On-Site PD Testing: A Field Workflow from Arrival to Sign-Off

23 September 2026

1. On-site PD testing is a sequence, not a measurement

Partial discharge (PD) testing fails on site far more often through procedural error than through instrument error. An uncalibrated measuring circuit produces numbers with no scale. An unknown noise floor makes every reading ambiguous. An unidentified cable makes the whole exercise dangerous.

Each stage below exists to protect the validity of the next one. The sequence is method-independent: it applies to VLF (0.1 Hz) testing with PD measurement, to power-frequency resonant testing, and to damped AC methods, because all three share the same measuring chain and the same sources of error.

Nine stages, four hold points. If you remember only one thing: the test starts at the office, not on site.

2. Stage 1 — Pre-mobilisation and the test plan

The test plan is what makes a field day repeatable rather than improvised. It should be written before the equipment is loaded, and it should contain:

Cable data — insulation type, voltage class, length, route, and the positions of every joint and termination from as-built records
Test method and governing standard — for example IEC 60502-2 for extruded cables, or IEEE 400.3 for PD testing of shielded cable systems in a field environment
Voltage profile — the test steps, hold times at each step, and the maximum level the cable will see
Measuring circuit diagram — coupling capacitor, measuring impedance, detector, and the earth reference
Calibration method and the scale factor the circuit is expected to produce
Acceptance criteria — the apparent charge level that constitutes a failure, and the document it comes from
Safety plan, roles, and the communication link to the far end of the cable

Two items are routinely left out, and both cost more than they save. The first is joint positions: without them a discharge reading cannot be converted into a location, and the asset owner receives a number with no address. The second is the acceptance criterion, which is a contractual term rather than a field judgement — deciding it on site, after seeing the result, is how disputes begin.

The hardware side of this preparation is covered in PD Testing Equipment for Commissioning, and the measurement agenda itself in PD Testing for Commissioning.

3. Stage 2 — Isolation, discharge and earthing

Nothing is connected until the cable is dead, discharged and earthed:

Isolate and lock off the cable at both ends, then prove dead at the test end
Discharge through a rated discharge rod, and leave the earthing applied
Keep the far-end earth under the test engineer's control — the key or lock stays with the person applying the voltage

Cable capacitance stores charge, and the bigger the cable the more of it. A cable switched out of service days ago can still deliver a lethal discharge, and a cable tested at VLF can retain charge for hours after the source is shut down. Induced voltage from parallel circuits is a separate hazard: isolation removes the supply, but it does not remove induction.

This stage is a hard hold point. There is no version of the workflow in which the measuring circuit is assembled first and the earthing sorted out afterwards.

4. Stage 3 — Cable identification and route verification

Energising the wrong cable is the most serious field error in this discipline, and it is always the result of trusting a label. Positive identification before connection is not optional:

Confirm the target cable at both ends using an identification or tracing method, not the numbering on the sleeve
Verify that no load, transformer or customer installation remains connected
Walk or trace the recorded route and compare it with the physical route

Routes drift. Networks are reconfigured, sections are re-terminated, and the drawing that was accurate at handover may be two decades out of date. Where the record and the ground disagree, resolve the disagreement with an instrument before applying voltage — never by picking the more convenient option.

5. Stage 4 — Measuring circuit assembly and calibration

The measuring chain has four functional blocks, and every one of them can silently degrade the result:

Block	Function	Field failure mode
Test source	Raises the cable to the specified voltage at the specified frequency	Source-generated interference mistaken for discharge
Coupling capacitor	Bypasses the discharge current from the high-voltage node into the measuring branch	Insufficient capacitance at the test frequency — reduced sensitivity
Measuring impedance	Converts the discharge current pulse into a measurable voltage pulse	Response not matched to the cable capacitance — wrong apparent charge
PD detector	Amplifies, digitises and resolves pulses against the AC phase	Bandwidth or processing mismatched between two instruments

Calibration follows IEC 60270 and is done on the complete loop, at the cable terminal, with a PD calibrator injecting a known charge. Calibrating at the instrument input instead of the test terminal measures an idealised circuit that does not exist on site, because it excludes the coupling capacitor, the connecting leads and the cable itself.

Two rules follow from that. First, record the scale factor in the report — an apparent charge figure without a stated scale factor is not reproducible. Second, if any part of the loop changes after calibration — a lead is re-routed, a connection is re-made, the coupling capacitor is moved — calibrate again. A calibration is a property of the circuit, not of the instrument.

6. Stage 5 — Background noise survey

Before the cable is energised, establish what the site sounds like electrically. With the source running at test level but the cable not yet connected — or with the measuring circuit in place and the cable at zero — record the discharge level the circuit reports with nothing to detect.

That figure is the noise floor, and it sets the discrimination threshold for everything that follows. It tells you the smallest discharge this circuit can meaningfully report on this site. Typical contributors are adjacent high-voltage equipment, radio and switching sources, and the test source itself.

The practical consequence: on a substation site with a high noise floor, a low-level discharge may be undetectable, and the report must say so. A result of "no PD detected" means nothing unless it is accompanied by the threshold below which detection was not possible.

7. Stage 6 — Voltage conditioning and PD measurement

With the circuit calibrated and the noise floor recorded, the cable is energised and raised to the specified test level in the steps defined by the test plan. Measurement happens at each step, not only at the top.

Record on the way up and on the way down:

PDIV — the voltage at which discharge inception occurs
PDEV — the voltage at which discharge extinguishes as the voltage falls
Apparent charge at every hold step, and the PRPD pattern at each level

The gap between inception and extinction is diagnostic in itself. A defect that keeps discharging well below the voltage at which it started is a different proposition from one that stops almost immediately, and the hysteresis between the two readings is only visible if you measure in both directions. Collecting only a single figure at maximum voltage throws that information away.

The four-question framework for turning these readings into a verdict — magnitude, pattern, trend, position — is set out in How to Interpret Partial Discharge Test Results.

8. Stage 7 — Discrimination and repeat measurement

The first number the detector shows you is a candidate, not a result. Before it enters the report it has to survive three checks:

Reproducibility — a real defect produces the same reading on a repeat hold. A pulse that appears once and does not recur is usually not the cable.
Phase behaviour — discharge clusters in the PRPD plot at characteristic positions relative to the AC cycle; much site interference does not.
Correlation with the noise survey — a pulse whose amplitude and character match the pre-test background is background.

Pattern families and their interpretation are covered separately in Interpreting PD Patterns to IEC 60270; the measurement-side errors that create false positives are catalogued in Common Mistakes in On-Site PD Testing.

9. Stage 8 — Localisation when PD is found

A confirmed discharge without a location is an unfinished job. Localisation is a separate technique layer applied after detection, and it depends on data gathered in Stage 1:

Time-of-flight methods — compare arrival times of the pulse at two measurement points to place the source along the length
Reflectometry — inject a pulse and read the reflection from the impedance change at the defect
Correlation with as-built records — a calculated distance only becomes a location when it is matched against joint and termination positions

This is why the equipment and the planning have to match. A reflectometry result expressed in metres of cable is only actionable if someone knows which joint sits at that metre. It is also why on-site time expands: localisation is the stage most likely to run past the planned day, and the reasons are examined in PD Test Setup Time on Site.

10. Stage 9 — Documentation and sign-off

A PD test that cannot be reconstructed from its records has limited value, because its real worth is comparative: the next survey is only useful if this one is documented. The report should carry:

Cable identification, section length and the positions of joints and terminations
Test method, governing standard, voltage profile and hold times actually applied
Calibration record with the scale factor, and the noise floor measured on site
Apparent charge at each step, PDIV and PDEV, and PRPD plots
Discrimination notes — what was excluded as interference and why
Comparison against any previous test on the same cable
A clear conclusion against the agreed acceptance criterion, signed by the responsible engineer

The comparison line is the one that compounds in value. A single reading describes a cable today; a series of consistent readings over years describes its rate of deterioration, and that is what supports a replacement decision.

11. The four hold points

A hold point is a place where the workflow stops until a condition is verified. Four of them are non-negotiable:

Hold point	Condition to verify	If not met
Before connection	Cable positively identified, isolated at both ends, discharged and earthed	Do not connect. Nothing downstream matters until this is satisfied.
Before raising voltage	Loop calibration valid and noise floor recorded	Re-calibrate or move the circuit. Results would be uninterpretable.
On discharge detection	Reproducible, phase-correlated, and above the established noise floor	Classify as interference or investigate before reporting it as cable PD.
Before sign-off	Calibration, noise floor, patterns and voltage profile all recorded	The result is not reproducible and has no comparative value.

12. Roles on site

The workflow above assumes a minimum of three people for any test where voltage is applied to a cable:

Role	Owns
Test engineer	The measuring circuit, calibration, voltage profile, and the recorded result
Far-end attendant	Continuous watch at the remote end, earth control, and voice contact with the test position
Safety supervisor	Isolation and lock-off discipline, discharge procedure, and the barrier around the test area
Asset owner's representative	Confirmation of the acceptance criterion and acceptance of the result

Single-person testing should be treated as an exception requiring an explicit, documented justification — the far-end attendant cannot be replaced by a phone call to someone who is not standing at the cable.

FAQ

Can PD be measured on a cable that is in service? Not with this workflow. The stages above require the cable to be isolated, discharged and earthed, and the measuring circuit is referenced to a known test voltage. Online PD monitoring exists as a separate discipline with permanently installed sensors; it answers a different question and uses different acceptance logic.

How long should an on-site PD test take? It depends far more on preparation and on whether localisation is required than on the measurement itself. The breakdown of where the hours actually go is analysed in PD Test Setup Time on Site.

Is there a universal pass/fail pC threshold? No. Acceptance criteria depend on the cable type, voltage class, the governing standard and the owner's specification, and they vary between commissioning and condition assessment. What matters procedurally is that the criterion is fixed in writing before the test, not selected after seeing the result.

Does PD testing replace a DC hipot test? They answer different questions. A withstand test establishes that the insulation survives a defined voltage; PD measurement characterises the defects that are present. The choice between test methods for commissioning is compared in VLF vs DC Hipot, and PD's relationship to tan delta in Partial Discharge vs Tan Delta.

What if the noise floor is too high to measure anything useful? Say so in the report, with the measured threshold. A result reported as "no discharge detected" without the noise floor is misleading, because it implies a sensitivity the test did not have. Where possible, re-site the coupling branch or repeat at a quieter time before accepting a high floor.

Should a discharge found at commissioning stop the energisation? That is a decision for the asset owner against the agreed criterion, and it should be made with a location in hand rather than a bare magnitude. The value of completing Stage 8 before escalating is that it converts "PD was detected" into "PD was detected at this joint", which is what makes a repair decision possible.

Item ditambahkan ke keranjang Anda