A client called us after their third UPS failure in two years. Two units had been replaced under warranty, and the supplier was suggesting the model was unsuitable. It was not. The problem was a neutral-to-earth voltage of eleven volts caused by an earthing arrangement that had never been correct, and it had been quietly destroying the input stage of every unit installed there.
UPS units do fail. But when the same site kills several in a row, the cause is almost always in the four items below, and every one of them is measurable.
1. Earthing and bonding
This is the most common cause we find and the least often checked. Symptoms include random shutdowns with no logged fault, communication cards failing repeatedly, and inexplicable data corruption on connected equipment.
What to measure. Neutral-to-earth voltage at the UPS input under load. Anything above two volts warrants investigation, and above five volts is doing damage. Also measure earth resistance at the electrode, and confirm there is a single bonding point rather than several competing earths.
In multi-tenant buildings we frequently find the server room bonded to a different earth from the rest of the floor, which creates a potential difference that current happily flows across, usually through your data cabling.
2. Harmonic distortion and generator interaction
Modern switch-mode power supplies draw current in pulses rather than smoothly, which distorts the waveform. A UPS on a generator sees this twice over: the generator's alternator has a much higher source impedance than the grid, so the distortion it produces under a non-linear load is significantly worse.
Symptoms. The UPS refuses to accept generator power, or accepts it then drops to battery repeatedly during a mains failure, which is precisely when you need it most. Sites often discover this during their first real outage rather than during commissioning.
What to measure. Total harmonic distortion of voltage at the UPS input, on mains and on generator, under realistic load. Above five per cent needs attention. Check the generator is sized for the non-linear load, not just for the kilowatt total, which usually means oversizing it by a factor of one and a half to two.
3. Battery temperature
Valve-regulated lead-acid battery life halves for roughly every eight to ten degrees Celsius above twenty-five. A battery cabinet in a room running at thirty-five degrees will not deliver a five-year design life. It will deliver about two, and it will do so without warning because capacity degrades gradually and then collapses.
Symptoms. Runtime that has quietly fallen from thirty minutes to four. The UPS reports the batteries as healthy right up until the load transfers and it drops immediately.
What to do. Monitor battery cabinet temperature separately from room temperature, because the cabinet is usually warmer. Run a real discharge test annually rather than relying on the self-test, which typically checks voltage under a brief load and tells you very little. Track internal impedance if the platform supports it, as it is a far better early indicator than voltage.
4. Load profile and sizing
Two failure patterns here, and they are opposites.
Overloaded. The room has grown. Nobody recalculated. The unit spends its life near capacity, runs hot, and the electronics age accordingly. Meanwhile the runtime everyone believes in was calculated for a load that has since doubled.
Underloaded. A 60 kVA unit protecting an 8 kVA load is inefficient, and its batteries can behave unpredictably at very low discharge rates. Oversizing "for the future" carries a real cost, which is one of the strongest arguments for modular units that grow with you.
What to measure. Actual load over a full week, including the peak when everything starts at once after an outage. Inrush is what trips units, not steady state.
The commissioning test people skip. Pull the mains under full production load and watch what happens end to end: does the UPS transfer cleanly, does the generator start and take the load, does the UPS accept generator power without kicking back to battery, and does everything come back cleanly when mains returns? Do it at a planned time with everyone present. Discovering the answer during an unplanned outage is considerably more expensive.
A survey checklist
- Neutral-to-earth voltage at the UPS input, under load
- Earth electrode resistance and a check for multiple bonding points
- Voltage total harmonic distortion, on mains and on generator
- Generator sizing against the non-linear load, not just the kilowatt figure
- Battery cabinet temperature logged over at least a week
- A real discharge test, with the result recorded
- Measured load profile including post-outage inrush
- Maintenance bypass present, labelled and actually tested
- Battery installation date recorded and a replacement date budgeted
- UPS alarms reaching a monitoring system, not just an LCD panel nobody looks at
What this costs you to ignore
A replacement UPS is an expense. A UPS that fails during a mains outage is an unplanned shutdown of everything behind it, and unplanned shutdowns are how storage arrays develop bad days. The survey above takes a competent engineer most of a day. It is the cheapest insurance in the room.