Firstly, I am a software developer that has been delegated the task of finding a UPS for my team's 20-core test-suite machine. It is powered by a 1kW-rated Corsair RM1000X power unit. I live in South Africa where, as of last year, we have been having regular power outages. My team would like to invest in a UPS that can keep the machine running through the longest possible power outage (~4hrs worst case). Even though it is a dedicated test-suite machine, we regularly use it to run medium-sized simulations (we're a computational fluid dynamics research group) overnight. It is mainly for this purpose that we would like to ensure that the machine can run through a power outage, uninterrupted.

I have been doing research on how to select a UPS, but I am very confused about the difference between the W-rating and VA-rating that's given for all UPSs, and I'm unsure how to select one that will keep the machine running for as long as ~4hrs. Can anyone advise me on this? Is a UPS what I should be looking for in the first place? Or should I be looking at inverters instead (if so, please recommend how I should select an inverter too)?

  • 16
    Also, one sad fact of life is that some (many, most?) UPS manufacturers don't actually specify the kWh rating of their devices; instead, they usually rate them at minutes at maximum and half load. However, this "up to 4 kWh" requirement is rather high; getting a UPS like this would likely be very costly, as many hardly touch a 1 kWh capacity. Do consider an option of even a small a backup generator alongside a cheaper UPS. Be sure to connect the machine up to the UPS, so it can gracefully shut down once the UPS runs out of juice if there's nobody around to run the generator.
    – Richard the Spacecat
    Commented Mar 12, 2020 at 9:27
  • 17
    Be sure to measure how much the system actually consumes under max load (with a kill-a-watt or similar device) to get a better estimate of what your actual required UPS capacity would be. This could mean that the machine would only consume, say, 300W, which would lower the required UPS capacity to ~1.2 kWh and then some for a good measure, making it possibly way cheaper to buy.
    – Richard the Spacecat
    Commented Mar 12, 2020 at 9:27
  • 12
    Definitely generator plus short lived UPS. Cheaper to buy, cheaper to run.
    – tomnexus
    Commented Mar 12, 2020 at 11:21
  • 10
    Making this as a comment - before you invest in a long-duration UPS, be fully sure that your test-suite machine can run completely independently from anything outside itself. No internet access, no shared servers, no license server for test tools, no network storage, no other test aggregation tools. With no power to those other devices if you need them, you'll be dead in the water anyway.
    – Milwrdfan
    Commented Mar 12, 2020 at 23:29
  • 6
    Also consider moving your test server to a cloud provider. This might turn out to be a LOT cheaper than implementing a UPS based solution.
    – Tejas Kale
    Commented Mar 13, 2020 at 13:47

6 Answers 6


There are a lot of things at play with the market for uninterruptible power supplies. Most people who buy smaller ones have little knowledge, and just want a sense of security. Correspondingly you have to be careful to interpret the marketing information appropriately.

The following are some notes and suggestions "suitable for software engineers", which will get you close enough to some answers for your situation that you can then have a reasonable conversation with suppliers. Do not make a purchase or installation based solely on these notes.


Volta-Amps vs Watts: The difference between VA and W can be complex and depends on the type of load. But as a simplification for computer equipment you can use a power factor of 0.67, ie if your AC mains power is 240 VAC and you measure it at 4 A, you have 960 VA, which is approx 643 W. (From APC documents.) Some power supplies have a power factor as high as 0.99, such as this one from HP, [thanks grahamj42]. You'll see that, either way, this a relatively small amount compared to the other headroom requirements.

Online? You need to distinguish between "online" and "offline" UPS: an online one feeds your computers with AC generated from its batteries, and constantly charges them when the mains AC when it's available. An offline UPS switches between the mains AC and the AC generated from your batteries, perhaps taking 10 ms or so to do so. The online variety is considered considerably superior for computers, probably essential for a server like you describe.

Basic Scaling. If you have 1,000 W for 4 hours, you need 4,000 Wh. A typical car battery is approx 500 Wh (Wikipedia), so your basic needs are at least eight car batteries worth of storage. But note that four hours is considered a lot. [Thanks commenters for picking up my arithmetic errors.]


  1. UPS running time is extremely non-linear on load.
  2. A computer marked 1,000 W doesn't use anything like 1,000 W most of the time (for starting estimate call it 500 W, but this is really worth measuring.)
  3. A device marked 1,000 W may well use considerably more during immediate startup (for starting estimate call it 2,000 W)
  4. The batteries age and degrade as they do so (call it 50% at end of life)
  5. Large battery sets have their own issues (high-current DC cables, fire safety, etc)
  6. As soon as you have your server on UPS, you find you need your ethernet switches, your routers, your monitors, your desktops your fans and ... oh no ... your air conditioning. Make sure you include everything you actually need.
  7. You are likely to trim your requirements once you see prices

Which means a 1,000 W computer needs a UPS which can briefly deliver 2,000 W, but we can estimate its running time at 500 W.

Because of the non-linearity, most UPSs are specified by peak load, and have a load-vs-time calculator or graph. The following graphs show a 10 kVA UPS (curve A) and the same one with various additional batteries which give longer times.

enter image description here
From APC datasheet

Some Suggestions

Battery Replacement Do not buy a UPS without understanding that you are going to be replacing the batteries every few years. Manufacturers typically say three to five years (APC note.) but conditions in the field may well give you less.

Test, test, test Do not rely on a UPS unless you test it. If you don't test it you won't find out if everything that's supposed to be plugged into the UPS is actually plugged into the UPS. (I recommend different coloured cables and sockets.) And you won't know if accidentlly you have a 3 kW heater plugged in and nothing like the run-time you expected. Or your batteries are dead. Or removed! Or blown fuses. (I've seen all of these in real life: usually failures are something simple, high-reliability operation is difficult because you have to find them all.) You should schedule periodic tests: usually I've specified weekly or monthly test-trips and annual calibrations.

Monitor, monitor, monitor Do not put your server on UPS without ensuring the server will shut down properly if power outage is imminent. (Means get the UPS with ethernet/USB/RS-232 interface, configure server to use it.)

Budget Just a budgetary comment. A 10 kW UPS has a list price of around US$ 7,500 (March 2020), and its calculator suggests about 1h10 at 1,000 W and 2h30 at 500 W. Adding extra batteries might get you to your target time. Remember you're going to be buying new batteries in a few years. I'm not saying this is a good idea, merely noting what the costs might be.

Generator? As suggested in comments, you might well be better served with a generator and say 10 minutes of UPS. Typically you want to overrate the current of the generator too (Tripp notes, Eaton notes) but not by too much (Hyundai notes). Basically you get your electrician to get you an automatic transfer switch, or you can do it manually with plugs if you are always manned. For budget purposes consider a 10 kW portable generator with electric start and interfaces for the transfer switch might cost US$ 1,000, and will run for 10 hours on a tank. I've seen high-quality purpose-build 5 kW backup generators (ie, automatic start, long run) for around US$ 2,000. A whole-house fixed generator of 20 kW might cost US$ 5,000, and can run on piped-gas if you have that (ie, indefinitely). A backup-generator will have automatic start and (important!) a charger for the starter battery. (Quick budget numbers from US supplier Home Depot and UK Hyundai web sites, March 2020.) Simple video showing a Hyundai 5 kW generator with "automatic transfer switch" on Youtube. In your application a UPS (and the computer!) would go where the lamp is in that video. Another video.

Documentation You will find that the major suppliers such as APC and Tripp have extremely good general documentation. (Eg Calculating Total Power for Data Centers at APC.) A quick search also shows a book from Kohler on the subject, also covers UPS-generator integration (Kohler).


If you really want four hours, you might use something like the following as a spec for what you need to get/do:

  • Check and recheck your power requirement, following assumed as 1 kW
  • Online dual-conversion UPS
    • with wide tolerances to accept generator power [thanks David Schwartz]
    • with ethernet interface, ideally SNMP (others prefer USB or RS-232 handshake) for single-server [thanks Peter Cordes]
  • Max power > 2 kW
  • 10 minute at 1 kW (long enough to refuel generator)
  • 5 kW diesel auto-start generator
  • Automatic transfer switch
  • Some method for server to see generator is operating, ideally with "low fuel" alarm, ideally SNMP
  • Colour sockets and cables for back-up mains
  • Colour ethernet cables for works-on-backup-power network
  • Testing regime

Of course with a generator you'll get much more than four hours, and can consider whole-working-day or 24-hour operation.

You may well find it's just too expensive and difficult.

Wiring is something along these lines:

                               +------------ Lights, non-essentials, etc
      Mains power ------ ATS --+-- UPS ----- Server, essential switches, etc
          Generator ------+

If you can make do with an hour, perhaps you can do it with a straight UPS and avoid the complexity of the generator. It's still expensive.

Costs For a starting guess, I'll write down capital cost of US$ 10,000 for a pure-UPS solution, and US$ 5,000 for a generator one. The UPS will have very significant battery replacement costs (perhaps 50% every three years); the generator will have more installation and maintenance costs. I'd guess the resell value of a UPS is very small, for a diesel generator you get a good fraction of the new price.

Opinions and alternatives

If it was my business, I'm pretty sure I'd be doing nothing like this -- but of course it depends a lot on exactly what business you're in. It's a lot of money for a UPS which does the job, and a generator is a lot of stuff and machinery. Either way it's thousands for something which gets used occasionally.

I'd certainly consider being minimal about what has to stay up, work on reducing its power requirements, and using some kind of pure DC solution. A system where there's no change in the primary system (power getting to computer) when there's an outage is much, much, much better than a system which has to start reliably but infrequently. See the excellent other answer about DC from Harper https://electronics.stackexchange.com/a/485951

I see in your comments the thing which has to stay up might well be just a self-contained compute server. If that's the case, a long-uptime UPS is almost certainly the wrong way to go. For my high-reliability systems I usually use 12 VDC-only computers with constant charging (exactly as Harper recommends as "home-brew lead-acid"), but they are always low-power CPUs. Nevertheless I hope this overview of conventional UPS+generator has been useful.

PS. Square brackets show information from kind commenters, thanks all.

  • 1
    Oh wow! Thanks for your detailed response. It's a lot to take in, and I hadn't anticipated that looking for a UPS would be such a tedious task, haha. But I will re-read your post a few times to absorb the pertinent information that will inform my decision :)
    – niran90
    Commented Mar 12, 2020 at 11:56
  • 1
    @Justin ... I'm sorry, it was a stupid error, and should have been 4,000 Wh. P watts for T hours = P * T watt hours. I have fixed the answer text.
    – jonathanjo
    Commented Mar 12, 2020 at 17:58
  • 3
    @niran90 They don't normally quote watt-hours because of the non-linearities. The best way to estimate time is to look up your load (in watts) against the curve of a particular UPS. The graph in my answer is very typical of the shape of the curve, though of course the power and time vary according to the given UPS.
    – jonathanjo
    Commented Mar 12, 2020 at 18:02
  • 2
    "you need 4,000 Wh. A typical car battery is approx 500 Wh (Wikipedia), so your basic needs are at least two car batteries worth of storage" Huh? Is that arithmetic correct?
    – user20637
    Commented Mar 12, 2020 at 20:00
  • 2
    $7500 for a UPS that doesn't even have the runtime!? Why not just install a subpanel with a Tesla PowerWall then? It's cheaper, and is actually made for long-term running, also the battery will last a LOOOT longer. Not like a "10kw" UPS with 1h10 at 1000W, obviously the battery is sized for 5 minutes to drop connections, settle databases, flush cashes and shutdown. Commented Mar 13, 2020 at 2:56

By StackExchange standards this is not a direct answer, but it's too involved for a comment.

I'd opt for a much stronger statement than "you might be better served with a generator". UPS units for computers are always intended to cover only short outages, until a generator ramps up to stable operation. This could be as little as 30 seconds, but seldom longer than a couple of minutes. Or, just long enough for a controlled, safe shutdown, probably automatically triggered when the UPS estimates a few minutes remaining runtime. Four hours is well beyond what a UPS is normally expected to carry a load. Ten to 30 minutes is a much more common range. Beyond that you will definitely need a unit that accepts extra batteries. That adds up-front cost, and bulk, and regular battery replacement costs.

As others have said, if you really expect to maintain work during an outage (this is called "business continuity") you need to keep all your network infrastructure, client machines and air-con running too. This is the norm for a commercial datacentre; rare in office environments. It's impractical with UPS units; it needs a generator, probably one tied into your building's electrical circuits. That's something your landlord's Facilities department, or Building Engineer, normally deals with.

You might start with collecting stats on the last few months' worth of outages you've had. If 80% of them have been under 15 minutes, size your UPS for that kind of runtime, and set up automatic shutdown if the power remains off longer than that. But if you have hours-long outages regularly, and you really want to keep working through them, a generator infrastructure is the sensible approach.

Further, if you're having frequent outages, choose an "online" model of UPS. These are wired so power is always being drawn from battery, which (while mains power is available) are constantly being recharged. The other kind, called "standby" or something else, have a switch that kicks in when the mains power quits, to begin drawing power from the battery. The issue is that these switches are rated for a surprisingly small number of operations. Where I work we do weekly generator tests, so the UPSs switch to battery and back every Wednesday. And those switches can fail in a year or two, 100 to 200 switching operations. (And regular generator tests are a really good idea.) Go with an online model.

  • 2
    This may be a dumb comment as I have little to no experience with UPSes, but wouldn't the online model reduce the lifetime of the battery significantly? One might not have to worry about replacing the switch every year, but then one might have to replace the more expensive battery a few years sooner than they otherwise would have. I'm just wondering here; it may even be that battery technology is such a battery that is continuously used lasts longer than one that isn't. I really don't know.
    – JoL
    Commented Mar 12, 2020 at 23:31
  • 5
    @JoL lead-acid batteries don't mind being floated, it's no worse than letting them sit full and occasionally topping off, as a standby UPS does.
    – hobbs
    Commented Mar 12, 2020 at 23:50
  • Yes, the batteries need to be replaced every year or two anyhow (manufacturer will recommend a schedule, or the UPS unit's self-test will determine it). When the switch fails, firstly, you only find out about it when your system goes offline at a power transfer event. Secondly, the switch doesn't get replaced; the whole UPS does. So it's really not a good strategy.
    – CCTO
    Commented Mar 14, 2020 at 3:35

TLDR: In this modern age, the most affordable practical bulletproof "won't get fired for recommending this" solution is a Tesla PowerWall or competitor. The "cheap at all costs" solution is a bank of batteries and a DC PSU (or if the server doesn't support that, an inverter). Gen and UPS solutions are simply not made for this, and force-fitting them will result in an ugly, high-maintenance misfit.

You don't want a UPS, per se

Because your requirement is for a 4-hour runtime, and that is not what UPSs do.

In fact, you really need to throw the book out, because new tech is here that solves the problem much more artfully than enormous UPSs or hokey generators.

Just look at the sizing of UPS's, as jonathanjo talks about - comparing their KW rating vs the number of minutes of power at that rating. Most of them are sized for ~10 minutes of runtime at rating. Just enough time for the bog-standard application of UPSs:

  • Ride for 20 sec. in case power comes right back (why shut down needlessly). Then,
  • signal the PC that it's time to shut down. The PC then
    • stops accepting new connections
    • wraps up existing connections
    • flushes database writes
    • flushes disk cache
    • gives some seconds for hard drives to flush internal cache
    • completes an orderly shutdown.

That is not what you want. You want to continue business as usual across an extended outage. You need a completely different product family for that.

When you try to force a UPS solution, you find the only way you can "buy" extended battery life is to get an insanely oversized inverter, and that's a game of diminishing returns, because bigger inverters have bigger standby losses, especially at a tiny fraction of rated load. Like the 10kw, $7000 inverter discussed (as a bad fit) by jonathanjo.

Also, UPSs use cheap lead-acid batteries, that have poor service life, and regardless, should not be dipped below about 30% DoD on a regular basis. (DoD=Depth of Discharge; 0%=full 100%=stone dead and damaged) This is an unfortunate characteristic of lead-acid batteries. So if the advertised 1KW runtime of the UPS is 13.3 hours, 30% happens at 4 hours. So you're actually looking for a 13.3 hour runtime, in lead-acid. That is a huge and stupendously expensive battery to be changing every 3 years.

Again, UPSs are just wildly inappropriate here because of the long runtime.

The 1980s era solution

Is a genny, transfer switch, UPS to bridge the gap, etc.

An automatic transfer switch detects the outage, calls for a generator to auto-start and spin up, then it automatically throws the transfer switch. However, this takes a minute or two, so you still need a UPS in the mix to carry the system for that time.

For a quality auto-startable generator with automatic transfer switch, wired in properly into a subpanel and all that, you're north of US$5000; and you still need a UPS also (albeit a modest one).

And worse, you now have 2 batteries to maintain: the UPS battery, and the generator start battery. Another failure point is fuel: if someone failed to top up the tank after the last time, you either get no-start or it quits after an hour.

It's much messier if you try to pinch pennies with a manual generator. Now your UPS must have a much longer battery - 30 minutes to give the crews 5-15 minutes to get the generator fueled, started, extension cords run and the UPS plugged into it -- and still have time for an orderly server shutdown if that doesn't happen. The longer the batteries must run, the more sensitive the UPS is to old lead-acid batteries with shrinking capacity. It might be a good policy to prophylactically change the lead-acid batteries every year or 18 months. SO we save some money but we spend more elsewhere. Ugly, ugly solution.

All in all, I consider generators to be an expensive and/or messy solution. We only consider them because nothing better exists, or to be more precise, existED.

The modern solution: commercial products

Tesla PowerWall and competitors.

When SuperStorm Sandy hit the American northeast, thousands of solar panel owners found out grid-tied solar inverters don't work when the grid is down. This lit off a variety of solutions for extended runtime off-grid backup power. These systems are designed exactly to provide low-draw (1kw-ish) loads for hours and hours. Which fits your use-case like a glove. *These things are new, which is why they're not everyone's first recommendation (and also considered "green" which has a liberal tone which some find distasteful).

The flagship product, of course, is the Tesla PowerWall. Let's take the new PowerWall 2, which tickets at US$7500 all-in, and has 13.5 kwh of power. So it'll run your 1kw load for (realistically) 11 hours. That's too big.

Try the original PowerWall for US$4000 all-in (cheaper still used), and get about 6 hours. Tesla has already "de-rated" the battery to avoid bottoming the lithium battery, so you can really get 6 hours reliably.

Now, remember that 10kw UPS that cost $7000 and only had a 2.5 hour runtime @ 1 KW (which did bottom the lead-acid battery, so really, only 40 minutes daily runtime)? LOL! This certainly illustrates what a misfit it was. Simply the wrong tool for the job. Tesla goes 10 x longer for 1/2 the price, and is lithium to boot.

Further, you can expect 15 years of useful use out of this battery, because Tesla has deep experience (and skin in the game, due to how Tesla car batteries are warranteed) at making lithium battery packs last a long time. The protective circuits are tip top.

Of course there are other manufacturers making PowerWall-like's.

It's funny, PowerWalls and UPSs contain the same nuts and bolts (batteries+inverters+chargers)... but they are sized very differently. That makes all the difference in the world.

By the way, I never mentioned solar panels; some of those systems will support bolting on a simple solar array. That could stretch your runtime while grid is down.

The modern solution: Homebrew on the cheap

The PC runs off batteries normally and constantly. When available, grid power tops up batteries and feeds the PC.

The first step here is see if you can obtain a low-voltage PSU for that server - one that intakes 12, 24, or 48 VDC from a battery - the kind you use in DC-based server farms. If it does, then you can eliminate AC-DC conversion losses in both directions. Otherwise, you will need an inverter rated for continuous use, e.g. 2kw.

Now, you lay up a bank of batteries big enough for your needs. You can go a couple of ways. You can go with lithium arrays such as a 25V/220AH pack out of a Tesla Model S, typically $1500. That'll get you 5.5KWH, and since you can deep-dip lithiums down to 75% DoD, you can safely get 4 KWH out of this. Exactly what you need.

Or, you can go lead-acid, remembering to derate for 30% DoD for daily use, so 13.3 KWH. Take a generic US$100 Interstate golf cart battery: 6 volts @ 210 AH = 1260 AH; ten of those and Bob's your uncle. Needless to say, you acquire lead-acid batteries locally; they are commodities available everywhere; shipping them is not even stupid.

That's it; that's your system; a server and a bank of batteries to power it. Except we add one more thing: a power supply/battery charger that replenishes it while the grid is up. So when the grid is up (say: 3/4 of the time), the charger powers the inverter, and tops up the battery if it needs that. When the power fails, the battery is already connected to the inverter/PC, and picks up the load seamlessly. The PC is not aware a change occurred.

"PC not knowing" is worth some attention. You may want to add a circuit to warn the PC when it passes 50% DoD, so it can blast out a 15-minute warning to wrap up your work, then do an orderly shutdown as per section 1.

  • 3
    These are excellent suggestions. I'd comment I wasn't proposing a US$ 7k UPS, I was saying it was available. If it was my money there's no way I'd regard it as sensible; but I was trying to keep my tone neutral in my answer. If you are stuck with infrastructure designed for reliable mains electricity supply but lack the reliability, generators and all the complexity are a good fit. But it might be better to design the whole thing away, exactly as you say. I've added some commentary along these lines to my answer.
    – jonathanjo
    Commented Mar 13, 2020 at 13:17
  • 1
    @jonathanjo I read it as meaning "[such a bad idea that it's] worse than stupid"
    – hopper
    Commented Mar 13, 2020 at 15:26
  • 1
    @jonathanjo Yeah, I'm sorry, I did make it sound like you were recommending that one, didn't I? That was my bad, I knew you weren't. But as far as generators, I've had it with them. They'll always have their applications, but I think the only reason they're still gold-standard is the PowerWall type products are too new to be widely accepted. But I see them as inevitable now. Commented Mar 13, 2020 at 21:11
  • 2
    Hey! My industry! 100% agree, a storage inverter with grid forming capabilites is really what you want here. Points to note: lithium battery lifetimes are ~10 years, and you've probably lost 20% capacity by that point. You'll also probably still need a UPS sadly, though a small one, because regulations essentially require a 5s downtime before they can kick back on (you may be able to find one that doesn't do this, if the person that explained that to me was wrong...)
    – mbrig
    Commented Mar 14, 2020 at 7:21
  • 1
    @Harper-ReinstateMonica they actually do, but there's still the 5 seconds... something about islanding detection or ride through requirements. It didn't really make sense to me, but I wasn't going to tell the head of their R&D how to do his job. Layout looks something like this: i.imgur.com/VEzUFnJ.png
    – mbrig
    Commented Mar 19, 2020 at 5:09

Watts (W) and volt-amperes (VA) are both representing electrical power, but they are not (necessarily) the same.

  • Power in watts refers to real power and represents how much energy is consumed/dissipated.
  • Power in volt-amperes refers to apparent power, which provides information how much current it will draw.
  • 1
    Thanks for your remark to my question, I'm intending to remove my answer (since it would be mostly a copy to yours). Commented Mar 12, 2020 at 9:21
  • 1
    Thanks for your response! Could you perhaps tell me which rating I should be paying attention to while I keep searching? Do I need to match the 1kW-rating of the machine's power unit with the "real" or "apparent" power?
    – niran90
    Commented Mar 12, 2020 at 9:36
  • 1
    @niran90 you need the apparent power rating of the UPS to be more than the peak apparent power of the load — which you probably aren't told! In the case of the RM1000x, I can guesstimate it from the spec sheet to be 1260VA (1100W peak load, 88% efficiency at 100% load, 99% power factor because it has PFC). But the best way is to measure.
    – hobbs
    Commented Mar 12, 2020 at 23:47
  • @hobbs Thanks for this information. I have seen many comments that suggest that I should add 50-60% to the peak load for "good measure". So I have been looking at UPSs with a 1500VA spec (1.2kW). Do you agree with this?
    – niran90
    Commented Mar 13, 2020 at 1:37
  • 2
    @niran90 yes. Over-sizing a UPS system by a bit will be good for the longevity of the components and your batteries.
    – hobbs
    Commented Mar 13, 2020 at 2:34

Been there, done that.

First, know your enemy.

Power drawn: a watt-meter and a test with a power-hungry task. Your 20-core rig may draw anywhere between 350 and 1200W from AC when loaded (1kW PSU is some 80-sh % efficient and it is not known how overengineered it is for your use case). It is WATTS that you should care about. VA ratings are something UPS manufacturers put in specs in order to boost the numbers that sell.

Does it reside in a closed, air-conditioned room? If yes, you either consider the additional power used by the AC system or at least test how hot it becomes when AC is off for the projected outage time.

The wrong solutions:

  1. Generator: it needs fuel (see your local fire code, but you can skip that and start crying right now), it needs exaust management (so you generally end up building something at a code-required distance to any other building, you do have a backyard, OK?) and it needs someone starting it (I have yet to see a generator auto-starting reliably). Yes, it also needs maintenance (test runs, oil change, etc...). No, you don't want a generator. Not unless you scale the task some 30-fold.

  2. Oversized (in terms of VA/watt ratings) UPS. Anything more than 2000W / 3000VA is overkill in your case - its self-consumption scales miserably and you don't get double the expected runtime by doubling the size of the UPS, esp. if you target runtime is in hours range. It also will be heavy and expensive.

The proper solution:

An UPS with external, adequately-sized batteries. You may opt for a commercially-available or a hand-made hack.

For off-the-shelf solution: manufacturers like to advertize "optimistic" (read: unreal) runtimes. In order to get the real picture, see the ratings of the spare batteries, multiple V to Ah numbers and allow for at least 50% age-derating of the batteries.

Lithium batteries: an overill. Expensive as hell, yes, long-lived (do you expect to run the same setup in 10 years?) and after these 10 years you won't find a replacement. Lead-acids are a commodity and even if you fail to find an original replacement batteries, you can still disassemble the packs and fit a standard pieces there.

What I did (for about 40 remote offices): APC-1500-SMART-something, sent its internal batteries for recycling and had the electrician to connect it to a pair of truck batteries (12v, 230Ah) connected in series in order to get the required 24v of the original battery pack. The batteries need to be replaced after ~5 years if you still want 4 kWh out of them (in my case the improved electrical grid and replacing the equipment with more energy-efficient allowed more than 10 years of use). Bonus: every truck driver knows where to get them for cheap and how to replace them.


So, you're in Suid-Afrika. Get a lot of lightning?

I did, growing up in Zap Central, the lightning capital of the United States, with > 90 days/year where you could depend on a lightning strike within a mile. The county was also flat and sub-tropical; the highest ground was five meters above mean high tide.

The highest building in the entire county at the time (the largest county east of the Mississippi, I might add) was the administration building of the county government complex, where the county's mainframe was. LOTS of lightning, lots of power outages, many extended. Oh, and did I mention hurricanes?

So, the axle of a motor-generator set was attached to a heavy flywheel. The motor spun the flywheel and the axle, the generator turned the rotation into clean steady power. When (not if, but when) the power fails, the flywheel would run until a diesel genny could be started to pick up the load.

No need for inverters, rectifiers, or batteries, which are all failure-prone and high maintenance. It was also the kind of thing a shade-tree mechanic could keep running easily.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.