Digital Partial Discharge Tester: Noise Rejection, On-Site Challenges, and Best Practices
On-site partial discharge testing differs fundamentally from laboratory measurements. Substation environments contain corona from energized hardware, power line communication signals, radio broadcasts, and switching transients—all of which can mask low-magnitude PD pulses. A digital partial discharge tester with advanced noise rejection capabilities is essential for obtaining reliable, repeatable results. This article examines common noise sources, rejection techniques, and field-proven best practices for successful on-site PD testing.
Common Noise Sources in On-Site PD Testing
Understanding the origin of interference is the first step toward effective rejection:
Continuous periodic noise: Radio broadcasts (0.5–30 MHz), power line carrier communications (40–500 kHz), and switching power supply harmonics. These appear as narrow-band spikes in the frequency domain.
Stochastic noise: Random pulses from corona on nearby live equipment, poor grounding contacts, or motor brush arcing. These mimic PD but lack phase correlation with the test voltage.
Pulse-train interference: Thyristor rectifiers, variable frequency drives, and welding equipment produce repetitive pulse bursts at predictable intervals, often mistaken for PD activity.
Impulse noise: Lightning strikes, breaker operations, or load switching generate high-energy transients that can saturate the digital partial discharge tester's front end.
Hardware Noise Rejection Techniques
A quality digital partial discharge tester implements multiple hardware-level noise mitigation strategies:
| Technique | Implementation | Effectiveness |
|---|---|---|
| Band-pass filtering | Selectable analog filters (e.g., 100 kHz–1 MHz or 1–30 MHz) | Removes out-of-band noise; essential but not sufficient alone |
| Differential input | Two probes measuring opposite polarity; common-mode noise cancels | Excellent for rejecting ground loop interference (40–60 dB rejection) |
| Galvanic isolation | Fiber optic links between sensor and digitizer | Complete elimination of conducted noise from test object to instrument |
| Adaptive threshold triggering | Dynamic trigger level based on background noise floor | Prevents false triggering during high-noise periods |
| Frequency hopping | Automatically select quiet frequency band by scanning spectrum | Highly effective for avoiding radio interference and carrier signals |
Software-Based Noise Separation and Filtering
Modern digital partial discharge testers employ sophisticated algorithms to separate genuine PD from interference after data acquisition:
Time-of-flight discrimination: Using two sensors at known distances, the tester measures pulse arrival times. External noise typically arrives from different directions than internal PD, allowing spatial separation.
Pulse shape analysis: Genuine PD pulses have characteristic rise times (1–100 ns) and decay patterns (exponential with time constant determined by circuit impedance). Noise pulses often have faster or slower edges, enabling classification.
Phase-resolved pattern recognition: PD activity follows the AC test voltage (0° and 180° regions). Noise appears randomly across the phase angle. Modern testers apply fuzzy logic or neural networks to identify PD-synchronous pulses.
Multi-channel correlation: PD pulses appear correlated across channels with predictable delays. Noise is often uncorrelated. Cross-correlation algorithms extract PD even when signal-to-noise ratio is below 0 dB.
Practical Field Best Practices
Even the best digital partial discharge tester requires proper on-site procedures for optimal results:
Pre-Test Survey and Planning
Identify all nearby energized equipment that could produce corona or arc discharges. Plan to de-energize them if possible, or maintain minimum safe distance.
Use a spectrum analyzer or the tester's built-in spectral display to map the noise environment before connecting to the test object. Select the quietest frequency band for measurement.
Check all grounding connections. Loose or corroded grounds are a primary source of intermittent noise. Measure ground impedance with a low-ohmmeter (<0.5 Ω recommended).
Sensor Placement and Connection
Install HFCT sensors directly on ground leads as close as possible to the test object. Every additional meter of lead adds noise pickup.
Use double-shielded coaxial cable (RG-223 or similar) for all sensor connections. Avoid running signal cables parallel to power cables.
For UHF PD testing, ensure couplers are fully seated and inspection ports are clean. Dielectric grease on threads improves contact.
When using acoustic sensors, apply couplant gel and press firmly to eliminate air gaps that reduce sensitivity.
Baseline Noise Measurement
Before energizing the test object, perform a background noise measurement with the digital partial discharge tester. Record the noise floor in pC or mV. If noise exceeds 10 pC (or 10 mV equivalent), attempt to reduce it. If reduction is impossible, document the noise level and set the PD alarm threshold at least 3 dB above the noise floor.
Step-Wise Voltage Ramp Testing
For offline tests (e.g., cable or motor acceptance), apply voltage in increments of 10–20% of rated voltage. At each step, hold for 2–5 minutes. Genuine PD appears gradually near inception voltage. Noise often appears abruptly at specific voltage thresholds (e.g., when a distant object begins corona).
Case Example: Overcoming Severe Substation Noise
During a recent GIS PD survey, a digital partial discharge tester initially recorded 150 pC of apparent PD. However, the pattern was phase-incoherent and pulse shapes were irregular. The operator switched to differential HFCT mode, which reduced the signal to 12 pC. Then time-of-flight analysis using two sensors 3 meters apart showed that 80% of remaining pulses originated from outside the GIS. After shielding a nearby lightning arrester that exhibited corona, the measured PD dropped to 3 pC—within acceptable limits. Without digital noise rejection, the GIS would have been incorrectly flagged for urgent maintenance.
Calibration of Noise Rejection Performance
When evaluating digital partial discharge testers, request a noise rejection demonstration using a calibrated PD source combined with simulated interference (e.g., a spark gap or signal generator). The tester should correctly report PD magnitude within ±20% of the calibrated value while rejecting interference at least 20 dB above the PD signal. Some manufacturers publish noise rejection specifications (e.g., 60 dB common-mode rejection ratio). Verify these claims with an independent test.
Limitations of Digital Noise Rejection
No technique is perfect. Operators must understand these limitations:
When noise occupies the same frequency band, has similar pulse shapes, and is phase-correlated, digital separation becomes impossible. In such cases, physically removing or shielding the noise source is the only solution.
Aggressive filtering can remove genuine PD if the filter passband does not match the PD pulse spectrum. Always validate filter settings with a known PD source.
Multi-channel correlation requires at least two sensors. This increases setup time and cost but is often worth the improved noise immunity.
Training and Competency Requirements
Using a digital partial discharge tester effectively requires more than reading the manual. Operators should understand PD physics, noise mechanisms, and the limitations of each rejection technique. Many manufacturers offer certified training courses. For critical assets, consider requiring operator certification (e.g., ISO 18436 for vibration analysis, adapted for PD). A trained operator recognizes when noise rejection is working correctly and when it is masking real defects.
Digital partial discharge testers have made on-site PD measurements routine, but only when operators apply the noise rejection techniques and best practices described here. The combination of hardware filtering, software discrimination, and careful field procedures yields reliable, actionable data that supports condition-based maintenance and prevents costly equipment failures. For high-voltage asset owners, investing in operator training alongside the digital PD tester delivers the highest return.
