Emergency power systems in hospitals and clinics are only as reliable as the last time someone proved they work. That proof comes from load bank testing — applying a controlled, measurable electrical load to a generator, UPS, or automatic transfer switch and watching how the system behaves under stress. On a medical site, though, the test itself becomes the problem: the load bank sits in a parking garage, a service yard, or a rooftop plant room, often within a few dozen meters of patient rooms, imaging suites, and operating theatres. A conventional load bank howling at 85–90 dBA is not an option at 11 p.m. in a hospital.
Over the past eighteen months our engineering team has run a structured comparison of low noise load banks intended specifically for healthcare sites. We measured acoustic output, harmonic distortion, load step resolution, transient response, and practical site-handling characteristics across three commercially representative units. This article reports what we found, how we measured it, and how to choose between the three configurations based on your site's constraints rather than on the datasheet alone.
Why "Low Noise" Means Two Different Things at a Medical Site
When a biomedical engineer or facilities manager asks for a low noise load bank, they usually mean the obvious thing: it should not be heard. But in a healthcare environment, noise has two separate dimensions, and a unit that is acoustically quiet can still be electrically "loud."
Acoustic noise is the dominant concern. Load banks dissipate energy as heat, and they reject that heat with forced air. Fan noise scales roughly with the cube of rotational speed, so a bank designed to run fans at full tilt produces a broadband roar dominated by blade-pass frequency and motor whine. In a hospital this creates three practical problems: complaints from patients and neighbours, restrictions on when testing can be scheduled, and — most importantly — interference with the very environment the testing is meant to protect. Testing that has to be done at 3 a.m. to avoid complaints introduces fatigue and scheduling risk into a maintenance task that deserves full attention.
Electrical noise matters because a load bank is a measuring instrument as much as a load. If the bank injects harmonic currents, the generator's voltage waveform distorts, the automatic voltage regulator responds to conditions that would never occur in real service, and the test no longer represents the facility's actual load profile. Hospitals carry a large share of non-linear loads — CT and MRI systems, linear accelerators, UPS rectifiers, LED drivers, IT equipment — so the load bank must be able to replicate both clean resistive demand and distorted, power-factor-displaced demand without contributing distortion of its own.
How We Tested: Method and Instrumentation
All three units were tested on the same concrete apron with a 500 mm clearance from any reflective wall and a background noise floor below 34 dBA. Acoustic measurements followed ISO 3744 principles using a Class 1 sound level meter compliant with IEC 61672-1, A-weighted, slow response, with five measurement positions at 1 m and five at 7 m from the enclosure, plus background correction. Reported values are the arithmetic energy average of the positions, rounded to the nearest 1 dBA.
Electrically, each bank was connected to a 1,250 kVA diesel generator through a calibrated power quality analyzer. We logged voltage, current, real and reactive power, total harmonic distortion of current (THD-i) and voltage (THD-v), and neutral current at 1-second intervals. Step resolution was verified by recording the smallest load increment the controller could reliably apply. Transient response was measured by tripping the full load in a single step and timing voltage recovery to within ±2% of nominal.
Two additional criteria turned out to matter more than expected: heat plume behaviour (whether the discharge air re-enters the generator's own radiator intake and causes derating) and cable management, because flexible cabling at 400–800 A is where most site safety incidents actually occur.
The Three Load Banks We Tested
Unit 1 — 400 kW Air-Cooled Resistive/Reactive Bank with EC Fans
The first unit represents the sweet spot for most district hospitals and ambulatory surgery centres. It pairs a resistive element stack rated 400 kW continuous with a 130 kVAr reactive section, giving a selectable power factor range of 0.8 lagging through unity to 0.8 leading. The defining feature is its fan pack: electronically commutated (EC) motors with closed-loop speed control that continuously modulate airflow to match the actual heat being dissipated rather than running at a fixed speed.
At full 400 kW load the unit measured 62 dBA at 1 m and 54 dBA at 7 m. At 25% load — the condition most relevant to routine monthly generator exercise — the fans dropped to roughly 40% speed and the unit measured 51 dBA at 1 m, which is quieter than the hospital's own generator radiator fan. That distinction matters: on most sites the generator itself becomes the noise floor, and a load bank that sits below it adds nothing to the acoustic footprint of the test.
Step resolution was 1 kW on the resistive side, with 5 kVAr increments on the reactive side, applied by a PLC controller through contactor banks with zero-crossing switching. THD-i contribution at full load was 2.1%, with THD-v held under 3% on the generator terminals. Transient recovery after a full-load dump was 60 ms, well inside the 100 ms window most generator AVRs specify.
Strengths: genuinely quiet at partial load, single-point lift and integral fork pockets, 1 kW resolution for accurate 25%/50%/75% NFPA 110 load steps, and no water supply required. Limitations: 400 kW ceiling, and outdoor installation requires attention to wind direction because the discharge plume is significant at full load.
Unit 2 — 1 MW Liquid-Cooled Load Bank
The second unit takes the opposite approach to noise: it removes the fan problem almost entirely by dissipating energy into a closed-loop water circuit with a remote dry cooler. The resistive elements are immersed in a dielectric-free water jacket, and the only significant noise sources are the circulation pump and the remote cooler fans, which can be located up to 40 meters away — outside the sensitive boundary.
Measured output at the load bank enclosure was 55 dBA at 1 m and 47 dBA at 7 m at full 1,000 kW, which is close to the practical limit of what our instrumentation could resolve above background in a live hospital yard. Step resolution was coarser at 5 kW, a limitation of the solenoid valve and element grouping architecture, and transient recovery was the slowest of the three at 80 ms.
The trade-off is infrastructure. The unit needs a water source, a means of dealing with the heat rejection on the remote cooler, and freeze protection if it stays outdoors through winter. On one site we tested, the remote cooler had to be placed on a service road, which required a temporary traffic management plan. When that planning overhead is acceptable — typically at large tertiary hospitals, data-centre-adjacent facilities, or sites with a permanent test pad — this is the quietest option available by a comfortable margin.
Strengths: lowest acoustic signature, 1 MW capacity in a compact footprint, minimal hot air discharge near the generator. Limitations: water supply and plumbing, 5 kW step granularity, higher maintenance burden, longer setup time.
Unit 3 — 2 MW Containerised Modular Bank with Variable-Speed Fans
The third unit is the heavy-lift option: a 20-foot containerised bank built from modular 500 kW resistive cassettes plus a 500 kVAr reactive module, with axial fans driven through variable frequency drives. Capacity is configurable from 500 kW to 2 MW by enabling cassettes, which lets a single asset serve both routine hospital exercise testing and the annual full-load commissioning of a new 2 MVA standby plant.
Noise performance depends entirely on how the unit is dispatched. At 2 MW with fans at 100%, we recorded 68 dBA at 1 m and 58 dBA at 7 m — louder than Unit 1 in absolute terms, but remarkable for the capacity. Running at 25% load with the VFDs at minimum, the same unit measured 53 dBA at 1 m. In other words, capacity and quiet are not in conflict here; the noise only appears when you actually ask for the megawatts, which on a medical site is a rare, planned, and usually daytime event.
Electrical performance was strong: 1 kW resistive steps, THD-i of 2.4% at full load, and the fastest transient recovery of the group at 50 ms thanks to short, heavy busbar runs inside the container. Cable entry through a rear gland plate with a copper busbar interface eliminated the flexible-cable congestion that plagued the other two units at high current.
Strengths: enormous headroom, modular capacity, best cable interface, excellent step resolution across the whole range. Limitations: requires a crane or heavy tractor unit for positioning, needs 40–60 m² of hardstanding with clearance for both air intake and discharge, and is over-specified for a 500 kW clinic.
Test Results at a Glance
| Parameter | Unit 1 (400 kW air-cooled) | Unit 2 (1 MW liquid-cooled) | Unit 3 (2 MW containerised) |
|---|---|---|---|
| Sound pressure, 1 m, full load | 62 dBA | 55 dBA | 68 dBA |
| Sound pressure, 7 m, full load | 54 dBA | 47 dBA | 58 dBA |
| Sound pressure, 1 m, 25% load | 51 dBA | 49 dBA | 53 dBA |
| Resistive step resolution | 1 kW | 5 kW | 1 kW |
| THD-i contribution, full load | 2.1% | 1.8% | 2.4% |
| Transient recovery after load dump | 60 ms | 80 ms | 50 ms |
| Power factor range | 0.8 lead to 0.8 lag | Unity to 0.8 lag | 0.8 lead to 0.8 lag
Continue Reading |
