UPS Systems

UPS units are classified by IEC 62040-3 by how they treat the incoming supply. An offline / standby (VFD) unit runs the load straight off the mains and only switches to inverter when the mains fails — cheap, with a few milliseconds of transfer gap, fine for a single PC, wrong for a server. A line-interactive (VI) unit adds an automatic voltage regulator (AVR) that bucks and boosts sags and surges without dipping into the battery — the sweet spot for small networks, routers and point-of-sale. An online double-conversion (VFI) unit continuously rectifies the mains to DC and re-inverts it to a clean sine wave, so the load never sees a transfer at all and is fully isolated from frequency and voltage disturbance.

For data rooms, medical equipment, lab freezers and PLC-controlled production, the answer is almost always online double-conversion. The grid in much of Kenya is not just outage-prone but dirty — sags, spikes, brown-outs and frequency wander that quietly age sensitive electronics. A line-interactive unit rides the sags but still passes the waveform through; only the online topology gives a genuinely conditioned output.

A UPS is rated in two numbers — VA (apparent power) and W (real power) — and undersizing happens when buyers read only the bigger VA figure. The ratio between them is the UPS's output power factor. Older units quoted 0.6–0.7, so a "1,000 VA" unit could only deliver 600 W. Modern IT loads (servers with power-factor-corrected supplies) draw at a power factor near 0.9–1.0, so a UPS must be matched to both the VA and the watt demand, whichever you hit first.

Then there is crest factor — the ratio of peak to RMS current that switch-mode power supplies pull in sharp spikes. A UPS that can't supply the crest current will distort or trip even when the average load looks well within rating. We size against measured load, add headroom for the inrush of any large supplies, and never load a UPS past ~80% of either rating so there is room for growth and for battery-charging current.

The single biggest surprise in UPS ownership is how runtime collapses as load rises. Battery capacity is quoted at a gentle discharge; pull it hard and you get proportionally less out — the essence of Peukert's law. Double the discharge current and you may keep far less than half the runtime. So a UPS that backs a 50% load for 20 minutes will not back a 100% load for 10 — it will manage rather less.

Heat compounds it. Battery life roughly halves for every 8–10 °C above 20–25 °C, and Kenyan comms rooms are often warm. This is why we specify the autonomy at the real load, site the batteries in the coolest practical spot, and schedule capacity tests — a UPS that has never been load-tested is a UPS whose runtime is a guess. When the autonomy needed runs to hours rather than minutes, the honest answer is a generator with a short-runtime UPS to bridge the start, not a battery room sized for the impossible.

Double-conversion isolation is not free — the rectifier/inverter chain dissipates energy as heat, so an online UPS running at 92–96% efficiency wastes a few percent of throughput continuously, and that heat then loads the room's cooling. Over a year on a large UPS that is real money. Eco-mode (and the newer multi-mode designs) bypass the conversion when the mains is healthy, lifting efficiency above 98%, then snap back to full conversion the instant the supply degrades — recovering most of the loss while keeping the protection.

For a facility manager the figure that matters is total efficiency including cooling: every watt the UPS wastes is a watt the air-conditioning must also remove. We weigh eco-mode against the sensitivity of the load — for a hospital theatre we keep full double-conversion; for a general office IT room, multi-mode is the sensible economy.

A single UPS is a single point of failure — and the day it fails or goes into maintenance bypass is the day you needed it. Critical facilities therefore design in redundancy. "N" is exactly enough capacity for the load; N+1 adds one spare module so any one can fail or be serviced with no loss; 2Nduplicates the entire system on independent feeds for the highest tiers. The right level follows the cost of downtime, not the cost of the UPS.

Redundancy only delivers if the rest of the chain respects it: dual feeds into dual-corded equipment, a maintenance bypass so the UPS can be serviced live, and a generator behind it for outages longer than the batteries. We design the UPS, the bypass and the genset transfer as one system — because a redundant UPS fed from a single failed changeover is not redundant at all.