Distribution Transformer Noise Diagnosis

---

The Most Dangerous Noises in a Low-Voltage Distribution Transformer (And What Each One Means). A field engineer's guide to diagnosing distribution transformer noise — buzzing, crackling, popping, and hissing sounds explained against IEC 60076-10, IEEE C57.12.90, and BNBC 2020, with a risk-based response matrix. Distribution transformer noise diagnosis, transformer buzzing sound causes, transformer partial discharge sound, IEC 60076-10 noise limits, transformer core saturation noise, Buchholz relay gas alarm, transformer overheating sound, dangerous transformer sounds, distribution transformer maintenance Bangladesh.

The Most Dangerous Noises in a Low-Voltage Distribution Transformer

Every distribution transformer hums. That is not a fault condition — it is physics. The moment an engineer stops treating hum as background noise and starts treating it as a diagnostic signal, the transformer's acoustic signature becomes one of the cheapest, fastest condition-monitoring tools available in the field — faster than a dissolved gas analysis (DGA) report, faster than a thermal scan, and available to any technician standing next to the unit with nothing more than a sound level meter or, in a pinch, a trained ear. This article works through the acoustic signatures that separate a healthy 11/0.4 kV or 33/0.4 kV distribution transformer from one that is heading toward insulation failure, core damage, or a tank rupture — what each sound means physically, what standard sound power/pressure limits apply under IEC 60076-10:2016, and what a practicing engineer should do about it before the next fault report writes itself.

Why Transformers Make Noise At All: The Physics Behind the Hum

A distribution transformer's baseline hum is not caused by current flow directly — it is caused by magnetostriction, the dimensional change that occurs in the grain-oriented silicon steel laminations of the core as the magnetic flux alternates. Each lamination expands and contracts fractionally with every half-cycle of the applied voltage, and because both the positive and negative half-cycles produce expansion (magnetostriction is a function of flux magnitude, not polarity), the mechanical vibration occurs at twice line frequency:

f_hum = 2 × f_line

• 50 Hz systems (Bangladesh, most of South Asia): dominant hum at 100 Hz, with harmonics at 200, 300, 400 Hz
• 60 Hz systems: dominant hum at 120 Hz

A second, smaller contribution comes from winding vibration — the Lorentz force (F = B × I × L) acting on current-carrying conductors under the leakage flux field, which also cycles at twice line frequency under balanced sinusoidal load but becomes broadband and load-dependent under harmonic-rich loads (VFDs, rectifier loads, LED street-lighting drivers — an increasingly common load type on Bangladeshi LT distribution feeders).

A third source, present only in oil-filled units, is cooling system noise — fan motors, oil pump cavitation, and radiator resonance — which is typically higher frequency (200 Hz–2 kHz) and load-independent, since fans cycle on thermostatic control rather than load current directly.

The engineering distinction that matters: normal hum is tonal, steady, and load-correlated within a narrow band. Dangerous noise is broadband, irregular, or load-uncorrelated. Everything in this article follows from that single distinction.

IEC 60076-10:2016 — The Baseline Every Field Reading Should Be Checked Against

IEC 60076-10:2016 (Power transformers — Part 10: Determination of sound levels) is the current edition, superseding the 2001 first edition. Engineers referencing older transformer noise literature should flag this — the 2016 edition introduced a formal definition of "distribution type transformer," added a mandatory sound level measurement specification clause, and — critically for anyone benchmarking factory acceptance test (FAT) data against site readings — changed the standard measurement distance from 0.3 m to 1 m for transformers other than distribution type. Distribution transformers retain the closer measurement geometry, so FAT reports and site readings on distribution units remain directly comparable; that is not true when comparing a distribution transformer's site reading against a power transformer's factory sound level guarantee measured to the 2016 geometry.

Table 1: Typical Rated Sound Power Levels for ONAN Distribution Transformers (IEC 60076-10 reference guidance)

Rated Power (kVA)	Typical Sound Power Level, L_WA (dB)	Typical Sound Pressure at 1 m, L_pA (dB(A))	Notes
≤ 100	45–50	35–40	Pole-mounted / pad-mounted, ONAN
100–315	50–58	40–48	Most common LT distribution range in Bangladesh
315–630	58–63	48–53	Substation-mounted, ONAN
630–1000	63–67	53–57	Industrial/commercial feeders
1000–1600	67–70	57–60	Larger distribution substations
>1600	70+	60+	Approaching power-transformer noise class

Sound power (L_W) and sound pressure (L_p) are related through the measurement surface area A: L_W = L_p + 10·log₁₀(A/A₀), where A₀ = 1 m². For a typical distribution transformer enclosure with a 1 m measurement surface area of roughly 8–12 m², expect L_W to read 9–11 dB higher than the L_pA reading taken at any single point.

The number that matters operationally is not the absolute dB(A) value — ambient substation noise, nearby traffic, and reflective walls all distort a single-point reading taken without a calibrated Type 1 sound level meter per IEC 61672. What matters is the delta: a transformer's noise level relative to its own historical baseline and relative to its rated load.

Rule of thumb for load-correlated noise growth:

• Normal: +3 to +5 dB rise from no-load to full-load
• Investigate: +5 to +7 dB rise — check tap position, loading records, and cooling fan operation
• Alarm: >7 dB rise, or any rise that is not proportional to load — strongly suggests loose windings, shifted core clamping, or incipient core-to-frame short

The Ten Diagnostic Noise Signatures (Ranked by Risk)

This is the working field reference. Each entry gives the acoustic character, the underlying physical mechanism, and the recommended engineering response.

1. Steady, Low, Tonal Hum (100/120 Hz) — Normal, No Action

The baseline magnetostrictive hum described above. Consistent in pitch and loudness across a shift, correlates weakly with load (a few dB rise from no-load to full-load), and does not change character day to day. This is the reference signature every other entry on this list is compared against.

2. Harsh, Higher-Pitched Buzz Layered Over the Hum — Loose Core Clamping (Moderate Risk)

When core clamping bolts loosen — through thermal cycling, transport vibration, or seismic/impact events — individual laminations begin vibrating independently rather than as a clamped mass. The result is a sharper, more metallic buzz superimposed on the normal 100/120 Hz tone, often with audible content in the 300–800 Hz range. This is mechanically progressive: loose laminations generate localized friction heating at contact points, which degrades interlaminar insulation (typically a thin varnish or oxide coating) and increases eddy current losses locally, which in turn increases local hotspot temperature — a self-reinforcing failure mode. Left unaddressed, this progresses toward core-to-core-bolt insulation breakdown and stray flux heating of structural steel.

Action: Schedule outage for core clamp torque check against manufacturer specification; thermal imaging of the tank/enclosure surface during the next load cycle to look for hotspots correlating with the buzz location.

3. Growling, Deep-Toned Hum That Rises With Load — Overload or Voltage Excursion (Moderate Risk)

A hum that deepens and intensifies proportionally with load current reflects increased Lorentz-force winding vibration and, at sustained overload, incipient core saturation as flux density approaches the knee of the B-H curve. This is functionally different from loose-core buzzing — it is load-correlated rather than constant, and it disappears (or drops back to baseline) when load is shed.

Action: Cross-check against nameplate rating, tap position, and IEC 60076-7 loading guide limits for the transformer's thermal class and ambient conditions. In Bangladesh's pre-monsoon and summer ambient temperatures (routinely 35–40°C), the effective continuous overload margin is materially lower than the nameplate rating assumes at 20°C ambient — a common source of "unexplained" noise complaints on urban LT distribution transformers during peak cooling-load months (March–May).

4. High-Pitched Whine or Screech, Independent of Load — Core Saturation / Overexcitation (High Risk)

A whine that persists or worsens at no-load, or that does not track with current, points to excess flux density rather than excess current — i.e., overvoltage, not overload. Since flux density is proportional to applied voltage and inversely proportional to frequency (Φ ∝ V/f), a sustained voltage rise of even 5–10% above nameplate — common on lightly loaded rural 11 kV feeders during off-peak hours, or where an upstream tap changer is misconfigured — pushes the core toward saturation, sharply intensifying magnetostriction and generating strong harmonic content (300 Hz, 500 Hz) alongside the fundamental hum.

Action: Verify incoming voltage against nameplate and tap setting immediately; sustained overexcitation accelerates core loss, insulation thermal aging (per the Arrhenius/Montsinger 6–8°C-per-halved-life rule), and — in extreme cases — leads to core-to-tank stray flux heating.

5. Rattling, Metallic Clicking, or Chattering — Loose Fittings, Fasteners, or Tap Changer Contacts (Moderate to High Risk)

Distinct from the tonal buzz of loose core laminations, this is a discrete, often intermittent mechanical rattle — loose enclosure panels, radiator fin resonance, loose bushings, or (most significantly) a deteriorating off-circuit or on-load tap changer (OLTC) mechanism. Chattering that occurs specifically during or immediately after a tap change operation, or that has a clearly cyclical/mechanical (non-tonal) character, warrants tap changer contact inspection — worn or misaligned OLTC contacts arc on transition, and that arcing is both a noise source and a DGA red flag (elevated acetylene, C₂H₂).

Action: Torque-check all accessible external fasteners first (cheapest, fastest elimination step); if the noise persists and correlates with tap operations, escalate to OLTC contact inspection and DGA sampling.

6. Crackling or Sizzling, Independent of Load — Partial Discharge (High Risk — Do Not Defer)

Crackling or a persistent, faint sizzling sound — often described by field technicians as sounding like frying oil or static — is one of the most serious acoustic signatures a transformer can produce. It indicates partial discharge (PD): localized dielectric breakdown in a small void, along a creepage surface, or at a sharp-edged conductor, occurring well below the insulation's full breakdown voltage. PD is progressive and non-self-healing in solid/oil-paper insulation systems — each discharge event carbonizes a microscopic path, and the discharge inception voltage of that path drops with each repetition, meaning the fault accelerates itself over time until full insulation breakdown occurs.

Action: This is not a "monitor and reassess" finding. Schedule an insulation resistance (megger) test and, where available, an ultrasonic PD survey or acoustic PD sensor placement to localize the discharge site. For oil-filled units, pull a DGA sample immediately — PD produces a characteristic gas signature dominated by hydrogen (H₂) and methane (CH₄) per IEC 60599 interpretation, distinguishable from thermal fault gas ratios.

7. Sharp Popping, Cracking, or Banging (Discrete Events) — Internal Arcing (Severe Risk — Immediate De-energization)

A single loud crack, pop, or bang — as opposed to continuous crackling — is consistent with a discrete arcing event: a flashover between windings, a bushing failure initiating, or a tap changer selector switch arcing under load beyond its design transition current. This differs from PD in both magnitude and urgency: PD is a slow-burning insulation degradation process; a discrete pop or bang is evidence that dielectric breakdown has already occurred at least once, at a level sufficient to generate an audible pressure wave — meaning fault energy, however momentary, has already been released inside the tank.

Action: Treat as a pre-trip condition. If the transformer is protected by a Buchholz relay (oil-filled, conservator type), check immediately for an alarm or trip signal — internal arcing generates gas rapidly enough to actuate the Buchholz float within minutes to hours, well ahead of a scheduled DGA sample. De-energize for inspection rather than waiting for the next scheduled outage window; a second, larger event is the realistic next step in this failure progression, and it typically does not wait for a convenient time.

8. Hissing or Continuous Air/Gas Escape Sound — Pressure Relief Device Operation or Gasket Leak (Severe Risk)

A continuous hissing sound, particularly if accompanied by a visible oil mist or a drop in oil level on the sight glass, indicates the pressure relief device (PRD) has actuated or is actuating — meaning internal tank pressure has risen enough to require venting. This is a downstream symptom of a fault already generating gas rapidly inside the tank (arcing, severe overload, or a rapidly developing hot-spot). A hissing sound from a gasket or valve without pressure relief involvement is a lower-urgency oil leak, but the two must be distinguished quickly, since one is a maintenance item and the other is an active fault indicator.

Action: Confirm PRD status and oil level before assuming a simple leak. If the PRD has actuated, treat identically to Signature 7 — de-energize and inspect; do not re-energize until DGA and insulation testing confirm the internal condition.

9. Sloshing or Gurgling — Low Oil Level, Moisture Ingress, or Gas Accumulation (Moderate Risk)

In conservator-type oil-filled transformers, a sloshing or gurgling sound during load or temperature transients can indicate low oil level (allowing free air space to move audibly), moisture-contaminated oil (which behaves differently under thermal cycling than dry oil), or gas bubbles migrating through the main tank toward the Buchholz relay — the latter being a precursor to a Buchholz alarm rather than a distinct fault in itself.

Action: Check oil level against the sight glass/gauge and dielectric strength/moisture content (Karl Fischer test) at the next scheduled sampling. Sloshing that develops suddenly, rather than being a long-standing characteristic of the unit, should trigger an out-of-cycle oil level and DGA check.

10. Complete Silence From a Previously Audible Unit — Do Not Assume "Fixed" (Context-Dependent Risk)

Worth flagging explicitly because field technicians sometimes read this as good news: a transformer that has gone unexpectedly quiet after a period of abnormal noise has not necessarily self-corrected. In rare cases this reflects a winding that has gone open-circuit (broken conductor, failed connection) rather than a resolved fault — the "noise" was arcing or vibration associated with a partial fault path that has now fully failed open. Any unexplained noise change, in either direction, on a unit with prior abnormal readings should trigger a load-current and voltage check across all phases before being logged as resolved.

Diagnostic Decision Matrix

Table 2: Noise Signature → Likely Cause → Risk Level → Recommended Response

Noise Signature	Likely Cause	Risk Level	Response Timeframe
Steady 100/120 Hz tonal hum	Normal magnetostriction	None	Routine monitoring
Harsh buzz over baseline hum	Loose core clamping/laminations	Moderate	Next scheduled outage
Deep growl rising with load	Overload / high ambient derating	Moderate	Load survey within days
High-pitched whine, load-independent	Overvoltage / core saturation	High	Voltage check same day
Rattling/chattering, tap-correlated	Loose fittings or OLTC contact wear	Moderate–High	Inspect within 1–2 weeks
Crackling/sizzling, continuous	Partial discharge	High	Megger + DGA within 48 hours
Sharp pop/bang, discrete	Internal arcing	Severe	Immediate de-energization
Hissing with pressure/oil signs	PRD actuation / active fault gassing	Severe	Immediate de-energization
Sloshing/gurgling	Low oil, moisture, or migrating gas	Moderate	Oil level + DGA within days
Sudden silence after prior noise	Possible open-circuit winding failure	Context-dependent	Phase current/voltage check same day

Field Diagnostic Procedure for Engineers and Technicians

1. Establish a baseline. Record sound character, approximate dB(A), load current, and ambient conditions during a known-healthy period for every LT distribution transformer under your responsibility — this single step makes every future noise complaint diagnosable rather than anecdotal.
2. Correlate with load, not just presence. Take readings at no-load and at typical peak load. A noise that scales proportionally with current is mechanically different from one that does not.
3. Localize before you diagnose. Use a mechanic's stethoscope or a listening rod against the tank at multiple points — core noise, winding noise, and cooling-fan noise localize differently across the tank surface.
4. Cross-check with thermal imaging. Any noise signature suspected to involve loose laminations, loose connections, or overexcitation should be paired with an IR scan during the same load condition; hotspots corroborate acoustic findings and help prioritize outage scheduling.
5. Escalate oil-filled units to DGA on any Signature 6–9 finding. Acoustic evidence of PD or arcing should never be the only data point driving a de-energization decision where DGA sampling is available — but it should never be ignored while waiting for lab results either.
6. Document tap changer correlation explicitly. Because OLTC-related noise is one of the most common "surprise" findings in urban Bangladeshi distribution networks (frequent tap operations to manage voltage on long LT feeders), always log tap position alongside noise readings.

Risk and Safety Instructions

• Never open an enclosure or approach an energized tank to "listen closer" without proper PPE and a maintained safe approach distance — arc flash risk applies even during routine noise investigation, since the investigation itself is often triggered by a developing fault.
• Treat Signatures 7 and 8 (discrete arcing sounds, PRD hissing) as an immediate trip condition, not a scheduling item — internal arc energy in a sealed tank can escalate to tank rupture or fire within a short window once fault gas generation begins.
• Do not silence or reset a Buchholz alarm without investigation. An alarm that coincides with a reported noise change should be treated as corroborating evidence, not a nuisance trip, until DGA confirms otherwise.
• Lock-out/tag-out (LOTO) applies in full before any physical inspection following a de-energization triggered by noise — do not rely on the noise having stopped as confirmation the unit is safe to approach.
• Coordinate with protection settings. Where differential or restricted earth fault (REF) protection is present, a noise-driven suspicion of internal arcing should prompt a review of relay event records for the same time window — corroborating electrical evidence strengthens (or rules out) the acoustic finding before committing to an outage.
• Follow IEC 60599 for gas-ratio interpretation once DGA results are available, and cross-reference findings against IEEE C57.104 for mineral-oil-filled equipment where the installation follows North American practice.

Conclusion: Treat the Hum as Telemetry, Not Ambience

A distribution transformer's noise is not incidental to its operation — it is a continuous, free, real-time output of the same electromagnetic and mechanical processes that DGA, thermal imaging, and insulation testing measure indirectly and periodically. The engineering discipline this article argues for is simple: stop treating "the transformer is making noise" as a binary complaint and start treating it as a signal with frequency content, load correlation, and a risk-ranked cause list attached. For utilities and industrial facilities operating large fleets of pole- and pad-mounted LT distribution transformers across Bangladesh's urban and semi-urban feeders — where scheduled DGA sampling intervals are often measured in months, not weeks — acoustic monitoring during routine feeder patrols is one of the highest-value, lowest-cost condition-monitoring practices available, provided the field team knows which sounds are physics and which sounds are a fault already in progress.

---

Have you diagnosed a transformer fault from its noise signature before confirming it with DGA or thermal imaging? Share the case in the comments — WAZIPOINT is building out a technical reference series on condition-based maintenance for LT/MV distribution equipment in Bangladesh.