- Myth 1 — "Same kVA on the nameplate means same usable capacity."
- Myth 2 — "A bigger engine just burns more fuel, so the SDMO is cheaper to run."
- Myth 3 — "Protection is protection; the controller is just a screen."
- Myth 4 — "We'll never parallel, so paralleling capability is wasted money."
-
The reality behind the four myths
Four Beliefs That Mis-Size a Hospital Standby Set: Cummins QSK vs KOHLER-SDMO D830 at 750 kW
A 600-bed hospital needs a standby plant that can hold the essential-systems bus through a utility outage and a generator that the estates team can actually live with for twenty years. At that load the honest comparison is a Cummins QSK in its lower band against a KOHLER-SDMO generator D830 — the SDMO unit rated about 750 kW prime / 825 kVA standby, the top of its D-series. Both are credible. But four beliefs that get repeated in tender meetings quietly push the sizing in the wrong direction. Here is each one, and the quantified tradeoff it hides.
Myth 1 — "Same kVA on the nameplate means same usable capacity."
"The D830 is 825 kVA standby and the Cummins generator set is 825 kVA standby, so on the essential bus they're interchangeable."
Mechanism
Usable capacity is not the nameplate number; it is the nameplate number minus everything the engine has to reject as heat plus everything the package consumes to stay alive. Heat rejection at this size is the sum of three streams — jacket water, charge-air (after-cooler), and radiator airflow — and the radiator fan that moves that air is a parasitic load drawn off the same shaft that feeds the alternator. Add alternator copper and iron losses and you get the gap between rated kW and the kW that reaches the switchboard.
Cummins specifies its QSK gensets at standby and prime ratings the way the standard intends, and the lower-band QSK reaches this class on a large, slow-turning displacement at 1500 RPM under ISO 3046, which lets the after-cooler and jacket-water loops run with thermal margin to spare. The D830 reaches the same kVA from a smaller, faster engine working closer to its thermal limit, so a larger share of fuel energy leaves as heat the cooling system has to chase.
Worked consequence — the derating you only see in August
Site the plant in a rooftop enclosure where summer intake reaches roughly 45 °C and the building sits at modest altitude. A diesel genset typically loses on the order of 1–1.5% of rated output per 5 °C above 40 °C and a similar slice per few hundred metres of altitude (illustrative, standard ISO 3046 derate behaviour). On a unit running with comfortable cooling margin that derate leaves usable output near the bus demand. On a unit already close to its thermal ceiling, the same ambient pushes the cooling system into limp before the alternator is full — the controller pulls power back to protect the engine.
Tradeoff in numbers: if your essential bus is 700 kW and you buy an 825 kVA set with a thin cooling margin, a 6–8% hot-day derate can drop usable output to roughly 660–700 kW — you are at or over the line on the worst day, which is exactly the day the utility fails in a heatwave. The set with margin holds 720–740 kW through the same ambient. Size on usable hot-day kW, not nameplate kVA.
When this reverses: in a climate-controlled basement plant room held at 25 °C with generous louvres, neither unit derates meaningfully and the nameplate-equals-nameplate intuition is essentially correct. The myth is only a trap where ambient and enclosure restriction eat the margin.
Myth 2 — "A bigger engine just burns more fuel, so the SDMO is cheaper to run."
"The Cummins displacement is huge; of course it drinks more diesel than a tighter SDMO."
Mechanism
Fuel burn is not set by displacement; it is approximately load multiplied by brake-specific fuel consumption (bsfc) at that load point. A hospital standby set does not live at full load — it lives on whatever fraction of the bus is energised, often 30–60% of rating during a real outage. The question that decides the fuel bill is therefore the shape of the bsfc curve at part load, not the size of the engine at full load.
A large, lightly stressed displacement with full-authority common-rail injection (Cummins MCRS) tends to hold a flatter bsfc curve down into part load, because each cylinder's injection quantity and timing are trimmed continuously. A smaller engine driven harder to make the same kVA, with a more conventional fuel system, typically climbs its bsfc curve faster as load drops below about half.
Worked consequence — the part-load fuel bill
Take a generator-transfer-test-plus-monthly-exercise regime, say 80 outage/test hours a year at an average 45% load on a 750 kW-class set — roughly 340 kW delivered. If the well-loaded large engine sits near 0.215 kg/kWh at that point and the harder-worked engine sits near 0.235 kg/kWh (illustrative bsfc figures), the burn is about 73 vs 80 kg/h. Over 80 h that is roughly 560 kg of diesel difference a year — small in isolation.
Tradeoff in numbers: the fuel delta is real but modest at standby duty. The myth only bites if you let it drive the capital decision. If the set runs as standby (tens of hours/year), fuel is a rounding error and should not pick the unit; if it ever runs prime or extended-outage (hundreds of hours), the flatter part-load curve compounds and the fuel argument flips toward Cummins, not away from it.
When this reverses: at sustained high load — a microgrid or a long grid-down event where the set runs near 80% for days — both engines sit in the efficient part of their curves and the per-kWh gap narrows. There the "bigger burns more" intuition becomes roughly true in absolute litres, though not in litres-per-kWh delivered.
Myth 3 — "Protection is protection; the controller is just a screen."
"Both have a digital panel with metering and breaker control, so the protection is equivalent."
Mechanism
The screen is not the protection; the relay logic behind it is. Cummins ships PowerCommand 3.3 with AmpSentry, an integrated protective-relay function matched to the alternator's actual thermal-damage curve, plus automatic remote start/stop and built-in paralleling logic. KOHLER-SDMO's D-series uses the APM-family panel (APM403 on the larger units) for manual/auto control, metering and breaker management. Both meter and both control; what differs is how the system behaves when the alternator is asked to clear a downstream fault.
On a hospital essential bus, selective coordination matters: a fault on one branch must trip that branch's breaker, not collapse the whole set. That requires the generator to hold fault current long enough for the downstream device to clear, while still protecting the alternator from cooking. A relay tied to the machine's damage curve can ride through the clearing time; a more generic over-current cut-out may trip the source first, dropping the entire essential bus to save one circuit.
Worked consequence — the fault that takes the whole floor dark
Imagine a bolted fault on an isolation-transformer feeder in an operating-theatre suite. With source-matched protection, the set sustains fault current through the few cycles the branch breaker needs, that breaker clears, and the rest of the essential bus never notices. With protection that trips the source on the same fault, the generator opens its main — and every theatre, not just the faulted feeder, loses the bus until the set is reset.
Tradeoff in numbers: the capital difference between the two control philosophies is a low single-digit percentage of set cost. The operational difference is the span between "one feeder out" and "every life-safety branch on that bus out during a fault." For a coordinated life-safety bus, buy the source-matched relay; for a single non-critical load with no selectivity requirement, a generic cut-out is acceptable and cheaper.
When this reverses: if the standby set feeds one simple radial load — a single pump house, a cold room — there is nothing to coordinate, the alternator-matched relay protects no extra value, and APM-class control is perfectly sufficient. Selectivity only earns its keep where a bus fans out to many independently-protected branches.
Myth 4 — "We'll never parallel, so paralleling capability is wasted money."
"One set, one bus. We're not building a power station, so isochronous paralleling is a feature we'll never use."
Mechanism
Paralleling is not only about running two sets at once; it is the same control fabric you need for the day the hospital grows. Cummins builds isochronous load-sharing and black-start into the standard PowerCommand platform, scaling from a single set to arrays of 2 MW up to 20 MW and beyond in N+1 or 2N. Designing the first set on that platform means the second set bolts onto the bus without re-engineering the controls.
Isochronous control also matters for a single set the moment you ever need concurrent maintainability: to service the only generator without a planned outage, you roll up a portable set and parallel it for the changeover. A platform built to parallel does that cleanly; one that is not has to drop the bus to swap sources.
Worked consequence — the expansion you didn't price
A hospital that adds a new wing in year seven typically needs N+1 redundancy on the enlarged essential bus — a second set sharing load isochronously so either can be serviced live. If the first set was chosen on a non-paralleling platform, the year-seven project pays to retrofit synchronising gear and controls, or to replace the original set entirely. If it was chosen on the paralleling platform, the second set commissions onto the existing bus.
Tradeoff in numbers: paralleling-ready control adds little to the first set's price. Retrofitting it later, or scrapping a serviceable set because it cannot share a bus, is a five-to-six-figure penalty against a capability that was nearly free at purchase. If there is any realistic expansion or concurrent-maintenance requirement in the asset's life, buy the paralleling platform; if the hospital footprint is genuinely fixed forever, the capability is indeed surplus.
When this reverses: a small fixed-scope facility with a guaranteed single set and an acceptable planned-outage maintenance window genuinely never parallels. There the integrated paralleling fabric protects no future value and the simpler platform is the rational buy.
The reality behind the four myths
| Belief | What actually decides it | Buy-side rule |
|---|---|---|
| Equal kVA = equal capacity | Hot-day usable kW after derate & parasitics | Size on worst-ambient usable kW |
| Bigger engine = costlier fuel | Part-load bsfc curve shape, not displacement | Standby: ignore fuel; prime: favour flatter curve |
| Protection is equivalent | Source-matched relay vs generic cut-out | Coordinated bus needs alternator-matched relay |
| Paralleling is wasted | Expansion & concurrent maintainability | Buy the platform if any growth is plausible |
None of these makes the D830 a poor machine; it is a sound, soundproofed, serviceable unit, and on a single climate-controlled radial load it can be the right and cheaper choice. The point is narrower: at the 750 kW hospital tier, the decision is not won on the nameplate. It is won on usable hot-day kW, the part-load fuel curve, the protection philosophy of the bus, and whether the asset will ever need a second set. Decide those four, and the kVA number on the front of the cabinet stops doing your sizing for you.
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. Cummins is a brand affiliated with this site; competitor names are used for identification only.
