UPS systems: what they do and what to check before specifying one

A UPS exists to cover a gap in time. When the utility fails, a standby generator needs somewhere between several seconds and a minute to start, stabilise and accept load. Nothing in a data hall tolerates that gap, so the UPS carries the critical load across it from stored energy. Everything else a UPS does, and there is a fair amount, follows from that one job.

Topology decides what you are actually protected against

The three common arrangements protect against different things, and the difference matters more than the badge on the front.

In a standby arrangement the load runs on utility power, and the inverter starts when the utility falls outside limits. There is a brief transfer, and conditioning while running normally is limited. This suits a single workstation or a small network cabinet, not a room.

A line-interactive unit adds voltage regulation, so it corrects sustained high or low voltage without moving to battery. On a feeder that sags under load this avoids discharging the battery several times a day, which matters because every discharge costs battery life.

In double conversion the load is fed continuously from the inverter. Incoming AC is rectified to DC and inverted back to AC, so the output is rebuilt rather than passed through. There is no transfer time to speak of when the utility fails, because the load was never on the utility directly. Input disturbances are isolated from the output. This is the arrangement almost all three phase data-centre systems use, and it is what people usually mean by “online”.

None of these is better in the abstract. A room with a generator and a clean feed needs different protection from a remote site on a long rural line.

Read the nameplate carefully

Two figures on a UPS nameplate are routinely confused, and the confusion produces systems that cannot carry the load they were bought for.

The kVA rating is apparent power. The kW rating is real power, and it is the one your equipment consumes. The ratio between them is the power factor the manufacturer assumed. Older systems were commonly rated at 0.8, so a 100 kVA unit delivered 80 kW. Many current systems are rated at unity, where the kVA and kW figures are the same number. If you size from a kVA figure and your load is measured in kW, you can be a quarter short before anything is installed.

Size from the real measured load rather than the sum of equipment nameplates, because nameplate totals overstate consumption substantially. State expected growth separately from present load, so the headroom is a decision someone made rather than an accident of rounding.

Runtime is a battery question

The UPS determines how the load is fed. The battery determines how long. Runtime falls away faster than load rises: a battery that supports a given load for some period will support twice that load for considerably less than half the time, because the discharge is less efficient at higher rates. This is why a runtime figure is only meaningful when quoted at a stated load, and why the honest source for it is the manufacturer runtime chart for that model and that battery configuration rather than an arithmetic estimate.

IEEE 1184 is the reference for selecting and sizing the battery system, and it covers installation, maintenance and testing as well as selection. It is worth reading before accepting a battery proposal, because it sets out the questions a sizing calculation should have answered.

The Canadian context

A UPS installed in Canada is a product covered by CSA C22.2 No. 107.3, the harmonized standard developed with UL as UL 1778. It applies to movable, stationary, fixed and built in systems for distribution up to 600 V AC. Certification to it is the baseline expectation for equipment going into a Canadian facility, and it is a reasonable thing to ask a vendor to evidence.

The installation itself falls under the Canadian Electrical Code, Part I, currently CSA C22.1:24, and under whatever the authority having jurisdiction requires locally. Battery rooms and energy storage in particular attract requirements that have been revised in recent editions, so a design checked against an older edition is worth rechecking rather than assumed to still comply.

What actually goes wrong

Three things account for most of what takes a UPS out of service, and none of them is the inverter.

Batteries age out. They are a consumable with a service life measured in years, and that life is strongly temperature dependent. A room run warm to save on cooling will consume battery life faster than the design assumed. Impedance readings taken regularly and compared against the commissioning baseline are how a weak block is found before it is discovered during an outage.

Capacitors and fans reach the end of their service life. Both are wear items with published intervals. Neither fails gracefully, and both are cheap relative to the load they protect.

The bypass is misunderstood. A static bypass carries the load when the inverter cannot, which is a protection. A maintenance bypass isolates the UPS for work, and while it is engaged the load is on raw utility power with no protection at all. Knowing which one is engaged, and what it costs you, belongs in the operating procedure rather than in an engineer’s head.

Repair or replace

A UPS whose batteries are at end of life is not a UPS at end of life. Battery replacement is routine and is the expected maintenance event, not a failure of the system. Replacement of the unit becomes the reasonable decision when parts are no longer available for the model, when the capacity no longer matches the load after growth, when the topology is wrong for the protection now required, or when the cost of returning it to a known good condition approaches the cost of a current system with a warranty and a supported parts path.

The one thing worth resisting is deciding this during an outage. The decision needs the load readings, the battery history and the parts availability, and none of those can be gathered while the room is down.