Skip to main content
Technical guide

Uninterruptible power supply
(UPS)

This guide explains how IEC 62040-3 classifies every static UPS, what each topology actually corrects, how to size the unit and its runtime, and where the safety boundaries sit. It is written for Canadian installations and stays vendor neutral.

Talk to a specialist

What is an uninterruptible power supply (UPS)?

An uninterruptible power supply (UPS) is a static converter that keeps a critical load powered without a break when the utility fails or drifts outside its limits, and that, depending on its topology, also regulates voltage and frequency while the utility is present. It is written here for IT managers, building engineers and facility managers. The decision it helps you make is easy to state and hard to get right: which topology, how much capacity and how much runtime your load actually requires. For the Eaton and legacy Powerware families specifically, see our Eaton UPS service and model guide.

1. Scope, terms and assumptions

This page covers static UPS systems, the kind whose converters have no moving parts. Rotary UPS systems, which use motor-generator sets, fall under a different family of standards and are not covered here.

The system boundary runs from the upstream connection point of the UPS to its receptacles or output panel. It includes the rectifier, the inverter, the battery, the static bypass and the maintenance bypass where one exists. It excludes the generator, the upstream automatic transfer switch, downstream distribution and room cooling, even though those elements frequently decide the real outcome.

Three terms recur throughout and need to be kept apart:

  • Normal mode: the UPS powers the load from the AC input while the battery is connected and either charging or fully charged.
  • High-efficiency normal mode: the UPS powers the load directly from the AC input to reduce losses. The battery stays connected. Bypass normal mode and power-factor-corrected bypass normal mode both belong to this category.
  • Stored energy mode: the UPS powers the load from the battery because the AC input is interrupted or outside its acceptable window.

Every UPS has a normal mode and a stored energy mode. Only some offer a high-efficiency normal mode, and nearly all of those switch between the two automatically depending on input quality [1].

One reading assumption: the figures quoted here are typical values published by a manufacturer or fixed by a standard. They do not replace the datasheet for the model you are evaluating, or a site survey.

2. The seven power problems and which ones a UPS solves

A UPS is not a universal corrector. Its class determines exactly which utility defects it neutralises and which it passes straight through to the load. This is the most common specification error we see in the field: an entry-level unit bought for a load that needs frequency regulation.

Table 1. What each input-dependency class corrects in normal mode. Derived from the class definitions in IEC 62040-3 and the notes in reference [1]: VFD means the output depends on input voltage and frequency, VI means voltage is regulated but frequency is not, VFI means both are regulated.
Power problem VFD VI VFI
Power interruption Corrected Corrected Corrected
Voltage sags and swells Not corrected Corrected Corrected
Voltage transients Only where surge suppression is fitted Corrected Corrected
Sustained over- and under-voltage Not corrected Corrected Corrected
Voltage waveform distortion Not corrected On-line types within the VI class only Corrected
Voltage fluctuations Not corrected On-line types within the VI class only Corrected
Frequency variation Not corrected Not corrected Corrected

The practical conclusion fits in one sentence: if your load is frequency sensitive, or if the site regularly runs on a generator, only the VFI class meets the requirement in normal mode. That is the typical case on generator-backed sites, where frequency drifts during load steps.

3. How IEC 62040-3 classifies every UPS

IEC 62040-3 exists precisely because marketing names mean nothing. It describes the output performance of a static UPS with a three-part code, AAA-BB-CC [1].

Part 1, input dependency (AAA)

Three values are possible. VFD, voltage and frequency dependent. VI, voltage independent. VFI, voltage and frequency independent. This is the part that drives everything summarised in Table 1.

Part 2, output waveform (BB)

Two letters describe the waveform, the first in normal mode and the second in stored energy mode. S means a sine wave, X a degraded sine wave, and Y a non-sinusoidal waveform such as square or stepped. A unit coded SY therefore delivers a sine wave while the utility is present and a square or stepped wave the moment it transfers to battery.

Part 3, dynamic output performance (CC)

Two digits describe behaviour during mode changes and during load steps. 1 is the most demanding class, 3 the most permissive. A unit coded 11 holds output voltage inside the tightest tolerance both on transfer and on a load step.

Two codes carry most of the meaning. VFI-SS-11 describes a double conversion UPS in on-line normal mode: sine wave in both modes, no interruption, voltage and frequency regulation. VFD-SY-33 describes a standby unit: no regulation, non-sinusoidal output on battery, transfer time in the order of 10 milliseconds [1].

Verification point. Ask for the full IEC 62040-3 code for every normal mode the unit offers. A unit advertised as VFI-SS-11 in on-line mode may be VFD-SS-31 when running in high-efficiency mode, which is a different level of protection.

4. The four static topologies and where each belongs

Four topologies dominate the market: standby, line interactive, hybrid conversion on-line and double conversion on-line. They differ in the path power takes in normal mode, and that path determines everything else.

Power paths of the four static UPS topologies Four block diagrams compared. Standby topology: the utility passes through a filter and a transfer switch to the load, while the battery and inverter only act when the utility fails. Line interactive topology: a tap-changing transformer regulates voltage and the converter stays connected to the output. Hybrid conversion on-line: an integrated rectifier and inverter powers the load continuously with a static bypass in parallel. Double conversion: the utility passes through a rectifier then an inverter, so the load is never connected directly to the utility in on-line mode, with a static bypass in parallel. A. Standby (VFD-SY-33) Utility Filter Transfer sw. Load Battery Inverter B. Line interactive (VI-SS-31) Utility Tap-changingAVR transformer Transfer Load Converter Battery C. Hybrid conversion on-line (VI-SS-11) Utility IntegratedAC/DC/AC converter Load Battery Static bypass D. Double conversion (VFI-SS-11) Utility Rectifier Inverter Load Battery Static bypass Normal path Battery path Bypass
Figure 1. Power path in normal mode for the four static UPS topologies. In topologies A and B the load stays connected to the utility through a switch; in topologies C and D it is fed continuously by the converter, which is why transfer time is zero. Original diagram by GDF Technologies, drawn from the topology descriptions in reference [1].
Table 2. Typical characteristics of the four topologies. Power ranges, transfer times, waveform and IEC classification taken from reference [1], tables 3 and 4. APC ranges are cited as an example of the manufacturer’s own published mapping, not as a recommendation.
Topology Typical power Transfer time Waveform on battery IEC class Example range
Standby 0.2 to 2 kVA under 10 ms Square or stepped VFD-SY-33 or VFD-SX-33 APC Back-UPS
Line interactive 0.5 to 5 kVA under 10 ms Stepped or sine VI-SX-31 or VI-SS-31 APC Back-UPS Pro, Smart-UPS
Hybrid conversion on-line 1 to 3 kVA in on-line mode 0 ms Sine VI-SS-11 APC Smart-UPS Ultra
Double conversion on-line 10 to 1500 kVA 0 ms Sine VFI-SS-11 Three-phase ranges

Line interactive remains the most common topology between 0.5 and 5 kVA because its tap-changing transformer corrects low and high line voltage without calling on the battery. That is a concrete reliability point: every unnecessary transfer to battery consumes battery life [1]. Above 10 kVA double conversion dominates, and it is the only class that is VFI in normal mode. For scalable architectures see our page on modular UPS systems, and for the large end of the range, high-capacity three-phase UPS.

5. Transfer time and waveform: what the load demands

A transfer time under 10 milliseconds suits the large majority of switch-mode IT power supplies, which hold enough capacitive reserve to ride through a gap of that length. It does not suit everything. Three cases require a zero transfer time, and therefore an on-line topology:

  • Strongly inductive loads, motors and compressors, whose inrush current on restart will trip an inverter sized too close to the load.
  • Laboratory and imaging equipment whose measurement cycle tolerates no discontinuity. See our note on UPS service for laboratories and pharmaceutical facilities.
  • Generator-backed sites, where frequency drifts during load steps and a VI or VFD class unit transfers to battery repeatedly.

Waveform deserves the same attention. A square or stepped wave, coded Y or X in stored energy mode, heats transformers, makes motors vibrate, and triggers refusal in some active power-factor-corrected supplies, which read the waveform as a fault and shut down. On a recent server the symptom is an abrupt shutdown at the exact moment the UPS takes over, which is the opposite of the intended result.

The specification rule that follows is simple. If the load includes an active power-factor-corrected supply, require a sine wave output in stored energy mode, that is a code ending in SS, and confirm it on the datasheet rather than in the brochure.

6. Sizing a UPS: from inventory to kVA and kW

Sizing almost always fails for the same reason: someone adds up volt-amperes and ignores watts, or the reverse. A UPS has two independent limits and the tighter of the two governs.

Five-step UPS sizing method A five-step decision chain: inventory the load in watts and volt-amperes, total the two figures separately, check both limits of the unit, add a growth margin, then choose runtime and verify inrush current at startup. 1. InventoryRecord W and VAfor every load 2. Two totalsSum of wattsSum of VA 3. Two limitsLoad W ≤ UPS WLoad VA ≤ UPS VA 4. MarginExpected growthover service life 5. RuntimeTarget durationand inrush Common trap A 1500 VA UPS is not a 1500 W UPS. The output power factor published by the manufacturer sets the watt limit, often between 0.6 and 0.9 depending on the range. The watt total is almost always the constraint that decides, not the volt-ampere total.
Figure 2. Five-step sizing method. The two limits in step 3 are independent and must be checked separately. Original diagram by GDF Technologies.

The output power factor published by the manufacturer links the two limits. A 1500 VA UPS with a published power factor of 0.6 delivers 900 W, not 1500 W. Another unit of the same VA rating with a factor of 0.9 delivers 1350 W. That 450 W difference is the difference between a unit that carries your rack and a unit that goes into overload the moment you add a network switch.

Table 3. Worked calculation for a small server room. The load values illustrate the method and must be replaced with the actual nameplate ratings of your own equipment.
Equipment Quantity Watts each Total watts
1U rack server 3 210 630
Storage array 1 180 180
48-port network switch 2 75 150
Firewall and router 1 60 60
Load total 1020 W
With a 25 percent growth margin 1275 W

With 1275 W to cover, a 1500 VA unit at a 0.6 power factor, that is 900 W, falls short despite its reassuring label. You need a unit whose watt limit exceeds 1275 W, which typically points to 2200 VA or to a range with a higher output power factor. For full sizing that also accounts for distribution and cooling, our team handles sale, consultation and installation across Canada.

7. Runtime and batteries: the wear part of the system

Runtime is not linear. Doubling the load cuts runtime by considerably more than half, because the effective capacity of a lead-acid battery falls as discharge current rises. Runtime quoted at full load and runtime at partial load are two different numbers, and it is always the manufacturer’s discharge curve that settles it, never a rule of three.

Useful service life depends first on temperature. Sealed lead-acid batteries are specified around 20 to 25 degrees Celsius, whereas the lithium-ion batteries used in UPS applications are often specified up to 40 degrees and degrade more slowly with heat [2]. A room at 30 degrees does not cancel the protection; it quietly shortens the replacement schedule.

Table 4. Typical service life of the replaceable parts of a UPS, from reference [3], table 4. These values are for budget planning, not a substitute for inspection.
Part or sub-system Typical service life Consequence of running past it
Battery strings 3 to 5 years Real runtime falls below displayed runtime
DC capacitors 5 years Increased ripple, stress on the inverter
Fan assemblies 5 to 7 years High internal temperature, accelerated ageing
AC capacitors 7 years Output distortion, nuisance trips
Internal power supply units 10 years Loss of control or display
Intelligence modules and controller boards 10 to 15 years Diagnosis impossible, parts scarce
Inverter assemblies 10 to 15 years End of life for the unit

This table is why a maintenance plan that only mentions batteries is incomplete. Capacitors and fans reach end of service well before the inverter does and are replaced at mid-life of the unit, before they cause the failure. For the battery itself see our APC UPS battery replacement and RBC cross-reference guide and our battery repair and replacement service. Replacement cartridges and packs for the common single-phase ranges are stocked at UPSPLUSBATTERY, APC Back-UPS replacement batteries.

8. Static bypass and maintenance bypass: two different paths

Confusing these two paths is the most expensive source of incidents we encounter, because the confusion surfaces at the worst possible moment, during an intervention.

The static bypass is internal to the UPS. The unit commands it automatically on overload, over-temperature or inverter fault, and it connects the load directly to the utility. It protects continuity, not quality: while it conducts, the load receives raw utility power with no filtering or regulation, and a utility outage takes the load down. A unit sitting in static bypass is no longer a UPS, it is an expensive power cord.

The maintenance bypass is external, manual, and exists for exactly one reason: to let the UPS be fully de-energised, safely, without dropping the load. Without one, replacing a capacitor, a fan or a board requires a planned load outage. It is an installation item, not a software option, and it has to be designed in.

Operating rule. A bypass alarm is a priority one alarm, even when the load is running normally. It means protection is suspended. Treat it as a failure in progress, not as a warning.

9. Canadian realities: code, climate and supply

Three local constraints routinely change a specification that looked correct on paper.

Installation and code

Installing a UPS and its battery falls under the Canadian Electrical Code, CSA C22.1, adopted with provincial amendments, including the Quebec Construction Code, Chapter V [4]. The points that come up at inspection are working clearance around the unit, identification of disconnecting means, and ventilation of the battery room. A maintenance bypass added after the fact almost always means rework of the wiring.

Climate

Canadian sites combine two opposing constraints. In winter an unheated technical room can fall below the operating range of lead-acid batteries, reducing available capacity at precisely the time of year when utility outages are most likely. In summer a room without dedicated cooling easily exceeds the 20 to 25 degree range that serves as the reference for lead-acid batteries [2]. The measurement that matters is not building temperature, it is the temperature at the battery.

Supply and end of support

A Canadian fleet almost always includes units the manufacturer no longer supports. At that point the lead time on an imported part, not its price, becomes the dominant factor in time to restore. That is the reason an independent multi-brand provider exists: GDF Technologies maintains and repairs APC, Eaton, Powerware, Tripp Lite, Schneider, Vertiv, Liebert, Mitsubishi and Delta ranges, including models withdrawn from the manufacturer’s catalogue. See the Eaton UPS service and model guide for an example of how legacy ranges are handled.

10. Failure modes, limits and safety

A UPS moves risk, it does not remove it. The failure modes that matter are few and predictable.

  • Undetected end-of-life battery. The unit displays a calculated theoretical runtime, not a measured capacity. Only a discharge test or an impedance measurement reveals the real state.
  • Ignored persistent bypass. The installation looks healthy because the load is running, while protection has been out of service for weeks.
  • Creeping overload. Successive additions cross the watt limit without anyone redoing the calculation in Table 3.
  • Single point of failure downstream. A redundant UPS pair fed through a single distribution bus does not deliver the availability claimed for it.
  • Frequency drift on generator. A VI or VFD class unit transfers to battery on every load step and exhausts its runtime before the outage ends.

Safety boundary

Two operations sit outside the scope of the operator. Replacing a cabinet battery string, where the DC string voltage can reach several hundred volts even with the unit switched off, and any operation on the maintenance bypass of a three-phase unit. That work requires an arc flash risk assessment and the corresponding protective equipment under CSA Z462 [5]. A plug-in cartridge in a desktop single-phase UPS does not fall into this category; a cabinet string does.

11. Decision checklist before you commit

  1. The load total in watts, growth margin included, is below the watt limit of the unit, and not merely below its VA rating.
  2. The IEC 62040-3 class is known for every normal mode the unit offers, not only for the most flattering one.
  3. The waveform in stored energy mode is sinusoidal if any active power-factor-corrected supply is connected.
  4. Does the site run on a generator? If so, the VFI class is required in normal mode.
  5. Target runtime is validated on the manufacturer’s discharge curve at the real load, not at full load.
  6. Battery room temperature is measured and sits inside the specified range.
  7. An external maintenance bypass is provided, or its absence is accepted explicitly along with the planned outage it imposes.
  8. Parts availability and end-of-support status are verified before purchase, not at the first failure.
  9. Alarms reach a person, by network or dry contact, and someone is named to respond to them.
  10. The wear-part replacement schedule in Table 4 is budgeted from the day of acquisition.

12. Records to keep

A complete equipment file shortens diagnosis and provides the evidence expected during an audit or a warranty claim.

  • Exact model, serial number, commissioning date and firmware version.
  • Dated load inventory in watts and VA, updated whenever equipment is added.
  • Battery installation date, exact cell or cartridge reference, and the result of the last capacity test.
  • Exported alarm history, including the dates of every transfer to bypass and to battery.
  • Single-line diagram showing the position of the maintenance bypass and the disconnecting means.
  • Previous service reports and parts replaced, with the measurements recorded.

These items are also the minimum a provider should require before pricing a contract. Our preventive maintenance programme establishes them on the first visit. When an alarm appears before that, the APC fault code guide and the Eaton alarm and fault code guide let you qualify the urgency before you call.

13. Sources

  1. Schneider Electric, Energy Management Research Center, The Different Types of UPS Systems, White Paper 1, version 10, 2024. IEC 62040-3 classification, the four topology descriptions, tables 3 and 4. se.com
  2. Schneider Electric, Data Center Science Center, FAQs for Using Lithium-ion Batteries with a UPS, White Paper 231, revision 1, 2018. Specified temperature ranges for lead-acid and lithium-ion. se.com
  3. Schneider Electric, Guidance on What to Do with an Older UPS, 2025. Table 4, typical service life of replaceable parts. se.com
  4. CSA Group, CSA C22.1, Canadian Electrical Code, Part I. Adopted in Quebec through the Construction Code, Chapter V, Electricity. csagroup.org
  5. CSA Group, CSA Z462, Workplace Electrical Safety. Arc flash risk assessment and protective equipment. csagroup.org
  6. IEEE, IEEE 1188, Recommended Practice for Maintenance, Testing, and Replacement of Valve-Regulated Lead-Acid Batteries for Stationary Applications. Replacement criterion based on measured capacity. standards.ieee.org
  7. International Electrotechnical Commission, IEC 62040-3, Uninterruptible power systems, method of specifying the performance and test requirements. webstore.iec.ch
  8. Schneider Electric, Comparing UPS System Design Configurations, 2016. Redundant configurations and availability. se.com

The load values in Table 3 illustrate the method and do not come from any particular installation.

Have your installation verified

GDF Technologies sizes, installs and maintains UPS systems from every major brand across Canada, including models the manufacturer no longer supports.

514 252-8324 · support@gdftech.com

Close Menu