UPS Transfer Time: Why a Few Milliseconds Usually Don't Matter (and When They Do)

What the number on the spec sheet measures

When utility power fails, a standby or line-interactive UPS has to do three things before the inverter carries the load: notice the failure, open its input relay (so battery power does not flow back out into the dead utility line), and connect the inverter output. The total is the transfer time.

Detection is the larger and more variable part. The UPS must distinguish a real failure from a brief disturbance, and how quickly it can do that depends on where in the AC cycle the failure happens. At 60 Hz one full cycle lasts 16.7 ms and a half cycle 8.3 ms. A failure near a zero crossing is harder to spot quickly than one near a peak, because the voltage is supposed to be near zero anyway. That is why many spec sheets give a typical figure and a higher maximum.

Hold-up time: the other half of the equation

Transfer time only matters relative to how long your equipment can survive without input. For computers, that is the power supply's hold-up time: how long its DC outputs stay within tolerance after AC input disappears.

A modern PC power supply with active PFC first converts AC to a high-voltage DC rail, typically around 380 to 400 V, stored on a large bulk capacitor. The DC-DC stage then converts that to 12 V, 5 V and 3.3 V. When AC input vanishes, the DC-DC stage keeps working from the energy in the bulk capacitor until the voltage drops too low to regulate.

The ATX12V design guide has long been cited as requiring roughly 16 to 17 ms of hold-up at full load, alongside a power-good signal that warns the motherboard shortly before the outputs fall out of regulation. Treat that as a design target, not a guarantee: requirements have been revisited in newer revisions of the guide, real products vary, and capacitors lose capacitance as they age.

Worked example: estimating hold-up from capacitor energy

The energy available from a capacitor as it discharges from V1 down to V2 is:

Take an illustrative 650 W power supply with a 470 µF bulk capacitor charged to 390 V, whose DC-DC stage can keep regulating down to about 300 V, and which is about 90% efficient.

1. Usable energy: 0.5 x 0.000470 x (390² - 300²) = 0.5 x 0.000470 x 62,100 ≈ 14.6 J.
2. Power drawn from the capacitor at full output: 650 W / 0.90 ≈ 722 W.
3. Hold-up at full load: 14.6 J / 722 W ≈ 0.020 s, about 20 ms.

Now repeat at other loads with the same capacitor:

The pattern is the point: hold-up time scales roughly inversely with load. An idling PC can ride through gaps many times longer than any UPS transfer. A PC running near its power supply's limit has the least margin, and that margin shrinks further as the capacitor ages and loses capacitance.

Why online units have zero transfer time

In an online double-conversion UPS, the inverter supplies the load all the time from a DC link. The rectifier feeds that link from utility power, and the battery is connected to it too. When utility power fails, the rectifier stops contributing and the battery carries the DC link instead. The inverter does not notice any change, so the output continues without a gap.

Two exceptions are worth knowing:

• Bypass transfers. If the inverter is overloaded or faults, the unit switches the load to utility power through a static bypass switch. These transfers are typically very fast, often well under a quarter cycle, but they are not zero, and while on bypass the load has no battery protection.
• Eco or high-efficiency mode. The unit runs the load on bypass to save energy and switches to the inverter when it detects a problem. In this mode it behaves more like a line-interactive unit, with a short transfer. Check how your model implements it before enabling it on sensitive loads.

When transfer time actually matters

Use this checklist. If none apply, transfer time is not a deciding factor for your purchase.

Two problems often get blamed on transfer time but have other causes. A PC that drops at transfer only on a simulated sine wave UPS is usually reacting to the waveform. A PC that drops after a few seconds on battery is usually seeing an overload or a weak battery. See UPS shuts off during outage for the diagnostic sequence.

Testing your own setup

A spec sheet cannot tell you about your specific power supply's condition. A practical test takes ten minutes:

1. Save your work and close anything that writes to disk heavily. Assume the PC might shut off.
2. Load the system realistically. Run the game, render or benchmark that represents your heaviest normal use, and note the load on the UPS display or in its software.
3. Trigger a transfer. Use the UPS self-test, or switch off the upstream breaker or a switched outlet strip feeding the UPS. Pulling the plug also works, but it disconnects the ground path, so the other methods are preferred.
4. Watch for a reboot or blank screen during the transfer and for 30 seconds afterward.
5. Restore power and confirm the UPS returns to utility mode and starts recharging.
6. Repeat at idle if it failed under load. Failure only under load points to hold-up margin, waveform or capacity rather than a dead battery.

Make this part of routine checks; the self-test and runtime calibration guide and the maintenance checklist cover scheduling it.

Reading the spec sheet

Look for a line labeled transfer time, switchover time or typical transfer time. Good spec sheets give both a typical and a maximum value. If a product does not state a transfer time at all, or describes it only as "fast", treat that as a reason to prefer another model for a desktop PC. For the other lines on the sheet, see UPS specifications explained.