Almost everyone sizes a UPS the same way the first time: they look at the recorder, find a wattage on the label, and buy something comfortably bigger. Then the power goes out, and the system drops within twenty minutes — because the recorder was never the load. The cameras were, and they draw from the same battery through the switch, and nobody added them up.
The reason this matters more for cameras than for a desktop is structural. Power over Ethernet puts the power on the data cabling itself, so the cameras are not separate devices hanging off separate outlets — they are a load on the same supply as everything else. The standard describes the mechanism as raising the power available to the powered device "by utilizing all four pairs in the specified structured wiring plant". One cable, and one load to account for.
There is a second misconception worth clearing up before any arithmetic. The number printed largest on a UPS is its VA rating, and that number has almost nothing to do with how long it will run. VA tells you what the unit can supply at an instant. Runtime is set by the battery inside it — and two units with the same VA rating can differ in runtime by a factor of three.
What follows is the arithmetic most guides skip: adding the whole chain into one load, converting that load into a battery capacity, sizing for the three different outages people actually experience, deciding what can be dropped, and getting the recorder to shut down in an orderly way. It closes with two complete designs, for four cameras and sixteen.
The Short Answer, in One Paragraph
Add up the recorder, its drives, the switch and every camera, and size the battery for that total rather than for any single box. A four-camera system lands near 50 W and a sixteen-camera system near 190 W, so runtime is decided by battery capacity, not by the VA rating on the front. Buy enough battery for the outage you actually expect — seconds for a generator handover, an hour or two for a typical fault, several hours if outages are long where you live — and make sure the recorder can shut the drives down cleanly before the battery empties.
What a Backup System Actually Has to Carry
Start with the four things drawing power, because three of them are easy to forget and one of them is easy to overstate.
The loads that matter
| Component | Typical draw |
|---|---|
| Eight-channel recorder, electronics only, no drives | 10–15 W |
| Larger recorder, electronics only, no drives | 30–50 W |
| 3.5-inch surveillance hard drive, recording | about 5–6 W |
| Fixed 4K PoE camera, daylight | 3–8 W |
| Fixed 4K PoE camera, infrared on | 6–12 W |
| PTZ camera with motors active | 15–30 W |
| Eight-port PoE recorder, whole unit including PoE output | up to about 90 W |
Two things in that table deserve attention. The first is the recorder label. A recorder advertised as drawing 20 W is usually quoting its electronics, and the drives are additional — a point that trips up almost every first attempt at this calculation. The second is the camera range. A camera rated at 15 W almost never draws 15 W continuously; it idles much lower and rises when the infrared illuminators come on. Since outages often happen at night, size the cameras at the infrared figure rather than the daylight one.
Why sizing them separately misses the point
A single battery has to cover the whole chain, because that is how the current actually flows. Power enters the recorder, passes through to the camera ports, and leaves along the Ethernet cables. There is no arrangement in which the cameras continue running while the recorder's battery does not.
This is also why a separate PoE switch matters to the calculation. Many larger systems use a recorder without PoE ports plus a switch that provides them — and in that layout the switch is a second powered device with its own draw, in addition to passing through every camera's power. Its own consumption is modest next to the cameras, but it is not zero, and leaving it out is one of the common reasons a system falls short of its estimate.
What the numbers look like for a small system
Take a common four-camera installation built on a recorder with built-in PoE ports. Fifteen watts for the recorder's electronics, six for its drive, and four cameras at eight watts each, sized generously for night operation with infrared running.
That is 15 + 6 + 32, or about 53 W. Round it to 50 W for the arithmetic that follows. It is a small number, and that is the useful part of the exercise: at this load, almost any battery will do the job, and the choice comes down entirely to how many hours you want.
Sizing the Battery Rather Than the Box
This is where most buyers go wrong, and it is not their fault. The industry sells these units by VA.
VA is not watts
| Rating | What it actually describes |
|---|---|
| VA | Apparent power — voltage multiplied by current |
| W | Real power, the portion converted into work |
| Power factor | Watts divided by VA, between 0 and 1 |
| Modern IT equipment with power factor correction | 0.9 or better |
| Older equipment | 0.6 to 0.75 |
| A common small unit rated 1000 VA | 700 W, implying a power factor of 0.7 |
| A common 1500 VA unit | typically 900 to 1050 W |
The practical consequence is narrow but important. Because camera-system hardware uses switched-mode supplies with power factor correction, the watts are close to the VA — which means the VA rating is very unlikely to be your constraint. A 1500 VA unit rated at 900 W can deliver roughly seventeen times the load of our four-camera system before it runs out of capacity.
What that unit cannot do is run for very long, because its internal battery is small. This is the trap: buying a bigger VA rating to get a longer runtime buys you nothing at all.
The capacity formula, including the steps most guides skip
Battery capacity in amp-hours converts to energy in watt-hours by way of the battery voltage. Two further factors reduce what you can actually use: how deeply the battery is discharged, and how much the inverter loses in the conversion.
Battery energy (Wh) = amp-hours × battery volts
Usable energy (Wh) = amp-hours × battery volts × depth of discharge × efficiency
Runtime (hours) = usable energy ÷ load in watts
Required amp-hours = (load in watts × hours) ÷ (battery volts × depth of discharge × efficiency)
Two notes on the assumptions. Inverter efficiency is commonly planned at around 0.8 for this class of unit — it appears as an assumption in published runtime calculators rather than as a measured specification on the box. And depth of discharge is not free: Sandia National Laboratories' work on sealed lead-acid batteries is explicit that "the deeper batteries are discharged on average, the shorter their cycle life," with a working life of roughly 300 to 500 cycles and four to five years of dependable standby service. Planning for a deep discharge every outage is how a battery pack reaches the end of its life in a couple of years.
The examples below use an inverter efficiency of 0.8 and a depth of discharge of 0.8, which together give a divisor of 7.68 at 12 volts. Treat both as planning assumptions and note that making them gentler — a shallower discharge to protect battery life — raises the amp-hours you need in proportion.
Worked runtime for a small and a larger system
| Desired runtime | Four-camera system, 53 W | Sixteen-camera system, 190 W |
|---|---|---|
| 1 hour | 7 Ah at 12 V | 25 Ah at 12 V |
| 4 hours | 28 Ah at 12 V | 99 Ah at 12 V |
| 8 hours | 55 Ah at 12 V | 198 Ah at 12 V |
For the sixteen-camera system, the same energy at a higher battery voltage is worth noting, because it is how larger installations keep the numbers manageable: 4 hours at 48 volts needs about 25 Ah rather than 99 Ah, since amp-hours are a measure of charge, not of energy.
Why published runtime charts read low
Manufacturer runtime figures look alarmingly short next to these numbers, and there is no contradiction. A unit rated 1000 VA and 600 W with a single 12 V 9 Ah battery publishes roughly 40 minutes at 100 W and 17 minutes at 200 W. Read the second figure: doubling the load more than halves the runtime, which is normal for lead-acid batteries under high drain. At 50 W, the same battery behaves far better than the chart suggests — but the chart was written to be honest about its worst case, not generous about yours.
Sizing for the Outage You Will Actually Have
Runtime is a decision, not a specification, and it should be driven by the kind of interruption that actually happens where the system is installed.
The short outage
Most interruptions are brief — a fault on the local distribution network, a tripped breaker, a utility switch that restores service in seconds or minutes. For this case almost any unit works, because the load is small and the duration is short. A single modest battery will carry a four-camera system for well over an hour.
The generator switchover gap
Where a standby generator serves the building, the outage duration is decided by the transfer switch rather than by the utility. A residential or light commercial generator with an automatic transfer switch typically detects the failure, starts, and transfers the load in about 10 to 30 seconds. NFPA 110, the US standard for emergency and standby power systems, requires a Level 1 system — the category covering life-safety loads — to take load within ten seconds, and the National Electrical Code permits up to 60 seconds for legally required standby systems.
Run any of those durations through the formula and the battery requirement is almost nothing. Ninety seconds at 190 W is 0.62 Ah at 12 volts. A bridge of this length costs pennies of battery, which is why the generator gap is the easiest part of the whole problem — the battery is there to cover the handover, not the outage.
The nuance worth keeping is that NFPA's own commentary notes the generator is not obliged to be online quickly when the critical loads are already backed by a UPS. In other words, the battery is not a substitute for the generator; it is the reason the generator is allowed to take its time.
The multi-hour outage
Long outages are the case that actually sets the battery size, and they are regional. Where service interruptions of several hours are routine, plan for several hours: 99 Ah at 12 volts for the sixteen-camera system, or about 25 Ah at 48 volts. Where they are rare, a single-hour target is a reasonable compromise, with the understanding that the system will simply stop recording when the battery is exhausted.
A runtime target framework
| Situation | Sensible target | What drives it |
|---|---|---|
| Generator on site with automatic transfer | 2 to 5 minutes | Covers the handover plus the recorder's shutdown window |
| Urban or suburban service, rare long faults | 1 to 2 hours | Rides through typical faults with margin |
| Rural service, frequent multi-hour faults | 4 to 8 hours | Matches the observed outage distribution |
| Site where recording is contractual or life-safety | 8 hours plus generator | Beyond battery economics; backup power becomes the primary source |
What Must Stay Up and What Can Drop
A battery of a fixed size runs longer if you ask less of it — and the list of things worth cutting is longer than most people expect.
Essential loads
Three things must be on the battery: the recorder and its drives, the switch providing power to the cameras, and the network path that carries the recordings if remote access matters. If the connection to the outside world is a modem or optical terminal with its own supply, that belongs on the battery too — otherwise you keep recording and lose the ability to see any of it.
Loads that can be dropped
Locally attached monitors and displays are the classic example, and they are often among the largest draws in the room. PTZ heaters, wipers and illuminators are worth shedding if the outage is long, accepting reduced camera capability. Non-essential cameras — a view of a car park nobody needs at 3 a.m. — can go dark to extend runtime for the cameras that matter.
The practical mechanism is the outlet layout. Better units separate their outlets into a battery-backed group and a surge-only group, and some add remote switching for individual outlets. Plugging by priority rather than by convenience is a free upgrade: it costs nothing and can extend runtime meaningfully.
Protecting the Drives: Clean Shutdown
This is the most mythologised part of the subject, so it is worth separating what actually happens from what gets repeated.
What a sudden power loss actually does
Modern surveillance drives are built for continuous operation, and in the ordinary case the loss of power does not physically damage one. The realistic risks are on the data side: recordings still in flight when the power drops, index entries that never get written, and a file system left inconsistent if the loss interrupts a journal write. The reported symptoms are correspondingly mundane — lost segments of the most recent footage, a recorder that boots with a warning, or, as users have documented on at least one recorder platform, camera configuration reset to defaults after the power came back.
That is a real problem for a surveillance system, where the last minute before an incident is precisely the footage you wanted. It is a data-integrity problem rather than a hardware-survival problem, and the fix is orderly shutdown rather than more battery.
The signal path from battery to recorder
For the recorder to shut down before the battery empties, it has to know the battery is running low. That conversation happens one of three ways: over USB, over a serial connection, or over the network using a management card in the battery unit and SNMP.
The first two are typical of smaller units and need the two devices physically close, with the recorder supporting the software that ships with the battery. The third works over distance and is the usual arrangement in a rack installation, but it requires a battery unit with a network card rather than a plain USB port.
The settings that enable it
The setting to look for goes by names like "UPS" or "power failure shutdown," and it typically lets you choose the battery level at which the recorder initiates shutdown — commonly somewhere around 20 to 30 percent remaining. The recorder then stops recording, flushes its writes, and powers off while the battery still has enough charge to complete the sequence.
Where the honest answer is uncomfortable
Support for this is uneven, and it is gated by firmware rather than by price. On at least one widely deployed recorder platform, earlier firmware had no UPS shutdown capability at all; the feature arrived later, and only over SNMP with a management card in the battery unit. Some recorders never expose the setting.
Before designing around clean shutdown, confirm three things: that your recorder has the setting, that its firmware is new enough to expose it, and that your battery unit has the right interface. If it does not, the honest fallback is to size the battery generously and accept that the drives will lose power — the data risk is real but bounded, and a longer runtime means the recorder reaches the end of its battery less often.
Redundancy: What It Protects and What It Does Not
Redundancy and battery backup get discussed together so often that they seem interchangeable. They solve completely different problems.
| Mechanism | Protects against | Does not protect against |
|---|---|---|
| Two power supplies in one switch | Failure of one supply or one feed | Site-wide power loss; a single feed wired to both inputs |
| A recorder with two network ports | Failure of one uplink switch or cable | Power loss; a single switch feeding both ports |
| Link aggregation across two links | One cable or port failing, with added bandwidth | Power loss; both links failing together |
| Spanning tree or RSTP | Layer-2 loops where multiple paths exist | Power loss; a switch failing outright |
Inside the box
A switch with two power supplies survives the death of one supply, one breaker, or one feed. That is genuinely useful in a rack, and it is also where a cost hides: on some platforms the strictest redundancy mode holds half the power supply's capacity in reserve, so a switch with a large power budget can supply only part of it. Read the redundancy mode as a capacity decision, not just a safety feature.
Outside the box
A recorder with two network ports, or two links aggregated together, survives a failed cable or a failed upstream switch. Both arrangements assume the failure is in the path rather than in the power, and neither helps if the single upstream device they both depend on is the thing that failed.
Why redundancy never replaces battery
None of the mechanisms above does anything about a site-wide power loss. A switch with dual power supplies connected to two dead circuits is a switch that is off. Redundancy addresses equipment and path failures; a battery addresses the loss of the supply itself. Systems that matter usually need both, and neither is a substitute for the other.
Two Designs, Worked Through
Both designs below use the same method: list the loads, sum them, decide the runtime, and convert. The component figures come from the table earlier; the totals are arithmetic on those figures, not published system ratings, and the inverter and discharge assumptions are stated so you can swap them for your own.
A four-camera system
| Item | Figure |
|---|---|
| Recorder, eight-channel, no drives | 15 W |
| One 3.5-inch surveillance drive | 6 W |
| Four fixed 4K cameras at 8 W, sized for night operation | 32 W |
| Total load | about 53 W |
| Battery for one hour at 12 V | about 7 Ah |
| Battery for four hours at 12 V | about 28 Ah |
A consumer unit with a single 9 Ah battery will carry this for around an hour and a half; adding an external battery module or moving to a unit with a larger internal pack takes it comfortably past four hours. Because the load is small, a single modest battery is enough — there is no need for a large tower.
Kit-based systems make this calculation shorter, because the ports and the power sit on the recorder itself. The 8 and 16 channel PoE systems from 4COVR are built that way, which reduces the list of things to add up to the recorder plus the cameras rather than a stack of separate supplies.
A sixteen-camera system
| Item | Figure |
|---|---|
| Recorder, larger chassis, no drives | 40 W |
| Two 3.5-inch surveillance drives | 12 W |
| PoE switch, its own consumption | about 10 W |
| Sixteen fixed 4K cameras at 8 W | 128 W |
| Total load | about 190 W |
| Battery for one hour at 12 V | about 25 Ah |
| Battery for four hours at 12 V | about 99 Ah |
| Battery for four hours, at 48 V | about 25 Ah |
The switch figure is the one to treat with caution. Vendors publish per-port budgets far more often than their own quiescent draw, so ten watts is a planning assumption rather than a specification, and a safer habit is to add a margin for it rather than to trust the number.
At this load, a single consumer unit is not a design — it is a stopgap. The realistic options are a unit with an external battery pack, or a rack-mounted battery bank, and the choice between 12 and 48 volts follows from how much amp-hours you are willing to buy.
Questions People Ask Before They Buy
How long will a UPS keep my camera system recording?
It depends entirely on the load and the battery, not on the VA rating. Work out the total wattage first — recorder, drives, switch and every camera — then divide the usable battery energy by that number. A four-camera system at about 53 W will run around an hour and a half on a single 9 Ah battery; a sixteen-camera system at about 190 W will run roughly a fifth as long on the same battery.
What size UPS do I need for an NVR and cameras?
Size by total watts and desired runtime, then check that the unit's watt rating covers the load with headroom. A common small unit rated 1000 VA and 600 W carries either system described here comfortably on watts, so the deciding factor is how much battery capacity the unit has or can accept.
Do I need a UPS for the recorder, the switch, or both?
Both, and they need to be on the same battery. The cameras draw their power through the switch, and the switch draws through its own supply, so a battery protecting only the recorder leaves the switch — and therefore every camera — unpowered. In a system where the recorder provides the PoE ports, one battery covers everything.
Will my cameras keep recording during a power outage?
Only if the battery is sized for the outage and the recorder has somewhere to write. Recording continues on battery power until the battery is exhausted or the recorder shuts down. Remote viewing is a separate question: if the modem or optical terminal is not on the battery as well, the recording continues but you cannot watch it.
Do I still need a UPS if I have a generator?
Yes, though a small one. The transfer switch takes roughly 10 to 30 seconds to bring the generator online, and during that window there is no power at all — enough to drop the cameras and interrupt recording. Ninety seconds of runtime at a sixteen-camera load is well under one amp-hour, so the battery that bridges the gap is small; its real value is covering the handover and giving the recorder time to shut down cleanly if the generator fails to start.
Can a UPS damage the recorder if the shutdown is not clean?
The risk is data rather than hardware. Surveillance drives are built for continuous operation and are not commonly damaged by a power loss, but the footage still in flight and the file system's index can suffer, and some recorders have been reported to come back with their camera configuration reset. The protection is a recorder that supports orderly shutdown on a low-battery signal, which is not available on every model.
Where This Leaves the Decision
Three decisions decide this, and they happen in order.
First, add up the whole chain — recorder, drives, switch, cameras at their night figure — rather than the box with the label on it. That total, near 50 W for four cameras and near 190 W for sixteen, is the number everything else depends on.
Second, choose a runtime from the outage you actually experience, not from the chart on the packaging. If a generator is on site the gap is seconds and a small battery covers it. If service is unreliable where you live, the battery is the design and 12 versus 48 volts becomes a real choice.
Third, confirm the recorder can shut its drives down before the battery empties, and check the firmware version that enables it. Where that support does not exist, buy a longer runtime instead — the data you save is the minute before the incident, which is usually the only minute anyone ever needed.