How Does a UPS Work? Inside Double Conversion

A double-conversion UPS works by converting incoming AC to DC, holding that DC on an internal bus, and converting it back to AC to feed the load. The battery connects to the same DC bus. Because the load is always fed from the inverter, a mains failure removes the input to the rectifier but changes nothing the load can see, which is why the transfer time is zero rather than merely short.

The path the power takes

Follow a single watt through the unit and the design explains itself.

Mains AC arrives at the rectifier, which converts it to DC and regulates the voltage on the DC bus. The inverter takes DC from that bus and synthesises a clean AC waveform for the load. The battery is connected to the bus through a charger, kept topped up while the mains is present.

When the mains fails, the rectifier stops contributing and the battery immediately supplies the bus instead. The inverter carries on doing exactly what it was doing. Nothing switches, nothing transfers, and the load never sees an event. This is the single most important consequence of the topology.

Why there is a bypass, and when it is used

The dashed path in the diagram is the static bypass. It connects the input directly to the output, and it exists for two situations.

The first is a fault or an overload. If a downstream short circuit draws more current than the inverter can supply, the UPS transfers to bypass within milliseconds so that the mains, which can deliver far more fault current, clears the protective device. Without a bypass, the inverter would current-limit and the faulty circuit would take down the whole load rather than just itself.

The second is maintenance. A maintenance bypass, usually a separate manual arrangement rather than the static one, lets the UPS be isolated and worked on while the load continues to run from raw mains. Any installation above a few tens of kW should have one, and adding it after the switchboard is built is expensive.

What happens in the first fifteen seconds of an outage

1. t = 0. The mains fails. The rectifier stops supplying the DC bus.
2. t = 0. The battery takes over the bus. There is no measurable interruption at the output, and no transfer to detect.
3. t = 1 to 3 s. The generator, if there is one, receives a start signal and cranks.
4. t = 5 to 15 s. The generator reaches stable voltage and frequency, and the transfer switch moves the building to generator supply.
5. t = 15 s onward. The rectifier accepts generator power, resumes supplying the bus and begins recharging the battery.

This sequence is why five to fifteen minutes of autonomy is a normal specification on a site with a generator. The battery is not there to run the room; it is there to cover the gap in step three and four, with margin for a generator that fails to start on the first attempt.

Efficiency and eco mode

Converting twice costs something. Current three-phase designs run at or above 96 per cent in full double conversion, meaning roughly four per cent of the input becomes heat that you pay for twice: once to generate it, once to remove it with cooling. At 400 kW, one efficiency point is around 4 kW.

Most units offer an eco or high-efficiency mode in which the load runs from the bypass while the mains is within tolerance, with the inverter on standby. Efficiency rises to around 99 per cent. The trade-off is that a transfer now exists, typically under two milliseconds, and the load loses the conditioning that double conversion provides. Whether that is acceptable depends on the equipment and on the quality of the incoming supply.

What the battery is doing while nothing happens

For almost all of its life a UPS battery is float-charged and idle, and it degrades anyway. Valve-regulated lead-acid cells lose capacity with temperature and with each discharge cycle, and a string that tested fine two years ago may not carry the load today. This is why battery monitoring and periodic discharge testing exist, and why a UPS that has never been tested under load is an assumption rather than a protection.

Lithium-ion changes the arithmetic. It tolerates higher temperatures, occupies roughly half the footprint for the same energy, lasts two to three times longer and reports its own state of health. It costs more up front, and the comparison that decides it is total cost across the life of the installation rather than price per kWh.

Parallel operation and redundancy

Two or more units can be paralleled, and it is worth being precise about what that buys. Paralleling for capacity means the combined output carries a load neither unit could carry alone. Paralleling for redundancy means each unit could carry the load alone, so one can fail or go into maintenance with no effect.

These are different purchases at different prices, and the decision has to be made before the single-line diagram is drawn. In a modular frame the same idea applies at module level, which is why N+1 in a modular UPS costs one module rather than a second system.

Sizing, in the order that works

Step four catches more mistakes than the other three combined. A 30 kW UPS cannot be fed from a 32 A single-phase circuit, and that is discovered at installation rather than at quotation more often than it should be.

Related pages

Our UPS systems page lists the families we supply and the capacity each covers. If the room also needs source transfer or rack-level distribution, transfer switches and rack PDUs are the two stages either side of the UPS in the same power chain.