Power Transformer Sizing and Selection: 8 Questions I Learned the Hard Way (and What They Cost Me)

I've been handling transformer and switchgear orders for nine years, mostly industrial and commercial retrofit work. In that time I've personally made — and written down — 14 significant mistakes that cost roughly $47,000 in rework, expedited freight, and one avoidable return. Most of them weren't technical failures. They were spec gaps.

These are the eight questions I wish someone had handed me in year one.

• How big does the transformer actually need to be?
• Step-up or step-down — can't I just wire it backwards?
• Pole-mounted or pad-mounted?
• When is a 3-phase dry-type the right call?
• What makes a photovoltaic distribution transformer different?
• What do I check before I sign for delivery?
• Is it worth ordering just one unit?
• Nobody ever told me about percent impedance. Why does it matter?

1. How big does my power transformer actually need to be?

This one costs the most money, and it's almost never about the transformer itself. It's about the load calculation.

People size in kW. Transformers are rated in kVA. The gap between those two is power factor, and on a plant full of induction motors running around 0.8 PF, that gap is 25%. A 400 kW load is a 500 kVA transformer. Not a 400 kVA one. I learned that in 2017 on a compressor room job where the quote said '400 kVA is plenty' and nobody had asked about power factor.

Everything I'd read said oversize for growth. In practice, oversizing has a cost nobody prices in: no-load loss. Core loss runs 24 hours a day whether the transformer is loaded or not; load loss scales with the square of the load. Under the DOE efficiency rules in 10 CFR Part 431, the efficiency tables assume operation near 50% load. Run a 750 kVA unit at 90 kVA and you're paying for iron losses for 20 years to cover load you might never add.

Ask for a 24-hour load profile, not a peak number.

2. Step-up and step-down transformers — can't I just wire one backwards?

Kind of. Three things bite you.

Taps first. Tap changers usually live on the high-voltage winding, so if you reverse-feed a step-down unit, your taps are now on the output side. That's usable, but it isn't what the nameplate drawing implies, and it's how you end up running 4% high for six weeks without noticing.

Inrush second. Energizing from the low-voltage side means the primary winding has fewer turns and lower impedance, so inrush relative to the winding rating is higher. Your upstream breaker may nuisance-trip. (We had to swap a 100 A breaker for a 150 A one on a temporary reverse-fed setup in 2021. Not a fun call to make.)

Protection third. Grounding changes sides. If the low-voltage winding was solidly grounded and you back-feed it, your ground fault scheme now sits on the primary.

A transformer is electrically bidirectional. The installation around it is not.

3. Pole-mounted or pad-mounted?

Mostly a site question, and often a utility question before it's an engineering question. If the transformer is utility-owned, you usually don't get a vote.

Pole-type units are overhead, cheaper per kVA, take no ground space, and are standard up to a few hundred kVA. Pad-mounted units sit at grade in a locked enclosure — ANSI C57.12.28 covers the tamper-resistance and enclosure integrity requirements for those compartments. Pad-mounted costs more, needs a concrete pad, clearances, drainage, and ideally bollards if anything with a bumper can reach it.

The mistake I made in 2020: assumed a pad-mounted unit in a parking area was fine because it met clearance. It met clearance. It got hit by a truck in February. (Note to self: bollards are not optional near delivery routes.)

If you're in a floodplain, on a tight urban site, or under overhead lines nobody wants to climb near, that decision mostly makes itself.

4. When is a 3-phase dry-type transformer the right call?

When you can't put oil inside the building. That's the honest answer.

NEC Article 450 governs transformer installation, including the restrictions on liquid-filled units indoors — you're looking at fire-rated vaults or listed less-flammable fluids, and most projects don't want that complexity. Dry-type units let you put the transformer on a mezzanine or in an electrical room without a vault.

The trade-offs are real. Dry-types are physically larger for the same kVA, they shed a lot of heat into whatever room they sit in, and they need airflow. Open-wound units don't love dust or humidity; cast-coil handles harsher environments better. Above about 1,000 m elevation, IEEE C57.12.01 calls for derating. Sound levels are covered by NEMA ST 20, and a dry-type in a room next to offices will be heard.

My rule now: if it's indoors and air-conditioned, dry-type. If it's outdoors on a pad, or over a few MVA, liquid-filled usually wins on economics.

5. What makes a photovoltaic distribution transformer different from a regular distribution transformer?

Three things, and they all come from the inverter.

Harmonics. Inverters are nonlinear loads, and harmonic currents cause extra eddy-current and stray losses in windings and structural parts. That's why K-factor ratings exist — K-4, K-9, K-13, and K-20 are common on dry-type units under UL 1561.

Reverse power flow. A PV transformer has to work in both directions, and the utility may require specific winding connections and grounding for the interconnection.

Duty cycle. These units sit at no load overnight and swing hard during the day. That thermal cycling is a design consideration, not a footnote.

Honestly, I'm not confident enough to recite the current DOE efficiency exemption list for PV and drive transformers from memory. It exists, and it has changed. Check 10 CFR Part 431 directly instead of trusting a vendor's summary slide — I've seen two quotes in the same week disagree about it.

6. What do I check before I sign for delivery?

Nameplate against the purchase order. Every line: kVA, primary and secondary voltage, tap configuration, vector group, percent impedance, temperature rise, cooling class, fluid type, sound level. I once accepted a unit that was correct in every way except the connection diagram showed Dyn1 while the yard drawing assumed Dyn11. It didn't matter until six months later when someone tried to parallel it.

Then the paperwork. Routine tests should be documented per IEEE C57.12.90 for liquid-immersed units or IEEE C57.12.91 for dry-type: turns ratio, polarity and phase relation, winding resistance, no-load loss, load loss, impedance, applied potential, induced potential. Liquid-filled units should come with oil test results as well.

Then physical: leaks, nitrogen or dry-air positive pressure, impact recorder, desiccant.

On site, before energizing, read insulation resistance, verify phase rotation, and torque the bus connections. That's a megohmmeter, a clamp meter, and about 20 minutes. Skipping the rotation check means a crane, a truck, and a week when you find out the hard way.

7. Is it worth ordering just one unit?

Yes. And if a supplier makes you feel small for asking, that tells you something.

Single-unit pricing is genuinely higher. Winding setup, routine testing, and freight all get spread over one transformer instead of twenty, and no honest quote can pretend otherwise. That's fine.

What shouldn't change is the test report, the nameplate accuracy, or whether anyone answers your email. In 2019 I ordered one 75 kVA dry-type for a tenant build-out — a rounding error in that vendor's year. They sent the full routine test sheet, called to confirm the tap setting, and shipped on schedule. They've since done about $140,000 of our business.

The vendors who treated my one-unit orders seriously are the ones I still call. Small doesn't mean unimportant. It means potential.

8. Nobody ever told me about percent impedance. Why does it matter?

Because %Z sets your available fault current, and almost nobody checks it before ordering.

Rough math on a 500 kVA, 480 V unit: secondary full-load current is about 600 A. At 5% impedance, available fault current is roughly 12 kA. At 4%, it's about 15 kA. If the downstream panel is rated 10 kA, that difference is the difference between a compliant installation and a replacement project.

It also decides whether you can parallel. Matching voltage ratio and vector group isn't enough — impedance has to be close, within roughly 10%, or the load splits unevenly and one transformer runs hot while the other loafs.

And I'll be honest: I've never fully understood why impedance tolerances vary so widely between manufacturers for nominally equivalent units. My best guess is that it falls out of core and winding design choices more than anyone wants to admit in a quote package.

So ask. Get it in writing on the nameplate spec. And check it against your protection study before the unit ships.