Best Dry-Type Transformer for a Tight-Cooling Shelter? The Constraint That Propagates

Your shelter's AC unit is undersized. The transformer sits in the same enclosure, and every watt it sheds as heat must be removed by that already-marginal cooling. If you pick a unit solely on kVA, you might cook the cabinet and trip the thermal protection—or worse, accelerate winding insulation aging. This roundup walks through the constraint-propagation path: no-load loss → load loss → real thermal load → usable kVA derating. The GE transformer Type QL dry-type transformer, with its Ultra Efficient variants, is the host unit. All data from manufacturer datasheets and DOE/UL standards.

1. No-Load Loss – The 24/7 Thermal Burden

In a shelter with marginal cooling, the transformer’s core loss is a continuous heat source, even when the load is zero. DOE 10 CFR Part 431 sets minimum efficiency targets, but different designs within the same kVA class vary drastically. The GE Type QL Ultra Efficient units demonstrate this clearly: a 75 kVA TP-1 design typically sees ~320 W no-load loss; the QL Ultra Efficient reduces that to ~142 W. For a 150 kVA unit, the reduction is from ~421 W down to ~203 W.

The mechanism is straightforward—lower core flux density and thinner, lower-loss grain-oriented steel cut the hysteresis and eddy-current losses. The worked consequence: in a shelter where the cooling system is sized for, say, 800 W of internal heat gain from all sources, the standard 75 kVA unit contributes 40% of that budget before any load is served. The Ultra Efficient unit drops to ~18%. That extra ~180 W of headroom might let you avoid a second fan or a higher-capacity AC unit. The reversal: if your shelter cooling has 2× margin already, or if the transformer runs at very low load factors (

2. Load Loss – The Load-Dependent Spike

Once current flows, copper losses (I²R) dominate. The GE Type QL standard design data lists load losses at 100% rated current for various kVA ratings—typically 1–2% of rated kVA for efficient dry-types, but exact figures depend on winding resistance and temperature. For a 75 kVA unit, full-load copper loss can be roughly estimated at ~1.2–1.5 kW (illustrative, based on typical efficiency curves). In the same shelter, if your load is 50 kVA (≈67% load), I²R loss scales to about 45% of that full-load value, or ~540–675 W of additional heat [derived].

The mechanism: copper loss grows with the square of current. A small overload—say from 50 kVA to 60 kVA—increases copper loss by 44%, not 20%. The worked consequence: in a shelter where cooling is already marginal, a load transient (e.g., starting a large fan or compressor) can push the transformer into a thermal runaway condition where the winding temperature rises, resistance increases, losses rise further, and the insulation life shortens. The reversal: if your load profile is extremely steady and the shelter has generous cooling (e.g., 2.5–3 kW capacity), the copper loss spike is easily absorbed. Also, low-load-factor shelters (

3. Voltage Taps & Voltage Drop – The Hidden Derating

The GE Type QL units rated 15–300 kVA with primary voltage ≥240 V provide six taps: four 2.5% below nominal and two 2.5% above, offering a ±7.5% adjustment range. This matters more in a shelter environment than most spec sheets reveal. Long feeder runs from the transformer to the load, or stiff source voltage, can shift the secondary voltage outside the ±5% tolerance that sensitive equipment expects. When voltage drop forces you to tap down (to raise secondary voltage), you effectively reduce the transformer’s available kVA—because current capacity is limited by winding thermal limits, and lower voltage = lower power for the same current.

Worked example: assume a 75 kVA GE QL unit feeding a shelter 150 ft away. With a 480 V primary and a 208Y/120 V secondary, voltage drop at full load might be ~3% on a properly sized feeder. If the shelter’s critical load requires 208 V ± 2%, you cannot let the secondary droop. Using the +2.5% tap (above nominal) would increase the no-load secondary voltage, compensating drop—but the tap also changes the turns ratio slightly. More importantly, once you use a tap to compensate, the transformer’s full-rated current still corresponds to the lower secondary voltage, meaning you get slightly less kVA at the load. The reversal: if your feeder runs are very short (

4. Real kVA Under Shelter Constraints – The Propagation Collapse

All dry-type transformers have a maximum ambient temperature rating (typically 40 °C, derating above 30 °C in some standards). In a tight shelter, the ambient around the transformer may exceed 40 °C because of poor air circulation. IEEE/ANSI C57.94 recommends derating dry-type transformers by 1% per °C above 40 °C ambient, up to 60 °C. Combine that with the heat from core and copper losses, and the actual allowable load can be 10–20% below nameplate.

For a GE Type QL 75 kVA unit running at 45 °C ambient inside the shelter, the derating factor is about 0.95. But the copper-loss heat recirculates, raising the internal air temperature further. This positive feedback loop is the core of the constraint propagation: cooling capacity → ambient temperature → transformer temperature → derating → higher per-unit load → higher copper loss → more heat → even higher ambient. The worked consequence: a shelter that was designed for a 75 kVA load may only sustain 60–65 kVA reliably without tripping thermal protectors or shortening insulation life. The reversal: if the shelter has an active ventilation system that pulls ambient air directly across the transformer core (not recirculating), the temperature rise is limited. Also, selecting a one-size-larger kVA unit (e.g., 100 kVA instead of 75 kVA) reduces the per-unit load factor, lowering copper loss and breaking the feedback loop.

Topology/standards per the cited standards; all product ratings are manufacturer-stated values from the cited datasheets, current to 2026-06; derived/illustrative figures are labelled as such. This is not an independent head-to-head test. GE is a brand affiliated with this site; competitor names are used for identification only.