Sensors 2017, 17, 2595 10 of 14

interference immunity even without any above mentioned signal processing, which contributes to a
relatively higher SNR in the actual application. The good EM interference immunity of the SiPM-
based detection is benefit from the insensitivity to magnetic field of silicon-based solid state sensor.
When comparing with SiPM, PMT is easily vulnerable to EM interference because the photoelectron
multiplication process of PMT is sensitive to the electric and magnetic field bias.

(a) (b)
Figure 9.PD outputs of PMT and SiPM sensors in identical electromagnetic (EM) noise conditions.
(a) Signals of SiPM and PMT in time domain; (b) FFT spectra of SiPM and PMT outputs. A high-
frequency electromagnetic switch in operation is used as the actual noise source and placed near the
PMT and SiPM sensors with a distance of approximately 1.0 meter.

3.3. Quantitative Relationship between the SiPM Output and the Current Pulse
A monotonic relationship between the signal response and the strength of a physical object is
the premise for building a measurement system. The intensity of the light emitted from the discharge
is positively associated with the quantity of PD charge and depends on the ratios among the
probabilistic cross sections of the excitation, ionization, recombination, and attachment during a
complete discharge process [27]. In general, the radiation spectrum and light intensity are closely
related to the specific electric field distribution and the involved interfaces of the dielectrics. Figure
10 shows the relationships between the light intensity detected by SiPM and the apparent PD
magnitude that was detected by a HFCT for all of the investigated types of PDs. For different types
of insulation defect and for the same defect, but under different polarities of applied voltage, the PDs
have the different active ranges in PD charge as well as in light intensity. For point discharge,
although the PD active range under negative half cycle of the applied voltage (tens of pC) is
significantly different from that under the positive half (hundreds of 100 pC), the light pulse
magnitude is linearly proportional to the increase of apparent PD magnitude for both negative and
positive half cycles, as shown in Figure 10a. For surface discharge, PDs are active in notably wide
ranges from several pC to hundreds of pC for both negative and positive half cycles. The relationships
between light intensity and PD charge at positive and negative half cycles remain consistent in a
relatively low active range, but present different trends as the PDs are more active, as shown in Figure
10b. This result is considered as related to the variations of ratios among the probability cross sections
of discharges involved in photon emission processes, which are caused by the transitions of
mechanisms underlying PD activities. When compared to the former types of PDs, the floating
discharge is more active once the applied voltage exceeds the PDIV, particularly in the negative half
cycles, which causes the very intensive light emission. Statistically, the light pulse magnitudes in both
half cycles linearly increase with the increase in apparent PD magnitude in the entire PD active range,
as shown in Figure 10c. In general, the magnitude of the SiPM output is positively correlated with
the apparent PD magnitude, which indicates that the SiPM-based PD detection can accurately
characterize the PD severity.
Sensors 2017, 17, 2595 11 of 14

(a) (b)

(c)
Figure 10. Light intensity detected by SiPM vs. apparent PD magnitude. (a) Point discharge; (b)
Surface discharge; and, (c) Floating discharge. The tests are performed in air condition. Over-voltages
applied on SiPM and PMT are 2.0 V and 600 V, respectively.

3.4. Stochastic PD Detection

When an alternating current (AC) electric field is applied, the discharge activities periodically
vary in terms of time lag and magnitude, based on which the PRPD analysis was proposed for PD
diagnosis and defect recognition in insulation systems under AC HV [28]. The PD data in a certain
number of applied voltage cycles are plotted on the phase axis in one voltage cycle. This phase-related
depiction of PDs provides additional information, including the physical properties and stochastic
behaviors of the discharges. Similarly, the PRPD pattern can be used to analyze stochastic light
pulses. In this case, the applied voltage peak, U, corresponding phase degree, ϕ, and light pulse
magnitude, I, are simultaneously recorded for thousands of AC cycles with optical detection to plot
the PRPD patterns. Figure 11 shows the obtained PRPDs for different types of defects at the PD initial
phases and PD intensive phases. By comparison, the PRPD results of the SiPM and current pulse
measurement are considerably consistent with each other in terms of both profile and distribution
intensity; the consistency is even higher than that between the PMT and the current from the HFCT.
This good performance of SiPM-based PD detection results from the high pulse time resolution and
high linearity with the PD magnitude of SiPM, as mentioned in Sections 3.2 and 3.3. It indicates that
the main distribution ranges of the SiPM PRPDs are consistent with those of current PRPDs,
especially for the defect models of surface discharge and floating discharge. Consequently, the
conventional PD diagnosis and defect recognition approaches can also be applied to SiPM-based PD
monitoring.
Sensors 2017, 17, 2595 12 of 14

Figure 11.Phase-resolved PD pattern (PRPD) patterns of PD light intensities detected by HFCT, PMT
and SiPM at different levels of PD activity. (a) Point discharge; (b) surface discharge; (c) floating
discharge. The tests are performed in air condition. The over-voltages applied to SiPM and PMT are
2.5 V and 800 V, respectively. The noise level of the PD current pulse detection is below 0.5 pC.

4. Conclusions
In this paper, the SiPM sensor is introduced for PD detection. The basic performances of SiPM
in PD detection were investigated by comparisons with the current pulse detection and conventional
PMT detection. Based on the experimental studies, the main conclusions are as follows.
The PDE (i.e., sensitivity) and DCR (i.e., device noise) of the SiPM sensor should be balanced by
selecting an appropriate applied over-voltage to achieve an acceptable sensitivity and a relatively
low noise level in room temperature conditions. The increase in coupling impedance can increase the
sensitivity and reduce the required sampling rate, but the pulse time resolution is also decreased.
Weibull parameters of the pulse intervals recorded by SiPM-based PD detection, including the
shape (β) and scale (α), are highly consistent with those of HFCT and PMT, which indicates that SiPM
can be applied to stochastic PD detection with a high-accuracy in PD event recognition.
Because the high quantum efficiency covers a wide response spectral range, the SiPM-based PD
detection almost has the identical capability to vacuum PMT in responding to notably weak PDs with
no additional signal amplification.
Sensors 2017, 17, 2595 13 of 14

In general, the magnitude of the SiPM output is positively correlated with the apparent PD
magnitude. In the PD active ranges of PDs, the relationship between the light relative intensity and
the apparent PD magnitude can be roughly described by a single or piecewise approximate linear
form, which implies that the SiPM-based PD detection can accurately characterize the PD severity.
The SiPM-based and HFCT-based detections have considerably consistent PRPDs in terms of
both profile and distribution intensity; this consistency is even higher than that between the PMT-
based and HFCT-based detections. This good performance results from the high pulse time
resolution and high linearity with the PD magnitude of SiPM.
Because of the additional advantages of SiPM-based PD detection, such as the small size, low
power supply, and low cost, the built-in optical PD monitoring for gas-insulated system can be put
into practical use in the near future.

Acknowledgment: The authors would like to thank the National Natural Science Foundation of China (Grant
No. 51507130 and 51777157), the National Key Research and Development Program of China (2017YFB0902701),
the China Postdoctoral Science Foundation (Grant No. 2014M560777 and 2016T90921). The authors also want to
thank the Spanish Ministry of Economy and Competitiveness (project DPI2015-71219-C2-2-R).

Author Contributions: In this work, Ming Ren conceived this study and drafted the manuscript. Jierui Zhou
conducted the related experiments with helps of Song Bo and Chongxing Zhang. Ming Dong designed the
experiments. Ricardo Albarracín designed the experiments and was also involved in drafting the manuscript.

Conflicts of Interest: The authors declare no conflict of interest.

References
1. Metwally, I.A. Status review on partial discharge measurement techniques in gas-insulated
switchgear/lines. Electr. Power Syst. Res. 2004, 69, 25–36.
2. Hu, X.; Siew, W.H.; Judd, M.D.; Peng, X. Transfer function characterization for HFCTs used in partial
discharge detection. IEEE Trans. Dielectr. Electr. Insul. 2017, 24, 1088–1096.
3. Tenbohlen, S.; Denissov, D.; Hoek, S.M.; Markalous, S.M. Partial discharge measurement in the ultra high
frequency (UHF) range. IEEE Trans. Dielectr. Electr. Insul. 2008, 15, 1544–1552.
4. PD IEC/TS 62478:2016–High-Voltage Test Techniques: Measurement of Partial Discharge by Electromagnetic and
Acoustic Methods; International Electrotechnical Commission: Geneva, Switzerland, 2016.
5. Gupta, N.; Ramu, T.S. Estimation of partial discharge parameters in GIS using acoustic emission
techniques. J. Sound Vib. 1998, 69, 25–36.
6. Albarracín, R.; Robles, G.; Martínez-Tarifa, J.M.; Ardila-Rey, J. Separation of sources in radiofrequency
measurements of partial discharges using time-power ratio maps. ISA Trans. 2015, 58, 389–397.
7. Robles, G.; Fresno, J.M.; Martínez-Tarifa, J.M. Separation of Radio-Frequency Sources and Localization of
Partial Discharges in Noisy Environments. Sensors 2015, 15, 9882–9898.
8. Albarracín, R.; Ardila-Rey, J.A.; Mas’ud, A.A. On the use of monopole antennas for determining the effect
of the enclosure of a power transformer tank in partial discharges electromagnetic propagation. Sensors
2016, 16, 148.
9. Judd, M.D.; Farish, O. High bandwidth measurement of partial discharge current pulses. In Proceedings of
the IEEE International Symposium on Electrical Insulation, Arlington, VA, USA, 7–10 June 1998; Volume
2, pp.436–439.
10. Schwarz, R.; Muhr, M. Modern technologies in optical partial discharge detection. In Proceedings of the
2007 Annual Report—Electrical Insulation and Dielectric Phenomena, Vancouver, BC, Canada, 14–17
October 2007; pp. 163–166.
11. Julio P.R.; Jose, A.G.S.; Jesus R.S. Fiber optic sensor for acoustic detection of partial discharges in oil-paper
insulated electrical systems. Sensors 2012, 12, 4793–4802.
12. Wang X.; Li, B.; Roman, H.T. Acousto-optical PD detection for transformers. IEEE Trans. Power Deliv. 2006,
21, 1068–1073.
13. Wang, X.; Li, B.; Xiao, Z.; Lee, S.H.; Roman, H.; Russo, O.L.; Chin, K.K.; Farmer, R.F. An ultra-sensitive
optical MEMS sensor for partial discharge detection. J. Micromech. Microeng. 2005, 15, 521.
14. Gamal, A.; Paul, B.; Michael, C.; Roland, H.; Johann, M.; Herbert, O.; Ken, S. Application of geiger-mode
photosensors in cherenkov detectors. Nucl. Instrum. Methods Phys. Res. 2011, 639, 107–110.