevant PD Measuring Methods for FAT and SAT .............................................................................. 17
3. Electrical method according IEC 60270 and its calibration .................................. 18

3.1 Calibration according to IEC 60270........................................................................................................ 18
3.1.1 Induced and apparent charge ............................................................................................................ 18
3.1.2 PD measurement setup ..................................................................................................................... 20
3.1.3 Frequency band selection .................................................................................................................. 22
3.1.4 Step 1: Checking the noise spectrum ................................................................................................ 22
3.1.5 Step 2: checking the calibration spectrum ......................................................................................... 24
3.1.6 Step 3: analyzing the real PD pulse spectrum ................................................................................... 25
3.1.7 Real field measurement example ...................................................................................................... 27
3.1.8 Summary of IEC 60270 calibration procedure ................................................................................... 28
3.2 Acceptance criteria according IEC 60076-3 .......................................................................................... 29
4. Electromagnetic Method (UHF) and its Calibration ............................................... 32

4.1 UHF sensors ............................................................................................................................................ 32
4.1.1 Valve UHF sensors ............................................................................................................................ 33
4.1.2 Window UHF sensors ........................................................................................................................ 33
4.1.3 Specialized UHF sensors .................................................................................................................. 35
4.2 Sensor characteristics ............................................................................................................................ 35
4.2.1 Techniques for measuring the frequency response of UHF sensors ................................................. 35
4.2.2 Characterization of UHF sensors in oil .............................................................................................. 36
4.2.3 Comparison of valve UHF sensors and window sensors ................................................................... 38
4.2.4 Insertion depth of valve UHF sensors ................................................................................................ 39
4.2.5 Equivalence of different UHF PD sensor characterization methods .................................................. 39
4.3 Recommendation for Placement of UHF Sensors on Power transformers ........................................ 42
4.3.1 Placement of sensors for partial discharge localization ..................................................................... 43
4.3.2 Positioning on the walls along the length of the transformer .............................................................. 43
4.3.3 Positioning on the top of the tank....................................................................................................... 44
4.3.4 Positioning on the walls along the width of the transformer ............................................................... 45
4.3.5 Positioning near the corner of the tank .............................................................................................. 46
4.4 Recommendations for the Number of UHF Sensors on Power Transformers ................................... 46
4.5 Calibration of electromagnetic method ................................................................................................. 47
4.5.1 Calibration method for the measurement devices: KM-Factor ............................................................ 48
4.5.2 Calibration method for the UHF sensor: KS-Factor ............................................................................ 48
4.5.3 Simplification of AF ............................................................................................................................ 49
4.5.4 Calibration procedure of the entire UHF measurement system: KUHF-Factor..................................... 50
4.6 Pre-conditions for UHF Calibration ........................................................................................................ 51
4.6.1 UHF measurement systems .............................................................................................................. 51
4.6.2 UHF sensors ...................................................................................................................................... 51
6
4.7 Comparison of electrical and UHF measurement procedure .............................................................. 51

4.8 Performance check of systems and sensors ........................................................................................ 53
4.8.1 Single port performance check .......................................................................................................... 53
4.8.2 Dual port performance check ............................................................................................................. 53
4.8.3 Performance check with pulse generator ........................................................................................... 54
5. Technical considerations of PD measurements for FAT and SAT ...................... 56

5.1 Emission spectra of PD........................................................................................................................... 56
5.1.1 Electrical emission ............................................................................................................................. 56
5.1.2 UHF emission .................................................................................................................................... 56
5.2 Most common sources of noise and measures to optimize SNR ....................................................... 63
5.2.1 Public broadcasting and mobile communication ................................................................................ 63
5.2.2 Industrial noise................................................................................................................................... 64
5.2.3 Other stochastically varying noise ..................................................................................................... 64
5.2.4 HV facilities – corona discharges ....................................................................................................... 64
5.2.5 Measures to optimize SNR by optimized measuring setup ................................................................ 68
5.2.6 Measures to optimize SNR by optimized voltage source ................................................................... 69
5.3 Influence of transmission path ............................................................................................................... 70
5.3.1 Electrical method according IEC 60270 ............................................................................................. 70
5.3.2 UHF method ...................................................................................................................................... 72
5.4 Correlation between UHF signal and IEC 60270 quantities ................................................................. 76
5.4.1 Simultaneous measurements of UHF and IEC quantities .................................................................. 76
5.4.2 Correlation between UHF and IEC quantities .................................................................................... 76
5.4.3 Conclusion ......................................................................................................................................... 78
5.5 Applicability of UHF Sensitivity Check for GIS to Transformers ......................................................... 78
6. Case Studies ............................................................................................................ 80

6.1 Case Study 1 (SAT) 333 MVA – Substation Transformer ..................................................................... 80
6.1.1 Conclusion ......................................................................................................................................... 83
6.2 Case Study 2 (FAT) - Generator Transformer ....................................................................................... 83
6.3 Case Study 3 (FAT) - Single Phase Transformer .................................................................................. 84
6.4 Case Study 4 (FAT) - Repaired 123 kV Substation Transformer ......................................................... 86
7. Measurement campaigns of The Working Group .................................................. 89

7.1 Initial Laboratory Investigation of different UHF sensors .................................................................... 89
7.2 On-site Investigation of different Measurement systems and signal generators .............................. 89
7.2.1 Performance check (dual port) .......................................................................................................... 90
7.2.2 Conducted measurements ................................................................................................................. 90
7.2.3 Impulse generators and measuring devices used.............................................................................. 90
7.2.4 UHF signals in time- and frequency -domain ..................................................................................... 90
7.2.5 Comparison and summary of measurement results .......................................................................... 91
7.2.6 Conclusion ......................................................................................................................................... 92
7.3 Calibration Method in Laboratory .......................................................................................................... 92
7.3.1 Calibration of measurement device and cables ................................................................................. 92
7.3.2 Calibration of UHF sensors ................................................................................................................ 93
7.3.3 Laboratory setup using an artificial UHF signal (PuM RF) ................................................................. 94
7.3.4 Measuring results of an artificial UHF signal (PuM RF) ..................................................................... 95
7.3.5 Measuring results of a real pd source with different calibrated UHF PD measuring systems ............ 95
7.3.6 Conclusion for UHF calibration .......................................................................................................... 97
7.3.7 Outlook for UHF testing ..................................................................................................................... 98
8. Recommendations for improved PD acceptance tests ......................................... 99
7
9. Outlook for further improvements .........................................................................101

9.1 Criteria Overview ................................................................................................................................... 101
9.2 Decision Tree ......................................................................................................................................... 102
9.2.1 Factory Acceptance Test (FAT) ....................................................................................................... 102
9.2.2 Site Acceptance Test (SAT) ............................................................................................................ 104
9.3 Outlook ................................................................................................................................................... 104
10. Conclusion ..............................................................................................................105
11. References ..............................................................................................................107

A.1. Introduction ............................................................................................................................................ 111
A.2. Summary of measurement principles .................................................................................................. 111
A.3. Traceability of the Measurements ........................................................................................................ 112
8
Figures and Illustrations

Figure 3-1 The stray capacitances in a transformer and a rotating machines stator winding .............................. 19
Figure 3-2 The pulse spectrum of a PD source stimulated at different winding positions [21] ............................ 20
Figure 3-3 Schematic of the PD measurement setup ..................................................................................... 20
Figure 3-4 Sensitivity of the measurable charge in dependence of the Ck to Ct ratio ......................................... 21
Figure 3-5 Typical frequency spectrum of various signal sources .................................................................... 22
Figure 3-6 White noise (top) and its FFT (bottom) ........................................................................................ 23
Figure 3-7 Switching noise (top) and its FFT (bottom) ................................................................................... 23
Figure 3-8 Sinusoidal noise (top) and its FFT (bottom) .................................................................................. 24
Figure 3-9 The frequency spectrum of an ideal calibration pulse (top) and with resonance at 350 kHz (bottom) . 25
Figure 3-10 The calibration pulse spectrum (top) and spectrum of the real PD (bottom) measured on a stator
winding ..................................................................................................................................................... 25
Figure 3-11 The real PD pulse (top) and its frequency spectrum (bottom) measured on a power transformer .... 26
Figure 3-12 The calibration pulse spectrum (top) and spectrum of the real PD (bottom) measured on a test object
with winding.............................................................................................................................................. 27
Figure 3-13 Q(t)&U(t) diagram (left) and PRPD pattern (right) @ 100 – 600 kHz ............................................. 27
Figure 3-14 Q(t)&U(t) diagram (left) and PRPD pattern (right) @ 100 – 250 kHz ............................................. 28
Figure 4-1 Principal of UHF PD Measurements in power transformers [5] ........................................................ 32
Figure 4-2 Potential valve types for retrofit of UHF sensors; valves with straight-through opening/duct [12] ...... 33
Figure 4-3 Oil valves without straight opening/duct; retrofit of valve-type UHF sensors not possible [12] ........... 33
Figure 4-4 Example of an UHF sensor for DN50/DN80 gate valves [31] ........................................................... 33
Figure 4-5 Drawing of stainless steel flange and dielectric window [12] .......................................................... 34
Figure 4-6 Examples of UHF Plate Sensor with stainless steel flange and dielectric window compatible with TB662
[12] [32] ................................................................................................................................................... 34
Figure 4-7 Example of UHF Sensor and interface flange without dielectric window ........................................... 35
Figure 4-8 Example of UHF sensor mounting on a flat dielectric window ......................................................... 35
Figure 4-9 An air-filled GTEM cell used to measure the sensitivity of UHF PD sensors for GIS. Note that the sensor
mounting facility is reproduced on a test plate specific to each type of sensor ................................................. 36
Figure 4-10: a) GTEM cell internal view with absorbers and septum before oil filling b) GTEM cell external view [37]
................................................................................................................................................................ 37
Figure 4-11 Transmission measurement (S21) for AF determination a) UHF sensor direct mounted to
the cell without oil valve b) UHF sensor installed via DN50 oil valve to the cell [37] ........................................ 37
Figure 4-12 Antenna factor (AF) of a UHF sensor measured in GTEM cell illustrating the influence of the different
mounting arrangements as per Figure 4-11 a) and b) [38] ............................................................................ 38
Figure 4-13 Comparison of the sensitivity between valve UHF sensor at 50 mm insertion depth and window sensor
(and combined in-oil UHF & acoustic PD sensor) [40] [41]............................................................................. 38
Figure 4-14 Dependency on the insertion depth using drain valve UHF sensors [40] ........................................ 39
Figure 4-15 Comparison of the effective height and antenna factor representations of a 25 mm monopole frequency
response [43] ............................................................................................................................................ 40
Figure 4-16 (a) Comparison of unadjusted effective height measurements for the 25 mm monopole sensor in oil
and air. (b) Comparison after the data for the air GTEM had been scaled for both wavelength and dielectric constant
[43] .......................................................................................................................................................... 41
Figure 4-17 Frequency response of a 62.5 mm monopole showing the measurements in air before and after scaling
for wavelength and dielectric constant: (a) Plotted as effective height He. (b) Plotted as antenna factor (AF) [43]
................................................................................................................................................................ 41
Figure 4-18 Electromagnetic wave propagation in the tank (Side view) [31] .................................................... 42
Figure 4-19 Sensor positioning for PD detection (View from the HV side) [31] ................................................. 43
Figure 4-20 Sensor positioning for PD localization (View from the HV side) [45] [31] ...................................... 44
Figure 4-21 Sensor positioning on the cover plate of the tank (View from the top) [31] .................................... 44
Figure 4-22 Sensor positioning on the tank wall with sensors installed on the top for a total of 4 sensors (View from
the HV side) [31] ....................................................................................................................................... 45
Figure 4-23 Sensor positioning on the tank wall with sensors installed on the top for a total of 6 sensors (View from
the HV side) [31] ....................................................................................................................................... 45
Figure 4-24 Suggested placement for 4 UHF sensors on a transformer [47] .................................................... 46
Figure 4-25 Suggested placement for 6 UHF sensors on a transformer (the pattern of 3 sensors shown on these
views would be repeated on both HV and LV sides) [47] ............................................................................... 46
Figure 4-26 Calibration Setup: measurement device & cables to derive KM factor [41] ...................................... 48
Figure 4-27 Example of simplifying the AF to derive KS [38] ........................................................................... 49
Figure 4-28 Calibrated measuring point of the UHF Sensor – independent from transformer [41] ...................... 50
Figure 4-29 UHF Measurement setup at a transformer [49] ........................................................................... 51
Figure 4-30 Comparison of PD measurement setups and their individual calibration point [49] .......................... 52
Figure 4-31 Performance check setup with correctly installed UHF sensors [41] ............................................... 54
Figure 4-32 Failed performance check with one sensor installed in a rising tube at the oil valve acting as an
electromagnetic shielding for the UHF sensor [41] ........................................................................................ 54
9
Figure 4-33 a) 60 V impulse of UHF impulse generator b) 250 V impulse of EMI impulse generator c) both
impulses in frequency domain [55] .............................................................................................................. 55
Figure 5-1 Measured UHF signal of probe “UHF 2” to prove broadband emission [5] ........................................ 57
Figure 5-2 Amplified (×20 voltage gain) UHF signals recorded during overvoltage testing of B-phase at 1.6 UN with
a measured PD level in the region of 20 – 30 pC [56] ................................................................................... 57
Figure 5-3 Frequency spectrum of the signal from UHF sensor No. 3 in Figure 5-2 [56] .................................... 58
Figure 5-4 Overview of PD patterns, average-/max-hold frequency spectrum of the three defects obtained with the
UHF technique. Frequency plots show noise (grey) and discharge (black) [57] ................................................ 58
Figure 5-5 Electrodes configuration and Electromagnetic wave frequency spectrum from PD of each models [58]
................................................................................................................................................................ 60
Figure 5-6 Discharge sources: (a) oil/air corona discharge, (b) creep discharge, (c) surface discharge, (d) turn-to-
turn insulation discharge [59] ..................................................................................................................... 61
Figure 5-7 Point-on-wave patterns of typical defects: (a) oil-corona, (b) air corona, (c) creep discharge, (d) surface
discharge, (e) turn-to-turn discharge [59] .................................................................................................... 61
Figure 5-8 Four types of artificial defect models: (a) Corona-in-oil discharge model (b) Surface discharge-in-oil
model (c) Gas-cavity discharge model; (d) Floating-discharge-in-oil model [60] ............................................... 62
Figure 5-9 Waveforms and normalized power frequency spectra of UHF PD signals: (1) Tip-in-oil discharge; (2)
Surface discharge-in-oil; (3) Gas-cavity discharge; (4) Floating-discharge-in-oil [60] ........................................ 62
Figure 5-10 left) Digital video broadcasting DBV T2 right) Mobile communication, GSM 900, example of one provider
[61] .......................................................................................................................................................... 63
Figure 5-11 Narrow band disturbances during online measurements using amplification [5] ............................. 64
Figure 5-12 PRPD pattern of three phase corona of close by overhead lines hindering sensitive electrical PD
measurements [5] ..................................................................................................................................... 64
Figure 5-13 210 MVA Grid-Coupling Transformer with copper wire (right) on phase L3, 110kV to stimulate corona
discharges [5] ........................................................................................................................................... 65
Figure 5-14 Comparison of PRPD without (left) and with Corona (right) [5] ..................................................... 65
Figure 5-15 (a) Laboratory-based 10/50 kV transformer for PD/corona measurement (UHF sensors and IEC 60270
system); (b) corona wire at HV bushing; (c) PRPD pattern at 25 kV [62] ........................................................ 66
Figure 5-16 (a) Time-domain corona signals from UHF sensors at condenser bushing; (b) spectral density; (c) UHF
signals from internal PD; (d) spectral densities. Note the different amplitude and frequency scales [62]. ........... 67
Figure 5-17 Test set up measuring external corona discharge by conventional method compare with UHF method
[63] as the condenser core bushing acts as a lowpass filter in between .......................................................... 67
Figure 5-18 (left) Wire attached to condenser bushing as corona source; (right) PRPD patterns detectable by
conventional / UHF method: Corona detectable by IEC but not by UHF method [63] ........................................ 68
Figure 5-19 Test arrangement for verification of UHF-PD behaviour of a voltage source for FAT based on power
electronics [64].......................................................................................................................................... 69
Figure 5-20 UHF spectra: left) noise level right) Converter in operation [64] ........................................ 69
Figure 5-21 Signal transmission for conventional PD measuring method.......................................................... 70
Figure 5-22 a) Test setup for both electrical PD measurement (IEC 60270) and UHF measurement b) maximum
measured UHF levels of UHF sensors 1-5 [55] .............................................................................................. 71
Figure 5-23 Apparent charge of electrical PD measurement at upper and lower end of winding; dotted line: Actual
charge of the source [55] ........................................................................................................................... 71
Figure 5-24 Signal transmission path for an unconventional PD measuring system (acoustic, UHF) .................... 72
Figure 5-25 Location of the 18 artificial PD sources inside the transformer model. (a) Top view; (b) Side view [66]
................................................................................................................................................................ 73
Figure 5-26 Comparison of Line of Sight distance and propagation distance-dependent attenuation in the
transformer model (a) All artificial PD sources; (b) PD sources inside winding; (c) PD sources outside winding [66]
................................................................................................................................................................ 75
Figure 5-27 Simultaneous measurement of UHF and IEC 60270 for correlation of one electrical PD pulse to its
related UHF signal...................................................................................................................................... 76
Figure 5-28 Rod-plane PD sources, 50-300 pC, 5-10 kV [10] .......................................................................... 77
Figure 5-29 Investigated PD locations in transformer tank with full active part; 10kV/380 V, 630 kVA, DN80 oil filling
valve of UHF probe [5] ............................................................................................................................... 77
Figure 5-30 PD source “glass ampoule”: Correlation between UHF amplitude and apparent charge QIEC [5] ........ 77
Figure 5-31 Sensitivity Check; same measurement readings for original PD source and artificial UHF signal source
by adequate adoption of the signal generator output voltage ......................................................................... 78
Figure 6-1 a) Electrical PD Measurement; fc = 1 MHz, Δf = 1500 kHz, 22 seconds b) UHF PD measurement at 310
MHz, Δf = 1500kHz, 22 seconds [5] ............................................................................................................ 80
Figure 6-2 333 MVA transformer showing positions of UHF sensors and acoustic sensors [5] ............................ 81
Figure 6-3 Electrical PD Measurement in test facility according IEC 60270 on 333 MVA single phase auto-
transformer; fc = 300 kHz, Δf = 400 kHz [5] ................................................................................................ 82
Figure 6-4 Deteriorated paper insulation on leads at the tap changer ............................................................. 82
Figure 6-5 Generator transformer with UHF sensors for FAT. In addition to visible sensors, sensor No. 1 is located
on the back (HV) side of the tank opposite the position of sensor No. 3 [56] ................................................... 83
Figure 6-6 Amplified (×20 voltage gain) UHF signals recorded during overpotential testing of C-phase at 200% with
a measured PD level in the region of 5 – 10 pC [56] ..................................................................................... 84
10
Figure 6-7 Frequency spectrum of the signal from UHF sensor No. 1 [56] ....................................................... 84
Figure 6-8 Performance check with two sensors opposite to each other with 50 V, 100 V, 200 V and 400 V signal
generator output voltage ............................................................................................................................ 85
Figure 6-9 UHF PD measurement during FAT without indication of electrical PD measurement .......................... 85
Figure 6-10 Installed UHF PD-Sensors (left: bottom of tank (sensor 1), right: top of the tank (sensor 2) ........... 86
Figure 6-11 PRPD of conventional PD measurement according to IEC 60270 showing strong noises due to switching
of thyristors of power converter .................................................................................................................. 87
Figure 6-12 PD-inception at 110 kV (Sensor 1 (left), Sensor 2 (right)), logarithmic scaling; acquisition time: 1 min
................................................................................................................................................................ 87
Figure 6-13 PD-inception at 110 kV (Sensor 1 (left), Sensor 2 (right)), linear scaling, acquisition time: 1 min ..... 87
Figure 6-14 PRPD-pattern at 145 kV (Sensor 1 (left), Sensor 2 (right)), linear scaling, acquisition time: 10 min .. 88
Figure 7-1 Positions of sensors at 800/3 MVA Transformer for comparative UHF measurements [70] ................ 89
Figure 7-2 Example of injected and coupled UHF signals in time- and frequency-domain .................................. 91
Figure 7-3 Comparison of measurement systems [38] ................................................................................... 93
Figure 7-4 AF(f) and corresponding KS of UHF sensors no1 and no2 [38] ........................................................ 93
Figure 7-5 UHF valve UHF sensor mounted with a DN80 gate valve. [38] ........................................................ 94
Figure 7-6 Schematic of the laboratory setup with two UHF sensors mounted on a transformer tank [38] .......... 94
Figure 7-7 PRPD of measured PuM RF signal [38] ......................................................................................... 94
Figure 7-8 UHF PRPDs of the same PD source with different visualizations. Left: MS3 using linear scale /nV; Middle:
MS2 linear scale / mV (phase shifted);right: MS4 logarithmic scale / % [38] ................................................... 95
Figure 7-9 UHF PRPDs of real PD source 1 after rescaling (calibration) with linear y-axis on MS3 and MS2 [38] .. 96
Figure 7-10 PRPDs of the second real PD source 2 after rescaling (calibration) with logarithmic y-axis on MS 4 and
MS 2 [38] .................................................................................................................................................. 97
Figure 9-1 Decision Tree for future combined electrical and UHF FAT/SAT measurements .............................. 103
Tables
Table 1 Description of terms related to calibration ........................................................................................ 15
Table 2 Requirements and tests for different categories of windings [1] ......................................................... 30
Table 3 Requirements and tests for different categories of windings [1] ......................................................... 30
Table 4 Recommended number of UHF sensors for power transformers .......................................................... 47
Table 5 Comparison of different measuring methods ..................................................................................... 91
Table 6 Measured voltage, used K factors, and calibrated UHF reading of all devices ...................................... 95
Table 7 Measured voltage, used K factors, and calibrated UHF reading at Sensor 1 for PD source 1 .................. 96
Table 8 Measured voltage, used K factors, and calibrated UHF reading at Sensor 2 for PD source 2 .................. 97
Table 9 Overview on criteria for FAT and SAT (green: criteria defined, yellow: criteria missing so far) ............. 101
11
1. Introduction
This TB deals with possible improvements to partial discharge (PD) measurements for factory
acceptance test (FAT) and site acceptance tests (SAT) of power transformers. In the joint working group
JWG A2/D1.51, it is common understanding that a sensitive PD measurement is in the common interest
of the transformer manufacturer as well as the buyer and end user of the transformer. It ensures high
quality and consequently low operational costs during technical lifetime of the transformers.
Hence, the main task of PD measurements during Factory Acceptance Test (FAT) or Site Acceptance
Test (SAT) is to evaluate the quality of the insulation of a transformer and to minimize the risk of
incidents during operation. An investigation can also be triggered by suspicious findings from dissolved
gas analysis (DGA), which may undermine the reputation of the manufacturer and cause clarifying
efforts to the buyer or end user of a transformer. With that regard, it is neither the intention of this
brochure to change acceptance criteria allowing easier passing of FAT/SAT, nor to tighten the
acceptance criteria by introducing an additional measurement method measuring signals without a clear
indication of critical PD activity.
PD in an electrical insulation system indicate flaws that may or may not develop into a full insulation
breakdown depending on the type of insulation system and the nature of PD. International standards
such as IEC 60076-3 [1] include electrical tests and electrical measurements to detect PD as routine
tests [2]. The apparent charge QIEC and respective limits as pass/fail-criteria have therefore become an
important indicator of transformer quality. Despite the fact, that the measured level of apparent charge
might depend rather strongly on the position of the PD within the active part of the transformer, care
also has to be taken concerning the measurement bandwidth of calibration and actual measurement.
In recent years, alternative methods for PD measurement, such as acoustic and electro-magnetic ultra-
high frequency (UHF) measurements, have been developed. These methods have proven to be sensitive
in detecting PD and can assist in determining the location of PD faults. Compared to electrical
measurement methods, the UHF method is more immune against external noises and disturbances in
most on-site applications. Hence, it allows an easier differentiation of external and internal PD. This
makes the UHF method suitable for both, its use in the manufacturer’s test laboratory (low ambient
noise with e.g., shielding and noise blocking efforts) and on-site after transportation and installation of
the transformer (usually elevated ambient noise level).
Unconventional PD measurements themselves need to be reliable and reproducible methods in order to
become reliable and easily accessible indicators for transformer’s quality useable in both, FAT and SAT.
Calibration of the sensors and the measurement set-up; system sensitivity regarding real PD defects in
a transformer; robustness against noise; number of and placement of sensors; comparability with
electrical measurements and procedures for acceptance tests in the factory and on-site are the topics
to be addressed in the brochure.
1.1 Objective of Brochure

As every PD measurement method needs to produce repeatable and reproducible results, the brochure
focuses on calibration procedures. Especially for UHF PD measurements, the term calibration was diffuse
or misinterpreted in the past. Hence, the method was missing a proper calibration. The brochure focuses
on that misunderstanding and provides a calibration procedure for UHF PD measurements.
The objective of any calibration is to ensure a general comparability and reproducibility of measurements
independent of used devices and setup conditions. It must be noted that the UHF calibration should not
be misinterpreted as an attempt to correlate the readings of the UHF method (calibrated measurements
in V/m) with the electrical method according to IEC 60270 in terms of pico-Coulomb (pC). A
measurement system for radiated signals cannot be calibrated by a system for conducted signals and
vice versa.
The brochure introduces two main factors required for UHF calibration. The particular measurement
device (including accessories such as cables, filters, attenuators, pre-amplifiers, etc.) is addressed in
the first step of the calibration by introducing the calibration factor KM, see chapter 0. The influence of
the UHF sensor, essentially its sensitivity and its conversion from an electric field strength to a voltage
signal, is included using calibration factor KS, see chapter 4.5.2. Using the proposed UHF calibration
method, different UHF measurement systems (including UHF sensors, cables, amplifiers, attenuators,
12
filters, and measuring devices) become comparable when used at the same measurement position in a
transformer.
Furthermore, the brochure provides additional basic information for a deeper understanding of PD
measuring by reviewing the conventional electrical method according to IEC 60270 and comparing it to
the UHF method. To get the best usage for both PD FAT and PD SAT, the brochure discusses and
evaluates given and alternative methods and concludes with recommendations for individual
improvements.
1.2 Structure of Brochure

After the introduction basic definitions are given, like the definition of partial discharge and the
definitions of FAT and SAT. Furthermore, the term calibration is discussed and its use within the
brochure is explained.
In chapter 3, the conventional electrical method is reviewed in detail with a special focus on its
calibration and later usage as acceptance criteria.
Chapter 4 discusses the general calibration procedure and describes the development of a calibration
method for the UHF PD measuring method. Therefore, background information is given like the
characterization of UHF sensors using the antenna factor. After a short overview about different types
of UHF sensors, the subsequent chapters provide advice on the proposed number of sensors to be
installed and recommended locations. The chapter compares the UHF calibration procedure to the
electrical calibration procedure and ends with an explanation of the necessary performance check of
UHF PD measurement technology.
To advise on improved measurements in test laboratories, for FAT as well as SAT, chapter 5 collects
information about common sources of noise and how to optimize UHF measurements for improved
sensitivity. It will explain the necessary knowledge for interpretation of measurement results like
emission spectra and the influence of the transmission path on measurable signals.
Chapter 6 presents some case studies used by the working group to obtain knowledge and experiences
on the sensitivity and calibration possibility of UHF signals in practice.
As the working group performed research by these experimental case studies for developing and
verifying the UHF calibration procedure, chapter 7 collects the information from certain measurement
campaigns to allow readers to understand and reconstruct the calibration procedure for other research
groups. During the measurement campaigns, several further improvement possibilities have been
identified and are documented.
The UHF calibration procedure and all related calibration factors are summarized in chapter 8 to give
effective recommendations for improved PD acceptance tests by improved reproducibility and
comparability of UHF measurements.
Chapter 9 presents an outlook for future work and research. It incorporates a discussion about the first
steps towards threshold definitions of UHF levels in terms of possible acceptance criteria. Thoughts and
ideas for this goal are summarized.
Finally, the brochure ends with a conclusion in chapter 10.
13
2. Definitions And Initial Considerations

2.1 Definition of PD
The definition of PD from IEC 60270 [2] of the International Electrotechnical Commission (IEC) is:
Partial discharge (PD) is a localized electrical discharge that only partially bridges the insulation
between conductors and which may or may not occur adjacent to a conductor.
The definition of PD from IEEE 100 [3] of the Institute of Electrical and Electronics Engineers (IEEE) is:
Partial discharge (PD) is an electric discharge which only partially bridges the insulation between
conductors, and which may or may not occur adjacent to a conductor.
The definition of apparent charge from IEC 60270 [2] of the International Electrotechnical Commission
(IEC) is:
Apparent charge q of a PD pulse is that charge which, if injected within a very short time
between the terminals of the test object in a specified test circuit, would give the same reading
on the measuring instrument as the PD current pulse itself. The apparent charge is usually
expressed in Pico coulombs (pC).
NOTE The apparent charge is not equal to the amount of charge locally involved at the site of the discharge, which
cannot be measured directly.
2.2 Definition of Calibration

According to the “International vocabulary of metrology” [4] the term “Calibration” is defined as follows:
Calibration: operation that, under specified conditions, in a first step, establishes a relation
between the quantity values with measurement uncertainties provided by measurement
standards and corresponding indications with associated measurement uncertainties and, in
a second step, uses this information to establish a relation for obtaining a measurement
result from an indication.
NOTE 1 A calibration may be expressed by a statement, calibration function, calibration diagram, calibration curve, or
calibration table. In some cases, it may consist of an additive or multiplicative correction of the indication with associated
measurement uncertainty.
NOTE 2 Calibration should not be confused with adjustment of a measuring system, often mistakenly called “self-
calibration”, nor with verification of calibration.
This definition has now to be transferred to the process of partial discharge measurement, see following
paragraphs and Table 1.
According to IEC 60270 the calibration of a measuring system in the complete test circuit is made to
determine the scale factor k for the measurement of the apparent charge. The calibration of a
measuring system in the complete test circuit is carried out by injecting short-duration current pulses
of known charge magnitude q0 into the terminals of the test object. That is, the current pulse with
known charge magnitude q0 is the quantity value provided by the calibrator (measurement
standard). The indication of the PD measuring instrument is corrected by multiplying by the scale
factor in order to display the apparent charge (measurement result).
The apparent charge is not equal to the amount of charge locally involved at the site of the discharge,
which cannot be measured directly [2]. Hence it must be noted that this calibration process according
to IEC 60270 gives no information regarding the correlation between real charge of the PD and the
measurement result (apparent charge) at the line terminals. As described and shown in chapter 5.3.1
the correlation between real charge and apparent charge is unknown for all test objects including test
objects with windings like power transformers due to the influence of the transmission path from PD
location to the measurement terminals.
14
Table 1 Description of terms related to calibration
Term Description Definition/ Description/ Definition/ UHF

/IEC IEC UHF
Quantity Apparent charge pC Received V/m
value electrical field
Measurement Freq. range, set-up, Freq. range, set-
uncertainties equipment, up, equipment,
environmental, environmental,
“calibration” “calibration”
Measurement Signal Generator of Apparent See chapter 4.5 Electrical Field
standards e.g., 100 pC Charge QIEC in strength EUHF in
connected to test pC V/m
object terminals
Indications Reading of measuring pC Reading of EUHF
device before measuring device
(across Zm) (sensor output)
calibration before calibration
Measurement Reading of measuring Apparent Reading of Measurable UHF
results device after calibration Charge: measuring device field strength
after calibration EUHF / V/m
QIEC / pC
(also including
sensitivity of
sensor)
Hence the calibration of a new PD measurement method (like the UHF method, which measures
radiated signals) cannot be interpreted in a way that allows calculation of a specific apparent charge
of the conventional measuring method, which employs conducted signals. Furthermore, neither method
quantifies the physical (real) charge that is displaced at the site of PD. Hence, the “calibration” of an
alternative PD measuring method in terms of Pico coulomb (pC) is physically not possible in any general
sense.
Beside the wrong use of the term “calibration,” a considerable number of experiments have been carried
out to investigate a correlation (or scaling method) between the output signal of a UHF PD sensor
with the measured apparent charge [5] [6] [7]. This work has demonstrated the strong influences of
the type and position of PD and variations caused by the PD current pulse shape and the signal
transmission path on both the conducted and radiated signals. Therefore, while correlations between
apparent charge QIEC and the apparent field strength EUHF of the UHF sensor can be observed, they are
always strongly dependent on the specific geometrical arrangement of PD source, sensor, and test
object and therefore cannot be used to determine the absolute magnitude of any PD pulses.
Nevertheless, many investigations have confirmed that a useful statistical correlation exists between
changes in measures of UHF signal magnitude and apparent charge values, which is important for
correct identification of trends in the levels of PD activity.
2.3 Definition of Factory Acceptance Test (FAT)

In general, the intention of both Factory Acceptance Tests (FAT) and Site Acceptance Tests (SAT) is to
check whether a transformer fullfils contractual requirements. The FAT often takes place in the
manufacturer’s factory at the end of the production process to verify that the manufactured product
fulfills internal and external quality standards as well as the contractual requirements of a specific order,
e.g., guaranteed losses. As power transformers are often designed according to customer specific
requirements, FAT normally consists of type and routine tests to verify both design and manufacturing
quality of a transformer. Characteristics of FATare as follows:
• Performed in the supplier’s factory
• Performed at the end of the production process
• May involve routine, type and special tests
• To verify design, functionality and guaranteed values (e.g., losses)
• To show that a product is compliant to a specification and ready to be shipped and installed on
site
15
2.4 Definition of Site Acceptance Test (SAT)

A Site Acceptance Test (SAT), is performed at the final location as part of the commissioning procedure.
SAT is intended to verify that the transformer’s condition has not changed since it left the factory. IEEE
C57.152-2013 divides “field transformer testing” into mainly two categories associated with stressors
that are unique to each period in the unit’s life cycle [8]:
• Transportation and installation
• In-service aging and Maintenance
The transportation and installation phase in the life of a large power transformer is brief but may present
significant structural and environmental challenges. SAT is conducted at the customers operating place
and is usually a part of the substation commissioning. The main objectives of SAT are to:
• Identify changes to the unit’s integrity after leaving the factory or storage
• Validate the correct configuration of the transformer and its accessories and identify incorrect
assembly
• Verify the functionality of transformer by routine and advanced tests(for repaired transformers)
• Confirm the proper processing and liquid filling of the tank
• Assure the suitability for initial energization of the equipment
• Establish a baseline for future condition assessments
SAT is conducted on the fully assembled transformer and is intended to assure that operation of tested
transformer is within applicable standards and manufacturer's tolerances, and the installation is in
accordance with the relevant design specifications and no damage has occurred during transportation
or installation. Upon agreement between parties, SAT is typically performed either by the equipment
manufacturer, contractor building the substation, or a 3rd party testing firm.
Note: In the context of this brochure, a SAT means an additional onsite high-voltage (HV) test for
measuring e.g., PD activity after FAT tests. It does not refer to standard low voltage (LV) tests like
measurements of resistances (of windings, core to ground, …), capacitances (between windings,
winding to core, …), oil quality measurements as well as testing secondary protection and control
functions.
Some utilities already decided to perform additional HV testing of transformer as standard procedure
on site for SAT purposes.
2.5 Motivation for improvement of PD measurements during FAT and

SAT
Defined FAT acceptance criteria concerning PD activity are currently only available for IEC 60270
measurement results, although the measurable apparent charge does not allow a determination of the
real charge turned over by the actual PD source and therefore it gives no evidence of whether a
measured PD is critical or not. Sometimes acoustics and/ or electromagnetic methods (UHF) are used
to support diagnostic and localization purposes after failed FAT to start effective repair process or
corrective actions. It should be noted that also the acoustic and UHF measurements do not allow a
determination of the real charge turned over by the actual PD source.
The purpose of this brochure is to give recommendation for improved PD FAT & SAT in future.
A complementary usage of the electromagnetic method during FAT might be helpful due to the following
reasons:
• Effective discrimination between internal and external sources (for FAT & SAT conditions)
• Verification of electrical measurements (& DGA measurements)
• Improvement of electrical measurements by gating
• SAT measurements with only electrical measurements might see certain drawbacks due to
heavy noises
• UHF PD fingerprint measurements from FAT can be used during SAT; goal is UHF
measurement only for SAT condition (like GIS community)
• Preparation as diagnostic tool for PD monitoring & PD localization (maybe by triggering of
acoustic measurements)
16
2.6 Relevant PD Measuring Methods for FAT and SAT

Due to the following phenomenological reasons, there are different possible methods for detecting and
measuring PD:
➢ Chemical reactions → Gas analysis (dissolved in oil or free gas)
➢ Emission of light → Optical measurements
➢ Release of sound → Acoustic measurements
➢ Transportation of charges → Electrical measurements
➢ Electromagnetic (EM) wave radiation → EM measurements in UHF range
Dissolved gas analysis (DGA) are commonly performed on transformer fleets with time intervals ranging
between years and weeks, depending on the severity of an expected fault or the importance of the
respective transformer. Results are often an indication of internal faults and trigger further diagnostic
methods such as PD measurements. DGA are also common to be performed by the manufacturers in
order to confirm no active defects during FAT.
Optical measurements are useful for external PD like corona discharges and are used to control HV
measuring setups on PD activity. Optical measurements are not workable within transformers due to
the insulation oil absorbing emitted light.
The acoustic PD detection method is based on the emission of acoustic sound waves for each PD event.
No standard for that method exists, and it is not possible to calibrate the method or to find any
correlation to electrical measurable quantities. The acoustic method is normally used on transformers
for PD localization purposes in combination with other PD detection methods [9]. As the acoustic PD
measurements are usually triggered by electrical or UHF measurements, it could generally be regarded
as less sensitive. Furthermore, the sensitivity is strongly dependent on the travelling path of the emitted
waves and internal structures of transformers may block the transmission of acoustic waves completely.
Further information about that method can be found in [10]. In current edition of IEC 60076-3, the
acoustic method is not used as acceptance criteria but only for diagnosis. It remains important for
localization of PD in case the acceptance criteria are not met (Annex A.4 of IEC 60076-3). It is well
described and covered in publications by the CIGRE [11], [12] IEC [13] and IEEE [14], and is not
regarded further in the brochure at hand.
The last two measurement techniques, i.e., the electrical and electromagnetic (UHF) measurements,
are identified as relevant for the brochure on hand and are introduced and discussed in more detailed
in the following chapters.
17
3. Electrical method according IEC 60270 and

its calibration
Partial Discharge (PD) measurement performed according to IEC 60270 has become an essential tool
for a product’s quality assurance [2]. The main goal of IEC 60270 is to standardize and unify PD
measurement to get comparable results for tests performed at various locations, with different
equipment and by different operators. To achieve this goal IEC 60270 has defined a complex set of
processes and key parameters, which are essential to be followed carefully. The original IEC 270
standard (1980’s) is designed for testing lumped capacitive test objects. Nevertheless, from long-term
experience, it has been found that good and plausible results can be obtained for test objects with
windings such as transformers and electrical machines as well. However, especially in case of testing
more complex RLC systems certain rules need to be followed and particular limitations should be
considered.
The purpose of this chapter is to provide clear and unambiguous guidelines on how to perform
repeatable, plausible and comparable PD measurement according to the IEC 60270. It describes briefly
the differences between the definition of the so called induced and apparent charge. It also discusses
in detail functionality and suitability of the commonly used measuring test setup and its components
including the coupling capacitor, measuring impedance, PD calibrator, PD detector and potential filtering
elements. In addition to this, the selection and explanation of proper test evaluation parameters and
PD detector settings are described including selection of a suitable frequency measurement range, which
is a key factor when it comes to PD testing. Last but not least, a robust but simple performance and
verification procedure for the test setup and PD detector settings is introduced and explained.
The electrical method was established for Transformers in 1980. It is a well proven quality assessment
method and used throughout the industry for FAT conditions within shielded laboratories. There are
certain drawbacks like ambient noise during SAT conditions, although its application is possible [12].
3.1 Calibration according to IEC 60270

3.1.1 Induced and apparent charge
Since the 1930’s several models have been developed to improve the physical representation of partial
discharge accurately [15]. The most popular and widely accepted among them is the 3-capacitor model
also referred to as the a-b-c model. Its simplicity enables the easy comprehension of discharge scenarios
by representing the defect as a lumped capacitor. Though several other models (more complex) were
proposed through the years, namely the 5-capacitor model in the 1960’s [16] and later the dipole model
in the 1980’s [17], the a-b-c model remained the most popular. Based on this model, IEC 60270 defines
the term “Apparent charge" which is the charge measurable at the terminals of the test object. The
message that the IEC conveys through the term apparent charge is that the value of the measured
charge is not exactly equal to the charge involved at the site of discharge. That is, if Cc is the capacitance
of the defect, Cb is the capacitance of the healthy dielectric in series with the defect and qc is the charge
at the defect location then the charge measured at the terminals (qa) is given by
𝑞𝑎 = 𝑞𝑐 (𝐶𝑏 /𝐶𝑐 ) (1)

under the assumption that Cc>> Cb, qa<<qc. Due to this reason this term has been misconstrued to be
an arbitrary value which has no relation to the value of the real discharge. However, this is not true,
the value of the measured charge is directly representative of the charge involved at the discharge site
as confirmed by E. Lemke in the 2010’s [18], [19] based on Pedersen’s Dipole model [17]. He introduced
the term “Induced charge” in an attempt to redefine the value of the measured PD at the test terminals
and dismiss the delusion involved with the usage of the term apparent charge. According to this theory
the measured charge is proportional to the real charge through a continuous dimensionless positive
scalar function λ which differs based on the geometry and location of the defect
𝑑𝑞 = −𝜆 ∙ 𝑑𝑄 (2)
where dq is the measurable charge and dQ is the charge involved at the defect location.
18
Henceforth, the charge value measured in pC as prescribed by the IEC is a quantity that is well correlated
to the PD severity. Presently, there have been discussions on “calibration” and “sensitivity checks” for
PD measurements using unconventional techniques. Though the term sensitivity check is still acceptable
in these cases, the process of calibration is rather complex and does not seem to be meaningful.
Unconventional PD measurement techniques depend on the response between the sensor position and
the failure location and not just the response of the HV test object as is in the case of the IEC
measurements. The results of such measurements (HF above IEC range, VHF, UHF, acoustic and other
radiated field measurements) change dramatically based on the PD test system configuration. For
instance, neglecting all the errors from external factors and considering an acoustic measurement of a
transformer, a single defect can generate signals of several different amplitudes. Depending on the
propagation path and sensor position, the generated signal might be completely damped/reflected, i.e.,
no signal reception. In addition, it is impossible to define this 3-D space in which the radiated waves
are measured.
Figure 3-1 The stray capacitances in a transformer and a rotating machines stator winding
It can be argued that the case is similar to the electrical measurement on a large test object or a test
object with windings where the PD calibration over the test leads is unlike the real PD event, and the
measured value is arbitrary and does not correlate to the real discharge. However, for instance, in case
of conventional electrical PD measurement on a large transformer, the Low Frequency (LF) components
of the PD source located anywhere deep in its winding partially “by-passes” the LC filtering effect of the
winding inductance and the inter-winding and tank stray capacitance as shown in Figure 1. Similarly, in
the case of rotating machines, the stray capacitance in between stator bars at the stator end-winding
provides good cross-coupling of the LF components enabling the location of faults deep inside the
windings. This has been explained in detail through simulation and measurement in [20], showing the
frequency response of several large test objects with complex RLC configurations.
The study published in [11] simulated a PD source at different locations on a transformer winding and
measured the pulse spectrum of a real PD pulse as shown in Figure 3-2. From this plot one can notice
the better stability of the spectrum at frequencies lower than 300 kHz; this will ensure the stability of
measured PD values and preserve the validity of the quasi-integration process (discussed in Section
4.2.2). The displayed PD pulse spectra indicate that in case of correct and IEC 60270 compliant settings,
charge reading would be consistent through the complete winding. This confirms the recommendation
of IEC 60270, Section 4.3.4, Note 2 which states that for test objects with windings the upper cut-off
frequency f2 shall be reduced to a few 100 kHz or even below.
19
Also, IEC narrow-band measurements can lead to significant errors. For example, a narrow-band
measurement at 500 kHz center frequency would lead to a measurement error of 30 dB (factor 32) in
case of the PD being 80 cm away from the calibration point, “deep” inside the winding. (Figure 3-2,
green spectrum).
Frequency
Figure 3-2 The pulse spectrum of a PD source stimulated at different winding positions [21]
In [22] and [23] it is reported that in a three winding transformer the PD signal was detected as
significantly damped to a level of about 1/35 depending on the occurrence position of the PD.
Consequently, it should be noted that, where the PD is detected in the actual test, the magnitude of
the PD that has actually occurred may potentially have been about 35 times larger than the
measurement value.
3.1.2 PD measurement setup
Partial discharge measurements for high voltage AC equipment have been defined by the IEC 60270
standard which specifies the general requirements of the measuring and calibration system. The partial
discharge measuring system is divided into two major subsystems, the coupling unit and the measuring
instrument. This section describes the preliminary step towards performing the PD test by describing
the selection of the circuit components and their functionality.
Measurement Circuit
The PD measurement circuit is shown in Figure 3-3 and the various components of the measurement
circuit are described in the following sections.
Figure 3-3 Schematic of the PD measurement setup
20
HV Filter
The use of a blocking impedance (HV filter) is strongly advised. The correct choice of a blocking
inductance is imperative to the outcome of the partial discharge test. The functionality of the HV filter
can be listed as follows:
• Noise and interference originating from the power supply side will be blocked and will not
interfere with the PD measuring loop.
• The HV filter is also called blocking impedance because it confines the High Frequency (HF)
partial discharge signal coming from the test object in the measuring loop and avoids signal
leakage thru the power source capacitance and hence improve the sensitivity of the
measurement.
In many cases, while performing PD acceptance tests close to the maximum voltage rating of the supply
transformer, discharge from the HV source is very likely. To prevent such errors the PD test system
should be tested regularly for discharges from the source. The attenuation of PD from the source
increases with increasing filter order.
Coupling Capacitor
The coupling capacitor (Ck or Cc) provides a low-impedance path for the HF discharge current to
circulate. The coupling unit needs to be carefully selected considering the following points:
• The ratio of the coupling capacitance (Ck) and the test object capacitance ( Ct) determines the
sensitivity of PD measurement. The relationship of the measurable charge (Qm) and the charge
at the test object terminals (Q) is depicted in Figure 4. By increasing the value of the coupling
capacitance, it is possible to achieve greater measuring sensitivity and SNR (Signal to Noise
Ratio).
• However, increasing the size of the capacitance also necessitates a higher power supply to
provide increased load current.
• It is also interesting to think in terms of the circuit’s time constant. A larger coupling capacitor
will increase the time constant thereby increasing the width of the discharge pulses.
• The coupling capacitor commonly serves as the primary of the voltage divider in which case the
capacitance should have sufficiently linear voltage characteristics to preserve a constant divider
ratio.
Figure 3-4 Sensitivity of the measurable charge in dependence of the Ck to Ct ratio
Measuring Impedance
The measuring impedance converts the input current signal of the partial discharge to an output voltage
signal while effectively blocking and separating the excitation frequency signal (power frequency during
normal operation or excitation frequency during induced voltage tests). This ensures that there is neither
excitation frequency signal nor its harmonics in the measured PD voltage signal.
21
Finally, the measuring impedance has to be equipped with over-voltage or flashover protection in case
of breakdown. This is a vital feature for this application.
PD Detector
The PD detector serves as the signal acquisition unit and it must fulfil the regulations defined by IEC
60270. The permissible filter ranges must be correctly and precisely implemented in the detector in
addition to the PD magnitude calculations (in terms of pC) which are currently defined by the pulse train
response. The detailed description of the filter settings of the detector are presented in Section 4 of IEC
60270.
3.1.3 Frequency band selection
Selecting the right frequency band for PD measurement is the most important step towards obtaining
reliable test results with a high level of reproducibility. In earlier times, obtaining the frequency spectrum
of a pulse required special oscilloscopes able to compute the Fast Fourier Transform (FFT) of the
acquired pulse or complex spectrum analyzers requiring skillful operators. However, it is now possible
with a unique single click feature using “Trigger by level” or “Trigger by position” enabled on the fully
digital state of the art PD detector [24], [25]. The typical frequency spectra of various signal sources
are shown in Figure 3-5.
Optimized measuring band
12 0
10 0
80
60
40
20
1 10 100 1000 10000

f(khz)
Partial discharge FFT RF broadcast FFT Tyristor pulse FFT
Figure 3-5 Typical frequency spectrum of various signal sources

Tips on checking the noise floor and discriminating interference pulses as well as on selecting the correct
bandwidth for PD calibration and measurement are described in this section.
3.1.4 STEP 1: CHECKING THE NOISE SPECTRUM
When it comes to electrical PD measurements there are three common categories of noise:
Random noise
Random noise otherwise referred to as white noise is the stochastic noise which randomly varies around
its mean value which equals to zero. The FFT/spectrum of such noise is characterized by a constantly
low dB level and no prominent peaks at specific frequencies as shown in Figure 3-6.
22
Figure 3-6 White noise (top) and its FFT (bottom)
Switching Noise
Switching noise is a high frequency radiated noise. The most common sources of this kind of noise
are equipment with electronic switching such as frequency converters (HV controls, heating ovens,
lifts etc.). The spectrum of an electronic switching pulse looks uneven with prominent resonant
peaks. The FFT of a switching pulse is similar to the FFT of a PD event when performing tests on
large test objects or test objects with windings such as transformers or rotating machines. These
pulses are synchronized to the 50 or 60 Hz power cycle and occur at a specific frequency (typically
the switching frequencies of a few kHz). One such example is shown in Figure 3-7.
Figure 3-7 Switching noise (top) and its FFT (bottom)
Sinusoidal noise
Sinusoidal noise is one other category which could hinder the PD test. It is a continuous wave
which might originate from radiated AM and FM signals or other communication bands. This noise
if not recognized can have strongly negative impact on the PD test results. Sinusoidal noise is
reflected in the noise spectrum as a narrow peak at a very specific frequency such as shown in
Figure 3-8.
23
Figure 3-8 Sinusoidal noise (top) and its FFT (bottom)
Therefore, one of the first and most vital steps while performing PD tests is to check the prevailing
noise floor in the testing field/laboratory. Once the operator has complete knowledge of the noise
conditions in the field test he/she goes on to select a frequency band for PD measurement trying
to exclude the noisy peaks. In case of standard (e.g. IEC/IEEE) compliance tests, the operator
needs to keep in mind the test specifications defined in the respective standards and ensure that
the filter settings are within the required range.
3.1.5 Step 2: checking the calibration spectrum

The calibrator is connected across the test object and a PD pulse of known charge value is injected.
The detector is then calibrated. The following tips on calibration are highly recommended.
General Tips for Calibration
• Always inject the calibration pulse across the leads of the test object and not the coupling
capacitor.
• Make sure all circuit components are connected, and the circuit is of the final form before
doing the calibration.
• Any alterations made to the circuit after calibration will require recalibration.
• Any alterations made to the PD detector settings (frequency range settings etc.) will require
recalibration
• To avoid calibration errors, remove the ground rod from the circuit.
• Always check the noise spectrum before calibration (see chapter 3.1.4).
• Check linearity of the calibration by injecting multiples of the originally injected charge value,
e.g., 50% and 200%.
• Noise level check can be recommended at 10-20 % of applied voltage. Reason: noise
level/character can change once voltage applied
The calibration is invalid in case the frequency spectrum is not flat over the filter band set by the user.
In this case the quasi-integration process is invalidated, yielding wrong charge values. Such errors are
more likely when testing large test objects or test object with windings, which are comprised of a
complex RLC network generating resonances over the prescribed measuring range. One such example
is shown in Figure 3-9 with a resonance at 350 kHz. The correct choice of filter settings for this
calibration is to restrict the higher cut-off frequency to < 300 kHz.
24
Figure 3-9 The frequency spectrum of an ideal calibration pulse (top) and with resonance at 350
kHz (bottom)
Quasi-integration
In the time domain methodology of charge estimation, the peak of the filter response is
proportional to the value of apparent charge of the PD pulse. During PD measurement the
proportionality constant is calculated by first injecting a known value of charge during calibration.
However, this proportionality is only valid if the filter extracts the PD pulse energy where the
spectral density is constant. This process is known as quasi-integration. In case of resonances in
the measuring setup/test object, the selected frequency spectrum can sometimes fall on non-flat
regions making the charge estimation invalid.
3.1.6 Step 3: analyzing the real PD pulse spectrum

The final confirmation check that is recommended for advanced users in order to ensure the
correctness of the measured charge value is to acquire a real PD pulse and check its spectrum. In
case of large test objects or test objects with windings this becomes crucial. It is possible that the
frequency response of the test object during calibration is different from that during a real PD event
as shown in Figure 3-10. This is because the frequency response depends on the pulse travelling
path.
Figure 3-10 The calibration pulse spectrum (top) and spectrum of the real PD (bottom) measured
on a stator winding
In the case of the stator presented in Figure 3-10. The setup had to be recalibrated placing the
filter between 100 and 400 kHz instead of the initial 100 to 700 kHz.
25
The purpose of checking the spectrum of the discharge pulse is to confirm the validity of the
calibration. However, this matter shall rather be taken care of by the application standards
(“vertical” standards) of the IEC which have been developed to deal with specific test objects and
define more specific criteria for those objects. It is a precautionary measure that can be exercised
in order to improve the reliability of test results. The PD test standards, IEC 60270 is a basic
standard (“horizontal” standard) and defines the basic allowable measuring bands up to 1 MHz.
However, the application standard IEEE Std. C57.113-2010 which deals with PD measurements on
power transformers recommends a maximum upper frequency limit of 300 kHz in order to perform
a charge estimation (integration) with sufficient accuracy. The PD event recorded from one such
PD test performed on a power transformer rated for 80 kV is shown in Figure 3-11. It clearly shows
that the energy of the PD pulse is confined to the lower frequencies and there is no apparent
reason to measure at higher frequencies.
Figure 3-11 The real PD pulse (top) and its frequency spectrum (bottom) measured on a power
transformer
In case of cable measurements “artificial resonance” peaks appear in the frequency spectrum due
to the reflected pulses from the cable end. However, this effect can be neglected. A more detailed
study on the frequency spectrum of cable networks is described in [26]. The latter investigates the
drastic attenuation of the PD pulse at frequencies above 1 MHz and shows an error of 28% for the
estimated charge value when using a filter in the frequency band of 1 to 20 MHz while it is only
4% when measured over the prescribed IEC band (50 to 500 kHz). Hence, selecting a low
frequency range prevents measurement errors due to increased cable attenuation at higher
frequencies resulting in more stable and accurate PD level measurement.
In case any PD activity is detected, the fault location processing needs to take advantage of the
full bandwidth (unfiltered signal). Properly to check the pulse spectrum and filter settings, it is
recommended to connect the calibrator at the cable end and to record a single impulse. In this
way, any resonance in the path from the cable termination through the coupling capacitor and to
the measuring impedance can be detected. The pulse shape recorded during calibration can be
used for comparison with the pulse shape recorded from real PD activity.
Multiple pulse reflections along the cable can make it difficult to check the pulse spectrum when
PD activity arises because one single pulse should be considered for spectrum verification. Anyway,
in the worst case the recorded real PD events are supposed to exhibit the same limited spectrum
and pulse shape like the PD calibrator pulses injected at the cable end.
Therefore, it is always recommended to remain in the lower frequency range, especially for large
test objects and test objects with windings and for advanced users to take the opportunity to
confirm the correctness of the measured charge values by cross-checking the spectrum of the real
PD event.
26
3.1.7 Real field measurement example

Medium size test object with winding has been tested for PD activity at the nominal voltage. By default,
the measuring bandwidth of 100 – 600 kHz has been selected to reach good signal to noise ratio (SNR)
and hence good background noise. However, as per description above this frequency is not commonly
suitable for PD measurement including windings. As recommended Step 2: Checking the calibration
spectrum, the calibration spectrum has been checked and found flat within the selected frequency range
– see Figure 3-12 – top.
Figure 3-12 The calibration pulse spectrum (top) and spectrum of the real PD (bottom) measured
on a test object with winding
Measurement results using measuring frequency range 100 – 600 kHz

According to the recommended Step 3: Analyzing the real PD pulse spectrum , the real PD pulse
spectrum has been checked – see Figure 3-12 – bottom. The spectrum gradually decreases over the
frequency and difference in amplitude between 100 kHz and 600 kHz is around 20 dB. This results in
invalid quasi-integration process and hence wrong charge reading.
Measurement results using measuring frequency range 100 – 250 kHz
After adjusting the frequency measuring range to the correct range (flat spectrum response), it can be
seen that the charge amplitude Qiec increased by factor 2, see Figure 3-14 in comparison to Figure 3-13.
Figure 3-13 Q(t)&U(t) diagram (left) and PRPD pattern (right) @ 100 – 600 kHz
In this particular case the PRPD pattern has not been significantly negatively affected in the mean of
shape (only in the mean of amplitude). However, this does not have to be a rule and differences in the
mean of PRPD pattern shape can be also recorded.
27
Figure 3-14 Q(t)&U(t) diagram (left) and PRPD pattern (right) @ 100 – 250 kHz
This example demonstrates the importance of either checking the real PD pulse spectrum or following
the IEC and IEEE recommendation and keeping the higher cut-off frequency f2 at lower ranges –
especially for test object with windings – typically f2 ≤ 300 kHz – as per multiple examples in this
chapter.
3.1.8 Summary of IEC 60270 calibration procedure
The general conclusions of the chapter before can be summarized as follows.
• Especially for large test objects and test objects with windings it is recommended to
measure in the lower frequency range (upper cut-off frequency < 200 to 300 kHz).
• The nature/geometry of the test objects creates shunt-paths and by-passes caused by the
stray capacitances across conductive parts which allows the LF components of the PD pulse
to appear at the test object terminals with the least attenuation possible.
• Conventional PD measurement at higher frequencies is less meaningful since the frequency

response of the test connections and the measuring loop curtail the maximum frequency
of the PD pulse to less than a few MHz (~ 2 MHz).
• Resonances in the PD measuring band invalidate the calibration based on quasi-

integration.
• The main purpose of the IEC standardization process is to define test procedures that
provide repeatability and inter-comparison.
• Measuring at frequency ranges outside the IEC spectrum brings risk of greater error and
defeats the purpose of standardization.
• Checking the frequency spectrum of the real-PD pulse is highly recommended as it

increases the reliability of the measurements.
• The term “apparent charge” defined by IEC 60270 well reflects the strong relationship
between the real charge (at the PD origin) and the measured charge (at the test object
terminals). Renaming the term “apparent charge” to “induced charge” might be considered
as proposed in [18] and [19].
During the FAT, the quality of the device under test needs to be ensured, so it is mandatory strictly to
follow IEC 60270 requirements. As described according to the current existing knowledge electrical
measurement according IEC 60270 is the only measuring procedure which enables user to conduct
plausible, repeatable and comparable PD measuring measurement. In case of SAT, the situation might
be complicated due to high background noise level, which cannot be accepted during the FAT. In that
case alternative methods might be recommended, as per chapter 7.3.
28
However, it needs to be considered that measuring on the particular resonances does not have reveal
all PD activity since various PD defects have various PD pulse distribution paths and hence different
resonance frequencies.
In addition, it can be strongly recommended to measure beyond the basic requirements of the IEC
60076-3 in the mean of conducting the PD measurement only at the HV terminals (above 72.5 kV).
During the FAT, this approach does not have to reveal all the PD activity – especially if the PD is in the
region of the core, the earthing sections or LV windings. In that case, PD signals are attenuated by a
measuring network acting as a low-pass filter and may not be measurable at all. Hence, it is strongly
recommended to measure on HV and LV winding and ground if applicable, to ensure there is no PD at
the nominal voltage.
3.2 Acceptance criteria according IEC 60076-3

In the 1st edition of IEC 76-3 in 1980 PD measurement was established as one of two alternatives for
transformers with a rated voltage Um of 300 kV and higher. The test could be done in single phase or
three phase connection and the duration for the PD measurement was 30 min.
Alternative 1 was an ”Induced overvoltage test“, which means only withstand test without any PD
measurement, whereas alternative 2 was an ”Induced overvoltage test with PD measurement“. Limits
for the PD value could be either 300 pC at 1.3 Um or 500 pC at 1.5 Um and the test program was handled
in different ways depending on agreement, e.g., recording of the PD level at voltage steps of 20%.
The 2nd edition of IEC 60076-3, which was established in 2000, extended the idea of PD measurement
to lower voltages (see Table 2), but only based on short duration measurements.
For transformers with rated voltages Um above 170 kV a 30 or 60 min (depending on Um , 60 min for
Um above 300 kV) PD measurement was foreseen, which could be done in single phase or three phase
connection (called ACLD). Limits for PD values were 500 pC at 1.5 Um for the test duration plus 100 pC
at 1.1 Um for 5 minutes.
For transformers with rated voltage Um from 72.5 kV to 170 kV a 20 min measurement was to be done
with different voltage steps, each 5 min. This test consists of a three phase test for all transformers and
an additional single phase test for transformers with non-uniform insulation and was more a withstand
test with additional PD measurement than a long term PD test like the ACLD. Especially the single phase
test was a replacement for the applied voltage withstand test, which cannot be carried out at the test
level of the line terminals at transformers with non-uniform insulation. As the applied voltage test does
not include a PD measurement the PD test would also not be necessary during the single phase ACSD
test which was corrected in the 3rd edition of IEC 60076-3, where the single phase induced test, called
LTAC, is now without PD measurement. [1]
29
Table 2 Requirements and tests for different categories of windings [1]
In the 3rd edition of IEC 60076-3, valid since 2013, the only test with PD measurement is now the
“induced voltage test with PD measurement”, IVPD (see Table 3). This test is applicable for all
transformers with rated voltages above 72.5 kV and the test duration is now always 60 minutes.
Table 3 Requirements and tests for different categories of windings [1]
Limit for the continuous level of PD was reduced from 500 pC to 250 pC during the one hour test period.
Nevertheless, it is mentioned in Annex A.2:
This method of calibration provides for measurement of apparent charge at the terminal, which
is the basis of this standard, but it does not give the real value of partial discharge occurring
away from the terminal because the transmission path between the discharge and the terminal
is not included in the calibration.
30
This means the real value of the partial discharge is unknown and so the measured apparent charge is
a weak criterion for the severity of a PD source, as very different PD sources might result in the same
apparent charge at the terminal. Also, the same source at different location inside of a transformer
might result in different measured apparent charges at the terminal.
Therefore, the following acceptance criteria are defined [2]:
Test acceptance criteria (IEC 60076-3 chapter 11.3.5):
The test can only be considered valid if the measured background PD level does not exceed
50 pC at both the beginning and the end of the test. For tests on shunt reactors a background
PD level of up to 100 pC may be accepted.
The test is successful if all the following criteria are fulfilled:
a) No collapse of the test voltage occurs;
b) None of the PD levels recorded during the one hour period exceed 250 pC;
c) The PD levels measured during the one hour period do not exhibit any rising trend and no
sudden sustained increase in the levels occur during the last 20 min of the test;
d) The measured PD levels during the one hour period do not increase by more than 50 pC;
e) The PD level measured at a voltage level of (1.2 × Ur )/√3 after the one hour period does not
exceed 100 pC.
If criteria c) or d) are not met, the one hour period may be extended, and these criteria will be
considered to have been met if they are fulfilled for a continuous period of one hour.
As long as no breakdown occurs and unless very high partial discharges are sustained for a long time,
the test is regarded as non-destructive. Failure to meet the PD acceptance criteria shall therefore not
warrant immediate rejection, but lead to consultation between purchaser and manufacturer about
further investigations. Suggestions for such procedures are given in Annex A [1]
31
4. Electromagnetic Method (UHF) and its

Calibration
PD related radiation of electromagnetic (EM) waves was first investigated and used as a PD detection
method on gas insulated switchgear (GIS) [27], [28]. EM radiation of PD sources was also found for PD
under oil within power transformers [5], [29]. The principle of the method is expressed in Figure 4-1.
Figure 4-1 Principal of UHF PD Measurements in power transformers [5]

At the beginning of the JWG UHF PD measurments were not well defined. There were only non-
characterized sensor at hand. These were inserted somehow into the transformer and some impulses
could be measured with e.g., an oscilloscope. The interpretation of the measurement results depended
on the experience of the person responsible for measuring. Comparable measurements were not
possible. Nevertheless, the method has been shown to be sensitive. Its application during FAT seems
possible and offers some advantages for SAT conditions like:
- No galvanic connection to high voltage necessary (safety)
- Easy applicable
- Often the transformer tank acts as a Faraday´s cage to shield against common disturbances
Consequently, in the current edition of IEC 60076-3 [1] the UHF method is not used as acceptance
criteria but only for diagnosis. It can be used as a trigger signal to support localization or even for
localization itself in case the acceptance criteria are not met (Annex A.4 of IEC 60076-3)
To achieve comparable measuring results, UHF measurement systems require a calibration including a
validation of the UHF sensor sensitivity. This brochure has the intention, to recommend a calibration
procedure to allow reproducible and comparable UHF measurements independently from model and
make of used measuring equipment including cable, accessories and sensors.
4.1 UHF sensors

Two different methods are commonly used for installing UHF Sensors on the transformers. In the first
and recent years, these UHF sensors are installed using oil drain valves, which are increasingly fitted to
new transformers specifically for condition monitoring purposes [30]. In the second method, a mounting
hole is prepared either on the main tank or a removable hatch cover plate and a robust dielectric window
is attached to it to maintain the oil seal. A planar antenna can be mounted on this dielectric window or,
in more advanced implementations, the window can take the form of a “pocket” for mounting the sensor
through the hole for better sensitivity [12]. The need to make some holes in the tank introduces severe
limitations to the implementation of this method for installed transformer but may be well established
during manufacturing for new transformers for FAT/SAT condition.
32
4.1.1 Valve UHF sensors

Most available UHF sensors are designed for standardized DN50 or DN80 gate valves, often referred to
as “drain valves”. The sensors also fit to guillotine and ball valves. In some cases, an adapter flange
may be required. Other valve types without a straight-through opening (e.g., globe, butterfly, and
diaphragm valves) are not useable for these UF sensors. In the following Figure 4-2 and Figure 4-3
valve types suitable and unsuitable for UHF valve sensor installation are shown, respectively.
Figure 4-2 Potential valve types for retrofit of UHF sensors; valves with straight-through
opening/duct [12]
Figure 4-3 Oil valves without straight opening/duct; retrofit of valve-type UHF sensors not possible
[12]
Valve UHF sensors (Figure 4-4) for e.g., DN50 or DN80 gate valves (and other straight through oil drain
valves) can be used for retrofit of transformers during operation.
Figure 4-4 Example of an UHF sensor for DN50/DN80 gate valves [31]
With these sensors, the insertion depth is crucial for the sensitivity. The positioning of the oil valves and
thus the sensors is given by other conditions (usually one valve for oil filling in the upper part of the
transformer, as well as one drain valve in the lower part).
4.1.2 Window UHF sensors
On new transformers, UHF sensors can be installed directly onto the tank wall; such UHF window type
sensors do not need oil valves for installation. They can be installed at a stainless steel flange with a
dielectric window at the transformer tank wall according to Figure 4-5. CIGRÉ already provides a
recommendation to use dielectric windows at new transformers in TB343 [30] and gives a
recommendation on its design in TB 662 [12].
33
Main advantages compared to valve UHF sensors are:

• Better high frequency earthing, which provides better suppression of disturbance signals
• Flat frequency response without resonance of tube of drain valve
• Compatibility with sensors that are designed according to TB662 recommendation
• Easier and safer installation without oil handling, simpler / robust sensor mechanics
(precondition: windows already installed at transformer)
Figure 4-5 Drawing of stainless steel flange and dielectric window [12]
Figure 4-6 shows two different UHF window type sensors (Type A, see left figure with blue case and
Type B, right figure with fully aluminium case. These sensors are compatibile to both, the dielectric
window and seal welding flange proposed in TB662 [12]. The exploded view in the second to right
picture illustrates in general the single components with the stainless steel welding ring on the
transformer tank (green), the inset dielectric window with two sealing rings and window sensor on top.
Figure 4-6 Examples of UHF Plate Sensor with stainless steel flange and dielectric window
compatible with TB662 [12] [32]
Another example of a designed UHF sensor for the installation in oil-immersed power transformer (not
compatible with TB 662 recommendation because of the dimensions and there is no dielectric window
window) is shown in Figure 4-7. This sensor does not use a dielectric window and therefore the sensor
is fully integral with the tank.
34
Figure 4-7 Example of UHF Sensor and interface flange without dielectric window
In Figure 4-8 another example of an UHF sensor for the installation at aforementioned flat dielectric
windows in oil-immersed power transformers, which is also not compatible with TB 662 drawing.
Figure 4-8 Example of UHF sensor mounting on a flat dielectric window

For safety reasons, dielectric windows need to be placed in regions with low electrical field strength. In
areas with high electrical field strength, the air inside the pocket of the dielectric window could lead to
PD. It is advisable to contact the transformer manufacturer and to follow these estimated guidelines
regarding the minimum installation distances with respect to HV potential in the active part. For rated
voltages of 420 kV, 245 kV and 123 kV, minimum distances from HV potential of 1.5 m, 1 m and 0.8 m,
respectively, are recommended.
4.1.3 Specialized UHF sensors
In cases where the oil valve type does not allow for insertion of a normal UHF probe shaft, it may be
possible to use a special sensor that can be “steered” through a more indirect path.
In certain cases, it can be possible to use existing gasket gaps between flanges on the tank to detect
UHF signals from PD inside the transformer. This depends on the transformer design as it is important
that the gaskets provide an insulated not metallic covered gap between adjacent bolts which can work
in the same principle as used for window sensors and will allow UHF waves to escape the tank and to
be captured by these specialized UHF sensors. Those special UHF sensors are for diagnostic purposes
and are not considered further in this brochure.
4.2 Sensor characteristics

It appears to all members of the JWG that a frequency range between 300 MHz and 750 MHz is sufficient
to detect PD under oil. The sensors which can be used for improved FAT and SAT should have attached
a documentation of the frequency response. There are different ways of achieving that, e.g., oil-filled
gigahertz transverse electromagnetic (GTEM) cell, air-filled GTEM, tank measurements.
4.2.1 Techniques for measuring the frequency response of UHF sensors
The application of UHF methods to transformers is a progression from their historical application to GIS.
Experience gained with GIS applications over several decades provides a platform on which to build
some of the procedures relevant to transformer applications. Most UHF sensors can be categorized as
electric field sensors and they usually consist of an open-circuit antenna electrode coupled into a 50 Ω
coaxial cable that connects to a measurement instrument. Such sensors generally have a capacitive-
type coupling characteristic at lower frequencies, but this changes to a more complex and potentially
35
resonant response at higher frequencies where the electrode dimensions are no longer insignificant
compared to the signal wavelength.
The National Grid Company in England and Wales recommended procedures and test limits for
calibrating UHF PD sensors for GIS more than 20 years ago [33], [34]. The method was based on
measuring effective height (He) of the sensor, defined as the sensitivity of the 50 Ω-loaded UHF sensor
in units of mV output per V/m of UHF electric field input, which causes the measurement unit of
sensitivity to be mm (hence the term effective height).
An important concept introduced at that time was a requirement to reproduce the sensor mounting
arrangement during the calibration process so that the sensitivity measurement gives a more accurate
measure of sensor performance in the mechanical arrangement as it will have to operate while in
service. For example, this applies to window-mounted UHF sensors, where the attenuation of the
window structure can have a dominant influence on the response of an otherwise sensitive UHF sensor
[35].
Since the dielectric constant of SF6 is very close to unity (as for air), UHF sensors for GIS have always
been calibrated in air, usually in a GTEM cell [36], as shown in Figure 4-9. However for sensors installed
on liquid-immersed transformers, the permitivity of the liquid will significantly change the response
compared with a measurement in air, since the dielectric constant is approx. 2.2 for mineral oil and 3.2
for ester liquids. An alternative GTEM cell has therefore been introduced, which can be filled with mineral
oil to provide a test environment that better reflects the operating condition of a UHF sensor on a
transformer [37]. At the same time, the measurement quantity antenna factor ( AF) has been introduced
to characterize the sensitivity as a function of frequency when the sensor is surrounded by oil. Since
the units of AF and He are interchangeable through a mathematical conversion, the measurements can
be directly compared.
Figure 4-9 An air-filled GTEM cell used to measure the sensitivity of UHF PD sensors for GIS.
Note that the sensor mounting facility is reproduced on a test plate specific to each type of sensor
However, the most important question concerns how the response of a sensor measured in oil compares
with its response measured in air and, most importantly, to understand whether a measurement in air
can be scaled to predict the response in oil. This task has been addressed by a subgroup of the JWG
and is discussed in chapter 4.2.5.
4.2.2 Characterization of UHF sensors in oil
The sensor sensitivity depends on its design in relation to the electromagnetic wavelength. Antenna-
type sensors are described by different characteristics, e.g., by the antenna gain or the antenna
aperture. For sensors which are not defined by a physical area, such as monopoles or dipoles, the
effective height He, or the antenna factor AF can be used. AF is defined as a function of signal frequency
f as follows:
E( f )
AF ( f ) = (3)
U( f )
Where U(f) is the voltage at the sensor terminals (with 50 Ω load) and E(f) is the electric field strength
incident on the sensor (with the electric field vector aligned in the direction of the sensor axis). A
36
specially designed oil-filled GTEM cell [37] is used for the evaluation of the sensor AF, as shown in
Figure 4-10.
Figure 4-10: a) GTEM cell internal view with absorbers and septum before oil filling
b) GTEM cell external view [37]
A GTEM cell is an expanded coaxial conductor where a defined electromagnetic field can be applied to
equipment under test (EUT), here the UHF sensor, without interference from the ambient
electromagnetic environment.
In the cell, a test volume is defined in which the sensor is situated. In this volume the cell ideally
generates a homogeneous electric field distribution Ehom and an orthogonal magnetic field of the TEM
wave. In addition, the electric field strength Ehom in the test volume has to be known for AF calculation
of the sensor. The AF of a UHF sensor can be determined using a transmission factor (S21) measurement.
The entire setup consists of the oil-filled GTEM cell with inserted UHF sensor and the vector network
analyzer (VNA), as shown in Figure 4-11.
Figure 4-11 Transmission measurement (S21) for AF determination

a) UHF sensor direct mounted to the cell without oil valve
b) UHF sensor installed via DN50 oil valve to the cell [37]
In this setup, the input port of the GTEM cell is excited with a sinusoidal frequency sweep from 300 kHz
to 3 GHz generated by the VNA. The second port of the VNA simultaneously measures the resulting
voltage at the output of the UHF sensor. The resulting transmission factor S 21 can be converted into the
AF of the UHF sensor if the electric field strength in the test volume is taken into account. Two different
AF measurements of a UHF valve UHF sensor are presented in Figure 4-12, showing the significant
influence of the sensor mounting arrangement.
37
Figure 4-12 Antenna factor (AF) of a UHF sensor measured in GTEM cell illustrating the influence of
the different mounting arrangements as per Figure 4-11 a) and b) [38]
That UHF sensor has the highest sensitivity in the frequency range of 200 MHz up to 900 MHz. The
measurement of the red curve is done without a standard oil valve at the GTEM cell. The influence of
oil valve on the sensor’s AF is not negligible, as Figure 4-12 shows. The highest influence of the valve
occurs at around 300 MHz (resonance). Compared with a real transformer setup, the GTEM cell
characterization measurement only considers the influence of the sensor itself, but not the surrounding
structure. Therefore, the sensitivities of different UHF sensors can be compared to each other, and
consequently corrected in a calibration process.
4.2.3 Comparison of valve UHF sensors and window sensors
Due to the influence of the oil valve on the sensor sensitivity, these valve UHF sensors are less sensitive
than window sensors integrated into the tank wall. This influence is reflected in a resonance in the
frequency range of a few 100 MHz, which is often used for UHF measurements, see Figure 4-12.
For new transformers, window sensor can be placed more unrestricted on the tank as valve UHF sensors.
Window sensor has a better high-frequency earthing than the valve UHFsensor, which leads to lower
disturbances from the surrounding. In addition, it has no negative influence on the sensitivity from the
pipe section of an oil valve, as at the valve UHF sensor. Figure 4-13 shows a comparison of the sensor
sensitivity between a valve UHF sensor (here without the influence of the oil valve) and a window type
sensor. Furthermore, the UHF output of a combined PD sensor (combined UHF and acoustic in-oil PD
sensor [39]) is compared. Drain valve and window type sensors show similar sensitivity
Drain Valve Combined PD Window

Adaption Sensor Sensor Sensor
flange
Outer
conductor
Dielectric
window
Septum
Outer
conductor
Figure 4-13 Comparison of the sensitivity between valve UHF sensor at 50 mm insertion depth and
window sensor (and combined in-oil UHF & acoustic PD sensor) [40] [41]
Note: This is only correct if the negative influence of the oil valve is neglected, at real installations at
power transformers, the valve UHF sensors are less sensitive than the window type sensors, due to the
resonance of the valve.
38
4.2.4 Insertion depth of valve UHF sensors

Various UHF performance check measurements on transformers (one exemplarily shown in Figure 4-14),
show that valve UHF sensors, which are still in the pipe section of the oil valve (pos. 0 & 1) are still
shielded against UHF signals from inside the transformer. In order to achieve sufficient sensitivity, the
sensor must protrude into the transformer tank (pos. 2). Further insertion (pos. 3 - 6) does not
significantly increase the sensitivity and can lead to safety risks.
Figure 4-14 Dependency on the insertion depth using drain valve UHF sensors [40]
The trade-off between safety and sensitivity leads to the general recommendation of 50 mm insertion
depth for valve UHF sensors. This insertion depth of the sensor is also achieved with the window
sensors, which are manufactured according to the design specification of the Cigré brochure TB 662
[12].
4.2.5 Equivalence of different UHF PD sensor characterization methods
Although oil is a more appropriate medium for testing transformer sensors, there are some inconvenient
handling issues (such as needing to clean the sensors after testing) and the requirement to lower the
oil and refill / air bleed the GTEM cell between each test (this drawback is only valid for valve UHF
sensors since UHF sensors using a dielectric window can be tested without oil contact in an oil-filled
GTEM). Furthermore, the impedance match of the GTEM cell absorber end is difficult to optimise in oil
for minimal internal reflections of the test signal. Hence it is useful to consider how a calibration
performed in air can be scaled using theoretical considerations to plot and measure the sensitivity in
oil. Sensor frequency responses represented as either AF or He are interchangeable in the following
manner [42]:
The effective height He of a UHF sensor is a measure of the output voltage that it delivers into a 50 Ω
load for a given electric field strength of the UHF signal inside the test object. A typical sensor might
have an average sensitivity of He = 10 mm, which means that it will deliver 10 mV of output signal for
1 V/m of incident electric field: 1 V/m × 10 mm = 10 mV.
The antenna factor AF of a sensor is a logarithmic representation of the inverse of effective height.
Therefore, AF defines the input signal (electric field, Vm-1) required to produce a certain output signal
(V). To convert (using units of mm for He), AF is defined as:
1000
𝐴𝐹(𝑓)[𝑑𝐵 1/𝑚] = 20 ∙ 𝑙𝑜𝑔10 ( ) (4)
𝐻𝑒 (𝑓) [𝑚𝑚]
The corresponding reverse conversion formula is:
1000
𝐻𝑒 (𝑓)[𝑚𝑚] = (5)
10(𝐴𝐹(𝑓)[𝑑𝐵 1/𝑚]⁄20)
Therefore, He = 10 mm corresponds to AF = 40 dB 1/m. Since the sensitivity of a UHF PD sensor will
be frequency dependent, the conversion formulae are applied to each data point in the frequency
response plot. The two different representations of the frequency response for a 25 mm monopole
sensor are compared in Figure 4-15 [43].
39
Figure 4-15 Comparison of the effective height and antenna factor representations of a 25 mm
monopole frequency response [43]
Since the main purpose of calibrating the response of internal UHF sensors for transformers is to
determine their sensitivity when immersed in oil, it would be convenient if the measurements in air
could be scaled to predict the frequency response in oil. There are two key factors that alter the response
of a UHF electric field sensor when the permittivity of the surrounding dielectric medium is changed by
a factor 𝜀𝑟 :
1. The wavelength of the electromagnetic field is reduced by a factor √𝜀𝑟 , which means that the
physical dimensions of the sensor are larger compared to the wavelength than they would be
in air. In this sense, the ratio of wavelength to sensor dimensions in oil at frequency 𝑓 would
correspond to the ratio in air at frequency 𝑓 √𝜀𝑟 . To scale from air to oil we must therefore scale
in the frequency domain by a factor 1⁄√𝜀𝑟 [5].
2. The electric flux density 𝐷𝑖 which the sensor experiences is scaled according to the relationship
𝐷𝑖 = 𝜀𝑟 𝜀0 𝐸𝑖 . Therefore, for the same incident electric field 𝐸𝑖 , the flux density (and therefore
the charge induced on the sensor) is scaled by the factor 𝜀𝑟 . In terms of the equivalent circuit
model presented in [43], the influence is to increase the sensor capacitance by the factor 𝜀𝑟 ,
which reduces the sensor’s voltage source impedance 𝑍 by the same factor. At frequencies
where the capacitive impedance is relatively large compared to 50 Ω, this results in an increase
in output voltage by a factor of approximately 𝜀𝑟 due to the reduction in 𝑍.
The preceding hypothesis concerning scaling factors can be illustrated through a comparison of
monopole-type electric field sensors, which were characterised in both the air- and oil-filled GTEM cells.
Note that the dielectric constant of the oil used was 𝜀𝑟 = 2.2.
The first sensor was a 25 mm monopole probe as defined in [43]. Figure 4-16(a) compares the two
responses before any adjustment was applied. An interesting observation is that the sensitivity in oil
appears higher than in air, which agrees qualitatively with the expectation that the induced electric flux
will be greater in oil. Figure 4-16(b) shows the effect of scaling the air measurement by compressing
its frequency scale by a factor √𝜀𝑟 and increasing its sensitivity by the factor 𝜀𝑟 . With this adjustment,
the two results are in much better agreement. Another example for a longer monopole that has a
resonant peak in sensitivity within the measurement band is shown in Figure 4-17
40
(a) (b)
Figure 4-16 (a) Comparison of unadjusted effective height measurements for the 25 mm monopole
sensor in oil and air. (b) Comparison after the data for the air GTEM had been scaled for both
wavelength and dielectric constant [43]
(a) (b)
Figure 4-17 Frequency response of a 62.5 mm monopole showing the measurements in air before
and after scaling for wavelength and dielectric constant:
(a) Plotted as effective height He. (b) Plotted as antenna factor (AF) [43]
The study reported in [43] validated an empirical scaling procedure for determining the response of a
UHF PD sensor under transformer oil from its response measured in air. The method accounts for the
effect of dielectric constant of the liquid insulating medium on the wavelength of electromagnetic waves
and the electric flux density which the sensor experiences when installed on an oil-filled transformer
and can be summarized as follows:
1. Sensitivity of the sensor in air (at each measurement frequency) should be increased by a scaling
factor 𝜀𝑟 .
2. Next, the scale of the measured frequency range should be compressed by a factor √𝜀𝑟 .
3. Finally, the increased sensitivity data is plotted on the compressed frequency range to produce the
frequency response characteristic of the UHF PD sensor in oil.
Note: Due to the frequency scaling stage, it is necessary to measure the sensor response in air up to a
maximum frequency that is a factor √𝜀𝑟 higher than the upper frequency required for its response in
oil.
The proposed approach will simplify the characterization of UHF sensors for transformers and should
also work for insulation liquids having different permittivity values, where it would otherwise be
necessary to design and construct a bespoke oil-filled GTEM cell for impedance matching using different
geometrical scaling factors for different types of insulating liquids.
41
4.3 Recommendation for Placement of UHF Sensors on Power

transformers
To be used as a tool for acceptance test, it is necessary that the sensitivity of a PD measuring technique
is sufficient to detect all PDs within a power transformer. Based on experimental results obtained from
a 300 MVA, 420 kV transformer, a single Ultra-High Frequency (UHF) sensor is unable to provide
sensitive coverage of the entire tank without the signal becoming noisy [44] [45]. Therefore, at least
two sensors are required to provide complete coverage of the tank. Also, for the performance check
procedure, two sensors are necessary.
Based on simulations carried out on a validated model of a power transformer [46], results show that
when a PD source is situated inside the windings, the electromagnetic waves cannot propagate through
the outer layer windings. Therefore, after hitting the inner surface of the outer winding, the waves are
reflected radially and axially. The axially reflected waves emerge from the top and bottom of the
windings and eventually propagate through the oil space in the tank. A similar phenomenon happens
with the radially reflected waves after reaching the opposite side of the windings. Figure 4-18 shows
the propagation of the waves. This phenomenon has implications on the positioning of sensors in that
the sensors should be ideally placed above and below the highest and lowest points of the windings,
respectively. The installation height thresholds are represented by the dotted lines in Figure 4-18.
Bushing Turret
Core
Artificial PD Source (Phase V Position 4)

Windings
Figure 4-18 Electromagnetic wave propagation in the tank (Side view) [31]
Results show that installing one sensor on each of the two larger side-walls of the transformer tank
provides complete coverage of the tank. The sensors should be ideally placed at the maximum possible
spatial distance from each other and in an area with low electric field stress, i.e., away from the windings
and HV lead exits.
Experiments performed on a transformer show that UHF sensors installed on the walls along the length
of the transformer tank perform better when the propagation of the electromagnetic waves from the
source to the sensor is direct, i.e., with minimal obstructions. If the PD occurs near the lead exits, then
the propagation path can be assumed to be direct. However, PD occurring inside the windings will
invariably have an indirect propagation path to such sensors. Therefore, sensor positioning should be
done based on the signal attenuation with respect to both signal propagation distance and propagation
path. Evaluation of the sensor performance shows that the sensors with the lowest signal attenuation
are located near the outer return limbs of the yoke.
Therefore, it can be stated that receiving sensors located at these positions have a relatively better
performance compared to the rest. Another advantageous aspect of placing a sensor close to the outer
return limb is that, since the sensor is not in close proximity to the windings, the electric field stress
encountered would also be low. Additionally, receiving sensors within a distance of approximately
100 cm from each other have similar performance.
42
Therefore, it can be safely assumed that two or more sensors installed in a circular region with a
diameter of approximately 100 cm will have similar performance.
Hence, it is preferable to position the sensors near the outer limbs of the core in a diagonal formation
on the opposite sides of the tank wall, as shown in Figure 4-19, where the general regions where
sensors can be installed are denoted by the blue circles, which have a diameter of approximately
100 cm.
y
O
L x
T
C
Figure 4-19 Sensor positioning for PD detection (View from the HV side) [31]
These positions also satisfy the requirements of the height thresholds, and two sensors placed at these
positions should be adequate for a Factory Acceptance Test.
4.3.1 Placement of sensors for partial discharge localization
The possibility of localization of a Partial Discharge (PD) sources is a benefit of the UHF method. Four
or more sensors are required to triangulate PD sources. The following requirements should be met for
best sensitivity: two or more sensors should not be placed close to each other and on the same tank
wall, and the four sensors should not form a geometrical plane. These factors will result in insufficient
deviations between the Time Difference of Arrival (TDOA) of the signal at the sensors when the location
of the PD source is changed, thus leading to lower sensitivity for localization.
4.3.2 Positioning on the walls along the length of the transformer
Based on the aforementioned observations from the measurement data, the optimal configuration would
be four sensors in a cross-diagonal arrangement, as shown in Figure 4-20. The cross-diagonal formation
maximizes spatial distance not only between sensors on the same side but also between those on
opposite sides. Additionally, the formation of a geometrical plane, which can lead to large errors in
localization if the PD source located at the geometrical centre of this plane, is prevented. Both factors
should aid in localization based on TDOA. Reliability can be improved by adding additional sensors. It
should be noted that depending on the design of the tank and the presence of flux shunts, it may not
always be possible to install the sensors at the proposed locations. Therefore, additional positions need
to be explored.
43
Figure 4-20 Sensor positioning for PD localization (View from the HV side) [45] [31]
4.3.3 Positioning on the top of the tank

Based on the phenomenon shown in Figure 4-18, it can be observed that installing sensors on the top
of the transformer tank will have benefits with respect to signal attenuation and propagation time when
compared to the sensors recommended in chapter 4.3.2. Additionally, there is the added convenience
of installation as there are fewer obstructions between the sensors and the active parts of the
transformer. The results obtained from simulations confirm that sensors on the top of the tank can
measure signals with the aforementioned advantages. The proposed positioning of such sensors is
shown in Figure 4-21.
Two sensors should be placed in a diagonal formation to maximize the spatial distance, and the sensor
positions are shown in Figure 4-21. A combination of two sensors on the wall and two sensors on the
top of the tank can be used. Alternatively, the two sensors on the top can be used in conjunction with
the four sensor positions recommended in in chapter 4.3.2.
Figure 4-21 Sensor positioning on the cover plate of the tank (View from the top) [31]
Considering the case where two sensors installed on the top of the tank and two on the tank wall for a
total of four sensors, the arrangement shown in Figure 4-22 is suitable because it allows for maximum
spatial distance between all sensors.
44
Figure 4-22 Sensor positioning on the tank wall with sensors installed on the top for a total of 4
sensors (View from the HV side) [31]
In the latter case, where two sensors are installed on the top in addition to the four on the tank wall,
i.e., a total of six sensors, the arrangement shown in Figure 4-23 is suitable because of the
aforementioned reasons. The decision to use either case depends on the value/priority of the
transformer in question.
Figure 4-23 Sensor positioning on the tank wall with sensors installed on the top for a total of 6
sensors (View from the HV side) [31]
4.3.4 Positioning on the walls along the width of the transformer

For sensors installed on the walls along the length of the transformer tank, there is a reduction in
performance when signals have indirect propagation. However, a sensor installed on the walls along
the width of the transformer shows similar performance for artificial PD sources on either wall along the
length of the transformer, the reason for which could be its approximately symmetrical position with
respect to sources on the HV and MV/LV side of the transformer tank. The signal attenuation is
comparable to signals with direct attenuation signifying that the propagation to this sensor is direct
when the PD sources are in the oil space of the transformer, e.g., lead exits. However, these sides are
generally inaccessible because of the cooling systems, and additionally, installation may be prevented
because of the presence of flux shunts. Therefore, it is not convenient to install sensors in these
locations.
45
4.3.5 Positioning near the corner of the tank

For signals with direct and indirect propagation, no noticeable difference can be observed in the
performance of a sensor installed at a distance of approximately 40 cm from the corner and the rest,
thus suggesting that there is no additional benefit or detriment in installing a sensor in such a position.
For PD location purposes, a minimum of 4 sensors are required, ideally each near a different corner of
the tank and not all lying in the same geometrical plane. Valves for PD sensors must be located in
position where a protrusion of up to 50 mm into the tank is permissible without any risk caused by
enhancement of the low power frequency electric field. Additionally, an area inside the tank 0.25 m
around the valve should be free of any obstructions such as frames, tank wall shunts or the corner of
the tank.
The three sensors near the top of the tank in Figure 4-24 give good PD sensitivity and location resolution
in the upper part of the volume where there tends to be more HV connections (such as the bushing
entry points). The side visible in this sketch is the side with more accurate coverage for PD location so
this should be chosen to coincide with the more complex internal construction if there is any significant
difference between the HV and LV sides in this respect (e.g., tap leads to OLTC).
Figure 4-24 Suggested placement for 4 UHF sensors on a transformer [47]

In case of critical it may be advisable to install 6 UHF sensors for the most accurate PD location, in
which case a preferable arrangement would be that shown in Figure 4-25.
OR
Figure 4-25 Suggested placement for 6 UHF sensors on a transformer (the pattern of 3 sensors
shown on these views would be repeated on both HV and LV sides) [47]
4.4 Recommendations for the Number of UHF Sensors on Power

Transformers
The execution of the performance check as described in chapter 4.8.2 is essential to receive meaningful
and reproducible UHF PD measurements. Hence at least 2 UHF sensors should be available in a
transformer tank. The most logical positions on the site walls may cause problems for the sensor
placement as often the OLTC is installed on one side of the transformer tank. This could negatively
influence the signal propagation.
Based on the measurements above and experience of the authors the recommendation is to place 4
UHF sensors on a transformer tank. Optionally, additional sensors can be placed for bigger size
transformers. See Table 4.
46
Table 4 Recommended number of UHF sensors for power transformers
Minimal configuration 2 sensors (minimum for performance check) (Figure 4-19)

(transformers < 100 MVA)
Recommended standard 4 sensors (minimum for localization purpose) (Figure 4-22)
configuration
High-end configuration 6 - 8 sensors (depending on size of transformer vessel and
importance of transformer) (Figure 4-23)
A reasonable placement of the UHF sensors depends on the position of the active part and other
components inside the transformer tank. The propagation paths of electromagnetic waves from the
signal source to the sensors are complex and usually multiple. It needs to be avoided, that internal
metallic parts like deflector plates are screening the sensor electromagnetically. Some considerations
regarding placement are:
• On the side walls of the transformer tank, the positions of flux shunts and tap changer must be
considered, which have the potential to screen UHF sensors and can lead to a reduction of their
sensitivity.
• The leads to the tap changer at the front and/or the back side have the potential to act as
unwanted shields the UHF sensors.
• By experience, transformers with special lead exits (turrets), extra care needs to be taken to
achieve a sensitive signal decoupling in these areas, as these areas are often prone to PD
activities [48].
In general, lower distances to corners are reducing the sensitivity of the sensors. Therefore, a sensor
has to be positioned in a way there is a fair distance to edges and corners of the tank (min 25 cm).
Figure 4-20 and Figure 4-23 show exemplary good distributions of the UHF sensors also considering
later localization approaches.
For safety reasons, dielectric windows need to be placed in regions with low electrical field strength. In
areas with high electrical field strength, the air inside the pocket of the dielectric window could lead to
PD. It is advisable to contact the transformer manufacturer and to follow these estimated guidelines
regarding the minimum installation distances with respect to HV potential in the active part. For rated
voltages of 420 kV, 245 kV and 123 kV, minimum distances from HV potential of 1.5 m, 1 m and 0.8 m,
respectively, are recommended.
As shown in chapter 4.1.2, window UHF sensors provide key advantages, including reproducible
sensitivity, more linear sensitivity over the frequency range as well as easy oil-free and safe handling.
Following the CIGRÉ recommendation [12] it is recommended to use UHF window sensors for new
transformers. If window sensors will not be installed at new transformers in the beginning, it makes
sense to provide at least dielectric windows (see CIGRÉ recommendation [30]), which will allow the
cost-efficient retrofit of UHF sensors for monitoring or diagnostic purpose (e.g., PD localization) at a
later point in time.
4.5 Calibration of electromagnetic method

The objective of any calibration is to ensure a general comparability and reproducibility of measurements
independent of used devices and setup conditions. It has to be noted, that the UHF calibration should
not be misinterpreted as an attempt to correlate the readings of the UHF method (mV or mV/m) with
the electrical method according IEC 60270 in terms of pico-Coulomb (pC). A measurement system for
radiated signals cannot be calibrated by a system for conducted signals and vice versa.
There are two main factors that must be addressed by UHF calibration. The particular measurement
device including accessories such as cables, filters, attenuators, pre-amplifiers, etc. is addressed in the
first step of the calibration by introducing the calibration factor KM, see next section. The influence of
the UHF sensor, essentially its sensitivity and its conversion from an electric field strength to a voltage
signal, is included using calibration factor KS, see the next section. Using the proposed UHF calibration
method, different UHF measurement systems (including UHF sensors, cables, amplifiers, attenuators,
filters, and measuring devices) become comparable when used at the same measurement position.
47
4.5.1 Calibration method for the measurement devices: KM-FACTOR

The measurement setup consisting of the recorder and connected cables, pre-amplifiers and other
Accessories is exposed with a defined calibration impulse, see Figure 4-26.
Figure 4-26 Calibration Setup: measurement device & cables to derive KM factor [41]
This type of waveform is called pulse modulated RF (PuM RF) and consists of a sinusoidal signal with
fsin = 500 MHz and a duration of Ton = 250 ns repeated two times per phase period (100 times per sec
for 50 Hz systems). A minimum time Ton allows the input state of UHF measurement devices to tune on
the correct signal amplitude. The signal is injected directly into the connector between antenna and
cable/pre-amplifier without the antenna being attached, see Figure 4-26. Hence, all effects of cables
and accessories are included.
From this device calibration, the calibration factor KM can be calculated:
𝑈0
𝐾𝑀 = (6)
𝑈𝑚
𝑈0 … sinusoidal reference amplitude
𝑈𝑚 … voltage reading provided by device
By introducing the KM calibration factor, comparability between different measurement devices using
arbitrary cables and accessories can be achieved. The point of calibration is represented by the
connection between cable connector and the UHF sensor.
Because this procedure incorporates only the measurement system, it is independent of any transformer
it is applied to. Hence, a recalibration is not required if the device under test changes. It is sufficient
only to calibrate a dedicated measurement system (recorder and cable setup) in standard intervals. The
comparability to other calibrated systems would be maintained for any installation on different
transformers, as long as no components of the dedicated system are changed (e.g., cables, etc.).
It has to be noted, that a later explained performance check (section 4.8.2) is additionally recommended
to demonstrate the functionality of the measurement chain on a case by case base.
4.5.2 Calibration method for the UHF sensor: KS-FACTOR
In order to include the sensor’s characteristic into the calibrated path, its frequency depended antenna
factor AF(f) is required. As in section 4.2 introduced, the AF describes the conversion between the
applied electric field strength and the resulting output voltage of an UHF sensor. Due to its definition, a
low AF yields a high antenna sensitivity:
𝐸(𝑓)
𝐴𝐹(𝑓) = (7)
𝑈(𝑓)
𝑈(𝑓) … voltage at the antenna terminals
𝐸(𝑓) … electrical field strength at the antenna of the incident electromagnetic wave generated by PD
48
For drain valve UHF sensors, the actual insertion depth has to be the same both during AF determination
and UHF PD measurement in the transformer. As shown in section 0, highest sensitivity for drain valve
UHF sensors is achieved at insertion depth of minimum d = 50 mm, which is recommended to use for
calibrated UHF measurements. For window type UHF sensors, the insertion depth is fixed.
The known transfer function provided by the AF(f) allows to incorporate the sensors sensitivity into the
calibration, which is shifting of the calibration point from the cable end to the inside of the transformer.
Hence, the AF needs to be specified for any UHF sensor, which shall be used of UHF measurements.
Therefore, a “calibration sheet” or “routine test report” including the sensor’s AF(f) from the
manufacturer is considered a precondition for UHF calibration. A UHF sensor without known AF(f) on
the other hand cannot be used for a calibrated electromagnetic PD measurement.
The commonly used UHF sensor types (drain valve and window type) consisting of a monopole antenna
provide the highest sensitivity in the frequency range of 300 MHz up to 750 MHz, where the AF is
lowest.
4.5.3 Simplification of AF
In order to simplify the calibration procedure for time-domain broadband measurement systems, the
frequency dependent AF(f) is reduced to a constant calibration factor KS, which is valid with sufficient
accuracy for the most commonly used UHF PD frequency ranges. It is proposed to calculate the constant
factor as the mean value of AF(f) from 300 MHz to 750 MHz. In a first step, the AF(f) is calculated in its
linear form AFlin(f).
𝐴𝐹(𝑓)𝑑𝐵⁄𝑚
𝐴𝐹𝑙𝑖𝑛 (𝑓) = 10 20 (8)
In a second step, the mean value in the corresponding frequency range is calculated, which yields
calibration factor KS.
𝐾𝑆 = 𝑚𝑒𝑎𝑛(𝐴𝐹(𝑓)𝑙𝑖𝑛 ), 300 𝑀𝐻𝑧 < 𝑓 < 750 𝑀𝐻𝑧 (9)

Figure 4-27 shows a general example of the correlation between frequency dependent AF(f) and the
calculated simplified antenna factor AFs = KS.
AF/
dB/m
KS
300 750 f / MHz

Figure 4-27 Example of simplifying the AF to derive KS [38]
Using KS of the sensor, the new calibration point can be shifted inside the transformer to the UHF
antenna, see Figure 4-28. Hence, the measurement results of UHF PD systems become calibrated and
comparable values of electric field strength at the sensor are measured in V /m.
Note: Often the effective antenna height 𝐻𝑒 (𝑓) is given in terms of [mm] for UHF sensors instead of
the 𝐴𝐹(𝑓). 𝐻𝑒 (𝑓) can be transferred to 𝐴𝐹(𝑓) by:
1000
𝐴𝐹(𝑓)[𝑑𝐵 1/𝑚] = 20 ∙ 𝑙𝑜𝑔10 ( ) (10)
𝐻𝑒 (𝑓) [𝑚𝑚]
Also, it has to be noted, that the proposed KS calculation is only valid for broadband UHF measurement
systems. For narrowband measurement systems, the actual linear AFlin at the center frequency
AFlin (fcenter) should be used for KS.
49
4.5.4 Calibration procedure of the entire UHF measurement system: KUHF-

FACTOR
The complete UHF calibration factor KUHF is calculated by:
𝐾𝑈𝐻𝐹 = 𝐾𝑆 𝐾𝑀 (11)
Using KUHF, an impulse Ui measured with the UHF measurement system can be recalculated and results
in a value correlated to the incident electrical field strength at the UHF sensor.
𝐸𝑈𝐻𝐹 = 𝐾𝑈𝐻𝐹 𝑈𝑖 [V/m] (12)

This value is named “measured UHF electrical field strength” EUHF and is measured in V/m.
Because this procedure incorporates the measurement system only, it is independent of any transformer
it is applied to. Hence, a recalibration is not required if the device under test changes. It is sufficient
only to calibrate a dedicated measurement system (recorder and cable setup) in standard intervals. The
comparability to other calibrated systems would be maintained for any installation on different
transformers, as long as no components of the dedicated system are changed (e.g., cables, etc.). It
must be noted that a performance check (see chapter 4.8.2) is additionally recommended to
demonstrate the functionality of the measurement chain on a case-by-case basis.
EUHF /
V/m
final
calibration
point
UHF
measurement
Amp.
device
Coax cable
UHF Preamplifier
Sensor (if used)
KS KM
KUHF
Figure 4-28 Calibrated measuring point of the UHF Sensor – independent from transformer [41]
As the structure inside the specific transformer cannot be considered when defining KS for the UHF
sensor by measurement in a GTEM cell, it is only an approximation. Because it is part of the tested
device (the transformer) and not of the measurement setup, it cannot be included in a calibration
procedure. Furthermore, the structure inside of the transformer will have the same influence on all UHF
sensors installed at the same position. Figure 4-29 shows the measurement setup at a power
transformer with UHF sensor installed at a drain valve.
50
EUHF / V/m
measurement
point
UHF
measurement
UHF device
Sensor
Figure 4-29 UHF Measurement setup at a transformer [49]
4.6 Pre-conditions for UHF Calibration

For UHF measurement instruments, used accessories, and UHF sensors, some pre-conditions need to
be assured for the calibration procedure.
4.6.1 UHF measurement systems
For measurement instruments and for accessories like pre-amplifiers, the most prominent pre-condition
is a linear input in terms of both amplitude and frequency.
A measurement instrument needs to prove that its input possesses of two pre-condition:
1) A linear behavior for the complete dynamic range
2) Frequency independency for a specified frequency range
This needs to be proven by the manufacturer supplying a calibration sheet with the device containing
measured data for both frequency and level behavior. As a practical proposal 100 MHz and 10 dB steps
can be tested for the complete dynamic ranges [41]. As an example, assuming 70 dB dynamic range
and the aforementioned frequency range from 300 to 750 MHz, this sums up to a testing matrix with
7 x 5 = 35 values. Further definitions are not part of this TB and should be investigated and discussed
in the future.
Until now, there is no standard available that defines the displayed PD value for UHF systems, as in the
case of electrical PD measurement as per IEC 60270 [50]. It supplies a definition and a calculation rule
for the displayed PD value. IEC 60270 also states this PD value calculation as a requirement for PD
measurement instruments. Not having a displayed PD value, which is the same for all devices makes
calibration a hard task. Without that, the displayed values cannot be used for calibration, as they might
be based on different calculations and consequently, lead to different readings for the same input signal.
As a pre-condition for UHF calibration, the implementation of a well-specified PD value is necessary,
which will then be the same for all measurement systems. A first approach could be to implement a
“largest repeatedly-occurring UHF magnitude” analogous to the rules of IEC 60270 [41].
4.6.2 UHF sensors
For UHF sensors that are intended to be used in a calibrated UHF measurement setup, one pre-condition
is needed: The AF or the effective height (He) of the sensor in oil needs to be supplied by the
manufacturer in a calibration sheet with the sensor. For simplification reasons, the manufacturer should
already include the KS factor to the calibration sheet for the described frequency range from 300 MHz
to 750 MHz.
4.7 Comparison of electrical and UHF measurement procedure

Preamble: numerous publications use “calibration of UHF sensors” as a faulty synonym for the
correlation between measured UHF antenna signals (in mV, dBm, …) and the apparent charge (in pC)
of the electrical measurement [7], [51], [52]. This general correlation is not possible, because both, the
signal propagation of electrical measurement and UHF measurement inside a transformer is intrinsically
unknown. Hence they are not generally comparable [5].
51
As a result, some publications wrongly conclude subsequently, that UHF calibration (but meaning:
correlation) is not possible [52]. Generally, it is important to emphasis, that correlation and calibration
represent two entirely different processes and must not be confused: Calibration aims to eliminate
influences of measurement devices and accessories for one specific type of measurement. It does not
yield any information about the comparability between two physically different measurement
techniques.
Nevertheless, one difference in the calibration set-ups of IEC 60270 and the JWG suggested UHF PD
system components calibration is that the Device under test (DUT)/Transformer is connected in
IEC 60270. Only its’ capacity is considered in the actual calibration. The Transfer function inside the
transformer remains unknown in both cases.Both values, the apparent charge (qapparent) of the electrical
PD measurement and the electrical field strength EUHF, are influenced by their individual unknown
propagation paths inside the transformer. This generally prohibits a direct conclusion from the measured
PD reading to the actual PD source level in both cases.
A general consideration of calibration methods for UHF and electrical measurements is determined in
the following.
Figure 4-30 shows both measurement procedures and all Transfer Functions (TF) which are included in
the individual calibration and the TF which cannot be included in both cases.
The propagation mechanisms of electrical and electromagnetic signals inside the transformer are
fundamentally different and so are the attenuations of the signals. The winding serves as conducting
propagation path for the electrical PD measurement and therefore the acquired frequency band may be
reduced down to a few 100 kHz and even below in order to prevent unacceptable attenuation [21].
TF of propagation path
TF of sensor TF of cable TF of recorder
inside transformer
known
q0 / pC calibration
electrical PD measurement (IEC 60270)
impulse
CK
PD electrical PD
ZM
instrument
included in included in included in

Not known calibration calibration calibration
UHF PD measurement
known UHF
calibration impulse
PD
U0 / mV
UHF PD
instrument
included in
sensor included in included in
calibration device device
Not known using its AF calibration calibration
(KS) (KM) (KM)
Figure 4-30 Comparison of PD measurement setups and their individual calibration point [49]
52
The propagation of the electromagnetic signals in the UHF range is a radiated emission in the entire
volume of the transformer, in oil and pressboard. Thereby, the electromagnetic wave is attenuated and
can be reflected by metallic parts. In both cases, the TF inside the transformer remains unknown.
Because it is part of the tested device (the transformer) and not of the measurement setup, it cannot
be included in a calibration procedure in both cases.
4.8 Performance check of systems and sensors

Generally, the calibration procedure for IEC 60270 measurements is regarded as proof of a sensitive
measurement capability. This is strictly not correct, as it may theoretically be possible to perform a
successful calibration at a transformer bushing without connected windings. In that case, no internal
PD of the disconnected winding are measurable at the bushing terminal. Due to the experiences with
electrical measurements and the measurable ambient noise, normally that special case does not need
to be accounted for, so no performance check of the electrical measurement setup is
necessary/performed.
For UHF measurement systems, proving and confirming the capability of detecting UHF signals from
inside the transformer tank is regarded as more essential. The following chapters describe several
different so called “Performance Checks” to demonstrate the general functionality and sensitivity to
capture UHF signals. Nevertheless, a failing performance check of UHF measurements may lead to
successful measurements afterwards since the UHF emission of an internal PD source may be of a
higher energy/amplitude than the UHF emission of a signal generator. Normally, a failed performance
check can be regarded as a strong indicator for a problematic setup.
4.8.1 Single port performance check
Some available UHF Sensors do offer the possibility of a “performance check” or “sensitivity check”
input, which is also incorporated in IEC TS 62478:2016 [53]. This method uses an artificial impulse
injected to this input port of the UHF sensor, which is measured at the output port of the same sensor.
It does only provide information of a working measurement setup itself (including sensor, cables, and
recorder).
However, be aware that this does not provide any information about the sensitivity of the UHF sensor
to internal PD sources or information about proper installation of the sensor at the transformer. For
example, a UHF sensor which is not installed at the power transformer at all may also show a positive
result at a single port performance check.
Conclusion: To determine if the sensor is proper installed at the transformer and not electromagnetically
blocked against PD signals from inside the transformer, only a “Dual Port Performance Check” (next
chapter) can be used.
4.8.2 Dual port performance check
With a second UHF sensor installed on the transformer, an artificial UHF impulse is injected at one of
the sensors. If this signal can be measured with the second sensor (and further installed sensors), the
performance check procedure proves that both (all) sensors are installed properly on the transformer.
It is not used as a part of the calibration procedure of the UHF measurement system, but needs to be
used to prove that the system is able to measure electromagnetic waves emitted by PD inside the
transformer.
In the case of steel plates in front of the UHF sensor or a standpipe mounted to the oil valve where a
UHF sensor is installed, the sensor is electromagnetically shielded. Hence a measurement system will
not be able to measure UHF PD signals from inside of the transformer and the performance check will
fail. Figure 4-31 shows an example of a passed and Figure 4-32 of a failed performance check,
respectively.
53
second
UHF sensor
EUHF / V/m
UHF
impulse
generator
UHF
measurement
UHF device
sensor
Figure 4-31 Performance check setup with correctly installed UHF sensors [41]
rising
tube
EUHF / V/m
UHF
impluse
generator
UHF
measurement
second UHF device
UHF sensor sensor
Figure 4-32 Failed performance check with one sensor installed in a rising tube at the oil valve
acting as an electromagnetic shielding for the UHF sensor [41]
A performance check on an individual sensor installed at a certain position at a transformer
complemented with the calibration procedure described further above will allow to establish a
comparability of the measurements in terms of amplitude and sensitivity independent of the used
equipment and sensors. It needs to be noted further that no official standards and/or definitions of a
minimum sensitivity regarding transformer UHF PD measurements is presently available.
A performance check on the individual sensor positions at a transformer complements the calibration of
the measurement setup. In other words: only a calibrated UHF PD measurement with a subsequent
performance check ensures the comparability and sensitivity of the individual setup.
4.8.3 Performance check with pulse generator
This procedure is mostly done with UHF impulse generators, designed for UHF sensitivity checks in GIS
systems [54] with an approx. maximum amplitude of Ûpulse =50...100 V (in 50 Ω domain) and an impulse
shape that can be seen in Figure 5.9 a). Experiences with small transformers with respectively shorter
propagation paths and accordingly lower damping of signal energy showed that amplitudes are
appropriate and performance checks can be performed successfully. Larger power transformers with
higher distances between the UHF sensors often provide higher signal attenuation and as a consequence
signals strength at the second (measuring) sensor is below noise level. Hence, the UHF sensitivity check
fails because the signal does not reach the measuring sensor and therefore cannot give information
about that sensor’s sensitivity.
Due to the lack of state of the art UHF impulse generators with higher amplitudes Ûpulse, a high power
pulse generator normally used for EMI/RFI/EMC-applications was tested at a large power transformer.
Its amplitude is approx. Ûpulse= 250 V (in 50 Ω domain), see Figure 4-33 b).
54
Figure 4-33 a) 60 V impulse of UHF impulse generator

b) 250 V impulse of EMI impulse generator
c) both impulses in frequency domain [55]
The impulse shape of the EMI impulse generator (Figure 4-33 b)) seems more suitable for UHF
performance checks at power transformers, as it shows higher amplitudes in UHF frequency range as
can be seen in Figure 4-33 c) between 100 MHz (108 Hz) and 1 GHz (109 Hz).
55
5. Technical considerations of PD
measurements for FAT and SAT
The following chapter includes a collection of general details of PD measurement with the focus on FAT
and SAT conditions. For successful PD measurements, sensors and systems must be designed for the
foreseeable measurable signals and users need to deal with the expectable drawbacks/limitations (e.g.,
noise) with different characteristics for the FAT and SAT conditions. Therefore, the chapter begins with
a collection of PD emissions measurable in the electrical or UHF range. The following chapter introduces
the most common sources of noise that can be experienced.
Important for further analysis of measurable signals is the knowledge, how the measurable signals may
have been influenced on their way from the source to measuring device. The subchapter 0 highlights
the influence of that transmission path of the signals.
Finally, the chapter explains that there is no general correlation between UHF signal quantities and IEC
60270 quantities and hence the application of the so-called “Sensitivity Check” from GIS community is
not workable for transformers.
5.1 Emission spectra of PD

The following subchapter demonstrates the reasonable frequency ranges for electrical and UHF
measurements. As an executive summary, it can be stated:
Electrical Measurements Stay in recommended frequency range of IEC 60270 with focus on
lowest possible frequency band
UHF Measurements Use Frequency range between 300 – 750 MHz
See detailed investigation in following subchapters.
5.1.1 Electrical emission
The electrical method according to IEC 60270 is well-established and there are numerous publications
at hand. Therefore, the [joint] working group decided against collecting new case.
From the broad number of publications it can be stated that:
- PD sources can be regarded as fast transient signal sources. Hence, their frequency range
exceeds the bandwidth of the measurement devices, which represent low pass filters. The
dedicated filter function can be seen exemplarily in Figure 3-2.
- The higher the measuring frequency range, the lower the sensitivity with increasing distance to
the measuring point
For further information consult [11] [12].
5.1.2 UHF emission
Following is a collection of emitted UHF signals from real PD located in transformers. As many published
case studies deal with a general analysis and often localization of a PD source, the challenge for the
working group was, to review those case studies and extract relevant findings for FAT and SAT
conditions. Therefore, this chapter focuses on the measurable signals in time domain with the goal to
demonstrate reasonable frequency ranges where UHF measurements are recommendable. Case study
signals, patterns and its interpretation are furthermore explained in a chapter 6 - Case Studies.
These results of different research teams show that a significant frequency range for UHF PD detection
in transformers is between 300 MHz and 750 MHz.
Case 1 Signals from case study for 333MVA transformer (see chapter 6.1)
In the first example, the measurable signals features frequency contents of up to 1 GHz, see Figure 5-1.
The frequency analysis of the measured signals of the installed UHF probes proved furthermore the
shielding characteristic of the tank, as narrowband noise, e.g., at around 500 MHz for digital video
broadcasting or around 900 MHz or 1800 MHz for GSM, see chapter 5.2.1, do not have relevant
frequency content in the signal. In Figure 5-1, the unamplified measured signal of a UHF probe is shown
with its frequency analyses (FFT).
56
Figure 5-1 Measured UHF signal of probe “UHF 2” to prove broadband emission [5]
The proposed frequency range from 300 to 750 MHz is well covered.
Case 2 Signals from case study (see chapter 6.2)

The test configuration for the UHF measurement was as follows: Each sensor was connected by a 10 m
RG213 coaxial cable with N-type connectors to a 26 dB gain UHF preamplifier. The amplified signals
were captured using an oscilloscope with an analogue bandwidth of 500 MHz and sampling rate of
5 Gsamples/s.
Figure 5-2 shows the largest UHF signals that were recorded during acceptance testing. The signal
(from sensor No. 3) is about 30 mV peak-peak after amplification. This corresponds to a signal level of
1.5 mV peak-peak directly from the UHF sensor. The spectral density of this signal is shown in Figure
5-3, as computed using the FFT method. Note the predominance of signal energy in the frequency
range 500 – 1000 MHz in this case.
No. 1
Amplitude ( 10 mV / div )
No. 2
No. 3
No. 4
0 10 20 30 40 50 60 70 80 90 100
time ( ns )
Figure 5-2 Amplified (×20 voltage gain) UHF signals recorded during overvoltage testing of B-phase
at 1.6 UN with a measured PD level in the region of 20 – 30 pC [56]
57
relative spectral density 0.5
0
200 400 600 800 1000 1200 1400 1600 1800 2000
frequency ( MHz )
Figure 5-3 Frequency spectrum of the signal from UHF sensor No. 3 in Figure 5-2 [56]
Again, the proposed frequency range from 300 to 750 MHz is well covered.
Case 3 Electromagnetic wave spectrum (example 1)

Partial discharge current is a nanosecond impulse signal with wide band frequency, which can generate
ultra-high frequency electromagnetic wave containing wide band frequency information [57]. The band
width of UHF signal is from dozens of megahertz up to gigahertz, so abundant frequency information
of PD can be detected using UHF PD measuring technology.
Partial discharges are detected by means of the electromagnetic waves radiated by the discharge source
with a frequency range which can go up to 3 GHz.
Figure 5-4 Overview of PD patterns, average-/max-hold frequency spectrum of the three defects
obtained with the UHF technique. Frequency plots show noise (grey) and discharge (black) [57]
58
The frequency spectra obtained show that signal spectrums of different defects can be detected in a
range up to approximately 1 GHz.
The frequency spectra of the three defects differ from each other, in particular when the frequency
spectra are post-processed using the average frequency spectrum and the max-hold frequency
spectrum of the defects. This gives information about the repetition rate of the discharges.
The average- and the max-hold frequency spectra of a surface discharge in Figure 5-4, have a similar
shape, but the amplitude is different. This can be explained due the fact that the discharge rate is high,
but the discharges are not continuously present.
Experiments of PD current pulse and phase resolved PD characteristics were performed using various
artificial defects in oil/pressboard composite insulation system in oil-filled transformer [58]. Artificial
defects: barrier insulation model-in-oil, oil wedge model, air void model, floating electrode model, and
oil gap model. As a result, it was found that the rise time, fall time, and pulse width of each PD current
pulse differed depending on each model. It confirmed the existence of PD-radiated electromagnetic
wave with frequency UHF band for the defects. It was shown from these results that UHF method is
applicable for detecting PD originating from the defects and identifying the defect types in oil-filled
transformer.
Figure 5-5 shows five kinds of electrode configuration used in the experiment: air void model, oil gap
model, floating electrode model, barrier insulation model and oil wedge model. Figures right show typical
results of spectra of PD radiated EM wave of each defect model together with background noise (BGN)
measured with UHF sensor. As shown in Figure 5-5 (b), (d) and (e), oil gap, barrier insulation and oil
wedge models have a wideband frequency spectrum up to 1GHz whereas the other two detect models
have mainly frequency spectrum between 300 and 500 MHz.
59
Figure 5-5 Electrodes configuration and Electromagnetic wave frequency spectrum from PD of each
models [58]
60

Various defects that can occur in a power transformer were simulated and studied [59]. Four of these
geometries are, as shown in Figure 5-6, air and oil corona discharges using a point-plane configuration,
creepage discharges by creating discharges along an oil-paper interface parallel to the electric field,
surface discharges using a pressboard sheet between two flat electrodes, and turn-to-turn discharges
using an insulation model of a transformer.
Figure 5-6 Discharge sources: (a) oil/air corona discharge, (b) creep discharge, (c) surface
discharge, (d) turn-to-turn insulation discharge [59]
(a)
(b)
(c)
(d)
(e)
Figure 5-7 Point-on-wave patterns of typical defects: (a) oil-corona, (b) air corona, (c) creep
discharge, (d) surface discharge, (e) turn-to-turn discharge [59]
61
These results of different research teams show that a significant frequency range for UHF PD detection
in transformers is between 300 MHz and 750 MHz.
The artificial defect models were placed into an organic glass container filled with transformer oil. The
UHF antenna was located in the vicinity of the testing models, with a distance of 50 mm between them.
Figure 5-8 Four types of artificial defect models: (a) Corona-in-oil discharge model (b) Surface
discharge-in-oil model (c) Gas-cavity discharge model; (d) Floating-discharge-in-oil model [60]
A digital oscilloscope with sampling frequency of 5 GHz was utilized to record the PD signals.
Figure 5-9 Waveforms and normalized power frequency spectra of UHF PD signals: (1) Tip-in-oil
discharge; (2) Surface discharge-in-oil; (3) Gas-cavity discharge; (4) Floating-discharge-in-oil [60]
62
5.2 Most common sources of noise and measures to optimize SNR

A low noise PD measurement in the UHF range is usually much easier than conventional PD
measurement according IEC 60270, especially for SAT conditions. However, PD measurement in the
UHF range can be influenced by noise. UHF noise is present in all areas of high population density where
radio application and industry are sources of interference.
Disturbance sources independent of test voltage:
- Neighboring high voltage
- Switching operations in other circuits
- Machines with commutator or thyristor power control
- Radio transmitters, mobile communication
- Self-generated noise of measuring set-up
- Etc.
Disturbance sources during high voltage testing:
- PD in the supply
- Harmonics in the test voltage
- PD on high voltage terminals
- Poor contacts
- Etc.
The noise level is very different depending on the location. The power of the noise source and the
transmission path to the measuring circuit will influence the level. Stable or stochastically varying noise
can influence the measuring results in different ways. Noise can be caused also by the test arrangement
itself. The test circuit and the measurement equipment need to be optimized for a sensitive PD
measurement also in the UHF range.
Most common sources of noise are discussed in the following. Again, this brochure focuses on the UHF
technology, as electrical measurements are regarded as well known.
5.2.1 Public broadcasting and mobile communication
Public broadcasting (video DVB and audio DAB) as well as mobile communication are characterized by
stable carrier frequency. Used frequencies are listed in frequency assignment plans of national and
international administrations [61].
Figure 5-10 left) Digital video broadcasting DBV T2 right) Mobile communication, GSM 900,
example of one provider [61]
UHF disturbances from outside the tank become measurable when measuring broadband in time domain
without filtering and using signal amplification (> 20 dB). The easiest way suppressing the measurable
disturbances in that case was by the application of a 300 MHz high pass filter. Disturbances are generally
emitted by walkie-talkies in test field areas, or mobile phones in the vicinity of the probe. External
corona discharges normally do not affect the UHF probes, as shown in chapter 5.2.4. The challenge is
now to identify whether the UHF impulses are originated by an internal PD or by external disturbances.
With a transient recorder with an analogue bandwidth of 3 GHz and with an amplification of 20 dB, UHF
signals are detectable on a 40 MVA transformer, although there have been no indications of internal
PD. A post processing and frequency analyses of the measured signal shows known narrowband
disturbances, see Figure 5-11.
63
a) Electrical measurements b) UHF measurements
Figure 5-11 Narrow band disturbances during online measurements using amplification [5]
Disturbances around 0.54 GHz are caused by the digital video broadcasting service (DVB-T), at around
0.9 GHz as well as 1.8 GHz are mobile phone disturbances (GSM), and at about 2.1 GHz signals of the
Universal Mobile Telecommunications System (UMTS) are detectable
5.2.2 Industrial noise
UHF noises caused in industrial enviroments is mostly an unwanted secondary effect. Compared with
broadcasting or mobile telecommunication, this industrial noise is varying stochastically depending on
the process. It can appear unexpectedly and of course negatively influence UHF PD measurements.
Therefore, a thoroughly reference sweep in advance is usually adviced to get information about these
noises.
5.2.3 Other stochastically varying noise
Other stochastically varying noise sources are collected as follows, without the intention of being
complete.
Radio sources: Amateur radio, private mobile radio, police radio, flight control systems
Other sources: Car ignition, electronic power supplies
5.2.4 HV facilities – corona discharges
Electrical onsite measurements during SAT condition might be affected by corona discharges on nearby
overhead lines. As can be seen in Figure 5-12, corona discharge easily dominates the measurable PRPD
and consequently limit the outcome of an electrical measurement.
Figure 5-12 PRPD pattern of three phase corona of close by overhead lines hindering sensitive
electrical PD measurements [5]
64
As corona discharges can be seen as the most common source of noise/interference for SAT condition,
the UHF technology is investigated on its resilience against them.
For purpose of demonstration, some experimental measurements were carried out on a transformer
planned for scrapping. Internal PD defects are confirmed by DGA results showing increased hydrogen
concentrations. The transformer investigated is a 210 MVA transmission transformer with a voltage
rating of 110/220 kV, see Figure 5-13. The on-site measurements presented were performed off-line,
and an external PD free generator step-up unit (110 kVA) was used to excite the transformer using the
tertiary side to the operating voltage.
Figure 5-13 210 MVA Grid-Coupling Transformer with copper wire (right) on phase L3, 110kV to
stimulate corona discharges [5]
In order to demonstrate the resilience of the UHF measuring method against external disturbances, a
copper wire was attached to the conductor of the bushing of the phase L 3,110. This wire represents an
external PD source (corona), see Figure 5-13 right. Simultaneous measurements according to IEC 60270
confirmed the external PD source activity with a magnitude of approx. 2000 pC.
At the same voltage level as the measurements without the external corona source, the UHF PRPD
almost does not change, see Figure 5-14 right.
Figure 5-14 Comparison of PRPD without (left) and with Corona (right) [5]
Additional UHF impulses emitted by the external corona discharge are not measurable. Hence the tank
wall and the condenser type bushing acting as a low pass filter, shield the UHF probe very well, and the
UHF method is not affected by external corona discharges.
The reason here is that the conductor and the capacitance of the bushings are presenting a low pass
filter with the cutoff frequency well below the UHF frequency range used for transformer UHF PD
measurement. While corona discharge itself is creating also frequencies well above 1 GHz (compare
Figure 5-4), it cannot enter the transformer tank as it works as a quasi-faraday cage and the bushings
itself representing a lowpass filter for any conducted signals. Hence, only lower frequency components
of corona discharges enter the transformer, which are irrelevant for UHF PD measurement. It has to be
noted, that transformers equipped with DIN bushings or non-condenser bushings (e.g. high current
bushings) might not work as (lowpass) filter. Hence conducted corona signals might enter the
transformer and can disturb UHF PD measurements. See also [52] where some of these aspects are
being briefly discussed.
65
Furthermore, time-domain comparisons between internal transformer PD signals and those coupled
from external corona have been presented in [62], as measured using external UHF sensors. The
pertinent information is summarized in Figure 5-15 and Figure 5-16, demonstrating the considerable
immunity of UHF sensors to corona signals coupled directly onto the bushing terminal, which can be
seen in both the signal amplitude and its frequency spectrum. Note that in Figure 5-16, different
amplitude and frequency scales have been used to show sufficient detail for each type of signal. In both
cases, the pre-amplifier with voltage gain ×20 was used. Hence the “raw” signals from the UHF sensors
are about 0.5 mV pk-pk for corona and about 5 mV pk-pk for the internal PD. Comparing the spectra of
the two signals in Figure 4-7(b) and (d), it is clear that the frequencies above 200 MHz of the external
corona signal are well filtered by the bushing and well shielded by the tank, while the internal PD signal
energy is predominantly at frequencies of 300 MHz and above is well represented. Therefore, even
greater discrimination between PD and external corona could be achieved by adding in-line 300 MHz
high-pass filters on the signal from each UHF sensor in this case. It needs to be mentioned, that different
bushings can have different cutoff frequencies due to the difference capacitance of the bushing
condenser core and different impedance of the bushing conductor building the lowpass.
HV 0V
HV 3 GHz
oscilloscope
S1
S2 dual UHF S1
pre-amp
+26
S2
dB
PD detector
transformer
coupling
capacitor
(a)
apparent charge ( pC )
~200 pC
~150 pC
~90 pC
0
0 90 180 270 360
phase ()
(b) (c)
Figure 5-15 (a) Laboratory-based 10/50 kV transformer for PD/corona measurement (UHF sensors
and IEC 60270 system); (b) corona wire at HV bushing; (c) PRPD pattern at 25 kV [62]
66
relative
amplitude
(10 mV / div) spectral density
1
S1
S1
S2
9 ns
S2 UHF band is “quiet”
0
0 50 100 150 200 250 0 200 400 600 800 1000
time ( ns ) frequency ( MHz )
(a) (b)
amplitude relative spectral

(50 mV / div) density
1
S1 S1
S2
S2
0
0 50 100 150 200 250 0 500 1000 1500 2000
frequency ( MHz )
time ( ns )
(c) (d)
Figure 5-16 (a) Time-domain corona signals from UHF sensors at condenser bushing; (b) spectral
density; (c) UHF signals from internal PD; (d) spectral densities. Note the different amplitude and
frequency scales [62].
Another test has been set up in a high voltage laboratory is to demonstrate that external corona
discharges, detectable by conventional PD measurement, being filtered out once a condenser bushing
is used to introduce a lowpass filter. Figure 5-17 shows the test set up which compares between the
PD measurement using conventional method (through coupling capacitor) and UHF method (through
UHF sensor and voltage applied through a condenser bushing).
Figure 5-17 Test set up measuring external corona discharge by conventional method compare with
UHF method [63] as the condenser core bushing acts as a lowpass filter in between
A copper wire was attached to the bushing terminal, and the test transformer supplied test voltages to
the bushing to generate corona discharges (noise). The PD indicator detects signals from coupling
capacitor (conventional measurement), and at the same time from UHF sensor (UHF measurement) as
shown in Figure 5-18. The result shows clearly that the high frequency components of corona PD can
be filtered by a condenser core bushing (working as a lowpass) and the UHF measurement is not
influenced anymore by external corona discharges, whereas the conventional method may be disturbed
in some conditions. This is valid for all discharges happen outside of the tank if the higher frequency
components are attenuated by a bushing type of lowpass and/or being screened by the tank.
67
Figure 5-18 (left) Wire attached to condenser bushing as corona source; (right) PRPD patterns
detectable by conventional / UHF method: Corona detectable by IEC but not by UHF method [63]
5.2.5 Measures to optimize SNR by optimized measuring setup

Proper earthing and shielding is essential for a sensitive UHF measurement. The tank of a power
transformer acts like a shielding but needs to be grounded (via low high-frequency impedance). A
separate earthing of the measuring circuit is recommended to avoid influence by UHF parasitic current
(low inductive earthing on the sensor).
Configuration of the test source:
- PD free HV source
- PD free HV connection
Configuration of the Measuring circuit:
- Suitable sensors and amplifiers (frequency range from 300 – 750 MHz)
- Window sensors (prepare tank during manufacturing) instead of valve UHF sensor, see chapter
4.2.3
- Optimize insertion depth for valve UHF sensors, see chapter 0
- Selective filters to suppress radio application (300 MHz High pass + 750 MHz Low pass)
- Cable with low damping and good shielding (e.g. RG 214)
- Demonstrate functionality by a performance check
- Selection of a low noise area between continuous working radio channels and stochastic noise
sources (low ambient interference signals, e.g., for passive radio astronomy reserved frequency
range 608 - 614 MHz)
- Use of resonance frequency in the measurement circuit (performance check)
68
5.2.6 Measures to optimize SNR by optimized voltage source

For FAT, mainly shielded laboratories and PD-free HV sources are commonly used. Therefore, the
members of the working group agreed that there is no need to advise improvements for FAT.
For SAT conditions, the most challenging part beside the measuring system is the voltage source.
Different voltage sources are used for feeding HV-tests facilities and on-site. Power station generators
(test facilities) and diesel generators (on-site) are in use whereas suitable frequency converters based
on power electronics are developed in the last 10 years and are state of the art for new or modernized
test setups. Such Electronic Power Sources (EPS) are used also on-site because of the compact design
and the technical parameters (suitable power, precise voltage and frequency regulation, harmonics
compensation, …). [64]
Based on the physical principles the IGBTs in the EPS produce high electromagnetic noise caused by
steep switching impulses and the switching frequency of the modulation. Therefore, EPS used for PD
testing have to be build according to special requirements. Compact design, consequent shielding and
earthing as well as several filters are necessary to avoid PD noise coupling into the PD measuring circuit.
For verification of the UHF behaviour, a test circuit with components of a transformer test field for FAT
condition was used. See in Figure 5-19 the EPS type HV source with build-in filters (1) feeding the step-
up transformer (2). The UHF sensor (3) was installed on the drain valve of the step-up transformer.
Figure 5-19 Test arrangement for verification of UHF-PD behaviour of a voltage source for FAT
based on power electronics [64]
Figure 5-20 shows measurement spectra of the background noise (left) and the voltage source in
operation. The recorded spectra in the UHF range of 100 MHz – 2 GHz are very similar in both operating
states of the power source. No influence of the power source can be seen.
Figure 5-20 UHF spectra: left) noise level right) Converter in operation [64]
The measurement spectra in Figure 5-20 (right) shows that a frequency converter does not influence
the basic noise of an UHF measurement carried out at a transformer in the range of 100 MHz – 2 GHz.
Using especially designed frequency converters for testing of power transformers have no limitations
regarding UHF measurements. Even conventional PD measurement according to IEC 60270 can be
performed where a PD level lower than 10pC can be achieved.
69
5.3 Influence of transmission path

Important for further analysis of measurable signals is the knowledge, how the measurable signals may
have been influenced on their way from the PD source to the measuring device. The subchapter 0
highlights the influence of that transmission path of the signals for the electrical as well as for the UHF
measurement method.
5.3.1 Electrical method according IEC 60270
Generally, the signal transmission path can be summarized with Figure 5-21.
Figure 5-21 Signal transmission for conventional PD measuring method

The influence of the location of PD on its electrical and UHF signals is determined using a laboratory
setup. A cylindrical steel tank is equipped with a winding and a stable, synthetic PD source, which is
adjustable in height along the winding. The entire winding is on high voltage potential, it can be
connected by the upper or the lower winding exit to HV potential. For UHF measurements, five UHF
sensors are inserted into to tank: four sensors attached through DN80 drain valves and one UHF plate
sensor welded directly on the top (sensor 5). Figure 5-22 a) shows the experimental setup. The synthetic
PD source consists of two copper plates connected through a capacitor and a gas-filled discharge tube
(GDT). One copper plate is galvanically connected to ground and the other coupled to the HV winding
through its stray capacity [55]. It provides a reproducible, phase stable, constant charge conversion.
The constant original PD charge is 1000 pC and it emits constant electromagnetic radiation and hence
can be used as UHF PD source.
During the experiment, the position of the PD source is changed and the distance between top of the
tank and source is stepwise increased. The electrical measurement is calibrated for each frequency
range used (see below for details). The broadband UHF measurement does not require recalibration
but needs knowledge of the different sensors’ antenna factors, see chapter 3.
The measured sensor voltage of each sensor at changing source location is shown in Figure 5-22 b).
The signal for sensors 1 and 2 decreases if the distance to the source increases (note that sensor 1 is
at the bottom of the tank) as damping is approximately 2 dB/m in a uniform propagation path. Due to
the geometry’s complexity introduced by the tank design, the propagation path damping differs from
linear damping depending on the sensor position, see sensors 3 and 4. One theory states that not the
PD source itself radiates the electromagnetic wave, but the surrounding structure (e.g., the winding)
acts as the active transmitting sensor [5]. Hence, propagation changes at each PD position by the
influence of the changing emitting sensor formed by the winding and the surrounding structure.
70
Figure 5-22 a) Test setup for both electrical PD measurement (IEC 60270) and UHF measurement
b) maximum measured UHF levels of UHF sensors 1-5 [55]
The electrical measurement is performed according to IEC 60270. Two frequency ranges are used:
• IEC 60270 broadband: fm = 300 kHz, Δf = 300 kHz
• IEC 60270 narrowband: fm = 1 MHz, Δf = 30 kHz
Apparent charges are measured using a calibrated system either on the upper or the lower end of the
winding. The results are shown in Figure 5-23 a) (broadband) and Figure 5-23 b) (narrowband). The
black line shows the original PD charge of the source at 1000 pC.
Figure 5-23 Apparent charge of electrical PD measurement at upper and lower end of winding;
dotted line: Actual charge of the source [55]
Determining the broadband measurement, several findings can be stated: Both the upper and lower
end of the winding measurement show a strong dependency considering the distance between the
respective calibration point and the source (coupling path) despite the constant source. In addition, the
correlation between distance and apparent charge is not trivial; there is no monotone or linear
dependency. At particular positions the apparent charge decreases strongly (e.g. at 80 cm for the upper
winding measurement and at 10 cm for the lower winding measurement). The upper winding
measurement shows a local minimum of the apparent charge at 60 cm, whose corresponding effect
cannot be found at the lower winding measurement. In conclusion, the electrical propagation is affected
by non-linear damping and is not reciprocal. The provided measurements show no correlation between
the measured data. The evaluation of the narrowband measurements (Figure 5-23 b) shows a
comparable behavior of upper and lower winding measurements.
The measured apparent charge strongly decreases within small distances between source and calibrated
measurement point (up to approx. 10 cm). At larger distances, the apparent charge stays low but does
not show any monotone or linear dependency.
71
Compared to the broadband measurement at lower frequencies, the narrowband measurement shows
a more expected behavior. Nevertheless, both results come to the same conclusion for the practical PD
measurement with unknown source position: the measured apparent charge cannot be correlated with
the actual charge of the source. Usually, the measured values underestimate the actual PD charge but
within broadband measurements an overestimation is also possible.
5.3.2 UHF method
The attenuation of the transmission path of electromagnetic PD measurement consists of several
elements (Figure 5-24). In this subchapter the influence of the insulation structure (medium) from the
PD location to the UHF sensors is discussed.
The active parts of the transformer, such as the core, winding etc. act as obstacles to the propagation
of UHF signals [65]. PD inside transformer windings, especially, the high voltage (HV) windings is one
of the most commonly occurring faults. Therefore, it is essential to investigate the propagation of UHF
signals when a PD occurs inside a winding. As there are no possibilities to arrange an artificial PD source
within a winding the investigation of the attenuation of the transmission path can just be performed by
a simulation method.
Figure 5-24 Signal transmission path for an unconventional PD measuring system (acoustic, UHF)
Here, the simulation model of the transformer consists of 18 possible PD source locations with 12 inside
and 6 outside the winding, as shown in Figure 5-25 [66]. The six sources outside the windings
correspond to PD locations at the lead exits. Considering a single-phase (U, V or W), the four sources
inside the winding are numbered from 1 – 4, and the two sources corresponding to the lead exits are
numbered 5 and 6.
72
5 5 5
2 2 2 5 6
1 3 1 3 1 3
4 4 4 2 4
U 6 V 6 W 6
(a) (b)
Figure 5-25 Location of the 18 artificial PD sources inside the transformer model. (a) Top view; (b)
Side view [66]
Before analyzing the distance-dependent attenuation, it is necessary to distinguish between Line-of-
Sight (LoS) distance and propagation distance. The LoS distance is simply a straight line drawn from
the source to the receiver and ignores any obstacles that prevent the propagation of the UHF signals,
whereas the propagation distance is the actual path travelled by the UHF signals. The propagation
distance is calculated from the Time of Arrival (ToA) of the UHF signals at the receiving sensors and the
speed of signal propagation in oil (assumed as two thirds the speed of light [5]). The ToA was
determined by using the energy criterion. The energy criterion takes the energy ( Si) of the UHF signal
at a time instant (i) and adds a negative trend (δ) to the value, thus, providing a quantity called the
partial energy (𝑆𝑖′ ). The time instant at which the partial energy plot reaches its global minima provides
the ToA of the UHF signals. The partial energy is calculated using the following equation:
𝑖
𝑆𝑖′ = 𝑆𝑖 − 𝑖𝛿 = ∑(𝑢𝑘2 − 𝑖𝛿) (13)

𝑘=0
where the negative trend (δ) is calculated from the following equation and uses the values of the total
energy of the signal (SN), length of the signal (N) and a parameter α = 1:
𝑆𝑁
𝛿= (14)
𝛼. 𝑁
Since the LoS distance-dependent attenuation is a simplification, it is necessary to evaluate whether it

provides an accurate representation of the attenuation inside the tank. All 18 sources (12 inside the
winding and 6 outside the winding) were simulated in the model of the transformer using the valve UHF
sensor model and the reference winding design. The attenuation curve was plotted, as shown in Figure
5-26(a). A quadratic function was used to fit the data points. It was observed that the LoS fitted curve
was a much poorer fit than the propagation distance fit curve on evaluating the sum of squares due to
error (SSE), which shows how useful the fit is for prediction. However, for both fit curves, there are a
lot of data points exhibiting significant deviation from the fit. Therefore, the data points were split into
two categories based on the location of the PD sources, namely “Inside Winding” and “Outside Winding”.
The former category consists of the data from the 12 PD sources inside the winding, and the latter
consists of the data from the 6 PD sources outside the winding from Figure 5-25. The fit curves for
these two categories are shown in Figure 5-26 (b) and (c).
73
It can be observed that the fit significantly improves for the “Inside Winding” category, see Figure
5-26(b). A clear demarcation can be observed between the data points representing each type of
distance measurement. However, the fit remained poor for the “Outside Winding” category, as shown
in Figure 5-26(c). In addition, for the “Outside Winding” category, the demarcation between the data
points is not as distinct as in the case of the “Inside Winding” category. There is more overlap in the
values suggesting that in the case of PD sources outside the winding, the LoS distance and the
propagation distance are approximately equal. Another observation is that for similar propagation
distances, the attenuation is higher when the source is inside the winding. It can also be observed from
Figure 5-26(b) that the attenuation curve for the propagation distance is effectively the LoS attenuation
curve shifted along the x-axis by approximately 120 cm, which corresponds to half of the winding height,
i.e., the distance that the UHF signal travels to exit the winding. In Figure 5-26 (a), (b), and (c), 0 dB
corresponds to the highest measured signal strength. Additionally, on comparing Figure 5-26 (b) and
(c), it can be observed that the attenuation of signals at comparable LoS or propagation distances is
lower when the sources are outside the winding.
74
(a)
(b)
(c)
Figure 5-26 Comparison of Line of Sight distance and propagation distance-dependent attenuation
in the transformer model (a) All artificial PD sources; (b) PD sources inside winding; (c) PD sources
outside winding [66]
75
5.4 Correlation between UHF signal and IEC 60270 quantities

As stated in chapter 4.5 the UHF calibration should not be misinterpreted as an attempt to correlate the
readings of the UHF method (mV or mV/m) with the electrical method according IEC 60270 in terms of
pico-Coulomb (pC). A measurement system for radiated signals cannot be calibrated by a system for
conducted signals and vice versa.
The following investigations demonstrate that there is no general correlation between UHF PD quantities
and IEC quantities. The general correlation means here a certain ratio, generally applicable for any test
object and measuring setup. While there is a general correlation that as electrical PD measurements
increase, so do UHF ones [5], however, this is beyond the scope of the investigations of this TB.
5.4.1 Simultaneous measurements of UHF and IEC quantities
For comparison of UHF signal amplitude and the apparent charge, the IEC measuring method is used
simultaneously with the UHF measuring method, see the setup in Figure 5-27. One channel of the Digital
Sampling Oscilloscope (DSO) measures the voltage signal of the measuring impedance ZM, representing
the quadrupole according to IEC 60270.
Figure 5-27 Simultaneous measurement of UHF and IEC 60270 for correlation of one electrical PD
pulse to its related UHF signal
The IEC channel has a bandwidth of 20 MHz. The bandwidth of the used quadrupole is limited to 15
MHz. The second channel with the full analogue bandwidth of 2 GHz simultaneously measures the
correlated UHF impulses from the UHF probe. With that setup, correlations of UHF signals and electrical
PD signals according to IEC 60270 can be measured and analyzed consecutively.
Additionally, a PD measuring system directly recorded the apparent charge level QIEC serving as a
reference measurement and voltage measurement. After calibrating the IEC measuring setup, the DSO
measures the apparent charge and the UHF amplitude of the same PD pulse. While recording, e.g. more
than 30 signals, the natural deviation of PD levels helps to find a correlation between the UHF signal
amplitude and apparent charge for the PD source used.
5.4.2 Correlation between UHF and IEC quantities
Two different sources of PD of the same rod-plane type are shown in Figure. The first PD source consists
of a glass ampoule with a rod-plane PD source under oil. The second PD source is another rod-plane
arrangement using the same oil volume as the transformer. The HV is applied by a coaxial cable (RG214)
with the length of approx. 5 m, which is PD free up to a voltage of approximately 12 kV. PD inception
voltage of the sources is about 5-10 kV.
76
Figure 5-28 Rod-plane PD sources, 50-300 pC, 5-10 kV [10]

The PD sources, see Figure 5-28, were injected into an active part of a transformer with voltage ratio
of 10 kV to 380 V and rated power of 630 kVA [5]. That active part was installed in a
10 kV/380 V, 1350 kVA distribution transformer tank. Through the available DN80 oil filling valve, an
UHF probe is inserted, receiving the emitted UHF waves from inside the transformer tank. The location
of the rod mounted PD sources can be changed to 20 different positions, defined in Figure 5-29.
DN80
9o 8o 7o
6o
10o
2o 3o 4o 5o
1o
9u 8u 7u
6u
10u
2u 3u 5u
1u
4u
Figure 5-29 Investigated PD locations in transformer tank with full active part; 10kV/380 V, 630
kVA, DN80 oil filling valve of UHF probe [5]
At the end of the coaxial cable of the PD sources, the IEC 60270 values are measured and recorded
with an oscilloscope as explained in the chapter before. During the measurements, the transformer tank
with its active part is closed and oil-filled. The measuring UHF probe is installed at the DN80 oil filling
valve and the PD source with the glass ampoule is used. The PD was fed with an excitation voltage
leading to 50-100 pC measured according to IEC 60270. For each measured UHF signal, the amplitude
and the apparent charge of the corresponding PD are determined with the simultaneous measurements
as presented before. In Figure 5-30, the measured UHF amplitude over the apparent charge Q IEC is
shown for the different PD positions.
Figure 5-30 PD source “glass ampoule”: Correlation between UHF amplitude and apparent charge
QIEC [5]
All measurements from 7 exemplarily chosen locations are coded with a distinct color. Different clusters
of points with the same color are noteworthy. These points show a linear correlation between the
apparent charge and the measured amplitude of the UHF signal. In order to examine this correlation
quantitatively, for the measured values regression lines were computed according to the method of the
smallest error squares, see Figure 5-30.
77
The following observations can be made:
▪ The regression analysis shows a linear correlation between the apparent charge and the measured
UHF signal amplitude.
▪ The correlation is linear for each position of the PD source. The correlation factor depends on the
position of the PD source and thus on the distance between the PD source and UHF probe.
▪ The signal/noise ratio decreases with increasing distance between the PD source and UHF probe.
The PD source shows - even with the constant apparent charge level, a large dispersion regarding the
UHF signal amplitude.
5.4.3 Conclusion
Measuring the maximum voltage of receivable UHF signals simultaneously with the apparent charge
according to IEC 60270 shows that there is a linear correlation. That correlation is only valid for a
specific distance and path of travel between a PD source and a UHF probe. A general correlation cannot
be determined.
It can be stated for one PD source, that if the apparent charge is increasing, the UHF reading quantity
also increases, i.e. monitoring and trend analyses are workable with both methods and will arrive at
the same conclusions.
If that general correlation in terms of a certain factor is missing, the so-called sensitivity check becomes
obsolete for transformers, see next chapter.
5.5 Applicability of UHF Sensitivity Check for GIS to Transformers

The general idea for the Sensitivity Check for the UHF measuring method on power transformers is
based on the procedure suggested for the UHF measuring method by GIS community [67]. Accordingly,
a real PD has to be placed into a transformer tank in a laboratory. Its level of e.g. 100 pC is defined
according to IEC 60270. That PD causes a UHF probe reading U1 by the emission of EM waves, see the
red lines in Figure 5-31.
Figure 5-31 Sensitivity Check; same measurement readings for original PD source and artificial UHF
signal source by adequate adoption of the signal generator output voltage
78
The same probe reading of U2 as measured with the real PD will be stimulated in a second step with an
UHF signal generator with variable output power using a second transmitting probe. That probe is
installed instead of the PD source at the same location of the former PD source. With an excitation
voltage of e.g. 10 V, this will lead to a ratio of the feeding amplitude to the apparent charge of 0.1 V/pC.
Installing the whole measuring system in an identical transformer with two oil filling valves may then
allow a sensitivity verification of the UHF measuring system. When at the second probe the artificial
UHF impulses becomes measurable at an excitation voltage of 10 V, then the UHF method is able to
detect PD effects with at least 100 pC. If UHF signals only become measurable with excitation voltages
of 25 V, then the sensitivity is at least 250 pC.
From the investigations in the previous chapters, it can be expected that the procedure of the Sensitivity
Check is not generally workable due to the main following reason:
▪ UHF signals experience an attenuation depending on the distance and the active part in the
transmission path between the source and UHF probe.
The dissertation [5] confirms and document the expectation that the Sensitivity Check is not workable
for power transformers. First, investigations are presented with a portable PD source at different
positions within a transformer in a laboratory. Second, a procedure is presented, applicable to
transformers in the field with a PD source mounted on a UHF probe. In both cases, a certain ratio
between UHF signals and IEC values is only applicable for a specific signal travelling path and not
generally valid. This is furthermore confirmed by collected results of other working groups and their
results.
Thus the Sensitivity Check, like for GIS suggested by [67], is not workable for transformers, and an
estimation of the apparent charge by measuring EM waves is not possible.
79
6. Case Studies
Existing and collected case studies from the members where electrical and UHF measurements are
performed simultaneously are mainly limited to diagnostic measurements and are not performed with
the only intention of pass or no-pass criteria during SAT/FAT. However, chosen diagnostic
measurements during SAT/FAT have been revisited and relevant information for SAT/FAT are presented.
6.1 Case Study 1 (SAT) 333 MVA – Substation Transformer

Because of increasing gas-in-oil values, a 400/220/30 kV, 333 MVA substation single-phase
autotransformer was tested onsite and online for PD. The high noise level at the site strongly disturbed
the conventional PD measurements, made according to IEC 60270, below 1 MHz frequencies.
Consequently, UHF PD measurements for PD detection in combination with acoustic measurements for
later PD localization were performed in order to get reliable results.
An electrical broadband measured PRPD-pattern at nominal voltage is shown in Figure 6-1 a). The
frequency range of measurements is between 350 kHz and 1.75 MHz, the measurement connection
point is the 400 kV bushing tap. The setup was calibrated according IEC 60270 procedure.
Figure 6-1 a) Electrical PD Measurement; fc = 1 MHz, Δf = 1500 kHz, 22 seconds b) UHF PD

measurement at 310 MHz, Δf = 1500kHz, 22 seconds [5]
80
During the measurements with a recording time of 22 seconds, external corona discharges of the 400
kV bus bar above the transformer were audible. Their PD pulses become visible in the pattern with a
level above 2000 pC and can be identified by the phase shift of 120° for the three different phases. The
pulses of the corona discharge occur at 90° (respectively 4 ms), 190° (respectively 10.5 ms) and 310°
(respectively 17 ms), see Figure 6-1 a).
Internal UHF probes use the tank wall as shielding against external disturbances and Figure 6-1 b)
shows the pattern of one internal UHF probe signals measured at 310 MHz with a recording time of 22
seconds. According to this pattern, only one internal PD source can be identified. That is a main
advantage of the UHF method against the electrical method in that case. The 120° shifted disturbances
are no longer detected. The provided screenshot have to be corrected and that software version for
experimental purposes will be changed to UHF measurements in terms of mV.
Three identical UHF Sensors were installed. Figure 6-2 shows the positions of the UHF probes (UHF 1 –
UHF 3). Two probes are opposite to each other at the top of either front end of the tank, and the third
(UHF 3) is located at the bottom in the middle of the transformer side tank wall.
Figure 6-2 333 MVA transformer showing positions of UHF sensors and acoustic sensors [5]
First, the so-called Dual Port Performance Check were attempted, see chapter 4.8.2. Artificial UHF
impulses were injected at each probe with a signal generator (60 V at 50 Ω). It was not possible to
detect the artificial impulses at any combination of emitting and receiving probe. The explanation by
the transformer manufacturer was that there are tubes behind the oil filling valves directing the oil flow
around the winding. According to the unsuccessful Dual Port Performance Check, it could be concluded
that the probes were electromagnetically decoupled from each other. Furthermore, they might also be
shielded against UHF pulses from internal PDs.
A further explanation might be that the maximum signal generator output voltage of 60 V is not
sufficient to transmit UHF waves through that transformer.
Nevertheless, at nominal voltage, UHF signals from internal sources were detectable with all three
probes, i.e. the internal PD causes UHF signals with higher energy content than the former artificial
impulses. It can be concluded that the Dual Port Performance Check was thus just a worst-case
estimation of the sensitivity. However, even though the Performance Check was not successful, sensitive
UHF measurements might still be possible.
For proof that the signals correlate to the PD, phase-resolved partial discharge patterns of the UHF
signals were recorded, synchronized to the voltage supply of the measuring unit. Phase stable UHF
pulses confirmed the occurrence of internal PD. Simultaneous IEC 60270 measurements were highly
disturbed by audible corona discharges.
After the online PD tests and the results of the PD localisation (in the vicinity of the tap changer), the
transformer was transported to the manufacturer and a PD measurement was repeated in the test field.
Figure 6-3 shows the electrical PD pattern according to IEC 60270 measured after calibration on the
400 kV bushing tap.
81
Figure 6-3 Electrical PD Measurement in test facility according IEC 60270 on 333 MVA single phase
auto-transformer; fc = 300 kHz, Δf = 400 kHz [5]
The measured PD level of approx. 8000 pC in a frequency range according IEC 60270 between 100 kHz
and 500 kHz can’t confirm the PD level of 300 pC determined by the gating process in field.
The transformer was de-tanked and a visual inspection of the active parts at the tap changer confirmed
the localization result [5], see Figure 6-4.
Figure 6-4 Deteriorated paper insulation on leads at the tap changer

After repair, the transformer passed the acceptance test without any indication of PD activity and was
put back into service.
82
6.1.1 Conclusion
The first step on-site should be the verification of the UHF PD measurement system’s functionality. The
Dual Port Performance Check proves that the installed UHF probes are sensitive against internal EM
wave emission. In cases where the transformer under test does not offer a second oil-flange or
appropriate Top Hatch probe, the Single Port Performance Check at least will show the functionality of
the probe, the measuring lines and the UHF measurement equipment.
UHF disturbances can be recognised by post processing of measured UHF signals or frequency analyses
(FFT). Disturbances like walkie-talkies, DVB-T or GSM are recognisable by their narrowband
characteristics. With e.g., 300 MHz high pass filtering or narrowband UHF PD detection systems,
measurements can be performed at frequency ranges unaffected by disturbances. UHF PRPDs can also
help distinguish between phase stable PD impulses and non-phase stable disturbances signals.
A crucial advantage of the UHF measuring method is its resilience against signals of external corona
discharges, which usually represent the largest disturbance potential for conventional measurements.
UHF PD impulses can be identified by their broadband characteristic up to 1.5 GHz and by their
occurrence correlated to the applied voltage. The UHF impulses can be monitored with adequate
measuring devices and a trend analyses seems possible and should be investigated further.
6.2 Case Study 2 (FAT) - Generator Transformer

These results were recorded during factory acceptance testing of a generator transformer unit which
successfully passed the overpotential testing with only minor and temporary PD activity. Nevertheless,
some small UHF signals were observed while the PD was active, with the advantage that the
corresponding IEC 60270 pC levels were recorded. Technical details of the transformer can be found in
[68].
Figure 6-5 shows the LV side of the tank where three UHF sensors were installed. The fourth sensor
was located on the HV side of the tank. The induced overpotential testing was carried out at 200 Hz,
one phase at a time, with the other two phases being earthed. Maximum potential applied was 2 U0 at
the LV winding, corresponding to 50 kV.
No. 4
No. 2 No. 3
Figure 6-5 Generator transformer with UHF sensors for FAT. In addition to visible sensors, sensor
No. 1 is located on the back (HV) side of the tank opposite the position of sensor No. 3 [56]
83
The test configuration for the UHF measurement was as follows: Each sensor was connected by a
10.0 m RG213 coaxial cable with N-type connectors to a 26 dB gain UHF preamplifier. The amplified
signals were captured using an oscilloscope with an analogue bandwidth of 500 MHz and sampling rate
of 5 GSamples/s.
Figure 6-6 provides a second example of UHF signals recorded during the FAT. In this case, the signal
from sensor No. 1) is about 22 mV pk-pk after amplification. This corresponds to a signal level of below
1 mV pk-pk directly from the UHF sensor.
No. 1
Amplitude ( 10 mV / div )
No. 2
No. 3
No. 4
0 10 20 30 40 50 60 70 80 90 100
time ( ns )
Figure 6-6 Amplified (×20 voltage gain) UHF signals recorded during overpotential testing of C-
phase at 200% with a measured PD level in the region of 5 – 10 pC [56]
The spectral density of this signal is shown in Figure 6-6, with a somewhat wider spread of frequencies
in the range 300 – 1200 MHz in this case.
1
relative spectral density
0.5
0
200 400 600 800 1000 1200 1400 1600 1800 2000
frequency ( MHz )
Figure 6-7 Frequency spectrum of the signal from UHF sensor No. 1 [56]
6.3 Case Study 3 (FAT) - Single Phase Transformer

During a FAT testing of a single phase transformers without IEC signals above noise level, a
simultaneous UHF measurement for fingerprint reasons was performed. A dual port performance check
demonstrated a linear correlation between the output power of the used signal generator and the
measurable UHF signal, see Figure 6-8. The ambient noise level is approx. below 2 mV. As the signal
generator was not synchronized to the measuring phase, here the supply voltage, with a certain
deviation a line across the phase angels appear.
84
Figure 6-8 Performance check with two sensors opposite to each other with 50 V, 100 V, 200 V and
400 V signal generator output voltage
Approximately a linear correlation between the increased output power and the measurable UHF signal
can be seen as expected, see table 6-1.
Table 6-1: Dual Port Performance Check – Signal Generator output and measurable UHF amplitude
Signal Generator 50 100 200 400

Output / V
Measurable UHF 15 40 90 220
amplitude / mV
During the 1 hour electrical PD testing some UHF patterns have been recorded, see Figure 6-9.
Figure 6-9 UHF PD measurement during FAT without indication of electrical PD measurement
They act as documented fingerprint measurements to be compared with future SAT PD measurements
with the UHF PD measurement technology.
As the IEC measurements reveals no PD activity the transformer successfully passed the FAT.
85
6.4 Case Study 4 (FAT) - Repaired 123 kV Substation Transformer

After internal flashover, the active part of a 15-year-old 123 kV substation transformer was cleaned
from deposits and carbon particles. After the repair an induced voltage test with partial discharge
measurement (IVPD) acc. to IEC 60076-3 was performed with a test level of 80% of the original test
level.
The conventional PD measurement according to IEC 60270 was performed by a commercially available
PD measuring system. The PD signals were decoupled from the bushing taps. The system was calibrated
in the frequency range from 40 kHz to 800 kHz.
For the UHF-PD measurement two UHF sensors were installed at DN80 drain valves. Sensor 1 was
installed at the bottom drain valve and sensor 2 was installed at a valve on top of the tank near the
OLTC (Figure 6-10). During the PD measurements both antennas were inserted 50 mm into the tank.
For the performance check, a broadband pulse was fed into sensor 1 and the response was measured
with sensor 2. The UHF signals were evaluated and recorded with a PD monitoring system. At an
insertion depth of 50 mm of sensor 2, a UHF amplitude of 1.5 mV (slightly above noise level) was
recorded when a 25 V signal was injected at sensor 1. When feeding sensor 1 with a 600 V signal, the
following signal amplitudes were recorded dependent on the insertion depth of the antenna:
• Insertion depth sensor 2: 50 mm: 180 mV
• Insertion depth sensor 2 20 mm: 90 mV
• Insertion depth sensor 2 0 mm: 40 mV
Figure 6-10 Installed UHF PD-Sensors (left: bottom of tank (sensor 1), right: top of the tank
(sensor 2)
When the converter-fed HV test transformer was switched on, the switching operation of the thyristors
for the IEC-compliant PD measurement caused a noise level of well over 3 nC, which increased to about
7.8 nC when the test voltage was increased. The PRPD pattern is presented inFigure 6-11, which shows
clearly six times the firing of the thyristors within one period. The detection and evaluation of internal
PD pulses by the conventional PD measuring method was therefore not possible.
When the converter-fed HV test transformer was switched on, the noise level of the UHF PD
measurement remained unchanged by below 1 mV (see Figure 6-12).
86
Figure 6-11 PRPD of conventional PD measurement according to IEC 60270 showing strong noises
due to switching of thyristors of power converter
At a testing voltage of 110 kV partial discharges incepted. Phase-resolved partial discharge patterns
generated by internal PD were clearly visible (see Figure 6-12 and Figure 6-13). There are only slight
differences between both sensors.
Figure 6-12 PD-inception at 110 kV (Sensor 1 (left), Sensor 2 (right)), logarithmic scaling;
acquisition time: 1 min
Figure 6-13 PD-inception at 110 kV (Sensor 1 (left), Sensor 2 (right)), linear scaling, acquisition
time: 1 min
After increasing the test level to enhancement voltage and lowering the test voltage to the one hour PD
measurement voltage of 145 kV, the PD activity showed a different pattern. Sensor 2 (top) shows a
87
pattern that indicates a void discharge. Increasing the acquisition time to 10 minutes, this pattern can
also be seen with sensor 1, but with a lower amplitude (about 10%) (Figure 6-14). Due to the larger
amplitude of the signal on sensor 2, the location of the PD can be assumed on the top side of the tank.
Figure 6-14 PRPD-pattern at 145 kV (Sensor 1 (left), Sensor 2 (right)), linear scaling, acquisition
time: 10 min
Because electrical PD measurement was not possible in this case, no information about PD level could
be obtained. Therefore, the transformer passed the test. For condtion assessement during operation an
online UHF PD monitoring system was installed.
88
7. Measurement campaigns of The Working

Group
7.1 Initial Laboratory Investigation of different UHF sensors
The first UHF measurement campaign of the Working Group was performed on a small transformer
including a real, reproducible PD source within a R&D laboratory. The first session was regarded as
learning experience and team building event. Many members participated in the measurements and the
performed work revealed a common understanding of the technology and measuring procedure. A
variety of different UHF sensors and measuring system was tested initially. The results can be concluded
as followed:
- All available UHF sensors are able to capture UHF signals
- All available UHF measuring systems are able to measure UHF signals
- By varying insertion depths of all sensors with different measuring systems, an insertion depth
of 50 mm was established as standard insertion depth
- Measuring results are not easily comparable (amplification, scaling, …)
Especially the experiences with the last point was the main motivation for continuing the work to develop
and introduce the calibration proposal as the main result of the working group.
7.2 On-site Investigation of different Measurement systems and signal

generators
For comparative UHF measurements, an 800/3 MVA single-phase transformer of the voltage level
420/247/33 kV was available. All measurements were carried out without high voltage and the
transformer disconnected from the grid. Figure 7-1 defines the positions of the 8 permanently installed
UHF sensors [69]. In addition, two UHF valve UHF sensors (see chapter 3) were temporarily inserted
through two oil drain valves.
Figure 7-1 Positions of sensors at 800/3 MVA Transformer for comparative UHF measurements [70]
89
7.2.1 Performance check (dual port)

The term performance check (see also chapter 4.8.2) describes the process where, an artificial pulse is
emitted with one sensor as transmitter into the transformer tank and measured with a second sensor
(or further sensors) as the receiver. This provides evidence that both sensors are installed properly,
meaning that they are able to receive UHF PD signals from the inside of the transformer. The
performance check corresponds to the second part of the CIGRÉ sensitivity test performed on GIS UHF
PD measurements [67]. The performance check would be negative if the sensor cannot receive any
signals due to electromagnetic shielding or high signal damping. For example, when a valve UHF sensor
is installed at an oil valve with a rising pipe inside or when metallic objects are mounted directly in front
of the sensor.
There are UHF sensors on the market that provide an additional input for a “single port performance
check” beside the actual measurement output. Here an artificial signal is fed in and directly coupled to
the actual UHF antenna. However, this test cannot give any indication whether the sensor is installed
correctly and not affected by electromagnetic shielding inside the transformer, since the coupling of the
signal is only inside the sensor itself.
7.2.2 Conducted measurements
The purpose of the measurements at the transformer described above is to compare the different used
measurement systems (time domain, frequency domain, commercial PD measuring devices) on the ten
installed UHF sensors and to give a recommendation for the optimal number and placement of the UHF
sensors based on the results of the measurements.
The following measurements based on the performance check procedure have been carried out for this
by using all different measuring systems and methods:
• inject constant artificial impulses into sensor 6 and record the sensor output amplitudes on each
of the other 9 sensors
• inject constant artificial impulses into sensor 5 and measure the amplitudes on all other sensors
• measure the amplitudes on sensor 5 and inject constant artificial impulses into all other sensors
7.2.3 Impulse generators and measuring devices used
Two different kinds of impulse generators respectively impulse shapes are being used for performance
checks on transformers, impulses in exponential, respectively capacitor discharge shapes and short
rectangular shapes. The exponential impulse shapes are basically limited in their frequency components
towards higher frequency ranges (above several hundred MHz), whereas short rectangular impulse
shapes are providing a flat frequency spectrum up to several GHz [71] [55]. For the comparative
measurements, carried out for this chapter, both types of impulse generators have been used.
Two different commercial PD systems, a digital sampling oscilloscope (to represent the time domain
measurements) and a spectrum analyzer (to represent the frequency domain measurements) have been
used to compare the different measurement principles. There are basically two different types of
commercially available UHF measurement systems: Conventional PD measuring instruments
(IEC 60270) with frequency converter accessories for UHF range and systems which are using directly
impulse detection in the UHF range. Depending on the system used or the settings, different bandwidths
and measuring frequencies can be used to generate phase-resolved PD patterns (PRPD, 2D, 3D). The
measuring systems for transformers typically cover a frequency range from some hundreds of MHz to
approximately 1.5 GHz.
7.2.4 UHF signals in time- and frequency -domain
To get information about the damping phenomena of UHF signals transferred through the active part
of the transformer, the responses of the defined rectangular pulse (50 V) injected at UHF sensor No. 6
at three UHF sensors placed at different positions on the tank, were recorded in time- and frequency-
domain. Using both a digital oscilloscope (4 GHz analog bandwidth, 40 GS/s sampling rate) and
spectrum analyzer (2 GHz analog bandwidth), all signals were recorded and analyzed. An example of
coupled signals in time- and frequency-domain to the UHF sensor No. 8, which is far from the injected
signal, is shown in Figure 7-2.
90
Time domain
Injection at Sensor No. 6
CH1 = Sensor No. A

(near the injection)
A = 20mV/div
t = 20 ns /div
CH2 = Sensor No. 8

A = 1mV/div
t = 20 ns /div
Frequency domain
f1 = 100 MHz
f2 = 1.8 GHz
AM = -26 dBm
Spectrum of Sensor No. 8
Figure 7-2 Example of injected and coupled UHF signals in time- and frequency-domain
The UHF signal measured at CH1 (near the injection) has a fast edge and a short back. In the contrary,
the signal at CH2 (far from the injection) shows a typical filtered longer impulse shape. In the frequency
spectrum, the PD signal is clearly visible between approx. 200 MHz and 1 GHz.
7.2.5 Comparison and summary of measurement results
The characteristics of the recorded spectra, measured with spectrum analyser and industrial PD
measurements systems, are very similar. The respective unaccounted amplification factors of the
preamplifiers or missing calibration result in an unquantified variation of the measured signal levels.
The following Table 5 shows the mean of the spectra given above, comparable to the CIGRÉ proposal
for Average Power (AP) in [67].
Table 5 Comparison of different measuring methods
50 V Injection in Sensor 6 Sensor 8 Sensor 5 Sensor 1
Oscilloscope 2.5 mV = -39 dBm 7 mV = -30 dBm 20 mV = -21 dBm
-79.8 dBm*1 -73.4 dBm*1 -68.5 dBm*1
-83.4 dBm*2 -77.9 dBm*2 -65.9 dBm*2
Spectrum Analyzer -35 dBm -30 dBm -25 dBm
Narrowband Commercial -89 dBm -84.5 dBm -76 dBm
Broadband Commercial -57 dBm -43 dBm - 30 dBm
91
The frequency range from 300 to 800 MHz was used for narrowband systems. In the broadband PD
system, the AP was used in the frequency range 300 MHz to 1500 MHz. In the case of the oscilloscope
time domain signal the "maximum value" was used and converted into dBm, additionally the time signal
was transferred into the frequency domain by an FFT, where the AP was also estimated to the above-
mentioned Frequency ranges (300 – 800 MHz*1, 300 – 1500 MHz*2)
The conducted transfer characteristic experiments show that for the UHF PD diagnosis on power
transformers, narrowband as well as broadband systems and measuring instruments in the time domain
as well as in the frequency domain can be used. For the selection of the most suitable system, the
requirements of the respective application must be observed. The quantitative results are comparable
in the form that all measuring systems clearly indicate an increasing attenuation with increasing distance
between signal source and measuring sensor. Differences in the attenuations result from the different
measuring principles, which are compared here uncalibrated. Among other things, it is the different
frequency ranges used that lead to a deviation in the signal power (AP), as shown in the example of
the oscilloscope. For a future comparability of UHF measurements with different equipment, a fixed
frequency range should be defined and documented.
7.2.6 Conclusion
The findings gained during the comparative measurement on a large power transformer led to the
conclusion that UHF signals can be sensitively measured with different measuring systems. The
experiment compared the qualitative results of time domain measurements by oscilloscope, frequency
domain measurements by spectrum analyzer, and two commercial PD measurement systems. The
qualitative results regarding the sensitivity of different sensors and the positioning of sensors are
summarized as follows, independent of the measuring method:
- UHF signals experience a path-dependent attenuation, which increases with increasing distance.
- Compared to valve UHF sensors, window sensors have the advantage of a better high-frequency
earthing and therefore lower disturbances. Further the insertion depth is fixed, which leads to better
reproducibility. In addition, the frequency response is more linear.
- A minimum UHF sensor configuration consists of at least 2 UHF sensors to perform the performance
check
- The recommended standard configuration consists of 4 sensors to assist in locating via UHF sensors.
- A high-end variant may consist of 6-8 sensors for very large tanks and / or critical assets.
- When positioning the UHF sensors, make sure there is sufficient clearance (at least 25 cm) to the
corners and edges of the tank. It must be avoided that internal oil guidance structures or metallic
structures shield the sensor electromagnetically from internal PD signals.
The comparability of different measurement methods was identified as the subject of further
investigations in order to investigate and quantify the influence of measurement results as a function
of the selected frequency range and the performance of narrowband or broadband measurements. In
general, the qualitative statements of the measurement methods used are comparable, so that a general
quantitative comparability by calibration [72] of the measurement methods seems possible. The UHF
signal transmission properties are likely to be significantly different for different transformer types and
tank sizes, which is why further experimental work is required to make more general and more specific
statements on transformer types.
7.3 Calibration Method in Laboratory

The proposed calibration procedure is tested in a laboratory setup for different measurement devices.
For simplification and better comparability, the used UHF frequency range for all used devices is set to
300 MHz – 750 MHz. The frequency band is provided by the application of appropriate commercially
available passive low pass and high pass filters, which are connected between the signal cable and the
measuring devices at any time.
7.3.1 Calibration of measurement device and cables
The measurements follow the proposed procedure in Chapter 0 to determine the calibration factor KM.
The introduced signal (PuM RF) from Figure 4-26 with known amplitude U0 is injected directly into the
four different measuring devices including the band pass filters and cables.
92
The specifications of the determined measurement systems are as following:

Osci: 4-channel DSO with 4 GHz analog bandwidth, 40 GS/s sampling rate and 8-bit
vertical resolution
MS1: 6-channel UHF PD monitoring system with 12-bit vertical resolution
MS2: 4-channel UHF PD module of a transformer monitoring system with 12-bit vertical
resolution
Spec: Spectrum analyser with 2.9 GHz bandwidth
All devices are equipped with an RG 214 coaxial cable of 20 m length. Only the “Osci 40m” measurement
is measured with the same cable type, but of 40 m length. The results are shown in Figure 7-3.
Figure 7-3 Comparison of measurement systems [38]

The x-axis represents the output power of the PuM RF signal in dBm. The y-axis represents the
measured peak amplitude of the different devices in mV. Plot U0 gives the calculated reference value
from the signal generators in dBm. When comparing the results, the expected influence of an additional
cable length of 20 m can be seen between the Osci 20m and Osci 40m results. At identical signal
generator output, the measuring reading is slightly lower due to cable attention over the entire input
range. All systems show a good linearity over the entire input range. Hence, a calibration using a
constant KM-factor is possible to correct the constant offset. The spectrum analyzer settings are the
following: 500 MHz, ZeroSpan, 2 MHz Receiver Bandwidth (RBW), 3 MHz Video Bandwidth (VBW). The
power measured at 500 MHz was used to calculate the corresponding peak voltage at 50 Ω. For the
actual KM calibration, the measured value at -25 dBm is used. Resulting KM calibration factors are
displayed in Table 6.
7.3.2 Calibration of UHF sensors
Exemplary, two valve UHF sensors are characterized in a GTEM cell (see chapter 4.2.2) and KS is
calculated in the used frequency range at 50 mm insertion depth according to chapter 4.5.3, see Figure
7-4.
Figure 7-4 AF(f) and corresponding KS of UHF sensors no1 and no2 [38]
93
7.3.3 Laboratory setup using an artificial UHF signal (PUM RF)

For a valid calibration, the same cables (for correct KM) and the same insertion depth of the UHF sensor
(for correct KS) must be used. Measurements are performed in a steel tank (Figure 7-5) equipped with
an active part of a 1350 kVA distribution transformer.
DN80 gate valve
UHF sensor
Figure 7-5 UHF valve UHF sensor mounted with a DN80 gate valve. [38]
Inside dimension of the transformer tank are 1720 mm in length, 760 mm in width and 1550 mm in
height, respectively. On the tank wall, there are two DN50 and two DN80 gate valves. In Figure 7-5 the
installed valve UHF sensor no1 with KS = 25,8 1/m is shown. For the measurements, a second UHF
sensor is used with the signal generator to get a stable impulse, which is identical for all measurement
devices, see Figure 7-6.
Second
UHF
Sensor
EUHF / V/m
signal
generator
UHF
measurement
UHF devices
Sensor
Figure 7-6 Schematic of the laboratory setup with two UHF sensors mounted on a transformer tank
[38]
In Figure 7-7 a PRPD of MS1 exemplarily shows the stable impulse of the excitation using the signal
generator in this laboratory setup.
Figure 7-7 PRPD of measured PuM RF signal [38]
94
7.3.4 Measuring results of an artificial UHF signal (PUM RF)

Measurements for the described devices are performed at a fixed sensor positions for both sending (as
artificial PD source) and receiving UHF sensor, as seen in Figure 7-6. Table 6 shows the measured Ui
values (peak) and the corresponding calibrated EUHF values using KM and KS calibration factors according
formula (1) and (4).
Table 6 Measured voltage, used K factors, and calibrated UHF reading of all devices
Osci Osci MS1 MS2 Spec

20m 40m 20m 20m 20m
measured 5,3 mV
peak value 11,0 mV 7,8 mV 12,1 mV 10,5 mV
(-35,5 dBm)
KM 1,36 1,83 1,36 1,51 2,81
KS 25,8 1/m 25,8 1/m 25,8 1/m 25,8 1/m 25,8 1/m
resulting
386 mV/m 368 mV/m 424 mV/m 409 mV/m 384 mV/m
EUHF
All calibrated values are at approximately EUHF = 395 mV/m with standard deviation approx. ±20 mV/m
and a maximum deviation below ±10%. Taking into account the various sources of errors, like manual
readings of measurement values form the screens of all devices, this deviation is considered acceptable
for that first investigation.
Further work is necessary to focus on reducing the various sources of errors by defining measurement
procedures, measurement equipment and measurement evaluation in the manner defined for IEC 60270
[2].
7.3.5 Measuring results of a real pd source with different calibrated UHF
PD measuring systems
An AC voltage in the range of approx. 12 kV is applied to the active part inside the laboratory steel tank
at different bushings and internal real PD sources are triggered/activated. To demonstrate that all
calibrated measuring systems result in the same or comparable electrical field strength EUHF, the applied
voltage is kept higher than the PD inception voltage (PDIV) and the source is tried to kept constant to
guarantee a constant PD reading while the measuring systems are exchanged.
Figure 7-8 shows the challenge, to identify a measurement reading for comparison by manually
evaluating UHF phase-resolved partial discharge pattern values. The screenshots show the phase-
resolved partial discharge patterns from different measuring systems which are not calibrated,
representing the variety of different visualization approaches and the reason, why UHF PD
measurements have been regarded as not comparable in the past. The various visualizations of phase-
resolved partial discharge patterns consist of:
- different scaling (logarithmic vs. linear)
- different y-axis values (nV, mV, dBm, dBµV, %, and even pC have been seen falsely)
- different color bar accumulation for PDs/sec (scaling, used colors, logarithmic vs. linear)
- …
Figure 7-8 UHF PRPDs of the same PD source with different visualizations. Left: MS3 using linear
scale /nV; Middle: MS2 linear scale / mV (phase shifted);right: MS4 logarithmic scale / % [38]
95
Note: The systems have been synchronized either to the applied test voltage, by their voltage supply,
or using the light inside the laboratory. Therefore, phase shifts are obvious between the systems.
The specifications of the determined measurement systems are as following:
MS2: 4-channel UHF module of a transformer monitoring system
MS3: UHF converter for electrical PD measurement device
MS4: Mobile UHF measurement device
After calibration and applying a comparable plotting setup, these PRPDs should look the same and give
the same EUHF reading on the y-axis. E.g., for the two linear scaled PRPDs phase shifts are corrected
and their y-axis are rescaled (which can be considered as applying the proposed UHF calibration
procedure). Figure 7-9 shows the result, which yields high comparability.
Figure 7-9 UHF PRPDs of real PD source 1 after rescaling (calibration) with linear y-axis on MS3 and
MS2 [38]
The results of two measuring systems are summarized in for the measurement of PD source 1 shown
in Table 7.
Table 7 Measured voltage, used K factors, and calibrated UHF reading at Sensor 1 for PD source 1
MS3 MS2
measured
peak value 10 mV 14 mV
KM 2.23 1.36
KS 31 1/m 31 1/m
Calibrated
691 mV/m 590 mV/m
EUHF
As a second example, the same process (scaling on y-axis and phase correction) was perforemd for a
second PD source, which is activated when appling voltage to a different bushing at the test tank. Again,
this source is measured using different UHF measurement devices. In this case, the PRPDs are plotted
with logarithmic scaling, see Figure 7-10. In both cases the color bar could not be changed, therefore
the patterns cannot look exactly the same.
96
Figure 7-10 PRPDs of the second real PD source 2 after rescaling (calibration)
with logarithmic y-axis on MS 4 and MS 2 [38]
However, to derive a calibrated UHF value, scaling in y-axis is of sufficient accuracy.

The results of the second measured real PD source are summarized in Table 8.
Table 8 Measured voltage, used K factors, and calibrated UHF reading at Sensor 2 for PD source 2
MS 4 MS 2
measured
peak value 7,6 mV 9 mV
KM 1,86 1,36
KS 26,6 1/m 26,6 1/m
resulting EUHF 376 mV/m 326 mV/m
The larger deviation of the calibrated UHF values in Table 7 and Table 8 compared to the results using
a stable artificial impulse in Table 6 can be explained mainly by the error in manual reading of the values
out of the real PRPDs. Figure 7-9 and Figure 7-10 show an expected similar PRPD shape after rescaling
the y-axis. These results indicate that a general calibration of the UHF method is possible if calibration
is done correctly and the various errors that occurred during this first test setup can be reduced to a
minimum.
One major issue to address is the manual reading of UHF peak values out of phase-resolved partial
discharge patterns. Like for the electrical PD measurement method an algorithm similar to the algorithm
described in IEC 60270 to display the “largest repeatedly occurring (UHF) PD magnitude” could be
applied to minimize erroneous manual readings [2]. In addition, all manufacturers need to incorporate
the facility to perform calibration in their software and rescale the visualization accordingly.
7.3.6 Conclusion for UHF calibration
In this measurement campaign, the proposed two-step calibration process has been tested: The first
step results in calibration factor KM and eliminates the influence of the signal recorder and additional
accessories like cable damping. This is achieved by injecting a defined impulse into the UHF
measurement system without UHF sensor. Hence, all deviations in the measurement can be corrected.
The second step results in calibration factor KS and includes the individual sensor’s characteristics,
namely its antenna factor (AF) into the calibration.
The basic feasibility of the proposed calibration method is demonstrated in a laboratory setup by using
different UHF PD measurement devices. In this early stage, the systems are not yet optimized for
calibration. However, using the proposed UHF calibration, the laboratory tests revealed that measuring
inaccuracy (resp. deviation) could be reduced to approximately ± 10% for the different measurement
systems in a setup with stabilized artificial “PD source”. In the same laboratory setup two real PD sources
are triggered using high voltage. With 5 different measurement devices and two different sensors, the
calibration proposal is again tested. The evaluation of the results shows higher errors (resp. deviations)
than using the artificial source which can be explained by different scaling of the systems, manual
reading of the values out of PRPDs, etc. These errors (resp. diviations) need to be reduced in the future
by the systematic implementation of the UHF calibration into the measurement systems.
97
7.3.7 Outlook for UHF testing

Like for the electrical PD measurement method an algorithm similar to the algorithm described in IEC
60270 to display the “largest repeatedly-occurring (UHF) PD magnitude” should be applied. Also, all
manufacturers need to incorporate the facility to perform KM and KS calibration in their software and
rescale the visualization accordingly to mV/m.
With UHF measurement systems that have incorporated the proposed UHF calibration procedure and
the automatic output of the “largest repeatedly occurring (UHF) PD magnitude”, a measurement
campaign similar to the presented laboratory test needs to be done.
With these systems, the community can then gain experiences in calibrated UHF measurements and
may identify criteria for UHF PD testing. With some years of experience using calibrated UHF
measurement at power transformers, it maybe gets possible to define acceptance levels for factory and
site acceptance tests (FAT and SAT) in the future.
The Performance Check can be further evaluated if it could be used as a kind of sensitivity estimation
of the following measurements or characterization of the unknown transformer internal travelling path.
98
8. Recommendations for improved PD

acceptance tests
The main improvement for Factory Acceptance Tests (FAT) and Site Acceptance Tests (SAT) provided
by this brochure is the possible achievement of general comparability of partial discharge (PD)
measurements in UHF range. This provides benefits for both transformer manufacturers and utilities.
In order to achieve comparability of FAT or SAT tests for UHF PD measurements, some requirements
have to be met. The used measurement system has to be calibrated, so all influences originated form
the system and to some extent from the setup can be eliminated.
Calibration requires the use of UHF sensors with a known antenna characteristic (chapter 4.2). The
characteristics can be provided either by the antenna factor AF, or in a simplified approach proposed in
this brochure by the KS factor of the sensor, see chapter 4.5.2. The influences of both the setup (e.g.,
cable damping) and the measurement system can be calibrated with the introduced factor KM, using a
standardized impulse as proposed in chapter 0. The measuring system must be able to incorporate the
calibration factors KS and KM in its signal processing software. In addition, the UHF measurement system
has to provide a frequency independent signal level accuracy over the entire dynamic range of the
system. If all aforementioned requirements are met, it is possible to cover the entire signal input range
with only one calibration impulse as proposed in chapter 4.5.4. The recommended frequency range for
UHF measurements by this brochure is 300 MHz to 750 MHz, as discussed in chapter 5.1.2.
In order to provide high coverage for PDs within the entire transformer, 4 UHF sensors are
recommended as standard configuration. At least two sensors should be used at absolute minimum for
a performance check. As a high-end configuration 6 to 8 sensors can be used, see chapter 4.4. It is
advised to use window UHF sensors (chapter 4.1.2) instead of valve UHF sensors, because they are
easier to use (insertion depth does not have to be set correctly) and have a higher sensitivity. Valve
UHF sensors should only be used for tests on repaired transformers which have no possibility for window
UHF sensors. It is recommended to run a dual port performance check on each combination of sensors,
see chapter 4.8.2. The check assures a sufficient sensitivity for UHF signal across the transformer and
indicates errors in the setup e.g., valve UHF sensors installed at unsuited positions or at an incorrect
insertion depth.
In general, documentation is the second important issue to assure the comparability of all
measurements, especially FAT vs. SAT measurements. Relevant points to be noted should be the used
sensor positions, used sensor types and their KS Factors, used measurement system and the obtained
KM factors for each measurement channels and notes about the performed dual port performance checks
between each possible sensor pair. For a FAT, occurring PDs should be noted by the applied voltage,
their measured levels and a phase-resolved partial discharge patterns should be included as fingerprint
reference for later SAT measurements.
Concerning conventional measurements according to IEC 60270, calibration is already established since
many years, meaning the basic requirement for comparability is already met. That does not mean, the
real charge of the PD is measurable. For the comparison between FAT and SAT, documentation is as
important for IEC measurement as for UHF. It should include the description of the setup, the capacity
of the coupling capacitor, the used calibration level (e.g., 5 pC), recorded noise levels after calibration,
the used frequency range and the used measurement device. In general, the center frequency chosen
within the IEC limits should be as low as possible: In general, low frequencies experience the lowest
signal attenuation and hence provide highest sensitivity for PD signals throughout the transformer.
99
If possible, PD measurements at FAT and SAT should be performed using both measurement according
to IEC and UHF sensors. Hence, the overall sensitivity to detect PD is the highest and the usage of
different physical propagation mechanisms provides the option to validate and confirm single
measurements. In most cases, UHF PD measurements are easier to perform at SAT compared to IEC
measurements. Therefore, a PD measurement validated using both methods at FAT could be confirmed
at SAT using at least one method.
FAT and SAT require an acceptance criteria or threshold value acting as “pass/fail” criteria. Due to
several reasons highlighted in the TB, the only focus of the apparent charge is not meaningful. It is the
common understanding of the members, that as state of today, it is meaningless to give
recommendations on acceptance thresholds for UHF measurements. Furthermore, there is no
correlation in terms of a certain ration between electrical and UHF measurements.
100
9. Outlook for further improvements

Measurements with a calibrated UHF measuring system may lead to acceptance criteria in the future.
CIGRÉ should collect experiences with the calibrated UHF measurements during FAT and SAT in the
future and decide which acceptance criteria may be derived. First ideas were collected in the working
group and are presented in the next chapter.
As far as this document concerns, a simple performance check has been discussed. Further work is
needed as well to establish a certain sensitivity requirement to be able to detect critical PD defects
whereas different parameter would play an essential role like location, type of defect, amplitude etc.
If electrical and UHF measurements are more often used simultaneously during FAT and/or SAT,
combined acceptance criteria can be developed by a future expert group.
9.1 Criteria Overview

Assumed that electrical and UHF measurements are used simultaneously, there are 4 possible cases
where a pass/fail criterion needs to be defined, see Table 9:
- IEC & UHF methods reveals no PD activity
- IEC & UHF methods reveals PD activity
- IEC method reveals PD activity, UHF method reveals no PD
- IEC method reveals no PD, UHF method reveals PD
One goal for future PD testing may be to focus on UHF PD measurements for SAT, therefore the
acceptance criteria need to be further developed. Table 9 collates the conditions where presently clear
criteria are defined and where criteria remain to be defined in the future. A “No” means here, that the
respective method didn´t reveal any PD activity according its criteria. The table collects open questions
for a future working group after collection of experiences with calibrated UHF PD measurements.
Table 9 Overview on criteria for FAT and SAT (green: criteria defined, yellow: criteria missing so far)
FAT SAT
IEC UHF UHF
Conclusion/Future criteria Conclusion/Future criteria
PD PD PD
No Good quality of transformer
No No Good quality of transformer Something new has appeared.

Yes
What can be advised?
No Any problem was fixed

Apply usual IEC test criteria.
Yes Yes Fingerprint comparison to FAT;
UHF as further diagnostic tool.
Is it worse than the level
Yes
accepted during FAT?
Conclusion?
No UHF was useless anyway
Apply usual IEC test criteria?
Yes No Can UHF proof the quality of Something new has appeared!?
Yes
the transformer? What can be advised?
No Any problem was fixed

Currently: Transformer passed
Fingerprint comparison to FAT;
test.
No Yes Is it worse than the level
What do we advise in future? Yes
accepted during FAT?
Use UHF as fingerprint?
Conclusion?
101
If both methods show a “No” indication for PD, the test is obviously passed. If both methods show
indication for PD, currently the IEC criteria should be used to define a pass or no pass action. In case
of a missing indication from the UHF PD measurements, it is the question if only the IEC test criteria
should be used or if the transformer “get a second chance” by a passed UHF PD test. Same applies the
other way round, if only UHF PD activity is revealed while the IEC measurements reveal no PD (this has
been experienced, as described in Chapter 6.3).
Assuming that only UHF PD measurements are performed during SAT, there are several possibilities
that are outlined in Table 9. Especially the yellow marked possibilities are presently undefined, so
manufacturer and customer cannot rely on any specific guidance. Before using UHF technology as an
acceptance criterion, those cases need to be investigated and experience gained so that any standard
developed in the future can cover those cases.
For example, if UHF measurements confirm again no PD, the SAT is obviously passed, and the
transformer should be in good condition. However, if UHF measurements reveal suspicious signals, what
should be the technical conclusion? Something new has happened, a new PD source is active, but how
critical is that for the transformer? What if IEC measurements are not possible, e.g., due to ambient
noise?
The working group concluded that it is too early to define pass or fail criteria for UHF measurement
before the establishment of the calibration procedure. Therefore, this topic needs to be addressed by
future expert groups.
9.2 Decision Tree

Another attempt to visualize and summarize the discussions in the working group is the so called
“decision tree”. The flow chart below in Figure 9-1 should give an overview how UHF measurements
can be used as a supportive tool for FAT and SAT. By intention, the decision tree is not showing details
of the measurements itself. In addition, basic conditions like calibrations, suitable test setups, availability
of equipment and minimum requirements on the transformer under test are considered as given. There
are no thresholds mentioned for the PD measurement. The evaluation of UHF signals is under
responsibility of the test crew, test engineer or transformer experts.
Starting point is the intention to check the integrity of a power transformer’s insulation system by
performing “High Voltage Testing”.
In advance to each test, the environmental condition is checked on electromagnetic disturbances.
If the electric noise within the IEC 60270 frequency range is below the approval level in IEC 60076-3,
the test can be performed according to both IEC standards. This is usually the case in shielded test
fields; therefore, this part of the decision tree is called the FAT part. On the other hand, in the presence
of electrical disturbances higher than approval level, which is often the case for SAT, the decision is
going to the SAT part of the chart.
9.2.1 Factory Acceptance Test (FAT)
The acceptance test will be performed according IEC 60067, including PD measurements according IEC
60270. As an optional supportive tool, two UHF antennas can be used for fingerprint measurements or
for a deeper investigation in case of active PD sources during the test.
As a minimum requirement for a successful performance check (see chapter 4.8) at least two UHF
antennas are necessary. If the performance check is successful, the UHF measurement is able as a
sensitive additional measurement during high voltage test. If not, due to the unknown propagation path
to the UHF antenna, the UHF measurement cannot give a reliable PD detection sensitivity to transformer
under test and UHF is not applicable. Nevertheless, even if so, the UHF antenna(s) can remain in the
transformer for learning / evaluation purpose.
Note: It is conceivable to use the UHF measurement during FAT as a fingerprint for SAT tests. However,
these assumption needs to be evaluated further and more experiences are needed to substantiate this
intention.
102
Figure 9-1 Decision Tree for future combined electrical and UHF FAT/SAT measurements
103
During test level of the induced voltage, PD measurements are monitoring the behavior of the
transformer under test.
• If a voltage breakdown appears, the transformer did not pass the test.
• If the electrical PD measurement shows no PD activity exceeding the acceptance criteria of IEC
60076, the transformer passed the test.
• In case of PD activity beyond the acceptance criteria and the UHF measurement will show
internal PD the transformer did not pass the test and further investigations will be needed.
• If the UHF measurement is clearly indicating only external PD, the test is passed.
9.2.2 Site Acceptance Test (SAT)
For SAT, it might be reasonable and necessary to change the frequency range of the electrical PD
measurement outside of the IEC 60270. The best signal to noise ratio at the lowest possible frequency
range would be recommended to use. It should be decided on individual base if the standard
IEC 60076-3 needs to be applied or not. In general, IEC 60076 is not intended for SAT and a good test
value for the induced test would be 110% nominal voltage.
In the unlikely case of high broadband noise over the whole range up to some MHz, the electrical PD
measurement cannot be used as a valid SAT criterion. In this case, maybe some improvements on the
shielding or the test setup should be considered.
For UHF PD measurements, at least two antennas are necessary to install on the transformer under test
in order to apply a performance check. If the performance check is successful, the UHF measurement
is able as a sensitive additional measurement during high voltage test. If not, due to the unknown
propagation path to the UHF antenna, the UHF measurement cannot give a reliable PD detection
sensitivity to transformer under test and UHF is not a good SAT criterion. Nevertheless, even if so, the
UHF antenna(s) can remain in the transformer for learning / evaluation purpose.
During test level of the induced voltage, PD measurements are monitoring the behavior of the
transformer under test.
If a voltage breakdown appears, the transformer did not pass the test.
For PD observation during test, both PD measurements should be applied simultaneously if possible.
In case of PD activity exceeding the acceptance criteria the results need to be discussed, as they will in
the case of PD signals received only from one measurement (either UHF or electrical PD).
If both measurements show no PD activity exceeding the acceptance criteria, the transformer passed
the test.
9.3 Outlook
After collecting UHF measurement results with a calibrated UHF measuring system, a following working
group may establish a way of defining threshold values or other acceptance criteria. Research teams
around the globe are further asked to perform calibrated UHF measurements as well to collect as much
comparable experience as possible with the “new” UHF technique.
104
10. Conclusion
The main improvement for factory and site acceptance tests provided by this brochure is the possibility
for standardization of UHF PD measurements by introducing a calibration method. Hence, UHF PD
measurements can become reproducible and comparable independent of the equipment setup or
instrument manufacturer. This standardization of UHF measurement procedures will provide benefits
for both transformer manufacturers and utilities.
Concerning conventional measurements according to IEC 60270, a calibration method has been
established for many years, meaning the basic requirement for comparability is already met. In relation
to larger test objects and/or test objects with windings, such as power transformers, electrical PD
measurements are recommended in the lower frequency range (upper cut-off frequency typically 200
to 300 kHz) for FAT scenarios. However, such settings might lead to noisier PD measurements under
SAT conditions. Examples given in this brochure show that the apparent charge value can differ
significantly dependent on the PD location, even when measuring in a frequency range according to
IEC 60270. Regarding this fact, it is questionable whether a pass/fail decision for FAT and SAT that
relies solely on measuring the apparent charge level as acceptance criteria can provide the optimum
quality assurance.
For the comparison between FAT and SAT, documentation is as important for IEC PD measurement as
it is for UHF. Regarding IEC, documentation should include a description of the setup, the capacity of
the coupling capacitor, the PD calibration level applied (e.g., 100 pC), recorded noise levels after
calibration, the frequency range used, and details of the measurement instrument. In general, the
frequency range chosen within the IEC limits should be as low as possible. The low-frequency range of
PD signals experiences the lowest attenuation and hence provides the highest sensitivity for PD signals
throughout the transformer. Measuring at frequency ranges outside the IEC spectrum or using narrow-
band measurements carries risks of greater error and defeats the purpose of standardization.
Conventional measurements offer the advantage of acceptance criteria for a clear pass/no pass decision
for FAT and SAT.
To achieve comparability for UHF PD measurements, this brochure has defined a general UHF calibration
procedure. That calibration requires the use of UHF sensors with a known antenna characteristic
(calibrated sensitivity). The characteristics can be provided either by the antenna factor AF, or in a
simplified, more practical approach proposed in this brochure by the KS factor of the sensor. The
influences of both the setup (e.g., cable attenuation) and the measurement system can be calibrated
with the introduced factor KM. A measuring system should be able to incorporate the calibration factors
KS and KM in its signal processing software. In addition, the UHF measurement system has to provide a
frequency-independent signal level accuracy over the entire dynamic range of the system. If all
aforementioned requirements are met, it is possible to cover the entire signal input range with only one
calibration impulse. A “frequency range” for UHF measurements and an “insertion depth” for valve UHF
sensors has been proposed. Nevertheless, it is recommended to use predominantly permanently
installed UHF sensors (such as the window type) for new transformers.
To provide high coverage for PD signal detection throughout the power transformer, 4 UHF sensors are
recommended as the standard configuration. At least two sensors should be used as an absolute
minimum configuration to enable a performance check. As a high-end configuration, 6 to 8 sensors can
be used, depending on transformer size and complexity. For new transformers, it is advised to install
window UHF sensors directly instead of valve UHF sensors, because they are easier to use (fixed
insertion depth, no moving parts, installation without oil handling) and have a higher sensitivity (better
earthing, lower disturbances measurable, flatter frequency response).
It is recommended to run a dual-port performance check on each combination of UHF sensor pairs. The
check ensures a sufficient sensitivity for UHF signal across the transformer as well as identifying
potential errors in the setup (e.g., valve UHF sensors could be installed at unsuitable positions or at an
incorrect insertion depth). Relevant points to be documented for SAT and FAT should be the UHF sensor
positions, used sensor types and their KS factors, details of the measurement system and the obtained
KM factors for each measurement channel and additional notes about the performed dual-port
performance checks between each possible sensor pair. For a FAT, any observed PD activity should be
recorded together with the applied voltage (e.g., PDIV), their measured levels and a PRPD pattern
should be included as fingerprint reference for later SAT measurements.
105
If possible, FAT and SAT should be performed using both PD measurements methods, electrical
according to IEC and UHF. Hence, the overall sensitivity to detect PD will be enhanced and the usage
of different physical propagation mechanisms provides the option to validate and confirm single
measurements. In most cases, UHF PD measurements are easier to perform at SAT compared to IEC
measurements due to the noisy electromagnetic environment. Therefore, a PD measurement validated
using both methods at FAT could be confirmed at SAT using at least one method.
The brochure has presented the most common sources of noise for especially SAT conditions. This has
demonstrated under which conditions the UHF PD measurements can be regarded as resilient against
the most common noise of external corona discharge representing the largest disturbance potential for
electrical measurements. Condenser bushings usually provide a helpful low pass filter behavior that
shields internal UHF sensors from corona disturbances. Knowledge of other noise sources and their
characteristics are presented to optimize the signal-to-noise ratio (SNR) for improved PD measurements.
Improvements can be achieved by applying the UHF calibration procedure and the use of calibrated
sensors and measuring systems as well as choosing an optimized low noise voltage excitation.
FAT and SAT require acceptance criteria or threshold values acting as “pass/no pass” limits. Due to
several reasons highlighted in the brochure, the only focus of the apparent charge is not meaningful. It
is the common understanding of the members, that as of today, it is not advisable to make any
recommendations on acceptance thresholds for UHF measurements due to the lack of practical
experience with calibrated UHF measurement systems.
In several measurement campaigns, the validity and feasibility of the proposed calibration method has
been demonstrated in a laboratory setup by using different UHF PD measurement devices. The tests
revealed a total measuring inaccuracy of only about 10% across 4 different measurement systems in a
setup with an artificial stable PD signal. In a second approach, the same laboratory setup was used with
two real PD sources at applied high voltage. With 5 different measurement devices and 2 different
sensor types, the calibration proposal was evaluated again. The results showed higher errors than using
the artificial source that can be explained by variation of the PD source, different scaling of the systems,
manual reading of the values out of PRPDs, etc. The identified error sources of the measuring system
need to be reduced in the future.
As with the electrical PD measurement method, an algorithm similar to that described in IEC 60270 to
display the “largest repeatedly-occurring UHF PD magnitude (EUHF in [V/m))” is proposed. Additionally,
manufacturers need to incorporate the facility to perform KM and KS calibration into their software and
rescale the visualization accordingly to V/m.
With UHF measurement systems that have incorporated both the proposed UHF calibration procedure
and the automatic output of the “largest repeatedly occurring (UHF) PD magnitude”, a measurement
campaign similar to the presented laboratory test needs to be performed: Using such calibrated systems,
the transformer community can gain experiences with UHF PD measurements and may identify criteria
for UHF PD testing. After collecting UHF measurement results with a calibrated UHF measuring system,
a future working group may develop rules for threshold values or other acceptance criteria for
acceptance tests regarding UHF. Research teams around the globe are asked to perform calibrated UHF
measurements as well to collect as much comparable practical experience as possible with the proposed
calibrated UHF technique.
106
11. References
[1] International Electrotechnical Commission, IEC 60076-3: Power transformers - Part 3: Insulation levels,
dielectric tests and external clearances, Geneva, Switzerland: IEC, 2015.
[2] International Electrotechnical Commission, IEC 60270: High Voltage Test Techniques - Partial Discharge
Measurements, Geneva, Switzerland: IEC, 2000.
[3] IEEE 100, The Authoritative Dictionary ofIEEE Standards Terms, New York: IEEE, 2000.
[4] JCGM 200:2012, International vocabulary of metrology - Basic and general concepts and associated terms
(VIM), JCGM, 2012.
[5] S. Coenen, Measurements of Partial Discharges in Power Transformers using Electromagnetic Signals,
Germany: BoD - Books on Demand, Norderstedt, , ISBN 978-3-84821-936-0, 2012.
[6] A. Reid, M. Judd, T. Fouracre, B. Stewart and D. Hepburn, "Simultaneous Measurement of Partial Discharges
using IEC60270 and Radio-Frequency Techniques," IEEE Trans. Dielectrics and Electrical Insulation, Vols.
Vol. 18, No. 2, no. April, pp. pp. 444-455, 2011.
[7] H. Jahangir, A. Akbari, P. Werle and J. Szczechowski, "Possibility of PD calibration on power transformers
using UHF probes," IEEE Transactions on Dielectrics and Electrical Insulation, pp. vol. 24, no. 5, pp. 2968-
2976, doi: 10.1109/TDEI.2017.006374, Oct. 2017.
[8] IEEE Standards Association, IEEE Guide for Diagnostic Field Testing of Fluid-Filled Power Transformers,
Regulators, and Reactors, C57.152, New York: IEEE, 2013.
[9] S. Markalous, S. Tenbohlen and K. Feser, "Detection and Location of Partial Discharges in Power Transformers
using Acoustic and Electromagnetic Signals," IEEE Trans. on Dielectrics and Electrical Insulation, vol. 6, no.
15, pp. 1576-1583, 2008.
[10] E. Grossmann, Akustische Teilentladungsmessung zur Überwachung und Diagnose vonÖl/Papier-isolierten

Hochspannungsgeräten, Stuttgart: PhD-Thesis of University of Stuttgart, 2002.
[11] CIGRE Technical Brochure 676, Partial Discharges in Transformers, Paris: CIGRE WG D1.29, 2017.
[12] CIGRE Technical Brochure 662, Guidelines for partial discharge detection using conventional (IEC 60270) and
unconventional methods, Paris: CIGRE WG D1.37, 2016.
[13] International Electrotechnical Commission (IEC), IEC TS 62478:2016, High voltage test techniques -
Measurement of partial discharges by electromagnetic and acoustic methods, IEC, 2016.
[14] IEEE Standards Association, IEEE Guide for the Detection, Location and Interpretation of Sources of Acoustic
Emissions from Electrical Discharges in Power Transformers and Power Reactors, C57.127, IEEE, 2018.
[15] A. Gemant and W. Philippoff, "Die Funkenstrecke mit Vorkondensator," Zeitschrift für Technische Physik, pp.
425-430, vol.13 no.9 1932.
[16] W. Böning, „Luftgehalt und Luftspaltverteilung geschichteter Dielektrika I. Untersuchung der Entladungen in
einzelnen Luftspalten bei äußerem Wechselfeld,“ Electrical Engineering (Archiv für Elektrotechnik) Springer
Berlin/Heidelberg ISSN 0948-7921 Print, 1432-0487 Online, vol. 48 no.1 1963.
[17] A. Pedersen, "Partial discharges in voids in solid dielectrics," in Conference on Electrical Insulation & Dielectric
Phenomena, USA, Gaithersburg, 1987.
[18] E. Lemke, "A critical review of partial-discharge models," IEEE Electrical Insulation Magazine, pp. 11 - 16, no.
6 vol. 28 2012.
[19] E. Lemke, "Analysis of the partial discharge charge transfer in extruded power cables," IEEE Electrical
Insulation Magazine, pp. 24-28, no.1 vol.29 2013.
[20] P. Mraz, "Innovative application of frequency response analysis for Partial Discharge measurement," in 19th
International Symposium on High Voltage Engineering, 2015.
107
[21] M. Siegel and S. Tenbohlen, "Comparison between Electrical and UHF PD Measurement concerning Calibration
and Sensitivity for Power Transformers," in International Conference on Condition Monitoring and Diagnosis,
Jeju, Korea, 2014.
[22] G. U. S. Okabe, “Partial Discharge Signal Propagation Characteristics inside the Winding of oil-immersed
Power Transformer - Using the Three-Winding Transformer Model in Air,” IEEE Transactions on Dielectrics
and Electrical Insulation Vol. 18, pp. 2024-2031, 2011.
[23] G. U. S. Okabe, “Partial Discharge Signal Propagation Characteristics inside the Winding of Oil-immersed
Power Transformer Using the Equivalent Circuit of Winding Model in the Oil,” IEEE Transactions on Dielectrics
and Electrical Insulation Vol 19, pp. 472-480, 2012.
[24] Haefely Test AG, "DDX 9121b - PD Frequency range settings," 2018.
[25] Haefely Test AG, "DDX 9121b Partial Discharge Detector".Haefely Test AG.
[26] S. Abdul Madhar,et al, "Frequency Response of a Real Cable Network and its Impact on Field PD
Measurements," in 25th International Conference on Electricity Distribution, Madrid, Spain, 2019.
[27] R. Kurrer, Teilentladungsmessung im Gigahertz-Frequenzbereich an SF6-isolierten Schaltanlagen,

Dissertation der Universität Stuttgart, 1997.
[28] R. Feger, Sensoren und Verfahren der UHF-Teilentladungstechnik an gasisolierten Schaltanlagen, Stuttgart:
Dissertation der Universität Stuttgart, 2001.
[29] S. M. Markalous, Detection and Location of Partial Discharges in Power Transformers using acoustic and
electromagnetic signals, Dissertation der Universität Stuttgart, 2006.
[30] CIGRE Technical Brochure 343, Recommendations for condition monitoring and condition assessment
facilities for transformers, Paris: CIGRE WG A2.27, 2008.
[31] S. Tenbohlen, C. P. Beura, M. Beltle and M. Siegel, “UHF Sensor Placement on Power Transformers for PD
Monitoring,” in Cigré Colloquium SCA2/SCB2/SCD1, India, 2019.
[32] S. Tenbohlen, M. Beltle and M. Siegel, “PD Monitoring of Power Transformers by UHF Sensors,” in
International Symposium on Electrical Insulating Materials (ISEIM), Toyohashi, Japan, 2017.
[33] National Grid Company, “Capacitive couplers for UHF partial discharge monitoring,” Technical Guidance Note:
TGN (T)121, vol. 1, 1997.
[34] M. D. Judd, J. S. Pearson and O. Farisch, “UHF couplers for gas insulated substations - a calibration
technique,” IEEE Proc. Science, Measurement and Technology, vol. 144, no. 3, pp. 117-122, 1997.
[35] M. D. Judd, O. Farisch, J. S. Pearson and B. F. Hampton, “Dielectric windows for UHF partial discharge
detection,” IEEE Trans. Dielectrics and Electrical Insulation, vol. 8, no. 6, pp. 953-958, 2001.
[36] M. D. Judd and O. Farish, “A pulsed GTEM system for UHF sensor calibration,” IEEE Trans. Instrumentation
and Measurement, vol. 47, no. 4, pp. 875-880, 1998.
[37] M. Siegel, M. Beltle and S. Tenbohlen, “Characterization of UHF PD Sensors for Power Transformers Using
an Oil-filled GTEM Cell,” IEEE Transactions on Dielectrics and Electrical Insulation, vol. 23, no. 3, pp. 1580-
1588, 2016.
[38] M. Siegel, S. Coenen, M. Beltle, S. Tenbohlen, M. Weber, P. Fehlmann, S. Hoek, U. Kempf, R. Schwarz, T.
Linn and J. Fuhr, “Calibration Proposal for UHF Partial Discharge Measurements at Power Transformers.,”
MDPI Energies, 12, 3058 2019.
[39] S. Coenen, M. Beltle, M. Siegel and S. Tenbohlen, “Combined In-Oil Sensor for Vibration Measurement and
Partial Discharge Detection using Acoustic and Electromagnetic Emissions,” in CIGRE SC A2 & C4 JOINT
COLLOQUIUM, Zürich, Switzerland, 2013.
[40] M. Siegel, M. Beltle, S. Tenbohlen and S. Coenen, "Application of UHF Sensors for PD Measurements at Power
Transformers," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 24, pp. 331-339, 2017.
[41] M. Siegel, Calibration Methods for Reproducible and Comparable Electromagnetic Partial Discharge
Measurements in Power Transformers, Stuttgart, Germany: BoD - Books on Demand, Norderstedt, ISBN
9783752628180, 2020.
108
[42] J. McLean, R. Sutton and R. Hoffman, “Interpreting antenna performance parameters for EMC applica-tions:
Part 3: Antenna Factor,” TDK RF Solutions Inc., available online at
http://tdkrfsolutions.com/images/uploads/brochures/antenna_paper_part3.pdf.
[43] M. Judd, M. Siegel and S. Coenen, “UHF PD Sensor Characterisation using GTEM Cells,” in VDE High Voltage
Technology, Berlin, Germany, 2018.
[44] C. P. Beura, M. Beltle and S. Tenbohlen, "Positioning of UHF PD Sensors on Power Transformers based on
the Attenuation of UHF Signals," IEEE Trans. Power Delivery (Early Access), 2019.
[45] R. Lebreton, G. Luna and S. Louise, "Detection and localization of partial discharges in power transformers
using four or more UHF sensors," in International Conference on Condition Monitoring, Diagnosis and
Maintenance, 2013.
[46] C. P. Beura, M. Beltle and S. Tenbohlen, "Attenuation of UHF Signals in a 420 kV Power Transformer based
on Experiments and Simulation," in ISH, Budapest, 2019.
[47] H. Mirzaei, A. Akbari, E. Gockenbach and K. Miralikhani, "Advancing New Techniques for UHF PD Detection
and Localization in the Power Transformers Advancing New Techniques for UHF PD Detection and Localization
in the Power Transformers in the Factory Tests," IEEE Transactions on Dielectrics and Electrical Insulation,
Vol. 22 2015.
[48] S. M. Hoek, A. Kraetge, M. Krüger and S. Körber, "D1-202: Application of the UHF technology to detect and
locate partial discharges in liquid immersed transformer," in CIGRE Session, Paris, France, 2014.
[49] S. Coenen, M. Siegel, M. Beltle, S. Tenbohlen, S. Hoek, P. Fehlmann, R. Schwarz, T. Linn, U. Kempf, M.
Weber and J. Fuhr, “D1-116: Proposal of a Calibration Methodology of UHF Partial Discharge Measurements
for Power Transformers,” in Cigré Session 48, Paris, 2020.
[50] International Electrotechnical Commission (IEC), IEC 60270: High Voltage Test Techniques - Partial Discharge
Measurements, Geneva, Switzerland, 2016.
[51] Liu, S.; Chen, J.; Deng, K.; Liu, H.T., Xiang J.; Du, L.; Xu, Z.; Zhang, L. Mao, W.; Li, P., “UHF Signals
Calibration for Typical Partial Discharge Defects in Transformer Oil,” in International Workshop on Wireless
Technology Innovations in Smart Grid (WTISG), Halifax, Canada, 2019.
[52] H. Jahangir, A. Akbari, P. Werle and J. Szczechowski, “UHF PD Measurements on Power Transformers -
Advantages and Limitations,” IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 24, No. 6, pp.
3933 - 3940, 2017.
[53] International Electrotechnical Commission (IEC), IEC TS 62478:2016: High voltage test techniques -
Measurement of partial discharges by electromagnetic and acoustic methods, 2016.
[54] Cigré Technical Brochure 654, UHF Partial Discharge Detection System for GIS: Application Guide for
Sensitivity Verification, Paris, France: Cigré, 2016.
[55] M. Siegel, S. Tenbohlen, S. Coenen, B. Dolata, G. Luna and S. Louise, "Practical Sensitivity of online UHF PD
Monitoring on Large Power Transformers," in CIGRE SC A2 COLLOQUIUM, Shanghai, China, 2015.
[56] M. Judd, G. Cleary and S. Meijer, “Testing UHF partial discharge detection on a laboratory based power
transformer,” 13th International Symposium on High Voltage Engineering Delft, 08 2003.
[57] R. A. Jongen, P. Morshuis, S. Meijer and J. Smit, “Identification of Partial Discharge Defects in Transformer
Oil,” in Conference on Electrical Insulation and Dielectric Phenomena, 2005.
[58] T. Katsuse, T. Kirishima, A. Morita, S. Ohtsuka and M. Hikita, "International Conference on Condition
Monitoring and Diagnosis,," in Partial Discharge Characteristics in UHF-band of Composite Insulation System
Including Various Artificial Defects in Transformer, Beijing, China, 2008, April 21-24.
[59] K. Raja, F. Devaux and S. Lelaidier, "Recognition of Discharge Sources Using UHF PD Signatures".IEEE
Electrical Insulation Magazine.
[60] J. Li, C. Wang and C. Cheng, “Optimization of UHF Hilbert Antenna for Partial Discharge Detection of
Transformers,” IEEE Transactions on Antennas and Propagation, 05 05 2012.
[61] Bundesnetzagentur für Elektrizität, Gas, Telekommunikation, Post und Eisenbahnen; Referat 221, Dienststelle
221-1b (FreqP), Frequenzplan gemäß § 54 TKG über die Aufteilung des Frequenzbereiches von 0 kHz bis
3000 GHz auf die Frequenznutzung sowie über die Festlegungen für diese Frequenznutzung, Bonn, 2016.
109
[62] M. Judd, G. Cleary and S. Meijer, "Testing UHF partial discharge detection on a laboratory based power
transformer," in 13th Int. Symp. on High Voltage Engineering, Delft, Netherlands, 2003.
[63] P. S. R. Gatechompol, “Partial discharge measurement in a transformer: comparing between conventional

measurement and UHF measurement,” High Voltage Testing Laboratory, Electricity Generating Authority of
Thailand, 2017.
[64] R. Gerritse, L. Dorpmanns and A. Thiede, "On-site Testing of Generator Transformers," in HIGHVOLT
Kolloquium, 2013.
[65] X. Zhao, Y. Cheng, Y. Meng, K. Wu and Y. Niu, “The propagation characteristics of UHF partial discharge in
power transformers with complex winding structure,” in IEEE Conference on Electrical Insulation and
Dielectric Phenomena, Montreal, QC, Canada, 2012.
[66] C. Beura, M. Beltle and S. Tenbohlen, "Study of the Influence of Winding and Sensor Design on UHF PD
Signals in Power Transformers," vol. Sensors MDPI, 2020.
[67] CIGRE Technical Brochure 654, HF Partial Discharge Detection System for GIS: Application Guide for
Sensitivity Verification, Paris: CIGRE WG D1.25, 2016.
[68] M. Wynd, “Changing Places,” Power Engineer, Vol. 21, No. 2, IET Publishing, pp. 26-31, 04/04 2007.
[69] D. Gautschi, T. Weiers, G. Buchs and S. Wyss, "Ultra high frequency (UHF) partial discharge detection for
power transformers: Sensitivity check on 800 MVA power transformers and field experience with online
monitoring, A2-115," in CIGRE Session, Paris, France, 2012.
[70] S. Coenen, M. Hässig, M. Siegel, J. Fuhr, S. Neuhold, T. Brügger, S. Hoeck and T. Linn, “Placement of UHF
Sensors on Power Transformers,” in ETG Conference, Berlin, Germany, 2018.
[71] S. Neuhold, H. Benedickter and L. Schmatz, "A 300 V Mercury Switch Pulse Generator with 70 Psec Risetime
for Investigation of UHF PD Signal Transmission in GIS," in ISH, London, 1999.
[72] M. Siegel and S. Tenbohlen, "Calibration of UHF Partial Discharge Measurement for Power Transformers," in
EIC Conference, Montreal, Canada, 2016.
[73] M. Judd, "Transient calibration of electric field sensors," IEE Proc. Science, Measurement and Technology,
vol. Vol. 146, no. No. 3, pp. 113-116, May 1999.
[74] C. Harrison, "The radian effective half-length of cylindrical antennas less than 1.3 wavelengths long," IEEE
Trans. Antennas and Propagation, Vols. Vol. AP-11, no. No. 6, pp. 657-660, 1963.
[75] H. Schmitt, C. Harrison and C. Williams, "Calculated and experimental response of thin cylindrical antennas
to pulse excitation," IEEE Trans. Antennas and Propagation, Vols. Vol. AP-14, no. No. 2, pp. 120-127, 1966.
110
APPENDIX A. Quality assurance of uhf partial

discharge sensor calibration measurements
A.1. Introduction
This Appendix outlines the theory, standards and measurement techniques that can ensure the validity
and traceability of measurements carried out on a GTEM UHF sensitivity calibration system for UHF PD
sensors.
The monopole probe sensors reported here (“Pa” and “Pb”) were calibrated relative to a third probe of
identical structure, which was certified according to international measurement standards at the
National Physical Laboratory in the UK and provided with a Certificate of Calibration documenting its
measured frequency response. The probes discussed in this document are intended for routine use in
checking and normalising the response of a UHF GTEM calibration system.
A.2. Summary of measurement principles

A UHF PD sensor calibration technique based on time-limited step response measurements was first
proposed in a scientific paper published in 1999 [73]. The fundamental basis of the measurement is
that a monopole sensor aligned perpendicular to a conducting ground plane has an accurately defined
frequency response that remains constant from one probe to another, provided that the dimensions of
the probe (length and diameter) remain the same. This principle was originally described and verified
in the 1960s [74], [75]. The equivalent circuit and parameters that define the frequency response in
[74] were for a relatively large monopole at lower frequencies. This data was scaled for smaller
monopole operating at higher frequencies in [73], resulting in the theoretical frequency response shown
in Figure C-1. This accurately defined response is the basis of the GTEM calibration method.
Figure A-1 Theoretical response of a 25 mm long, 1.3 mm diameter SMA monopole [C1].
The UHF sensitivity measurement process outlined in [73] is implemented in the calibration analysis
software which determines the ration of the computed frequency spectrum of the output voltage of a
sensor under test to that of the 25 mm reference probe. This data is then scaled by the accurately
known response of the probe shown in Figure A-1.
For example, if the voltage output of a sensor under test at 900 MHz was 5 times larger the
corresponding voltage spectral density of the reference monopole, the sensitivity of the sensor under
test at 900 MHz would be 5 × (effective height of the reference probe at 900 MHz) = 10 mm.
Requirements for this relationship to be valid are that: (i) the same incident electric field step must be
applied to the sensor under test and the reference probe, and (ii) the measurement cables and sampling
oscilloscope should be the same for both measurements.
Prior to beginning the process of UHF PD sensor calibration, a monopole probe (25 mm long × 1.3 mm
diameter) is mounted at the centre of the GTEM test plate and the measurement is normalised, which
results in the system displaying a response very similar to that shown in Figure A-1.
111
A.3. Traceability of the Measurements

Validity and traceability of the measurements is achieved through testing at an accredited laboratory,
such as the National Physical Laboratory (NPL) in the UK. The certified probe is tested in the same
configuration as it would be mounted in the sensor calibration system, that is, inserted through the
ground plane of a test plate, as shown in Figure A-2. 25 mm Aluminium
mounting
block
25 mm 25 mm
60 mm
Figure A-2 Mounting arrangement for the 25 mm monopole reference probe.

A comparison of calibration results between an NPL measurement and the UHF coupler calibration
system data can be seen in Figure A-3. Note that the tolerance of electric field sensitivity measurements
quoted by NPL is ±2.5 dB, which is a typical uncertainty level for electric field sensor measurements.
The comparison in Figure A-3 confirms that the probes “Pa” and “Pb” closely match the response of R5
(maximum deviation is ±0.2 dB over the frequency range 500 – 1500 MHz).
UHF probes S/N Pa & Pb compared with NPL certified reference probe S/N R5
100.0
reference probe S/N R5 (NPL Certified)

25mm reference probe S/N Pa
25mm reference probe S/N Pb
10.0
Effective Height ( mm )
1.0
0.1
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
Frequency (MHz)
Figure A-3 Comparison between the NPL calibrated reference probe R5 mounted as shown in
Figure A-2 and the responses of two other monopole probes “Pa” and “Pb”.
References for Appendix A

[C1] M D Judd, "Transient calibration of electric field sensors", IEE Proc. Science, Measurement and Technology,
Vol. 146, No. 3, pp. 113-116, May 1999
112
APPENDIX B. Recommended documentation

for FAT and SAT PD measurements
The following points should be documented to achieve comparable UHF PD measurements:
• Sensors: Types, Serial numbers, Sensor positions, Insertion depth (in case of Drain valve),
Picture of mounted sensors, KS
• Measuring system: Type, Serial number, used Accessories (Pre-amp, filter, cables), Photo and
diagram of set-up
• Measuring system: relevant Set-up: Frequency range, auto gain ranges, …
• Calibration: signal generator type, Serial number, used amplitude, KM
• KUHF (see chapter 4.5.4)
• Performance Check: pass/fail, signal generator type, serial number, used amplitude, measured
UHF signal,
• “Fingerprint” PRPD at 1.0 Un for later SAT / online, PRPD settings: recorded time, X-, Y- axes
settings (logarithmic, linear, maximum value) colour coding (logarithmic, linear, maximum
number)
113
ISBN : 978-2-85873-566-2
TECHNICAL BROCHURES
©2022 - CIGRE
Reference 861 - February 2022

861 Improvements To PD Measurements For Factory and Site Acceptance Tests of Power Transformers

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

861 Improvements To PD Measurements For Factory and Site Acceptance Tests of Power Transformers

Uploaded by

Copyright:

Available Formats

A2/D1

S. COENEN, Convenor DE T. LINN CH

2. Definitions And Initial Considerations ................................................................... 14

3. Electrical method according IEC 60270 and its calibration .................................. 18

4. Electromagnetic Method (UHF) and its Calibration ............................................... 32

4.7 Comparison of electrical and UHF measurement procedure .............................................................. 51

5. Technical considerations of PD measurements for FAT and SAT ...................... 56

6. Case Studies ............................................................................................................ 80

7. Measurement campaigns of The Working Group .................................................. 89

8. Recommendations for improved PD acceptance tests ......................................... 99

9. Outlook for further improvements .........................................................................101

10. Conclusion ..............................................................................................................105

11. References ..............................................................................................................107

Figures and Illustrations

1.1 Objective of Brochure

1.2 Structure of Brochure

2. Definitions And Initial Considerations

2.2 Definition of Calibration

Table 1 Description of terms related to calibration

Term Description Definition/ Description/ Definition/ UHF

2.3 Definition of Factory Acceptance Test (FAT)

2.4 Definition of Site Acceptance Test (SAT)

2.5 Motivation for improvement of PD measurements during FAT and

2.6 Relevant PD Measuring Methods for FAT and SAT

3. Electrical method according IEC 60270 and

3.1 Calibration according to IEC 60270

𝑞𝑎 = 𝑞𝑐 (𝐶𝑏 /𝐶𝑐 ) (1)

Figure 3-3 Schematic of the PD measurement setup

Figure 3-4 Sensitivity of the measurable charge in dependence of the Ck to Ct ratio

1 10 100 1000 10000

Partial discharge FFT RF broadcast FFT Tyristor pulse FFT

Figure 3-5 Typical frequency spectrum of various signal sources

Figure 3-6 White noise (top) and its FFT (bottom)

Figure 3-7 Switching noise (top) and its FFT (bottom)

Figure 3-8 Sinusoidal noise (top) and its FFT (bottom)

3.1.5 Step 2: checking the calibration spectrum

3.1.6 Step 3: analyzing the real PD pulse spectrum

3.1.7 Real field measurement example

Measurement results using measuring frequency range 100 – 600 kHz

• Conventional PD measurement at higher frequencies is less meaningful since the frequency

• Resonances in the PD measuring band invalidate the calibration based on quasi-

• Checking the frequency spectrum of the real-PD pulse is highly recommended as it

3.2 Acceptance criteria according IEC 60076-3

Table 2 Requirements and tests for different categories of windings [1]

4. Electromagnetic Method (UHF) and its

Figure 4-1 Principal of UHF PD Measurements in power transformers [5]

4.1 UHF sensors

4.1.1 Valve UHF sensors

Main advantages compared to valve UHF sensors are:

Figure 4-8 Example of UHF sensor mounting on a flat dielectric window

4.2 Sensor characteristics

Figure 4-11 Transmission measurement (S21) for AF determination

Drain Valve Combined PD Window

4.2.4 Insertion depth of valve UHF sensors

4.3 Recommendation for Placement of UHF Sensors on Power

Artificial PD Source (Phase V Position 4)

4.3.3 Positioning on the top of the tank

4.3.4 Positioning on the walls along the width of the transformer

4.3.5 Positioning near the corner of the tank

Figure 4-24 Suggested placement for 4 UHF sensors on a transformer [47]

4.4 Recommendations for the Number of UHF Sensors on Power

Table 4 Recommended number of UHF sensors for power transformers

Minimal configuration 2 sensors (minimum for performance check) (Figure 4-19)

-79.8 dBm1 -73.4 dBm1 -68.5 dBm*1

-83.4 dBm2 -77.9 dBm2 -65.9 dBm*2