Backup Power by Scenario: When an APC UPS Is Enough, and When You Need a Generator, Transfer Switch, or Solar

There is no single right answer, and anyone who gives you one is guessing

People ask me for 'the best backup power setup' the same way they ask for the best truck. The question cannot be answered until I know what you are hauling and how far. A 750VA rack UPS and a three-phase Galaxy unit are not competing products. They live in different universes.

I coordinate emergency power deployments for an electrical equipment distributor. I have handled 200+ rush jobs in 11 years — same-day APC UPS swaps in clinic server closets, 36-hour generator hookups for outdoor events, and at least a dozen calls that started with 'the power is already out, what do we do.' About 200 jobs — maybe 180, I would have to pull the log. The point is that almost every one of those calls got resolved differently, because the situations were different.

So instead of a product recommendation, here is how I actually triage. Four scenarios cover roughly 90% of what I see. Find yours, then read that section.

The four scenarios at a glance

• Scenario 1: A rack, a till, a single machine, or a handful of devices that must survive short outages.
• Scenario 2: A whole building or site that needs to keep running for hours.
• Scenario 3: A load big enough to require three-phase power and real redundancy.
• Scenario 4: No grid at all — off-grid or solar-backed systems.

If you are somewhere between two of these, that is normal. Read both. The overlap is where most expensive mistakes happen.

Scenario 1: Keep a rack, a till, or one critical machine alive

This is the most common case and the one people overthink the most. If your actual load is a network rack, a POS station, a single server, or a piece of lab equipment, the right answer is usually a line-interactive UPS in the 750VA to 1500VA range, not a data-center-grade online unit.

A good reference point is the APC Smart-UPS Line-Interactive 750VA 230V (the SMT750I family). It delivers roughly 500W of real power, has a sine-wave output, and gives you enough runtime to ride out a breaker trip or bridge the gap until a generator starts. For one rack with a switch, a firewall, and a small server, that is genuinely sufficient.

Here is the counterintuitive part: online double-conversion is not automatically better. I have watched people spend 2.5x on an online UPS for a closet with no air conditioning, then wonder why the unit runs hot and the fans never stop. Online units are typically 90-94% efficient, line-interactive units are closer to 95-98%. That difference becomes heat inside a locked cabinet. If your site has stable voltage and no serious distortion, line-interactive is the more sensible engineering choice, not the budget compromise.

In my first year doing this, I made the classic specification error: I treated VA and watts as the same number. A 750VA unit is not a 750W unit. It is a 500W unit with a 750VA apparent-power ceiling. I sized a UPS at 90% of its VA rating, the load tripped it on the first real outage, and we ate a return and a very awkward phone call. Now I size at 60-70% of the watt rating and sleep fine.

Two more things I always mention. First, the battery is the part that fails. Valve-regulated lead-acid batteries in these units are generally good for three to five years, and heat shortens that significantly — the old rule of thumb is that every 10°C above 25°C roughly halves service life. Second, check the discharge curve in the datasheet, not the marketing page. 'Runtime' on a product page is usually measured at half load, in a new battery, at room temperature.

Scenario 2: Keep the building running for hours

Once you are past a single circuit, a battery alone stops making sense. You need a generator, and a generator needs a transfer switch. This is where I see the most dangerous shortcuts.

In March 2024, a client called me 36 hours before a two-day event at a venue with a failing service panel. Normal lead time on the equipment they needed was 8-10 days. We found a rental 3500W-class inverter generator and a manual transfer setup through a partner, paid about $1,100 in rush and rental fees on top of a $2,400 base cost, and had them running the next afternoon. Their alternative was cancelling a paid event. The reason it worked is that their critical load was lighting, AV, and a small kitchen circuit — under 2kW — and a manual transfer was acceptable because they had a licensed electrician on site.

A 3500W inverter generator like the MInM 3500 is a reasonable fit for that kind of light commercial load. Inverter units typically hold total harmonic distortion under 3%, which is why they can run electronics that a conventional generator would make unhappy. Conventional portables often run 5-15% THD. But do not take that from me — pull the manufacturer datasheet and confirm the THD figure for the exact model, because it varies within a product line.

The transfer switch is the part people try to skip. A Square D transfer switch — manual or automatic, depending on whether anyone will be on site — does two jobs: it moves the load from utility to generator, and it prevents backfeeding. Backfeeding means energizing your panel from a generator through a regular outlet. It is illegal in most jurisdictions, it can electrocute a utility worker down the line, and it will destroy your generator when the utility comes back. If a quote does not include a proper transfer switch and an interlock, the quote is incomplete. Full stop.

On the code side, optional standby systems are addressed in NEC Article 702, and IEEE 446 is the standard reference for commercial and industrial standby system design. Those two documents are what a competent electrician will be working from. I am not going to give you wiring instructions — this is genuinely a licensed-trade job, and I have seen the results of DIY versions.

I went back and forth for two weeks on a similar job between a manual transfer switch and an automatic one. Manual was a third of the cost and simpler to install. Automatic meant nobody had to be on site at 2 a.m. We went manual because the client had overnight staff. That is the whole decision framework, honestly: who is physically there when the power drops?

Scenario 3: The load needs three-phase power

'APC Galaxy vs UPS' is a search I see a lot, and it usually comes from someone comparing things that are not comparable. Galaxy is not a different brand of the same product — it is a different class of machine.

The Galaxy family (Galaxy VS, VX, and the larger 7000-series units) is three-phase, modular, and built for centralized power protection from roughly 10kW up into megawatt territory. These are the units that sit in a dedicated electrical room, accept multiple battery strings, and support N+1 module redundancy so you can service a module without dropping the load.

A standard Smart-UPS is single-phase, rack or tower format, and lives in the kilowatt range. Putting a Smart-UPS in front of a 200kW chilled-water plant is not a solution, and putting a Galaxy in front of a four-server rack is a very expensive way to heat a room.

The actual decision points are: Is your load three-phase? Do you need module-level serviceability without downtime? Is there a realistic expansion path that would double the load in three to five years? If you answered no to all three, the Galaxy conversation is premature. If you answered yes to any of them, you are now in critical-power engineering territory, and you need a design firm, not a product page.

This is where I draw a line for our own team. We supply and deploy single-phase UPS systems and transfer equipment all day. We do not design paralleled three-phase systems, we do not do load-bank commissioning on megawatt installations, and we do not sign off on the short-circuit coordination study. When a client needs that, I tell them so and point them at firms that do it properly. The vendor who says 'that is not our strength, here is who does it better' is worth keeping for everything else.

Scenario 4: There is no grid to fall back on

Off-grid and solar-backed systems are a different discipline, and I will be upfront: we do not design them. But we do get asked the diagnostic question constantly — how to test a solar panel with a multimeter — and it is worth walking through because it is genuinely a five-minute check that tells you whether you have a panel problem or a controller problem.

Before you touch anything: isolate the panel from the charge controller and the load. A panel in sun is a live DC source and it will arc. Then cover it or do the test early morning.

For open-circuit voltage, set the meter to DC volts, put the red probe on the positive lead and the black on the negative, and read it. Compare against the Voc on the panel nameplate. Two caveats. Voc rises in cold weather — a panel rated 22V Voc might read 24V at 5°C, which is normal and not a fault. And Voc tells you the panel is producing voltage, not that it can produce current under load.

For short-circuit current, switch to the DC current range and measure across the leads with the panel isolated. Most handheld multimeters have a 10A maximum, so this only works on small panels. A 400W panel can push over 10A and you will blow the meter's fuse or worse. If you need to test a full-size array, use a DC clamp meter instead — safer and faster.

Someone asked me last quarter why their panel read a healthy 21V and the system still did nothing. The panel was fine. The controller had failed. Open-circuit voltage is a go/no-go test, not a health score. For anything past that — string sizing, MPPT selection, rapid shutdown requirements under NEC 690 — hire a solar contractor. The regulations changed materially in the 2020 and 2023 code cycles and vary by jurisdiction. Verify current requirements with your local authority having jurisdiction before assuming anything.

How to figure out which scenario you are in

Three questions, in this order. They resolve about 90% of the ambiguity.

1. What is the actual wattage you must protect? Not the nameplate on the panel, the measured or documented load. Add it up. Most people overestimate by 2-3x and then buy for the overestimate. A plug-in power meter costs $30 and settles the argument.

2. How long must it run? Seconds to minutes means a battery. Hours means a generator plus a transfer switch. Days means you are looking at fuel logistics, not equipment. That last one surprises people — a generator that runs 48 hours needs a fuel plan, and a fuel plan is often harder than the generator.

3. Who is on site, and who is licensed? Manual transfer assumes a person. Automatic transfer assumes a budget and a maintenance contract. Anything involving a panel, a transfer switch, or grid interconnection assumes a licensed electrician and probably a permit. If the answer to 'who is licensed' is 'nobody yet,' that is your first purchase, not the UPS.

If you can answer those three, the scenario picks itself. If you cannot answer question one, stop shopping. You are not ready to buy anything, and any vendor who sells you equipment without asking that question is selling you a return.

Where I stop being useful

I would rather do one thing well than six things adequately. We handle single-phase UPS selection and deployment, battery replacement, and transfer equipment supply. We do not do three-phase system design, we do not do solar array engineering, and we do not do generator service contracts. When a call lands in one of those buckets, I say so in the first five minutes and hand over a referral.

That costs us some revenue. It also means that when we say 'this will work,' nobody has to wonder whether we were just trying to make a sale. In emergency power, that distinction matters more than a line item on a quote. The setup that fails is rarely the one that was undersized — it is the one that was sold by someone who did not understand the load.