Our Response – Use of Resistive Temperature

Detectors as Partial Discharge Sensors in

Rotating Equipment
Claude Kane Igor Blokhintsev
Alexander Golubev Cal Patterson
Electrical Diagnostic Innovations, Inc Eaton Corporation - Predictive Diagnostics group
Plymouth, MN Minnetonka, MN and Burlington, ON

Abstract: Much debate and misinformation has been published in II. RESPONSE TO CONCLUSIONS OF THE 2006 PAPER
various technical journals and in the marketplace on the use of
Resistive Temperature Detector (RTDs) as a sensor of partial Table 1 reviews the conclusions from the 2006 paper and the
discharges (PD) in motors and generators. The authors have
our response to each point made.
more experience in this field of technology than any other
company in the world. With hundreds of applications working in
operating facilities and millions of data records, this technology is TABLE I
well proven. This paper addresses the concerns in using RTDs as RESPONSES TO THE CONCLUSIONS OF 2006 PAPER
a PD sensor through sound technical analysis and case studies. 2006 Paper Conclusion
Specific response to the paper titled “Investigations Into the Use Experiments have been conducted to determine the ability of RTDs in
of Temperature Detectors as Stator Winding Partial Discharge stator windings to detect PD. Tests were done on operating and off-
Detectors” presented at the 2006 IEEE International Symposium line stators from motors, hydro generators and a turbine generator, as
on Electrical Insulation is made. well as more controlled simulated coils in slots. It is clear that in most
circumstances PD can be detected by the RTDs. However, rather than
I. INTRODUCTION the RTD itself detecting the PD, it is the lead (or wires) from the
RTD that detect the PD signal.
Four new case studies are presented to once again
substantiate the use of this technology. One case study Response
compares signals from an RTD that is installed in the same slot This is a well known fact and has never been disputed. Our
as a Stator Slot Coupler (SSC). experience shows that both the RTD and the wiring will detect the
PD signals. What difference does it make what picks up the PD? The
A point by point response to the paper referenced in the RTD or the wire? A PD signal is a PD signal. Antennae are used by
abstract is not made, however, the conclusions of the 2006 many companies to detect PD in large power transformers,
paper is analyzed, and appropriate responses given. switchgear and other electrical equipment.
In general, points made in the 2006 paper do not reflect our If one does not understand the physics of sensing does it mean that
experience and do not recognize the main issues in making PD there is no sensing?
measurements in the high frequency ranges. In many cases the leads in a motor or generator act as an antenna.
The main issue is signal attenuation and having limited For example, when a pulse was injected into phase B neutral, the
highest pulse was detected on the Phase C output lead with opposite
sensors, only a very small portion of the winding can be
polarity.
monitored. Each individual sensor has a limited coverage area
and provides limited information as to the health of stator 2006 Paper Conclusion
insulation. Use of more sensors distributed throughout the Since the length of the lead wire, its proximity to high voltage stator
insulation will provide more information and the ability to coils, as well as whether the lead is shielded or not are all usually
provide a better assessment of the insulations health. unknown – it is very difficult to interpret the severity of any PD from
the magnitude of the detected signals.
Thousands of data records and dozens of calibrations show
Response
that RTDs are a valuable and inexpensive, ready-to-go
technology for the measurement of PD in rotating equipment. RTDs with spiral armor at times can attenuate the PD signals. This
Based on both scientific evidence and actual field experience again is not a new revelation and has been presented in technical
papers. Experience shows that in about 10-15% of the cases on older
this technology can not be rejected.
large generators and rarely on motors, the RTDs have this spiral
armor and in 2-3% of those machines the spiral armor attenuates the
signals as the writers state. It looks like the authors have a sample of
“One” and therefore draw a conclusion that this is true for “All
Machines”.
One can usually assess if the RTD wiring has this spiral armor during Response
the installation of the RTD Module/Sensor), therefore it is a known
Utilizing advanced signal processing techniques, the data is sorted
and not an unknown, unlike what the authors state. The responders
and presented showing the common two ‘clumps’. It seems the
have extensive experience in the use of RTDs and have calibrated
authors have no knowledge in using of such techniques.
many systems to verify results. Millions of data records are available
showing significant PD found deeper in windings detected by the Millions of data records from RTDs are available and many have
RTDs and not by coupling capacitors installed at the line terminals. been presented in technical papers that show the common two
‘clumps’ of data. As a matter fact the authors failed to use, any
Also, the health of the insulation should not be based on magnitude
reference or acknowledge any of the professional technical papers
alone. Pulse count, partial discharge intensity and phase resolved
that have been presented on this very specific topic.
data should be considered during evaluation. The same can be said
for traditional coupling capacitors. Without calibration, one has no The actual position of the pulses are not lost. A voltage reference is
idea about true magnitude. always used in order to determine phase position. Also, a set of
coupling capacitors are ALWAYS installed at the line terminals. The
2006 Paper Conclusion use of both types of sensors is complimentary and not mutually
This is verified by the on-line and off-line PD tests on 8 stators where exclusive.
the magnitude of the detected signals from the RTDs had no
2006 Paper Conclusion
correlation to either the detected PD using conventional sensors, nor
the known condition of the insulation. The conclusion of these investigations is that while PD may be
detected by RTD leads, even an expert with decades of experience
Response will have a great deal of difficulty making useful, accurate
This is exactly what the responders have been saying for nearly ten conclusions on the condition of the stator winding insulation.
years. The main reasons there is little to no correlation between the
Response
PD signals on the RTDs and the 80pF coupling capacitors are
twofold: Interpretation of data from the RTDs is no different than data from
traditional coupling capacitors. Magnitudes, pulse counts, phase
1. The 80pF coupling capacitors are not detecting the PD that the
relationships, polarity and trend are all items used in the diagnostics.
RTDs are since this PD is deeper in the winding. This
There is more data to look at since there are more sensors, but more
demonstrates that coupling capacitors see very little of the
information as to the overall health of the machine is also provided.
winding.
2. The authors of the paper used two types of sensors that have very
poor frequency response to PD signals when the tests were III. MAIN ISSUE OF FIELD MEASUREMENTS –
performed. A 10 turn HFCT used by the authors is 2.5 to 3 times SIGNAL ATTENUATION
less sensitive than used in the responders RTD-PD sensors. Even
direct connecting to the RTD with a scope probe has better It is well known in that PD signals have significant
sensitivity. This is still not as good as the sensor the responders
attenuation especially in inductive equipment such as motors
normally use for measuring PD from RTDs. Therefore the claims
being made about sensitivity are invalid.
and generators. Due to the attenuation of PD pulses while
propagating from the place of origin, all PD sensors in the high
The responders have performed dozens of calibration tests over the
frequency range have a limited zone of sensitivity. This feature
last ten years on a variety of machines. There is an excellent
knowledge base established as far as RTD sensitivity and allows the technology to localize a source of PD activity with
effectiveness is concerned. The same is true for 80pF coupling certain accuracy and take preventive measures such as locally
capacitors. One can not start comparing sensitivities of different repair the insulation, replace coils or bars and otherwise save
sensors without understanding the calibration of all sensors. The healthy stators from failure.
sensitivity of a coupling capacitor will vary between one that is
directly coupled to the bus and one that has a two or three foot lead Type of Sensor
length. It is well accepted that sensors installed at the line terminals
Some professionals feel calibration is not necessary since trending is of a machine can not cover the entire insulation system. The
the key component in PD diagnostics. If that is truly the case, then type of sensor and location of sensors are very important.
sensitivity plays no part. Figure 1 shows PD pulses obtained from a 13.8 kV motor that
2006 Paper Conclusion has three 80pF coupling capacitors and three radio frequency
These results also reveal other difficulties with interpreting PD current transformers (RFCTs) installed on the surge capacitor
results using RTD sensors. Since it is the lead that detects the PD, ground circuit. Traces 1, 2 and 3 are the 80pF coupling
and since in most motors the RTD leads take a circuitous path from capacitors , while traces 4, 5 and 6 are from the RFCTs. The
the slot around the stator – the lead is likely to detect PD from all location of the defect in the machine is known and is close to
three phases. Thus one does not normally see just two ‘clumps’ of the main termination box. It can be easily seen the RFCTs are
pulses per cycle (often indicative of PD), but 6 clumps that often
far more sensitive (ten times) to the defect than the coupling
merge into a mass of pulses across the AC cycle, which makes
separating noise from actual stator PD very difficult even for an capacitors. This substantiates the coupling capacitors see very
expert. In addition, the actual position of the pulses with respect to little of the winding. Due to the frequency response of the
the AC phase position is always unknown, and the pulse polarity RFCTs, they can ‘see” deeper into the winding.
information is lost when combined RTD/HFCT sensors are used.
 coupling capacitors. This demonstrates that each sensors is
1> detecting a different PD event.
2>

3>

4>
4

5>

6>

1) CCA: 2 Volt 500 ns

2) CCB: 2 Volt 500 ns
Figure 3 - Different patterns are shown on the RTDs and coupling capacitors.
3) CCC: 2 Volt 500 ns Most of the PD activity is phase to ground.
4) SCA: 2 Volt 500 ns
5)
6)
SCB:
SCC:
2 Volt
2 Volt
500 ns
500 ns
Crosstalk can be significant and depends on multiple factors,
including winding design, sensor location and the origin of a
Figure 1 - Comparison of RFCTs sensors on Surge Capacitors ground (traces
4,5 and 6) and 80pF coupling capacitors (traces1, 2, 3) on the same 13.8 kV PD pulse. In some cases, cross-coupling can be significant, and
motor mislead one about the location of the PD site. This is valid for
all types of sensors including coupling capacitors. In one case
Use of Multiple Sensors a signal was injected to the neutral of B phase of a small
One can expect to have higher magnitudes on the sensor generator stator, and one should expect to see highest
closest to the origin of the PD pulse. The response of six RTDs magnitude on B phase line lead. However, the highest
to a PD pulse originating in the vicinity of RTD#5 in a 13.8 kV magnitude was detected on phase C and pulse polarity was
motor are shown in Figure 2. As can be observed, RTD5 is the opposite to the injected pulse.
closest RTD to the defect since higher magnitudes are present. Calibration performed by injecting signals all over a winding
can provide valuable information on sensitivity and crosstalk.
1>
1 Figure 4 shows the response of three RTD sensors to a signal
injected near the RTD in slot #2 (Trace 1) at turbine end.
2>
RTD#6 (Trace 2) located +4 slots from RTD#2, while
3> RTD#52 (Trace 3) is located -3 slots at exciter end. The
4>
signals from RTD#2 are five times higher than RTD#6 and 15
times higher than RTD#52 located on other end.

5>

6>

1>
1) RTD1: 200 mVolt 100 ns
2) RTD2: 200 mVolt 100 ns
3) RTD3: 200 mVolt 100 ns
4) RTD4: 200 mVolt 100 ns
5) RTD5: 200 mVolt 100 ns
6) RTD6: 200 mVolt 100 ns
2>
Figure2 2 - Response of six RTDs to a PD pulse originating close to RTD 5. 2

Sorting out Cross Coupled Signals 3>

1) RTD2: 10 mVolt 100 ns
Utilizing proprietary algorithms and advanced signal 2) RTD6: 10 mVolt 100 ns
processing techniques, one can “sort” out cross-coupled 3) RTD52: 10 mVolt 100 ns

signals and only show pulses from the sensor that is most Figure 4 - Signals from RTD#2 are five times higher than RTD#6 and 15 times
sensitive to each individual pulse. The authors of the 2006 higher than from RTD#52.
paper apparently do not understand the use of these advanced
technologies and therefore are confused about the ‘6 clumps’ The differences are the combined effect of attenuation along
of data versus the expected ‘2 clumps’. the stator coils, along RTD wires and crosstalk in the stator
Figure 3 shows phase resolved PD distributions for different and between RTD wires. This simple example shows how PD
sensors that have different phase positions and patterns. One sensors are sensitive to a localized area.
can see PD distributions from six RTD sensors and two 80pF
Use of Multiple Sensors and How Many Sensors are Needed
To cover a complete stator on a large turbine or hydro
generator with equal high sensitivity one would require sensors
in every slot on both ends of the stator. Since this is not
realistic it is recommend that 12 RTDs located in the most
electrically stressed parts of the winding, and that are evenly
distributed on the turbine and exciter end of the stator winding
be used.
It can be argued that it is a difficult task to choose the proper
RTD sensor configuration, but reliability of PD measurements
must be taken very seriously and it is strongly felt that optimal
sensor location is the most important factor of the RTD Figure 6 - Multiple PD patterns detected in the stator of a hydro generator
showing complexity of pulse propagation and attenuation.
technology. Use of multiple sensors will provide the user more
data to review, but with a significant increase in reliability in Case Study 2 - Comparing RTD and SSC Measurements
determining the health of an insulation system. It is similar to a
When making an attempt to compare PD sensors one must
physician making a conclusion on a patients health based on
understand that sensors connected to different parts of an
temperature only compared to a physician using full lab tests
object, most probably, will detect different PD activity and
and a MRI.
simple correlation of the response of capacitors to RTD
sensors is just wrong. Figure 7 shows the response of a Stator
CASE STUDIES Slot Coupler (SSC) as well as a RTD that is located in the
same slot directly under the SSC on a 20 kV, 300 MW turbine
Case 1- Hydro Generator generator. The SSC is installed above the top bar (under the
Figure5 shows two snapshots from RTDs and 80pF coupling wedge) and the RTD is between the top and bottom bars.
capacitors on a hydro generator where one can see low or no Measurements show that both types of sensors have almost
response from the capacitors, while having high magnitudes on identical sensitivity.
RTD sensors and vice-versa. The stator had two sets of Figure 7 shows a snapshot of PD pulses detected
capacitors along the ring bus and 12 RTDs evenly distributed simultaneously on SSC72 and RTD72
along the stator.

1>

2>
2

1) RTD72: 50 mVolt 100 ns

PD pulse in vicinity of RTD#4 – no Phase A to phase B discharge 2) SC72: 50 mVolt 100 ns
response from CCs detected by CCA2 and CCB2
Figure 7 - Almost identical response of SSC and RTD sensors to PD pulse.
Figure 5 - Independent registration of two PD events by different sensors.

Figure 8 shows the PD patterns look almost identical for the

Figure 6 shows multiple and different PD patterns detected by three sets of SSCs and RTDs. Minor differences can be
the RTDs and 80pF capacitors. Even if the capacitors explained by the symmetrical position of RTDs between stator
connected to the same ring bus 90 degrees apart, have different bars and the asymmetrical position of SSC. The PD from the
PD patterns detected by those sensors, which confirm that the bottom bar will be shielded by the top bar and one can expect
sensors are local and picking up different partial discharges. lower sensitivity on the SSC to PD pulses originating in the
Attenuation of a calibration pulse along a quarter of the ring bottom bar.
bus is ~0.4 and one can expect only ~15% of the PD signal
coming from the most distant point on the ring bus to the
sensor.
Figure 8 - PD patterns by SSC and RTD sensors located next to each other.
Contrary to the conclusion of the authors of the 2006 paper
that RTDs having a spiral armor have low sensitivity, this case
study shows the RTD has a slightly higher sensitivity than the
SSC. RTDs in this case have a spiral metal armor as can be
seen in Figure 9.

Figure 10 - Fast aging of electrical insulation in different parts of the stator

after swapping line and neutral leads.

Case Study 4 – Capacitors and RTDs are Complimentary

It is strongly promoted and recommended that both capacitors
and RTDs be used to provide a reliable assessment of stator
insulation health. Using both types of sensors is
complimentary and not meant to be mutually exclusive.
The capacitors will be more sensitive to PD occurring near
the line termination area while the RTDs will be more
sensitive to PD occurring deeper in the winding. This provides
an opportunity for much more reliable diagnostics.
Figure 11 and Figure 12 shows the Partial Discharge Intensity
(PDI) distributions and phase resolved data for two distinctive
cases with PD activity is mostly detected only by one type of
sensor.
Figure 11 shows a case where the detected PD is deeper in the
windings and there are very low signals at the coupling
capacitors. Conversely, Figure 12 shows a case where the PD
Figure 9 - RTD termination compartment, one can see bundles of RTD wires is near the line terminals and the capacitors detect the majority
in spiral armor coming from the stator.
of the PD and the RTDs show fairly low PD activity. In this
Case Study 3- Swapping of Line and Neutral Leads case the defects are closer to the line terminal area.
This study shows the progression of aging insulation on a
machine where the line and neutral leads were interchanged.
Figure 10 shows both the similarities and differences of trend
from different sensors. The capacitors and RTDs each show
different trends illustrating a very interesting case of fast aging
of old insulation by electrical stress. The effect of CO2
cleaning of the end winding area is seen on the capacitors.
 misleading statements that reflect a lack of understanding the
measurement of PD in the higher frequency ranges.

It is time to stop debating the value of utilizing existing RTDs

as a PD sensors. With minimum effort from electrical
societies, manufactures and repair facilities, one can have an
excellent, natural technology for estimation of stator insulation
conditions on-line, verifying quality of repairs, saving costs on
maintenance, extending insulation life and preventing critical
failures.

REFERENCES
[1] S.R. Campbell, G.C. Stone “Investigations into the Use of Temperature
Detectors as Stator Winding Partial Discharge Detectors”, Conference
Record of the 2006 IEEE International Symposium on Electrical
Insulation, Toronto, ON, pp 369-375.
[2] IEEE Std 1434-2000 Guide to Measurement of Partial Discharges in
Rotating Machinery
Figure 11 - Low PD activity in line termination area are sensed by the
capacitors and high PD activity occurring deeper in winding are sensed by the [3] Z. Berler, I. Blokhintsev, A. Golubev, G. Paoletti, A. Romashkov, “RTD
RTDs. as a Valuable Tool in Partial Discharge Measurements on Rotating
Machines”, 67th Annual International Conference of Doble Clients,
March 27-31, 2000, Boston, MA
[4] C. Kane, A. Golubev, I. Blokhintsev, “Further Experience in the Use of
Existing RTDs in Windings of Motors and Generators for the
Measurement of Partial Discharges”, Conference Record of the 2004
IEEE International Symposium on Electrical Insulation, Indianapolis, IN,
pp 434-439.

Figure 12 - Significant PD activity in line termination area sensed by the

capacitors and low PD activity occurring deeper in the winding sensed by the
RTDs.

IV. CONCLUSIONS
Responses to the conclusions of the paper titled
“Investigations Into the Use of Temperature Detectors as
Stator Winding Partial Discharge Detectors” presented at the
2006 IEEE International Symposium on Electrical Insulation
were presented. Four new case studies also were discussed
showing that PD information obtained from multiple sensors
(RTDs located throughout the winding) is very valuable. Most
of the conclusions from this paper are a response to the