Electrical Transformer Design & Construction

1. Fundamentals of Transformer Operation

A transformer transfers electrical energy between two or more circuits through electromagnetic induction, with no moving parts and no electrical connection between windings (autotransformers excepted). An alternating voltage applied to the primary winding drives a magnetising current that establishes an alternating flux Φ in the laminated steel core. This mutual flux links the secondary winding and induces a voltage in it proportional to its number of turns.

1.1 The EMF equation

For sinusoidal excitation at frequency f, with peak core flux Φ_m (Wb), the RMS voltage induced in a winding of N turns is:

where B_m is the peak flux density (T) and A_c the net core cross-section (m²). This single equation drives the entire design: it ties the voltage rating to the product of turns, core area and flux density. The designer trades copper (turns) against iron (core area) — the classic volts-per-turn decision.

1.2 Ideal vs. real transformer — the equivalent circuit

An ideal transformer has V₁/V₂ = N₁/N₂ = I₂/I₁ and no losses. A real transformer adds four imperfections, each with a physical origin in the construction:

1.3 Key rating definitions

• Rated power (kVA/MVA): apparent power the transformer can deliver continuously at rated voltage and frequency without exceeding temperature-rise limits.
• Impedance voltage (u_k%): primary voltage, in % of rated, needed to circulate rated current with the secondary short-circuited. Typical: 4–6% (distribution), 8–20% (power). It governs fault current, voltage regulation and parallel operation.
• Vector group (e.g. Dyn11): the winding connections (D = delta, y = star, n = neutral brought out) and phase displacement in clock notation (11 = 330° lead of LV over HV).
• Tapping range: turns adjustment on (usually) the HV winding to regulate output voltage, e.g. ±2 × 2.5% off-circuit, or ±10% in 17 steps with an on-load tap-changer (OLTC).

1.4 The B–H curve and what it costs you

Choosing B_m is a three-way bargain: higher B_m → smaller core, less copper (shorter mean turn), cheaper — but more no-load loss and noise, less V/f headroom. The knee at ≈1.9 T is a hard ceiling; the 0.1–0.3 T below it is the designer's working room.

1.5 The per-unit system — the designer's native language

All transformer engineering is done in per-unit (pu): every quantity is expressed as a fraction of its own rated base. Base impedance Z_b = V_b²/S_b. The same physical transformer then has the same pu impedance seen from either side — the turns ratio disappears.

• Fault current through the transformer (stiff source): I_sc = 1/u_k pu = 20 × rated for u_k = 5%.
• Two transformers in parallel share load as S₁ⁿ/u_k1 : S₂ⁿ/u_k2 (see §13.3) — a one-line pu calculation.
• Converting between MVA bases: Z_pu,new = Z_pu,old × (S_new/S_old).

1.6 Measuring the equivalent circuit: open-circuit & short-circuit tests

The elegance: each test isolates one branch of the equivalent circuit because the other branch carries almost nothing. These are exactly the routine tests of §11 — the equivalent circuit is not an abstraction, it is what the test bay measures on every unit shipped.

1.7 The loaded transformer — phasors and the third winding

Three-winding transformers (HV + LV + tertiary) extend the model to a star of three impedances Z_H, Z_L, Z_T meeting at a fictitious common node, derived from the three pairwise short-circuit tests: Z_H = ½(Z_HL + Z_HT − Z_LT), and cyclically. A practical quirk worth knowing: one branch (often the middle winding's) can legitimately come out negative — it is a mathematical element, not a physical winding impedance. Power flowing H→L, H→T and L→T each see the sum of two branches, which is why tertiary loading affects the main-path voltage drop.

2. Classification: Power, Distribution, Dry-Type & Cast Resin

Transformers are classified by function in the network, insulation/cooling medium, and construction. The four families covered in this tutorial share identical electromagnetic theory but differ sharply in materials, manufacturing and accessories.

2.1 Choosing between oil and dry insulation

2.2 Oil preservation systems compared

The preservation system decides the unit's moisture destiny (§5, §16): a free-breathing 30-year-old transformer typically carries 2–3% moisture in paper; a healthy hermetic unit of the same age stays under 1% — half a lifetime of difference from one constructional choice.

3. Magnetic Core — Design & Construction

The core provides a low-reluctance path for the mutual flux. Its design fixes the no-load loss, no-load current and sound level for the life of the transformer — none of these can be improved after manufacture. Core cost is typically 20–25% of total material cost.

3.1 Core steel

Grain-oriented silicon steel (~3% Si) has its magnetic domains aligned with the rolling direction; flux must follow that direction, which dictates the mitred joint geometry below. Laminations are coated with an inorganic insulation (C5/C6 “carlite”) ~2–4 µm thick to suppress inter-laminar eddy currents.

3.2 Core-form vs shell-form

Three-phase core-form transformers normally use a three-limb core; five-limb cores (two unwound return limbs) reduce yoke height for rail/road transport clearance on large power transformers and provide a return path for zero-sequence flux.

3.3 Stepped (cruciform) cross-section

3.4 Mitred step-lap joints

Where limb meets yoke, flux must turn 90°. With grain-oriented steel, laminations are cut at 45° (mitred) so flux stays in the rolling direction. In a step-lap joint, successive lamination layers shift the joint position in 5–7 steps, so flux passing the air gap of one layer is carried by its neighbours:

3.5 Core building, clamping and earthing

🧱 Open the interactive 3D: build the core yourself — stack the step-lap top yoke, inspect the wooden clamps and insulation, then explore the complete core & coil assembly →

1. Slitting & cropping: coil steel is slit to the packet widths, then cut to length with 45° mitres on automatic cut-to-length lines (e.g. Georg, AEM); burr ≤ 20 µm to avoid inter-laminar shorts.
2. Stacking: laminations are stacked horizontally on a building frame, usually 2 laminations per layer, with step-lap offsets; modern lines auto-stack.
3. Clamping: yokes are clamped by steel frames with insulated tie-rods or bands (no bolts through modern limb laminations — banded with resin-impregnated glass tape to avoid holes that concentrate flux).
4. Raising: the core is raised upright; top yoke is removed for winding assembly and re-laid ("re-yoking") after the windings are lowered on.
5. Earthing: the core and clamping frame must each be earthed at exactly one point (usually via a copper strap to a tank-cover bushing or the clamp). One point only — two would create a closed loop for circulating current. A floating core partially discharges and shows up as gas in DGA.

3.6 Core noise — where the hum comes from

CRGO laminations physically stretch and shrink ~10⁻⁶ of their length twice per cycle (magnetostriction), so a 50 Hz transformer hums at 100 Hz plus harmonics (200, 300, 400 Hz — the harmonics carry most of the annoyance). Load current adds a smaller winding-force component at 100 Hz. Design levers, in order of effectiveness:

• Lower B_m — roughly −5 dB per −0.1 T, but directly trades against core size/cost.
• Better steel — Hi-B and domain-refined grades have inherently lower magnetostriction; step-lap joints alone are worth ~3–4 dB over conventional lap.
• Decoupling — anti-vibration pads between core feet and tank, flexible yoke-clamp pads, sound panels or double tank walls for city substations.
• Verification — sound power L_WA measured per IEC 60076-10 (sound-pressure walk-around or sound-intensity method); typical guarantees: 1000 kVA ONAN ≈ 58–62 dB(A), 100 MVA ONAF ≈ 75–85 dB(A). Fans add their own broadband noise — at low core levels the cooling, not the core, dominates.

3.7 The building factor — why a real core loses more than the steel datasheet

Epstein-frame datasheet loss (W/kg) is measured on perfect strips magnetised along the rolling direction. A built core adds: flux turning 90° at joints (local off-grain magnetisation), normal-flux eddy currents where laminations cross, burr-induced inter-laminar shorts, stacking stresses, and slitting-edge damage. The ratio of real to datasheet loss is the building factor:

A guaranteed P₀ missed at test means penalty or rejection — which is why core-cutting line condition (burr!) and stacking discipline are checked in any factory audit.

3.8 Amorphous and wound-core construction

Amorphous ribbon (Fe₈₀B₁₁Si₉, ~25 µm, no crystal structure) cuts no-load loss ~65–70% versus CRGO — but it saturates at ~1.56 T (design B_m ≈ 1.3–1.4 T), is brittle, and has a poor stacking factor (~0.86), so the core is bigger and the copper path longer. It is used almost exclusively in wound cores (ribbon wound into rectangular loops, then cut and re-lapped) for distribution transformers up to ~2500 kVA, where energy-efficiency regulation or loss capitalisation justifies the premium. Wound-core construction with CRGO ("unicore") is likewise common for small distribution units: lower building factor and cheaper automation, at the cost of repairability.

3.9 Core quality control — factory and field

4. Windings — Design & Construction

The windings carry load current and must survive three simultaneous demands: dielectric (rated + impulse voltage), thermal (I²R heating under load), and mechanical (electromagnetic forces during through-faults, proportional to I²). Winding design selects conductor type, winding type, insulation grading and cooling ducts to satisfy all three at minimum cost.

4.1 Conductors

• Round enamelled wire: small distribution LV/HV windings, up to ~3–4 mm dia.
• Rectangular paper-covered strip (DPC): the workhorse — copper (or aluminium) strip wrapped with 2+ layers of kraft paper; thermally upgraded paper for 65 K rise designs.
• Foil/sheet: full-winding-height copper or aluminium foil for high-current LV windings — inherently balanced axial current distribution, excellent short-circuit strength.
• Continuously transposed conductor (CTC): 5–85 enamelled strands transposed continuously and paper-wrapped as one cable. Cuts eddy loss in large windings and improves space factor; epoxy-bonded CTC adds mechanical rigidity for short-circuit withstand.

4.2 Winding types

4.3 Concentric arrangement and leakage duct

In core-form construction the LV winding sits innermost (cheapest to insulate from the earthed core), the HV outermost, separated by the main duct — pressboard cylinders and axial oil ducts. Tapping turns sit either within the HV or in a separate tap winding. The duct dimensions set the leakage reactance:

A taller winding or thinner duct lowers impedance; a fatter duct raises it. Because impedance is a guaranteed nameplate value (±7.5% or ±10% tolerance per IEC 60076-1), the designer fixes the duct geometry early.

4.4 Transposition

Where a turn comprises several parallel strips, strips at different radial positions link different leakage flux and would share current unequally, causing circulating-current loss. Strips are therefore transposed — rotated through every radial position along the winding length (standard transpositions at 25%, 50%, 75% of height for helical windings; continuous in CTC).

4.5 Short-circuit forces

4.6 Winding eddy loss — the h² problem

Leakage flux passing through a conductor induces eddy currents in it. For a strip of height h (the dimension facing the radial leakage flux) the local eddy loss relative to its DC loss is approximately:

• Halving strand height cuts its eddy loss ~16× — this, not current capacity, is why large windings use many thin strands (CTC) instead of one fat bar.
• Leakage flux is strongest and radially bent at the winding ends → the end discs run hottest. Designers reduce strand height there, add extra ducts, or shield with flux collectors.
• Eddy + circulating losses are why measured load loss exceeds calculated I²R by 8–20%; the stray balance heats clamps and tank (managed with magnetic shunts on the tank wall of large units).

4.7 How a winding is actually made

1. Mandrel & cylinder: the pressboard winding cylinder is mounted on a horizontal (distribution) or vertical (power) winding machine; hard paper end caps and key spacers placed.
2. Winding: conductor pays off tensioned reels through paper-lapping heads if not pre-covered; turns counted electronically; brazed joints (silver alloy) only where unavoidable, each insulated and staggered.
3. Transpositions made by hand at calculated positions (helical) — the craft step where strand order is rotated without nicking insulation.
4. Taps & exits: tap leads brazed and crimped, formed into cleated drops; exits reinforced with extra paper and snouts.
5. Sizing & stabilisation: winding pressed axially in a hydraulic press, dried (often pre-dried before assembly), pressed again to final height under the calculated clamping force — so it cannot loosen in service (a loose winding fails its first short circuit).
6. Quality gates: turn count, resistance per phase, dimensional checks, duct gauge sweeps, X-ray/visual on brazes — recorded before the winding is released to core–coil assembly.

4.8 Thermal design inside the winding — ducts and gradients

• Duct sizes: radial (horizontal) ducts between discs 4–8 mm; axial ducts 6–12 mm. Below ~3 mm a duct stops working — boundary layers meet and the oil stagnates.
• Oil velocity: natural circulation gives ~0.05–0.3 m/s in winding ducts; directed (OD) cooling 0.5–1 m/s. Above ~1 m/s static electrification becomes a real risk (charge separation at the oil–pressboard interface has destroyed large transformers — flow rate is capped, not maximised).
• Gradient budget: average winding-to-oil gradient g ≈ 8–15 K for ON cooling. The designer reaches the target by trading conductor current density, paper thickness (every 0.1 mm of paper adds ~1–2 K), duct count, and disc grouping (oil-guide washers force a zig-zag flow path through disc winding sections, multiplying duct utilisation).
• Where windings actually fail thermally: not at the average — at the ends (radial leakage flux → extra eddy loss, §4.6) and under the top oil where ducts exhaust warmest. Hence hotspot factor H = 1.1–1.3 and fibre-optic probes placed in the top discs.

5. Insulation System

The insulation system is what actually ages and dies in a transformer — “the life of the transformer is the life of its insulation.” Oil-immersed designs use an oil–cellulose composite; dry types use polymer films, Nomex® aramid and epoxy.

5.1 Oil–cellulose system (liquid-immersed)

• Mineral insulating oil (IEC 60296): dielectric strength > 70 kV/2.5 mm treated; also the coolant and a diagnostic window (DGA). Alternatives: synthetic ester (IEC 61099), natural ester (IEC 62770), silicone (IEC 60836) — chosen for fire safety or biodegradability.
• Kraft paper: conductor covering, 0.05–0.075 mm per wrap. Thermally upgraded (amine-stabilised) paper permits 65 K average winding rise instead of 55 K; higher-thermal-class conductor papers (a current Weidmann development focus) push hybrid designs further still.
• Pressboard (transformerboard): cylinders, angle rings, snouts, strips and blocks — Weidmann transformerboard is the industry reference, supplied either as sheets/components or as complete engineered insulation kits per winding design. Moulded barriers shape the oil gaps into many thin, sub-divided ducts — oil gaps withstand far more stress when subdivided by solid barriers.
• Wood/densified laminates: cleats, lead supports, winding support rings and core-to-coil packers, machined from beech-laminate/densified wood boards (specialist suppliers include Rancan and Lamtuf).

5.2 Insulation grading and impulse design

Insulation must withstand three voltage classes of stress, verified by test (Section 11): power-frequency (AC), lightning impulse (LI, 1.2/50 µs), and for ≥ 170 kV switching impulse (SI). Steep-fronted impulses distribute non-uniformly across a winding according to its capacitance ladder:

Design responses: reinforced paper on line-end turns, static end rings (stress shields), interleaved or shielded discs, and graded barrier counts toward the line end.

5.3 Thermal classes and life

Ageing is Arrhenius-type: per IEC 60076-7, every ~6 K above the reference hotspot roughly doubles the cellulose ageing rate. This is the basis of all loading-capability and loss-of-life calculations.

5.4 Processing: drying and impregnation

1. Winding stabilisation: windings are dried and axially pressed to final height before assembly.
2. Active-part drying: vapour-phase drying (kerosene vapour at ~110–125 °C under vacuum) for power transformers; vacuum hot-air or vacuum-nitrogen ovens for distribution; low-frequency (LFH) drying — current injected at 0.1–1 Hz heats the windings from inside while vacuum pulls the moisture — for fast, automated lines (specialist plant builders such as Hedrich supply this spectrum). Target: < 0.5% moisture in cellulose.
3. Oil filling under vacuum: degassed, filtered oil (< 10 ppm water) admitted under vacuum so no gas pockets remain in barriers and ducts.
4. Standing/impregnation time before HV testing, allowing oil to fully penetrate paper.

5.5 Insulation levels and the stress numbers designers carry in their heads

Working stress rules of thumb (mineral oil system, 1-minute design): subdivided oil ducts 5–10 kV/mm; large unsubdivided gaps 2–4 kV/mm; oil-impregnated pressboard 10–20 kV/mm; creepage along solid surfaces derated ~3× versus puncture (surface tracking is the weak mode — hence angle rings that force creepage paths to cross equipotentials at right angles). The barrier system's whole purpose is to convert one big weak gap into many small strong ones.

5.6 Moisture physics — the quiet life-limiter

• Where the water sits: cellulose holds 95%+ of the unit's water inventory. At equilibrium the split depends strongly on temperature (Oommen/Fabre-Pichon curves): heating drives water from paper into oil, cooling pulls it back — which is why a single oil ppm reading without its sampling temperature is almost meaningless. Always record top-oil temperature and load when sampling.
• Relative saturation matters more than ppm: 20 ppm in hot oil may be 15% saturation (fine); the same 20 ppm at 20 °C is ~35% (risky — free water possible on cooling). Modern moisture sensors read % saturation directly for this reason.
• The bubbling threat: a wet transformer (≥ ~2% in paper) that is suddenly overloaded can flash dissolved water into vapour bubbles at hotspot temperatures around 140–160 °C — instant dielectric weak spots. Dynamic-loading software embeds a bubbling-temperature check precisely because of this; it is the hard ceiling on emergency loading of aged, wet units.
• Measuring moisture in the solid: dielectric response methods — PDC (polarisation/depolarisation currents) and FDS (frequency-domain spectroscopy down to mHz) — fit a moisture model to the insulation's slow polarisation behaviour, giving % moisture in paper without opening the tank. The honest accuracy is ±0.5% class, enough to separate "dry" (<1%), "service-aged" (1.5–2.5%) and "wet — plan drying" (>3%).
• Drying economics: every drying technology (§20.4) ultimately fights diffusion: water leaves thick pressboard barriers orders of magnitude slower than thin paper wraps. Short site-drying runs dry the paper surface and the oil — and the barriers re-wet everything within months. Plan drying durations (or LFH technology) around the thickest insulation in the unit, not the oil figures.

6. Cooling Systems

Losses (Section 10) appear as heat in core and windings and must be moved to ambient air (or water). IEC 60076-2 designates liquid-immersed cooling with four letters; dry types use two (IEC 60076-11).

6.1 How the heat actually flows (ONAN)

Oil heated in the winding ducts becomes buoyant, rises to the tank top, flows out through radiators or fins where it cools and sinks, and re-enters the tank bottom — a thermosiphon needing no pumps. The driving head is the height difference between winding thermal centre and radiator thermal centre, so radiator mounting height is a real design variable.

6.2 Cooling stages and uprating

6.3 Temperature-rise limits (IEC 60076-2, mineral oil)

• Top-oil rise: ≤ 60 K; average winding rise (by resistance): ≤ 65 K; hotspot rise: ≤ 78 K — over the IEC reference ambient (yearly avg 20 °C, max 40 °C).
• Hot climates (GCC): with 50 °C peak ambients, purchasers specify reduced rises (e.g. 50/55 K) or accept derating per IEC 60076-2 Annex; sustained high ambient directly accelerates insulation ageing.

6.4 Worked thermal ladder — 1000 kVA ONAN at full load

Now repeat at a 50 °C GCC summer ambient: hotspot ≈ 118 °C — ageing ~8–10× reference rate while it lasts. That is exactly why Gulf utilities specify reduced rise limits (e.g. 50/55 K) or larger cooling: the transformer is bought oversized in thermal terms so its insulation lives a normal life in an abnormal climate.

6.5 How far can you load it? (IEC 60076-7 framework)

• Time constants do the work: oil heats with a 1.5–3 h time constant, windings in 5–10 min. A 2-hour peak rides mostly on stored thermal mass — which is exactly what cyclic-loading calculations exploit.
• What else limits overload: bushings, OLTC contacts and CTs are rated components too — IEC 60076-7 warns that above 1.5 pu, ancillary equipment is often the real ceiling, not the windings.
• Cooling plant rules of thumb: radiators reject roughly 450–600 W/m² at 60 K mean oil rise (ONAN); adding fans roughly doubles a radiator's duty; fan auxiliary power ≈ 0.3–0.5% of the heat removed. Dirty radiators in dusty climates can silently cost 10–20% of cooling capacity — a wash programme is the cheapest uprate there is.

7. Active Part, Tank & Accessories

The active part (core + windings + cleating + leads + tap-changer) is assembled, dried, and lowered into the tank. Everything bolted onto the tank afterwards — bushings, conservator, protective devices, cooling — is the accessory ecosystem that keeps the active part alive and observable.

7.1 Tank

• Construction: welded mild-steel; stiffeners sized for vacuum filling and internal pressure; designed leak-free over thermal cycling. Surface preparation (shot-blast SA 2.5) + epoxy/PU paint systems — C5-M marine grade common in the Gulf.
• Sealing systems: (a) free-breathing conservator with silica-gel breather; (b) conservator with rubber bag/diaphragm (air never touches oil); (c) hermetically sealed corrugated-fin tank (most pad-mount/distribution units — fins both cool and flex to absorb oil expansion); (d) gas-cushion sealed.
• Corrugated fin walls: on distribution transformers the pressed-fin walls are the radiator; fin depth/pitch are selected from the thermal calculation.

7.2 Bushings

• ≤ 36 kV: solid porcelain or epoxy (DIN 42531/-3, EN 50180/50386 types), plug-in elbow connectors on pad-mounts.
• ≥ 52 kV: condenser-graded — oil-impregnated paper (OIP) or, increasingly, resin-impregnated paper/synthetic (RIP/RIS) with composite silicone housings; capacitance tap provided for tan δ monitoring.
• Bushing failure is a leading cause of transformer fires — hence the shift to dry RIP/RIS technology and online monitoring.

7.3 Tap-changers

• DETC (off-circuit): ±2×2.5% typical on distribution units; operated only de-energised.
• OLTC (on-load): diverter-switch type with transition resistors (Reinhausen, Hitachi/ABB, HM); modern units use vacuum interrupters in the diverter — no arcing in oil, maintenance interval stretched to ~300,000 operations. Selector in main tank or separate compartment; motor-drive mechanism with AVR relay (e.g. 90 relay) for automatic voltage control.

7.4 Protection & monitoring devices

7.5 Assembly sequence (factory)

1. Core building → core erection (Section 3.5)
2. Winding manufacture on vertical/horizontal winders; stabilising press
3. Core–coil assembly: lower windings over limbs, fit end insulation, re-yoke
4. Cleating, lead dressing, tap-changer connection, ratio/resistance checks ("untanked tests")
5. Vapour-phase drying of complete active part
6. Final press, tanking, cover fit; vacuum oil filling
7. Fit accessories, standing time, routine tests (Section 11), dispatch preparation (oil drop + N₂ or full-oil transport, impact recorders)

7.6 Opening up two critical components

(a) Inside a condenser bushing

The core is a paper (OIP) or resin-impregnated (RIP/RIS) cylinder wound around the central conductor with 30–60 concentric aluminium foil layers at calculated radii. Each foil pair forms a series capacitor; stepping the foil lengths grades the radial voltage into near-equal shares, so neither the oil end nor the air end sees a damaging concentration. C₁ (conductor → last foil) is the working insulation; C₂ (last foil → flange) is small and earthed through the test tap. Monitoring lives off that tap: rising C₁ = punctured foil pairs (replace soon); rising tan δ = moisture/ageing. The air-end shed profile handles pollution creepage; the oil end is shaped to grade into the turret oil. Failure energy is enormous (the bushing is a charged capacitor bolted into an oil tank) — which is why tan δ trending and infrared on the top connection are non-negotiable on EHV fleets.

(b) One OLTC tap change, slowed down

1. AVR relay calls for a raise; motor-drive charges a spring (stored-energy mechanism — the switch must never stall mid-stroke).
2. Tap selector (in main-tank oil, no current breaking) pre-selects the next physical tap contact.
3. Diverter switch fires in ~40–60 ms: main contact opens → load current passes through transition resistor R₁ → bridging position: both taps connected through R₁ + R₂, circulating current limited by the resistors → second main contact closes on the new tap → resistors out.
4. Total interruption seen by the network: none. Energy dumped into the resistors: sized for it. Arcing (vacuum-bottle types: contained in vacuum; oil types: carbonises the diverter oil — hence its separate compartment and §20.2 inspection counts).

8. Dry-Type & Cast Resin Construction

Dry-type transformers replace the oil–cellulose system with solid polymer insulation cooled directly by air. Two technologies dominate the MV market: VPI (vacuum pressure impregnation) open-wound coils, and cast resin (CRT), in which the complete HV coil is vacuum-cast in filled epoxy — but they sit inside a wider family.

8.0 The dry-type family at a glance

8.1 VPI (open-wound, impregnated)

• Windings of Nomex®/glass-insulated conductor are wound with air ducts, then repeatedly dipped or vacuum-pressure-impregnated in polyester or silicone varnish and baked (a dedicated VPI plant line: autoclave, resin storage and conditioning, vacuum pumping sets).
• Thermal class up to 180 (H) or 220 — runs hotter than epoxy, so smaller for some duties; favoured in North America and marine applications.
• More tolerant of very large sizes; coils are repairable; but the winding surface remains exposed to humidity and dust → environmental class typically E1/E2 with heaters.

8.2 Cast resin (CRT)

• LV winding: almost always aluminium or copper foil with class-F prepreg interlayer, oven-cured into a solid cylinder.
• HV winding: wound from foil sections, strip or round wire; vacuum-cast in epoxy filled with silica flour (and often reinforced with glass fibre — “filament” or “band” reinforced) for crack resistance over thermal cycling.
• Core: identical CRGO step-lap technology as oil units, usually with rounded multi-step section; protected by resin coating against corrosion.
• Connection: delta-star Dyn11 typical; HV delta links between phase coils; elastic spacers allow differential thermal expansion between coil and core.

8.3 Classification per IEC 60076-11

Cast resin routinely achieves E2/E3 C2 F1 — the combination that qualifies it for tunnels, metros, high-rise risers and offshore platforms where oil is excluded.

8.4 Enclosures, ventilation and installation

• Open frame IP00 (in a dedicated room) or enclosure IP21/IP23/IP31; IP54 only with significant derating or forced cooling — louvre area matters: rule-of-thumb ≥ 0.25 m² inlet + outlet per 10 kW of loss with ≥ 1 m height difference.
• AN base rating; cross-flow fans (AF) add 40–50% for peak/emergency duty.
• Accessories: PT100 sensors in LV ducts (3 per phase typical) wired to a temperature relay (alarm/trip/fan control), anti-vibration pads, plug-in or busduct terminations, optional enclosure heaters.
• No fire bunding, no Buchholz, no breathers, no oil testing — the maintenance schedule reduces to cleaning, torque checks and IR scans.

8.5 Dry-type design specifics the datasheets assume you know

• Reduced impulse levels: air-insulated windings carry lower standard LI levels than oil units — IEC 60076-11 lists e.g. 12 kV class: 60/75 kVp (vs 75 kVp standard in oil), 24 kV: 95/125, 36 kV: 145/170. The lower "list" is common; insulation coordination with arresters matters more, not less, for dry types.
• Air clearances do the dielectric work: phase-to-phase and phase-to-earth distances in air (roughly 90–220 mm at 12–36 kV depending on BIL) drive the core window and enclosure size — a CRT of the same rating is bigger than its oil cousin mainly because air is a worse insulator than oil.
• Thermal cycling is the life test: the resin–copper expansion mismatch works the epoxy every load cycle; quality shows as C2-class crack resistance (−25 °C shock test). Glass reinforcement and flexible inner layers are what stop hairline cracks → PD → tracking years later.
• PD is the health metric: <10 pC at 1.1 U_m on every coil, for life. A site PD re-test after transport/installation is cheap insurance — cracked coils from rough handling are the classic CRT infant failure.
• Altitude & ambient derating: air cooling and air insulation both thin with altitude (derate above 1000 m); at 50 °C GCC ambients apply IEC 60076-12 loading rules — class F insulation does not make a 100 K-rise design immortal in a 55 °C electrical room with blocked louvres.
• Harmonic duty: data-centre/VFD feeders specify K-factor (UL) or factor-K (EN 50541/BS 7821) dry types: subdivided foil/strip conductors, lower flux density, 200% neutrals — exactly the §10.3 physics in air.
• Noise: no oil mass to damp the core, and a resonant enclosure can amplify it — specify L_WA with the enclosure fitted, and isolate with pads (the 100 Hz hum travels through building steel beautifully).

9. Worked Design Example — 1000 kVA, 11/0.433 kV, Dyn11, ONAN

This section walks the classical hand-calculation sequence a design engineer follows before optimisation software refines it. Numbers are rounded for clarity but self-consistent.

Step 1 — Ratings and currents

Step 2 — Volts per turn and turns

LV turns N₂ = 250 / 14.2 = 17.6 → round to 18 turns (turns must be integers; LV rounds first because it has fewest turns). Recalculate E_t = 250/18 = 13.89 V/turn.

HV turns N₁ = 11 000 / 13.89 = 792 turns at principal tap. With off-circuit taps at ±2 × 2.5%: extreme tap +5% → 832 turns; tap section placed mid-winding to limit axial force unbalance.

Step 3 — Core sizing

With 9-step section (K_s ≈ 0.93) and stacking factor 0.97, gross circle area = 379/0.90 ≈ 421 cm² → core circle diameter d ≈ 232 mm. Steel: Hi-B 0.23 mm, step-lap mitred joints. Estimated core mass ≈ 800 kg → P₀ ≈ 800 kg × 1.05 W/kg × 1.15 (building factor) ≈ ≈ 970 W no-load loss — comfortably inside a 1100 W guarantee.

Step 4 — Windings

J is chosen from the loss guarantee, not from thermal limits alone: lower J → more copper mass → lower load loss but higher first cost. Distribution transformers bought on capitalised losses (€/W) often end nearer 2.5 A/mm²; lowest-first-cost designs push past 3.5 A/mm².

Step 5 — Impedance check

Target u_k = 5.0% (±10% tolerance). Using the duct geometry formula of §4.3 with winding height ≈ 400 mm, HV–LV main gap ≈ 12 mm (with 2 pressboard cylinders), radial builds ≈ 25 mm each: calculated u_x ≈ 4.8%, u_r ≈ 1.05% → u_k = √(4.8² + 1.05²) ≈ 4.9% ✓. If low, deepen the main duct or shorten windings; both also raise load loss — the design loop iterates.

Step 6 — Load loss and temperature rise

• DC I²R loss (75 °C) from copper masses ≈ 8.9 kW; add eddy + stray (clamps, tank) ≈ 12–15% → P_k ≈ 10.2 kW vs 10.5 kW guarantee ✓.
• Thermal: total loss at rated load ≈ 11.2 kW → required fin-wall/radiator surface from ~500 W/m²·K effective; verify 60 K top-oil / 65 K winding rise margins (reduced for 50 °C ambient specs).

Step 7 — Mechanical and dielectric verification

• Short-circuit: I_sc = I_rated/u_k ≈ 20 × rated; peak factor 2.55 → radial force on LV checked against foil buckling criterion; end-thrust vs clamp design (IEC 60076-5).
• Dielectric: 11 kV class (U_m = 12 kV) → 28 kV AC 1 min, 75 kV LI. Layer insulation, end discs, and HV–LV barriers chosen accordingly.

Step 8 — What the finished design weighs and costs (indicative)

Sanity ratios a reviewer carries: distribution transformers run ≈ 2.8–3.5 kg/kVA total mass and ≈ 0.55–0.8 kg/kVA of conductor; if a bid is 25% lighter than the field, the losses or the short-circuit margin paid for it somewhere.

Step 9 — What the optimiser actually does

Modern design programs sweep the same loop §1–§9 walked manually: free variables (volts-per-turn, B_m, core diameter, winding heights, current densities, duct widths), hard constraints (guaranteed P₀/P_k/u_k ± tolerance, temperature rises, impulse margins, short-circuit stresses, transport envelope), objective = evaluated cost with the tender's A/B values. Thousands of candidate designs are costed per second; the engineer's judgement moves to vetting the winner: is the lowest-cost design also the one you'd want to build, test and stand behind for 40 years?

10. Losses, Efficiency, Regulation

10.1 Loss inventory

Because P₀ flows 8760 h/year regardless of load, utilities capitalise it heavily (typical evaluation 5–10 €/W for P₀ vs 1–2 €/W for P_k). Distribution transformers are therefore designed with √(P₀/P_k) ≈ 0.3–0.45, putting peak efficiency at 30–45% load — matching real feeder load profiles. Minimum-efficiency regulations (EU Ecodesign Tier 2, DOE 2016/2024, GSO equivalents in the GCC) now set the floor.

10.2 Voltage regulation

At 0.8 pf lagging a 5%-impedance unit drops ≈ 3.7% at full load — recovered by the tap-changer. Note regulation is dominated by u_x·sinφ: reactive load, not active, pulls voltage down.

10.3 Harmonics and K-factor

Non-linear loads (VFDs, UPS, data centres) raise winding eddy loss, which scales with h²·I_h². Specify K-factor (UL) or factor-K (BS 7821 / IEC 61378-relevant) rated units — they use subdivided conductors, lower flux density and oversized neutrals (triplen harmonics circulate in the delta but load the LV neutral at up to 1.73 × phase current).

10.4 Measuring losses honestly — corrections and uncertainty

• Temperature correction: losses are guaranteed at 75 °C (IEC) but measured at whatever the windings happen to be. The DC (I²R) part scales up with temperature — × (235 + 75)/(235 + T_meas) for copper (225 for aluminium) — while eddy and stray losses scale down by the inverse factor. The test engineer splits measured load loss into the two parts via the DC resistance, corrects each in opposite directions, and recombines.
• Why load-loss measurement is metrology, not just measurement: during the short-circuit test the power factor is 0.02–0.05 — the wattmeter reads a tiny active sliver of a huge reactive flow, so a 0.1° phase error in a CT/VT corrupts the loss figure by several percent. Test bays therefore run accuracy-class 0.01–0.05 measuring chains, calibrated end-to-end; at guarantee margins of 1–2%, measurement uncertainty is commercial money.
• Regulatory floors: minimum-efficiency rules now bound what may be sold at all — EU Ecodesign Tier 2 (2021) cut permitted losses to roughly P₀ ≈ 693 W / P_k ≈ 7600 W for a 1000 kVA oil unit (indicative), with DOE and GSO/Gulf equivalents following the same logic. Note our §9 design (970 W / 10.2 kW) meets a classic GCC utility spec but not EU Tier 2 — the same kVA is a different transformer per market, which is exactly how loss regulation reshapes designs.

11. Factory Testing

IEC 60076-1 divides tests into routine (every unit), type (one unit of a design), and special (by agreement). The test sequence matters: dielectric tests come after loss measurement, and PD/repeat tests follow any dielectric test that could have aged the insulation.

11.1 Three tests in detail

(a) Lightning impulse — reading the records

Failure detection is comparative, not visual: a shorted turn changes winding impedance subtly, so the digitised neutral-current trace of the full-voltage shot is overlaid on the reduced-voltage reference. Divergence anywhere in the first tens of µs fails the unit. Transferred surges are checked on non-impulsed windings.

(b) Temperature-rise (heat run) — how you heat a 100 MVA unit without 100 MW

By the short-circuit method: LV shorted, HV fed at the voltage that drives total losses (P₀+P_k) of current — the tank doesn't know the difference between real load heat and test heat. Run until top-oil rise changes <1 K/h, then inject rated current for 1 h, switch off and measure winding resistance decay; extrapolate the cooling curve back to switch-off instant to get average winding temperature. Gradients and hotspot follow per IEC 60076-2.

(c) SFRA — the transportable fingerprint

Sweep frequency response analysis injects 10 Hz–2 MHz into each winding and records the transfer function. Every mechanical detail — winding position, clamping, core contact — shapes the resonances. Identical-looking curves at factory and site = nothing moved in transport; a shifted resonance band localises the problem (low frequencies → core/magnetising path, mid → inter-winding movement, high → lead/tap geometry). It costs nothing and has saved many a silently-damaged unit from energising into failure.

11.2 Site testing & diagnostics through life

• Pre-commissioning: IR/PI, ratio, vector group, winding resistance, SFRA vs factory fingerprint, oil BDV/moisture, Buchholz/protection trip checks.
• In service: annual oil DGA (the single most informative test — H₂: PD; C₂H₂: arcing; C₂H₄: overheating; CO/CO₂: paper ageing), furans for cellulose life, tan δ of bushings, thermography, OLTC maintenance by operations count.

11.3 Partial-discharge measurement in practice

• The circuit: a coupling capacitor in parallel with the test object feeds a measuring impedance; apparent charge (pC) is read per IEC 60270. Before testing, a calibrator injects a known charge across the terminals to scale the chain, and the background level is verified (< 50 pC for oil units, much lower for CRT bays — a noisy bay cannot certify a quiet transformer).
• The profile (induced test with PD, ACLD type): raise to an enhancement level (~1.7–1.8 × U/√3) briefly to "ignite" any defect, drop to ~1.58 × U/√3 and hold (up to 60 min on EHV) while watching level and trend. Acceptance is both a ceiling (e.g. ≤ 100–250 pC oil, ≤ 10 pC cast resin at its own levels) and stability — a rising trend fails even below the ceiling.
• Reading patterns: phase-resolved PD patterns separate defect types — symmetrical rabbit-ear clusters: internal void; one-sided: surface/creepage; noise-like at fixed phase: external corona (fix the bay, not the transformer). Acoustic/UHF sensors triangulate the source inside the tank if the electrical channel says "real".
• Why it matters commercially: PD is the only routine test that catches workmanship invisibility — a slightly low oil level in a barrier pocket, a contaminated snout, a void in a casting. It is the last gate before a latent defect ships.

12. Standards Map

12.1 The IEC 60076 series, part by part

12.2 The IEEE C57 family, grouped the way it's used

12.3 Liquids, materials and oil-laboratory standards

12.4 Efficiency regulation & regional frameworks

12.5 CIGRÉ — the working literature between standards

Standards say what; CIGRÉ technical brochures (study committee A2) explain why and how. The ones every transformer engineer ends up reading: TB 642 (transformer reliability survey — the failure statistics quoted in §16.5), TB 445 (maintenance guide), TB 761 (condition assessment), TB 343/glossy guides on SFRA interpretation, TB 296 (DGA interpretation refinements), TB 659/779 (bushing reliability), TB 575 (OLTC reliability), and the A2 brochures on GIC, short-circuit design review and dynamic loading. They are purchasable references — for a service business (§20), the reliability and condition-assessment brochures are the technical backbone of the Green/Amber/Red report.

12.6 IEC vs IEEE — the practical differences that bite

• Temperature-rise conventions: IEEE liquid-immersed practice is built around 65 °C average rise over 30 °C average ambient (with 55/65 dual ratings historic); IEC's reference is 60/65 K over 20 °C yearly average. Same physics, different bookkeeping — never transplant a guarantee between systems without converting.
• kVA habits & impedance bases: IEEE specifies ONAN/ONAF dual ratings as e.g. 12/16/20 MVA OA/FA/FA stages and quotes %Z on the self-cooled base; IEC quotes on rated (usually maximum) power. A "9% vs 11.25%" disagreement is often the same transformer.
• BIL philosophy: IEEE ties a fixed BIL ladder to system classes and tests chopped-wave at 110% as standard; IEC offers LI level menus per U_m with LIC optional by agreement.
• Test defaults: induced-voltage with PD (ACLD) is routine in IEC ≥ 72.5 kV; IEEE relies on its own induced/applied sequence with different PD limits (C57.113). Sound: IEC 60076-10 sound power; IEEE/NEMA quote sound pressure at standard distance — numbers differ by construction of the measurement, not the noise.
• Vector notation: Dyn11 clock notation (IEC) vs angular-displacement diagrams (IEEE); North-American practice favours 30° lag (Dyn1-equivalent) — a paralleling trap when mixing fleets (§13.3).

Closing summary

Every transformer is the same five decisions wearing different clothes: a volts-per-turn that splits the work between iron and copper (§1, §9); a core whose steel grade and joints fix the no-load behaviour forever (§3); windings whose geometry simultaneously sets impedance, load loss and short-circuit survival (§4); an insulation-plus-cooling system matched to the environment — oil-cellulose for compactness, epoxy-air where fire safety rules (§5–§8); and a verification regime of guarantees and tests that makes the whole thing contractually real (§10–§12). Master those couplings and any datasheet, tender deviation or test report becomes readable at a glance.

13. Vector Groups, Connections & Parallel Operation

13.1 Why connection choice matters

13.2 Clock notation

The vector group symbol — e.g. Dyn11 — reads: HV delta (D), LV star (y), neutral brought out (n), LV phasor at clock position 11, i.e. the LV leads the HV by 30° (each clock hour = 30°). Standard groups: 0 (Yy0, Dd0), 6, 1 (Dyn1, Yd1) and 11 (Dyn11, Yd11). Dyn11 is the IEC/GCC default for distribution.

13.3 Conditions for parallel operation

1. Same vector group (or compatible after external phase swaps) — a Dyn11 cannot parallel a Dyn1: the 60° difference drives fault-level circulating current.
2. Equal voltage ratios at the operating tap — ratio mismatch drives circulating current ≈ ΔV% / (u_k1+u_k2) of rated.
3. Equal impedance voltages (within ~±10%) — units share load inversely to impedance; the low-impedance unit overloads first.
4. Similar X/R ratio (second-order effect); rating ratio preferably within 1:3.

13.4 Unbalanced load behaviour

Dyn and Yzn groups feed unbalanced LV networks gracefully: the delta (or zigzag) gives zero-sequence MMF a path, so the neutral holds. A plain Yyn0 three-limb unit on unbalanced load suffers neutral displacement and tank heating from zero-sequence flux returning through tank steel — the reason Yy is avoided for 4-wire distribution.

13.5 Zero-sequence impedance — the connection's hidden fingerprint

Protection engineers ask for Z₀ before anything else: it sets earth-fault levels, REF sensitivity and neutral-earthing-resistor sizing. Note that Z₀ is a core-construction property as much as a connection property — the same YNyn nameplate behaves completely differently on 3-limb vs 5-limb iron (§14.4 told the same story for DC flux).

14. Magnetising Behaviour, Inrush & Overfluxing

14.1 The magnetising current is not sinusoidal

Because the B–H curve saturates, a sinusoidal flux demands a peaky, harmonic-rich magnetising current (dominant 3rd harmonic). At rated voltage it is only 0.2–2% of rated current, but its waveshape matters: differential relays use harmonic signatures to distinguish magnetising events from internal faults.

14.2 Energisation inrush

Mechanism: flux is the integral of voltage, so the flux demanded after switching depends on the closing instant on the voltage wave and on the remanent flux left from the previous de-energisation. Worst case (closing at voltage zero with adverse remanence) demands ≈ 2Φ_m + Φ_r: the core saturates deeply and the magnetising branch briefly behaves like an air-cored reactor drawing 6–12 × rated current.

• Consequences: nuisance differential-relay operation (hence 2nd-harmonic restraint), voltage dips on weak networks, mechanical winding stress, sympathetic inrush in already-energised parallel units.
• Mitigation: point-on-wave controlled switching, pre-insertion resistors, core demagnetisation after testing, or simply grading upstream protection for the inrush I²t.

14.3 Overfluxing (V/f)

Flux ∝ V/f. Overvoltage or under-frequency (generator run-up, load rejection) pushes the core toward saturation: magnetising current and core temperature climb steeply and stray flux spills into clamps and tank. IEC 60076-1 requires 105% V/f continuously; beyond that, withstand time falls fast — GSU transformers always carry V/f (ANSI 24) protection.

14.4 DC bias and GIC

Even a few amperes of quasi-DC (geomagnetically induced currents, HVDC earth-return, DC traction leakage) offsets the core into half-cycle saturation: noise, hot spots, harmonic injection and reactive-power swings. Five-limb and shell cores are more susceptible than three-limb cores (whose zero-sequence DC flux must cross the high-reluctance tank gap). Mitigation: neutral blocking/resistive devices, monitoring, design margin.

14.5 Sympathetic inrush and energisation practice

• Sympathetic inrush: energising one transformer drives a DC-offset current through the system resistance, which slowly drags the cores of already-running parallel transformers into partial saturation — they begin "answering" with their own offset currents. The two inrushes can sustain each other for many seconds on weak networks, confusing relays on units nobody switched. Recognise it: the energised unit's inrush decays unusually slowly while a neighbour gasses up its differential restraint.
• Remanence management: after DC winding-resistance tests, the core retains substantial flux — demagnetise (decaying AC or reducing bipolar DC sweeps) before impulse testing or energisation, or expect spectacular inrush and even distorted SFRA traces.
• Good energisation practice: energise from the higher-impedance side where practical (smaller inrush per flux demand), use point-on-wave controlled closing on large/critical units (each phase closed at its own optimum instant, with remanence either measured or forced to a known pattern), confirm 2nd-harmonic restraint healthy, and treat repeated rapid re-energisations as mechanical abuse — every inrush hammers the windings with offset forces.

15. Special Transformer Types

15.1 Autotransformers

Series + common winding share one electrical circuit. The built (frame) size shrinks by the co-ratio (1 − V₂/V₁): a 400/220 kV interconnector (co-ratio 0.45) needs only ~45% of the equivalent two-winding frame — cheaper, smaller, lower losses. The costs: no galvanic separation, lower natural impedance (higher fault duty), direct surge transfer, and the neutral must be solidly earthed. Universal at 400/220/132 kV; the delta tertiary stabilises the star point and feeds reactive compensation or station supply.

15.2 Earthing (grounding) transformers

Zigzag (or Yd) units that manufacture a neutral on a delta-fed network, rated for short-time fault current, often with a neutral earthing resistor; frequently combined with an auxiliary LV winding for substation supplies.

15.3 Converter / rectifier transformers (IEC 61378)

Feed 6/12/24-pulse rectifiers, VFDs, electrolysis and HVDC. Designed for a declared harmonic spectrum (winding eddy loss scales ~h²·I_h²), phase-shifted secondaries (Dd0 + Dy11 for 12-pulse cancellation), reinforced mechanical duty and often interphase reactors. Data-centre and traction supplies are this family's fast-growing cousins.

15.4 Furnace transformers

Arc-furnace duty: tens of kA at 100–1500 V on the secondary, very wide regulation (many-step OLTC, often a series booster), violent load swings, flicker, water-cooled LV deltas closed right at the furnace. Built like a power transformer, stressed like a forge press.

15.5 Generator step-up (GSU)

Matched to one machine: full load continuously, unit-connected, exposed to generator transients and V/f excursions; OFWF/ODWF cooling common inside power stations. Renewables now cycle GSUs hard — a known ageing accelerant.

15.6 Phase-shifting transformers

Inject a quadrature voltage to steer active power between parallel transmission corridors — effectively a tap-changer that controls MW instead of volts. Single- or two-core designs; OLTC duty is severe.

16. Installation, Commissioning & Life Management

16.1 Transport and receipt

• Large units travel with oil dropped, under dry N₂ or dry air, with radiators, conservator and bushings removed. Three-axis impact recorders ride along — download before unloading and compare against the ~3–5 g limits.
• On receipt: core-insulation check via the test link, dew-point of the gas blanket (≈ ≤ −40 °C target), internal inspection for displaced blocking if the recorders show an event.

16.2 Site erection and oil work

1. Position on plinth/rails; anti-vibration pads under dry types; verify clearances for OLTC handle, bushing turrets and fire separation walls.
2. Fit radiators, conservator and bushings in clean, low-humidity conditions (<75% RH; cumulative active-part exposure typically limited to 8–12 h before re-drying is required).
3. Pull vacuum (≤ 1 mbar on large units), then fill with hot (50–60 °C) degassed oil through a mobile vacuum oil purification plant (degassing, dehydration, particle filtration) until moisture < 10 ppm and BDV > 70 kV/2.5 mm; observe standing time (≈24 h ≤ 72.5 kV, 72 h+ for EHV) before energising.

16.3 Commissioning test set (typical)

16.4 Condition monitoring through life

• DGA — the blood test (IEC 60599 / IEEE C57.104): H₂ → partial discharge; CH₄/C₂H₆ → low-temperature overheating; C₂H₄ → severe overheating; C₂H₂ → arcing (act immediately); CO/CO₂ → cellulose ageing. Trends beat absolutes; Duval triangles classify the fault.
• Furans (2-FAL) in oil correlate with the paper's degree of polymerisation: DP ≈ 1000 new → DP ≈ 200 = end of mechanical life.
• Online monitoring: multi-gas DGA, bushing sum-current, UHF/acoustic PD, fibre-optic hotspots, OLTC torque signatures — feeding fleet asset-health platforms.
• Oil care: reclamation (fuller's earth) reverses oxidation; online cellulose drying matters most — 95%+ of the unit's water sits in the paper, not the oil.

16.5 Failure statistics and end of life

CIGRÉ surveys put major-failure rates around 0.4–1% per transformer-year, led by OLTC mechanisms, bushings and winding insulation. Practical life is 30–50 years for well-loaded, dry, cool units — thermal history is destiny (recall the ~6 K doubling rule, §5.3). Repair-vs-replace usually pivots on capitalised losses: a 1980s unit can pay for its modern replacement from saved load loss alone.

16.6 Two field skills worth doing properly

(a) Taking an oil sample that means something

1. Sample from the bottom drain/sampling valve with the unit in service or just off load; record top-oil temperature, load and ambient at the moment of sampling (the moisture interpretation of §5.6 depends on them).
2. Flush the valve generously (1–2 L) into a waste container — you are sampling the tank, not the dead leg.
3. For DGA: glass syringe, no air bubble, no headspace, sealed and shaded in transit (sunlight and air strip the very gases you're measuring). For moisture/BDV: clean dry bottles, filled to the top.
4. One unbroken chain: same sampling point, same lab, same method, every year — DGA is a trending tool; switching labs mid-history halves its value.

(b) Storing a spare transformer so it's actually usable

• Best: stored oil-filled with conservator and breather active, sampled annually like an in-service unit; energised trickle/storage heating where humidity is brutal.
• If stored gas-blanketed (as shipped): log the N₂/dry-air pressure weekly — a falling gauge is a leak inhaling humid air; re-pressurise and fix before the insulation pays.
• Exercise the accessories: OLTC through-run quarterly (mechanisms seize, oil films drain), fan/pump test runs, valve cycling, dessicant changes.
• Before deployment: full §16.3 commissioning set against factory fingerprints — a spare that fails on energisation after 8 quiet years is a spare that was stored, not maintained.

17. Transformer Protection Schemes

A transformer is protected by layered electrical relays plus its own mechanical devices (§7.4). The aim: clear internal faults in < 100 ms (every extra cycle of arcing multiplies the damage and the oil-fire risk) while never tripping on inrush, OLTC operation or through-faults.

17.2 Putting numbers on the schemes

• 87T typical settings: minimum pickup 0.2–0.3 pu (covers CT and tap-position mismatch at low current); slope 1 ≈ 20–30% up to ~2 pu bias (rides ratio errors); slope 2 ≈ 60–80% beyond (rides CT saturation during heavy through-faults); unrestrained high-set ≈ 8–10 pu (clears terminal faults in one cycle, no questions asked). Harmonic logic: block or restrain at 2nd-harmonic content ≳ 15–20% (inrush), 5th ≳ 25–35% (overfluxing).
• Numerical relays earn their keep by computing the vector-group shift and ratio matching internally (no interposing CTs), tracking the actual OLTC position to shrink mismatch, and adding sensitive REF, thermal images and V/f in one box.
• CT homework: protection cores sized so saturation doesn't fake a differential during through-faults — class PX (knee-point specified) for high-impedance REF with the stabilising resistor set above the worst saturated-CT spill voltage; 5P20-or-better with generous burden margin for biased differential; remanence margins on reclosing feeders.
• Tripping matrix discipline: Buchholz surge, PRV, sudden-pressure, 87T, 87N → master lockout (86) tripping both breakers, no auto-reclose, no remote reset without inspection. Gas alarm, WTI alarm, low oil → alarm/SCADA only. The matrix is small; the discipline of keeping mechanical trips on the lockout path is what saves tanks.

18. Procurement, Tender Evaluation & Economics

18.1 Total ownership cost — the capitalisation formula

Transformers are bought on evaluated price, not purchase price. Losses are converted to present-value money:

• A applies the full 8760 h/year (core loss never sleeps); typical utility values 4–12 $/W.
• B is weighted by loading: B ≈ A × (load factor)² × loss-load factor; typically 1–4 $/W.
• Consequence: two units differing 20 kW in losses can swing an evaluation by >$100k — manufacturers tune designs to the A/B values published in the tender. High A → low-flux Hi-B cores; high B → low current density, more copper.

18.2 What a complete technical tender schedule contains

18.3 Factory assessment — what to look at in an audit

• Core line: burr height records, step-lap stacking discipline, steel storage (CRGO is stress- and rust-sensitive).
• Winding shop: humidity/cleanliness control, conductor handling, transposition workmanship, stabilisation press records.
• Processing: vapour-phase autoclave capability vs unit size, oil plant (ppm achievement records), exposure-time logging between drying and tanking.
• Test bay: impulse generator and PD background level, calibration traceability, loss-measurement uncertainty (matters at guarantee margins of 1–2%).
• Type-test pedigree: short-circuit tested designs in the same family — the single strongest differentiator between manufacturers.

18.4 Commercial life cycle

Order → design review (4–8 weeks) → drawings approval → manufacture (distribution 3–6 months; large power 10–20 months, EHV longer — core steel and bushings are the common long-lead items) → FAT with witnessed routine tests → transport → site erection & commissioning (§16) → warranty period with first-year DGA milestones. Penalties typically bite on losses, late delivery and failed guarantees; smart buyers also contract response times for warranty defects.

18.5 Deriving A and B — one worked capitalisation

Utility assumptions: energy cost c_e = 0.08 $/kWh, evaluation life N = 25 yr, discount rate i = 8%, transformer load factor 0.5, loss-load factor ≈ 0.3.

• A (no-load loss): 1 W costs 8.76 kWh/yr × 0.08 $ = 0.70 $/yr → A = 0.70 × 10.67 ≈ 7.5 $/W.
• B (load loss): weighted by the loss-load factor: 1 W of rated-load loss actually dissipates ~0.3 W on average → B ≈ 7.5 × 0.3 ≈ 2.2 $/W.
• Sensitivity is the lesson: at i = 4% (cheap capital) PWF ≈ 15.6 → A ≈ 11 $/W and low-loss designs sweep the evaluation; at i = 12% (expensive capital) A ≈ 5.5 $/W and cheaper, hotter designs win. Energy price moves A and B together; load factor moves only B. A procurement department that publishes its A/B values is literally telling manufacturers which transformer to design (§19 Q12 runs the comparison).

19. Glossary & Self-Test

19.1 Quick glossary

19.2 Self-test — sixteen questions

Answers are hidden under each question. Do them after reading the linked sections.

20. After-Sales & Service — The Second Life of Every Transformer

A transformer earns its price over 30–50 years, and everything after FAT is the after-sales domain: warranty management, preventive maintenance, condition assessment, oil and component services, repair, retrofit and end-of-life. For owners this is where reliability is won; for service providers it is a business larger and steadier than the new-build market.

20.1 The warranty phase — protect the claim

• Documentation discipline from day one: impact-recorder downloads, storage records (oil/N₂ pressure logs if stored), erection and oil-processing records, commissioning test set vs factory values — a warranty claim without this trail is a negotiation, not a claim.
• First-year milestones: DGA at energisation, 24 h, 1 week, 1/3/6/12 months. A rising combustible-gas trend inside warranty is the manufacturer's problem — but only if sampled properly (clean syringes, no air ingress, accredited laboratory).
• First service visit (~12 months): retorque accessible connections, gasket inspection, breather condition, OLTC first inspection per OEM book, protection re-test, thermography baseline under load.

20.2 Preventive maintenance schedule (oil-immersed, typical)

20.3 Condition-based services — reading the asset

• Oil laboratory panel: DGA (fault gases), moisture-in-paper estimate, acidity/IFT (oxidation), furans (DP estimate), corrosive sulphur, PCB screening on older fleets. One annual sample replaces a hundred opinions.
• Electrical assessment outage: SFRA, bushing tan δ, winding resistance, excitation current, dielectric response (PDC/FDS — measures moisture in the solid insulation directly). Together with the oil panel this grades each unit Green/Amber/Red for capex planning.
• Online monitoring retrofits: multi-gas DGA monitors, bushing sum-current units, fibre-optic hotspots, PD sensors — the §7.4 ecosystem retrofitted to legacy fleets, increasingly demanded by GCC utilities for critical substations.

20.4 Corrective services — the heavy end

20.5 Spares strategy

• Hold at site/fleet level: complete bushing set per voltage class, Buchholz + PRV + gauges, breathers and media, gasket kits, fan and pump motors, OLTC contact kits and drive spares, silica gel, touch-up paint system.
• Strategic: a shared spare transformer per critical rating class — at 12–20+ month new-build lead times, the spare unit is the only real insurance for single-contingency substations.
• Match spares to the installed accessory makes recorded at handover (§18.2) — the rating plate never tells you whose Buchholz is on the pipe.

20.6 Running after-sales as a business

The transformer service market is global and structurally growing: fleets installed in the 1970s–90s boom are ageing out everywhere at once, while new-unit lead times keep owners maintaining and life-extending instead of replacing. The same service portfolio sells in Houston, Hamburg, Mumbai or Dubai — what changes is the climate adjustment and the local delivery model.

• Contract forms: annual maintenance contracts (visit-based), condition-based fleet contracts (lab panel + assessment + ranked report), emergency-response SLAs with guaranteed mobilisation times, resident-engineer arrangements for industrial plants, and framework agreements with utilities covering whole fleets.
• The deliverable is the report: owners pay for a defensible Green/Amber/Red fleet ranking with budget recommendations — the measurements are the means, not the product.
• Adjust the OEM book to the duty: maintenance intervals written for temperate Europe shorten under 50 °C desert ambients, dust-loaded radiators and coastal salt (Gulf, MENA, coastal Asia), under heavy cycling (renewables, traction), and under harmonic-rich loads (data centres, EV charging) — and lengthen for lightly loaded indoor units. Reading the duty correctly is the service engineer's core skill, whatever the geography.
• Anchor partnerships: an accredited oil laboratory, an insulation materials and kit channel, site oil-processing and drying capability, and OEM service agreements for OLTCs and bushings — the supply chain behind every service above.

21. Bushings — Getting the Conductor Through the Tank

A bushing carries a live conductor through the earthed tank wall while holding off the full phase-to-earth voltage across a few hundred millimetres. It is deceptively critical: industry failure surveys (CIGRÉ TB 642) attribute roughly 30–40 % of catastrophic, fire-causing transformer failures to bushings — more than any other single component. For a 400 kV unit the bushing is also one of the most expensive bought-in parts. It deserves its own chapter. interactive Rotate and section a 400 kV bushing in the 🔌 3D Condenser-Bushing Cutaway.

21.1 Why a plain insulator will not do — condenser grading

If you simply passed the conductor through a porcelain tube, the radial field at the conductor surface and the axial field along the flange would both concentrate to the point of discharge. A condenser (capacitor-graded) bushing solves both at once: thin conducting foils are wound into the insulation at set radii, forming a stack of series capacitors between the conductor and the earthed flange.

The outermost foil is earthed at the flange; the last-but-one foil is brought to the test tap. This is the whole idea behind the concentric foils you can slice open in the 3D model.

21.2 Types — the family tree

21.3 Construction & ratings

Working from the top down: the top terminal and a corona/grading ring (round-off of the air-end field); the porcelain or composite air-side insulator climbing the creepage distance; the condenser core (foils + paper/resin); the mounting flange with the test tap; the shorter oil-side insulator; and the central conductor — either a fixed rod, a bottom-connected stem, or a flexible draw-lead pulled through after the bushing is fitted. A bushing CT sits in the turret around the oil end (see §11 / §17).

Governing standards: IEC 60137 and IEEE C57.19; oil-side interface per the transformer's own insulation design.

21.4 Failure modes & monitoring — why bushings burn transformers

The dominant failure chain is moisture ingress / gasket ageing → partial discharge in the condenser core → rising capacitance (a shorted foil steps C up) and rising dissipation factor tan δ → thermal runaway → flashover and, with OIP, tank rupture and pool fire. Because the process is gradual, it is monitorable:

• Offline: C1/C2 capacitance and tan δ (power factor) measured at the test tap at each outage; a > +5–10 % capacitance rise or tan δ climbing above ~0.7 % (OIP) is an alarm.
• Online bushing monitors: sum the three phase test-tap leakage currents — a healthy set nearly cancels; a developing fault unbalances the sum and shifts its phase angle, giving weeks of warning.

22. Tap-Changers — Regulating Voltage Under Load

The tap-changer is the transformer's only moving power part, and after bushings it is the second-largest contributor to failures (CIGRÉ) — precisely because it moves, arcs and wears. It is also core to your accessory business, so it gets its own chapter. interactive Watch a tap operation, snap the diverter and swap resistor↔vacuum in the 🔀 3D OLTC Cutaway.

22.1 On-load vs de-energised

22.2 How an OLTC transfers current without interrupting it

The trick is to split the job between two devices (see the 3D model): a tap selector that pre-selects the next tap off-load (so its contacts never arc), and a fast diverter switch that then transfers the current from the present tap to the pre-selected one in ~40–60 ms. During that brief transfer both taps are momentarily bridged; a transition impedance limits the circulating current and takes the switching arc:

• Resistor type (IEC world): a transition resistor carries the arc; a spring energy-accumulator snaps the diverter so arc energy is small and independent of how fast the operator drives the mechanism.
• Reactor type (traditional North American): a preventive autotransformer (reactor) allows continuous operation bridged across two taps.
• Vacuum type (modern): switching happens inside sealed vacuum interrupters — no arcing in oil, so the oil stays clean, maintenance stretches to ~300,000 operations and no diverter-oil filter is needed.

A change-over (reversing) selector flips the regulating winding in series-additive or series-subtractive with the main winding, doubling the range from one set of tap leads. On a star HV the regulating winding sits at the neutral end, where insulation to earth is lowest.

22.3 Control — holding the setpoint

An automatic voltage regulating (AVR) relay compares the measured voltage to a setpoint with a deadband and time delay, and commands raise/lower steps. Line-drop compensation lets it regulate a remote point by modelling the feeder impedance. Paralleled transformers need coordination — master–follower, circulating-current or negative-reactance methods — or they will fight each other and drive reactive current between units.

22.4 Maintenance & diagnostics

• Operation counter triggers time/'000-operation-based servicing of the diverter (contacts, springs, oil).
• DGA of the diverter oil is read separately from the main tank — the diverter arcs by design, so its gas signature is different; rising C₂H₂ beyond the arc baseline flags contact coking or a mis-operation.
• Motor-drive checks: torque, step completion, limit switches, heater.

Standards: IEC 60214 (OLTC) and IEEE C57.131; selection per IEC 60214-1 duty classes.

23. Impulse Response & Electrostatic Stress Control

Chapters 5 and 11 size the insulation for the BIL and test it. This chapter explains the missing physics in between: how a winding actually sees a lightning surge, and the hardware — static rings, end shields and interleaving — that keeps the line-end turns from being destroyed. This is the design detail that separates an EHV transformer engineer from a hobbyist.

23.1 A steep front is a capacitive problem, not a resistive one

A lightning impulse rises in ~1.2 µs. At that speed the winding inductance is momentarily "open" — the surge cannot yet flow as turns current. The winding instead behaves as a ladder of capacitors: series capacitance C_s between adjacent discs/turns, and ground capacitance C_g from each point to the earthed core and tank. The initial voltage therefore distributes non-uniformly, crowding at the line end.

The initial stress at the line end is roughly α times the average. A plain disc winding has α ≈ 5–15, so the first few discs see 5–15× their "fair share" — the classic cause of line-end failures.

23.2 The remedies — raise C_s, intercept C_g

• Static end rings (stress rings): a floating metal ring, at line potential, wrapped in insulation and placed just outside the first discs. It supplies the ground-capacitance current that would otherwise flow through the winding, so the first turns are relieved — the single most important line-end fix.
• Electrostatic / wound shields: shield turns tied to the line end that add capacitance to the outer discs, linearising the profile.
• Interleaved & intershielded disc windings: re-ordering the conductors so that turns far apart electrically sit physically adjacent multiplies the effective series capacitance C_s, pushing α toward 1–3 (near-linear). This is why big HV windings are interleaved (see §4).
• Grading of the major insulation and oil ducts to match the now-linear stress.

23.3 After the front — the oscillation

Once inductance re-enters (microseconds later) the winding swings from its non-uniform initial distribution toward the linear final one, overshooting and oscillating. The transient can drive mid-winding and inter-disc voltages temporarily higher than the applied crest — so the designer checks not only the line end but the whole envelope. The chopped-wave test (a flashover-chopped impulse, even steeper) is the most severe case for inter-turn stress.

24. Fire, Oil Containment & Environmental Protection

A 500 MVA unit holds ~95,000 litres of flammable oil beside 400 kV. When a bushing or OLTC fails, the arc can rupture the tank and ignite a pool fire that threatens the whole substation. Designing the envelope around the transformer — fire prevention, containment and environmental protection — is as much a part of the job as the active part, and is mandated by substation codes. This chapter pulls together what §16 and §7 only touched.

24.1 Pressure & arc protection (stop the rupture)

• Pressure-relief valve (PRV): spring-loaded, vents a sudden internal pressure rise and reseals.
• Sudden-/rapid-pressure-rise relay: trips on the rate of pressure rise from an internal arc, faster than Buchholz for a violent fault.
• Buchholz (§7): gas accumulation alarm + oil-surge trip.
• Nitrogen-injection fire-prevention systems (e.g. transformer "fire protector" units): on a confirmed internal fault they drain oil from the top to kill the expansion and inject nitrogen at the bottom to stir and cool below the fire point — preventing or extinguishing a tank fire. Increasingly specified for urban and critical GCC substations.

24.2 Fire suppression & separation (contain the fire)

• Deluge / water-spray systems per NFPA 850 for oil-filled units above threshold size or too close to other assets.
• Fire-rated separation walls (barriers) between adjacent transformers and to buildings, or code separation distances where walls are impractical — governed by IEC 61936-1 and local utility standards (DEWA/ADDC/SEC substation fire codes in the GCC).
• K-class (ester) fluids, fire point ≥ 300 °C, can remove the deluge requirement and enable indoor/rooftop installation — a growing driver for natural/synthetic esters.

24.3 Oil containment & the environment (contain the spill)

• Bund / oil-containment pit sized for 100 % of the oil volume plus firewater, with a fire-quenching stone/grating layer over a drain to a remote holding tank.
• Oil–water separator on the site drainage so rainwater can be released but oil is trapped.
• Kerbs, ramps and interceptors to keep a spill on-site and out of ground/surface water.
• Legacy hazards: PCB in old askarel-filled units (disposal controls); SF₆ in some bushings/GIS interfaces (greenhouse gas handling).

24.4 Sustainability — the losses are the footprint

Over a 40-year life the CO₂ from a transformer's losses dwarfs the embodied carbon of its steel and copper — which is exactly why the loss-capitalisation A&B factors (§18, §26) and the minimum-efficiency regulations (EU EcoDesign Tier 2, DOE, GSO/SASO) exist. Eco-levers: lower flux density and better core steel (amorphous cuts no-load loss ~70 %), biodegradable esters, and end-of-life recovery of copper, electrical steel and regeneration of oil rather than disposal.

25. Worked Design — 500 MVA, 400/220 kV Power Transformer

Section 9 sized a 1000 kVA distribution unit. Here is the EHV counterpart that sits behind the 🏭 500 MVA 3D model and the calculator's Power mode — a two-winding 400/220 kV grid transformer, YNyn0+d, 5-limb, ONAN/ONAF/ODAF. The method is identical; only the scale, the impulse grading and the cooling stages change.

Step 1 — Ratings & line currents

The 220 kV winding carries the higher current, so it is the heavier-conductor disc winding; both main windings are disc (not helical) at these voltages.

Step 2 — Volts-per-turn, flux & core

EHV design runs a high volts-per-turn to keep the turn count and hence the winding height sensible. Taking a Hi-B steel at B_m ≈ 1.7 T:

Solving A_net from E_t gives a core diameter around 0.85–0.95 m; a 5-limb core keeps the yoke — and therefore the transport height — within rail/road limits and gives zero-sequence flux a return path for the YN/yn connection.

Step 3 — Impedance & short-circuit

Grid units at this rating run high impedance to limit downstream fault levels:

Step 4 — Insulation & impulse grading

BIL 1425 kV (400 kV) / 1050 kV (220 kV). The line end gets static rings and interleaved discs (§23) to hold α near 2; major insulation is pressboard barriers with graded oil ducts; the neutral end runs at low insulation and carries the tap winding.

Step 5 — Cooling stages

Total loss ≈ NLL + LL ≈ ~1.3 MW at full load. Dual/triple rating is normal:

Step 6 — Losses, mass & economics

Indicative: NLL ≈ 250–350 kW, LL ≈ 900–1100 kW (at rated), total mass ≈ 340 t, oil ≈ 95 kL. Evaluated with GCC loss-capitalisation (§18): with A ≈ 6,000 $/kW and B ≈ 2,000 $/kW, the capitalised loss cost is on the order of the purchase price — so the optimiser spends iron (better steel, lower B_m) to buy back no-load loss.