Datacenter
UPS and STS commissioning hold points for data centers
Walk a UPS and static transfer switch from de-energized cold checks through energization, burn-in, and witnessed transfer tests that prove the critical load never drops.
Direct answer
A commissioning hold point is a witnessed checkpoint where work stops until a specific check passes and is signed off. For UPS and static transfer switch (STS) systems, hold points gate energization on closed cold checks and prove every power transfer carries the critical load with no break. The project spec and manufacturer control acceptance.
Key takeaways
- A commissioning hold point stops work until a specific check is witnessed and signed; cold checks gate energization, and burn-in gates transfer tests.
- A static transfer switch transfers in roughly 2 to 4 milliseconds, sub-cycle, using SCRs, invisible to server power supplies that ride through about 10 to 20 ms per the ITIC curve.
- Battery capacity acceptance is commonly held to about 90 percent or higher of rated capacity, but the exact threshold, rate, and end voltage come from the manufacturer and spec.
- Maintenance bypass is make-before-break (load path overlaps); the STS is break-before-make so it never parallels two out-of-phase sources.
- A charged battery string holds full DC voltage at its terminals regardless of breaker position; treat DC terminals as live and use insulated tools.
Commissioning hold points, and why critical power is built on them
A commissioning hold point is a defined checkpoint where the work stops until a specific verification is done, witnessed, and signed. Nothing past that point happens until the gate is cleared. On critical power, hold points are not a formality. They are the reason a UPS or a static transfer switch gets installed wrong on Tuesday and caught before it ever carries a server load.
Critical power is built around hold points because the whole job is to keep the load up when something upstream fails, and you cannot prove that after the fact. You prove it on purpose, in stages, before the room is live. The sequence runs cold to hot the same way the rest of the data center does: de-energized static checks first, then controlled energization, then the functional transfer tests that force the failures the system was bought to survive.
The thing that makes this trade unforgiving is that the equipment looks finished long before it is commissioned. A UPS that powers up, syncs, and carries a load looks done. It is not done until it has transferred to bypass and back under load with no break, run its battery down at design load and made its autonomy, and had every one of those results witnessed and recorded against the spec. A UPS that was energized but never had its transfers proven was not commissioned. It was just turned on.
UPS topologies: why data centers run double-conversion
A UPS is an uninterruptible power supply: it conditions incoming power and carries the load through a source disturbance from stored energy, usually a battery. There are three common topologies, and they are not interchangeable. Standby (offline) UPS runs the load straight off utility and flips to inverter only when utility fails, which leaves a short break on transfer. Line-interactive sits in the middle, regulating voltage with a tap-changing transformer and switching to battery on a real outage, still with a brief transfer. Double-conversion (online) runs the load through the rectifier and inverter continuously, so the battery is always in the path and there is no transfer break when utility drops.
Data centers run double-conversion almost without exception, because the load never sees the utility directly. The rectifier and inverter rebuild the waveform, so sags, swells, frequency drift, and harmonics on the incoming line do not reach the IT load, and the loss of utility is a non-event because the DC bus is already feeding the inverter. The cost is efficiency, since you are converting the power twice, and that is what eco-mode and the efficiency criteria later in this guide are about.
The other split is physical: modular versus monolithic. A monolithic UPS is one large frame with a fixed rating. A modular UPS is a frame populated with hot-swappable power modules, so capacity scales by adding modules and a failed module pulls without dropping the load. Commissioning a modular unit adds checks the monolithic one does not have: module-to-module load sharing, the redundancy logic when a module is pulled, and that the frame still carries rated load with a module out if the design claims N+1 inside the cabinet.
The UPS power path: rectifier, inverter, static bypass, maintenance bypass
You cannot commission transfers you do not understand, so know the path. Power enters at the rectifier, also called the charger, which turns incoming AC into DC. The rectifier does two jobs: it feeds the inverter and it floats the battery at its charge voltage. The inverter takes that DC and builds clean, regulated AC for the critical load. In a double-conversion UPS this path carries the load all the time, and the battery sits on the DC bus ready to hold it up the instant the rectifier loses its source.
The static bypass is the fast escape route. It is a solid-state switch, built on SCRs, that can move the load from the inverter to the raw bypass source in a fraction of a cycle when the inverter cannot support the load: an overload, an internal fault, or a downstream fault that needs more current than the inverter can deliver. Because it is solid-state, the transfer is sub-cycle and the load rides through it. For that to be true, the inverter output has to stay synchronized to the bypass source, which is one of the things commissioning verifies.
The maintenance bypass, sometimes called the wrap bypass, is the manual one. It is a set of switches, usually mechanically interlocked, that lets a technician feed the load directly from the bypass source and physically isolate the entire UPS for service. Its defining requirement is make-before-break: the bypass path is established before the UPS path is opened, so the load never loses power during the switch. Proving that sequence, step by step, with the load up the whole time, is a hold point on its own.
VRLA vs lithium-ion for UPS batteries
The battery is the stored energy that makes a UPS uninterruptible, and the chemistry changes how you commission and live with it. VRLA, valve-regulated lead-acid, has been the data center default for decades: lower first cost, familiar, and sealed so it needs no watering. The catch is life and behavior. VRLA design life is often quoted at 10 years but real service life runs shorter, it is sensitive to temperature, every 8 to 10 degree C rise roughly halves its life, and it can go into thermal runaway if overcharged or run hot, venting hydrogen as it does.
Lithium-ion, in UPS service usually lithium iron phosphate (LFP) for its thermal stability, has taken a large share of new data center builds. It lasts longer, commonly 10 to 15 years, takes far less floor space and weight for the same energy, charges faster, and tolerates higher ambient temperatures. It costs more up front and it comes with an integral battery management system (BMS) that is part of the protection, not an accessory. Commissioning a lithium system means commissioning that BMS: addressing, communication, and that its trips actually open the battery breaker.
Neither chemistry is plug-and-play at acceptance. VRLA wants its float and equalize voltages confirmed against the manufacturer's curve at the measured battery temperature, because the wrong float cooks it or starves it. Lithium wants the BMS limits and the UPS charger profile matched to the cell maker's spec, because the UPS and the BMS both think they own the battery and they have to agree. Either way, the autonomy you bought is only real once it is proven on a discharge test.
Battery commissioning and the capacity test
Battery commissioning proves the string is installed right, charged right, and able to deliver its rated energy, and the centerpiece is the discharge capacity test. You charge the string fully, discharge it at a defined constant current or constant power to a specified end-of-discharge voltage, and compare the delivered capacity to the rating. An acceptance test is commonly held to a percentage of rated capacity, with figures on the order of 90 percent or higher treated as acceptable and a result below that flagging cells that will not make it to the next test. The exact threshold, rate, and end voltage come from the manufacturer and the project spec, not a rule of thumb.
IEEE recommended practices frame the battery work by chemistry and topic. IEEE 1184 covers selection and sizing of batteries for UPS systems. For maintenance, testing, and capacity verification, IEEE 1188 covers valve-regulated lead-acid, the vented lead-acid practice covers flooded cells, and IEEE 1106 covers vented nickel-cadmium. Lithium follows the manufacturer's and BMS maker's procedures, with the IEEE 1679 series giving the characterization and evaluation framework. Cite the one that matches the chemistry actually on the floor.
Before the discharge, take and record the baseline: individual cell or unit float voltages, the ohmic value (impedance or conductance) of each cell as the reference for future trending, intercell connection resistance against the manufacturer's limit, electrolyte level on flooded cells, and battery and room temperature. The connection resistance check is the one crews rush. A high-resistance intercell link runs hot under discharge current, drops voltage where you need it most, and is the failure that shows up exactly when the utility drops and the string has to perform.
Battery thermal, ventilation, and safety verification
The battery is the one part of a UPS that stores enough energy to hurt you with the system completely de-energized, so its safety verification is its own hold point. A charged string sits at full DC voltage at its terminals no matter what any breaker is doing, and a large string can deliver a fault current that vaporizes a dropped wrench. Treat the DC terminals as live the entire time, use insulated tools, and follow the arc-flash and shock boundaries the study calls for.
Confirm the conditions that keep the chemistry safe. For VRLA and flooded lead-acid, verify the room ventilation actually moves air and holds hydrogen below the lower explosive limit, because lead-acid off-gasses hydrogen on charge and a sealed, unventilated battery room is an explosion waiting on a spark. Confirm temperature monitoring is live and the charger is temperature-compensated, since a hot string and a fixed float voltage walk each other into thermal runaway. For lithium, confirm the BMS temperature trips, the room detection the design calls for, and that the listing and fire protection match what was specified.
Verify the rest of the life-safety stack is real, not just present: the spill containment and neutralization for flooded electrolyte, the eyewash, the battery disconnect and its labeling, and that the room signage and the disconnecting means match the code and the AHJ. A battery monitoring system that trends cell voltage, impedance, and temperature earns its place here, because the cell that fails the next capacity test usually shows it in the impedance trend months ahead if anyone is watching.
What are the cold checks before a UPS is energized?
Cold checks are everything you verify with the UPS de-energized, and this is where most real defects get caught, because once the system is hot the cost and risk of fixing anything climbs fast. Start with install verification against the drawings: the unit is the right rating and configuration, it is anchored per the seismic detail, working clearances and egress match the listing and the code, and the cable entries and bus landings are where the design put them.
Then the electrical integrity. Torque every power connection to the manufacturer's value, follow the specified sequence, and put a witness mark on each joint so a later infrared scan can tell you if one moved. Confirm phase rotation and polarity, including the DC polarity to the battery, because reversed DC into a charger is destructive. Verify grounding and bonding: the cabinet, the battery rack, and the bypass are bonded to the equipment grounding system and the bonds are landed and tight, not just hanging there. Megger or insulation-test the sections the manufacturer allows, by their method.
Finish with the parts that make it a system, not a box. Ring out the control wiring point to point against the schematic, because a UPS that powers up with a control wire on the wrong terminal will mis-transfer later in a way that is miserable to find hot. Confirm the communications and monitoring integration, the SNMP or Modbus to the building management system or DCIM and the BMS link on a lithium plant, reports the right points. And verify the bypass interlocks mechanically before any power is present, because an interlock you prove cold is one you are not discovering hot.
The maintenance bypass interlock gets proven de-energized, by hand, before the system can be trusted. The interlock, often a Kirk-key or captive-key scheme or a hard mechanical sequence, exists to force the operator through the steps in an order that never opens the load path and never back-feeds a dead UPS. Walk the full sequence cold and confirm the keys, the mechanical stops, and the switch positions only allow the make-before-break order and physically block the wrong one. This is a step crews are tempted to wave through because it is tedious and the switches feel obvious. Do not. The order of operations on the door placard and the actual mechanical behavior of the switches have to agree exactly, and where they disagree you find out which one is wrong now, with the gear dead, not during a live bypass with the data hall up.
Energization, startup, and the burn-in period
Energization is a controlled, sequenced startup, not a breaker thrown by one person, and on most projects the manufacturer's field service engineer performs and documents the initial startup as a condition of the warranty. The cold checks have to be closed first. Bring the unit up in the maker's order: confirm the rectifier walks in softly instead of slamming the DC bus, the DC voltage comes up to the float setpoint, the inverter starts and synchronizes, and the unit carries its own internal checks before any real load is on it.
Burn-in is the soak that shakes out infant mortality. You run the UPS loaded, usually with a load bank, for a defined period to catch the components that fail early, the connection that was not quite tight, and the cooling that cannot hold temperature over hours. The duration is set by the spec and the manufacturer, with periods on the order of 24 to 72 hours of continuous loaded operation common on critical projects, sometimes longer. A unit that runs clean for an hour and a unit that runs clean for two days are different levels of confidence, and the soak is cheap compared to a failure in production.
Infrared comes after load, not before. Scan the power connections, the bus, the bypass, and the battery intercell links with the system loaded and warm, because a loose or high-resistance joint only shows its heat when current is flowing through it. Compare the witness marks from the cold torque check, and any joint reading hot against its neighbors gets re-torqued and re-scanned. A clean thermal scan at full load is one of the most useful pieces of paper in the turnover package.
What transfer tests prove a UPS is commissioned?
The transfer tests are the heart of UPS commissioning, because they prove the system does the one thing it exists to do: keep the load up through a disturbance. They are run live and loaded, and each one is a witnessed hold point with a pass criterion. The first is the loss-of-utility test: open the UPS input and confirm the load transfers to battery with no break, the unit annunciates the right alarm, and the output never breaks. Then return utility and confirm the rectifier walks back in and recharges without disturbing the load.
Next is the static bypass transfer and return. Force a condition that sends the load to static bypass, an overload or a simulated inverter fault per the manufacturer's test, and confirm the load moves to the bypass source sub-cycle with no interruption, then transfers back to inverter cleanly. This is where synchronization matters: if the inverter is not locked to the bypass source, the transfer is not bumpless, and the test exists to prove it is. Verify the unit does the right thing on a bypass that is out of tolerance, which is to stay on inverter or alarm, not to dump the load onto bad power.
The whole point is the no-break requirement, so you have to watch the output, not just the alarm panel. Put a power quality recorder or a scope on the critical bus and confirm the output voltage stays inside the ITIC (CBEMA) tolerance envelope through every transfer, with no sag or dropout that a server power supply would see. A transfer that the front panel calls successful while the output dipped below the ITIC curve is a failed transfer that lied to you. The waveform is the evidence; the alarm light is not.
The maintenance bypass make-before-break test
The maintenance bypass test proves you can take the entire UPS out of service and put it back without the load ever knowing, and it is run live with the critical load up. Working the interlock sequence you proved cold, transfer the load to the maintenance bypass source, confirm on the output that the load stayed up through every switch movement, then fully isolate the UPS, then reverse the steps and bring the UPS back into the path. The success criterion is simple and absolute: the load never dropped.
Make-before-break is the requirement that makes this safe, and the test confirms the hardware actually behaves that way and not just on paper. The bypass path is energized and carrying load before the UPS path opens, so there is overlap, not a gap. Out of sequence, you either back-feed a UPS that should be dead or you open the only live path to the load. The interlock is supposed to make the wrong order impossible, and this test, with the load watching, is where you confirm it does. Anyone running this for the first time on a live data hall without having proven the sequence cold is gambling the room on a placard.
How fast does a static transfer switch transfer?
A static transfer switch transfers the load from one source to another in roughly 2 to 4 milliseconds, which is sub-cycle, within a quarter of a cycle (a quarter cycle is about 4 milliseconds at 60 Hz). It is that fast because it is solid-state: the switching elements are SCRs (thyristors), not mechanical contacts, so there is no contact travel time. That speed is the whole reason an STS exists in a data center. A server power supply rides through a power loss on the order of 10 to 20 milliseconds per the ITIC curve, so a sub-cycle source swap is invisible to the load where a mechanical automatic transfer switch, taking many cycles, would drop it.
An STS sits between two independent sources, a preferred and an alternate, and feeds a downstream bus, commonly the input of a PDU or a rack. It monitors the preferred source and, when that source goes out of tolerance, transfers the load to the alternate within that few-millisecond window, then transfers back when the preferred source recovers and is healthy. Commissioning proves the transfer time meets the spec, the sense-and-decide thresholds are set right, and the unit does not chatter back and forth on a marginal source.
The operation is break-before-make between the two sources, and that detail drives a hard commissioning requirement. The STS opens the SCRs on the outgoing source before it closes the SCRs on the incoming source, because closing both would parallel two independent and possibly out-of-phase sources and dump cross-current between them. For the break-before-make to still look unbroken to the load, the two sources have to be synchronized within the unit's phase and voltage window, so commissioning verifies the sync-check and confirms the STS blocks or alarms a transfer when the sources are too far out of phase rather than transferring into a fault.
STS commissioning: sources, neutral, and retransfer
Beyond transfer time, an STS has a short list of checks that decide whether it helps the load or hurts it. Confirm the preferred and alternate source assignment and that both sources are correct, in phase rotation and within the configured tolerance, at the input. Set and verify the sensing thresholds, the voltage and frequency windows that decide a source is bad, against the spec, because thresholds set too loose let bad power through and too tight cause nuisance transfers. Verify the manual transfer function and that a transfer initiated by hand behaves like an automatic one.
Neutral handling is the trap on an STS. Most static transfer switches switch the three phases only and rely on a common, solidly connected neutral shared by both sources, because switching the neutral on a sub-cycle device creates its own problems. If the design uses sources with separately derived neutrals, that is a grounding and bonding question that has to be resolved before energizing, since a momentarily floating or doubly-switched neutral during transfer loses the voltage reference for the load. Confirm the neutral and grounding arrangement matches the design and does not create objectionable neutral current or a lost neutral on transfer.
Verify the retransfer logic and timing. After the preferred source recovers, the STS should wait a set, confirmed-stable period before transferring back, so it does not bounce the load onto a source that is recovering in fits. Set the retransfer delay per the spec, confirm the manual-versus-automatic retransfer mode, and prove the unit will not hand the load to a source it has not confirmed healthy. An STS that retransfers too eagerly turns one source disturbance into two transfers the load has to survive.
Load banks and the integrated systems test
You cannot prove a UPS or an STS carries its rated load by hoping, so commissioning uses load banks to apply real, measured, adjustable load through the burn-in, the transfer tests, and the battery discharge. A resistive load bank covers most of it; where the spec wants the unit proven against a realistic power factor and harmonic profile, a reactive or nonlinear load bank is used so the output voltage regulation and THD are tested under the kind of load IT gear actually presents, not a clean resistor. Load bank acceptance criteria, the steps, durations, and tolerances, are their own subject and are worth setting before the test, not during it.
Standalone testing is necessary and not sufficient. The last and most demanding gate is the integrated systems test, the IST, the scripted, loaded test of the whole power chain working together: utility, generators, switchgear, the UPS, and the STS as one system reacting to forced failures. The Uptime Institute frames this kind of full load-bank-backed integrated test as the proof of the design intent, and it is the same IST that closes out the generator and the larger electrical commissioning scope.
The IST finds the faults that each device passing alone will hide. The UPS rides through a utility loss fine on its own, the generator starts fine on its own load bank, and then under the real, scripted utility failure the sequence is a half second off, the generator does not accept the block in time, the UPS battery is carrying longer than planned, and the margins you thought you had get spent. You only catch that with the live, loaded, scripted failure, and a thin script produces a clean report that proves very little. That is exactly the cross-link to the power-QA scope of the project.
Verifying battery runtime at design load
Autonomy is the number the owner actually bought, and it only counts at design load. Runtime, the minutes the UPS can carry the load on battery before the source has to come back or the generator has to pick up, scales hard with load: a string that holds 15 minutes at half load may hold far less than half that at full load, because the discharge is nonlinear. So the autonomy verification runs at the design load the room will actually see, not at whatever load is convenient on the day.
Run the discharge with a load bank set to design load, time the battery from full charge to the end-of-discharge voltage, and compare the measured runtime to the design autonomy. In a data center the design autonomy is usually short on purpose, often in the range of a few minutes to around 15 minutes, sized to ride through the gap until the generators start and accept load rather than to run the building for an hour. Confirm what the spec sized the battery for and prove that number, because the battery exists to bridge to the generator, and if it cannot make the bridge the generator start time becomes a hole in the load.
Record the discharge curve, not just the endpoint. The shape of the voltage versus time tells you whether the string is healthy or whether a weak cell is dragging it down early, and it is the baseline future capacity tests trend against. A runtime that just barely makes the design number on a brand-new battery has no margin for the aging every battery does, so a marginal pass at acceptance is a finding, not a celebration.
Eco-mode and efficiency verification
Eco-mode is the efficiency play, and it changes what you have to prove. In double-conversion the load runs through the rectifier and inverter continuously, giving clean output at an efficiency commonly in the mid-90s percent. In eco-mode the UPS runs the load on the bypass line and keeps the inverter on standby, conditioning only when needed, which pushes efficiency up toward 99 percent. The savings are real, and so is the catch: in eco-mode the inverter is not already carrying the load, so on a disturbance the unit has to bring the inverter in fast enough that the load still rides through.
Commissioning eco-mode means proving that transition. Force a line disturbance with the unit in eco-mode and confirm it transfers the load to inverter quickly enough that the output stays inside the ITIC envelope, the same no-break test as everywhere else in this guide, just with the harder starting condition. Verify the efficiency figures at the load points the spec calls out, in both double-conversion and eco-mode, with calibrated metering. If the eco-mode transfer cannot hold the load on the disturbances the site will see, the right answer may be to run double-conversion and bank the reliability over the efficiency, which is a project decision, not a default.
What are the acceptance criteria for UPS and STS commissioning?
Acceptance is judged against measured performance and witnessed transfers, not against the fact that the system powered up. The UPS has to hold output voltage and frequency within tolerance under load, keep output distortion within limits on the specified load profile, transfer to and from battery and bypass with no break, make its rated efficiency, and deliver its design autonomy at design load. The STS has to transfer within its rated time, sense and decide at the set thresholds, and not transfer into an out-of-sync source.
The table below gives typical ranges so you know roughly where the lines fall and what to watch. They are not numbers you can cite as a standard. The voltage and frequency tolerances, the THD limit, the efficiency and eco-mode targets, the transfer times, and the autonomy all come from the manufacturer's data sheet and the project specification, and a Tier-rated critical facility may hold tighter than the generic figures here. Hold to the contract numbers, fall back to the manufacturer's published values where the spec is silent, and get any gap resolved in writing before turnover.
One criterion gets lost in the rush to a passing number: the system has to hold, repeatably, at the conditions the site will actually run. A transfer that worked once with the room cool and the load light is not the same as a system that transfers clean at full load after a burn-in soak. Where the spec allows, prove the transfers more than once and at load, because the failure you want is the one on the commissioning floor, not in production.
| Acceptance item | What it proves | Typical range (verify against spec and mfr) |
|---|---|---|
| UPS output voltage regulation | Inverter holds the bus under load | Often within about plus or minus 1 percent static |
| UPS output frequency | Output is stable and synced to bypass | Within a fraction of a Hz, synced when on bypass |
| Output voltage THD | Clean waveform under real load | Commonly under 3 to 5 percent per load type |
| UPS-to-battery and bypass transfers | Load rides through with no break | No dropout outside the ITIC (CBEMA) envelope |
| UPS efficiency, double-conversion | Conversion losses at rated load | Commonly mid-90s percent, per data sheet |
| Eco-mode efficiency and transfer | Savings without losing the load | Toward 99 percent, transfer within ITIC |
| STS transfer time | Source swap is invisible to the load | About 2 to 4 ms, sub-cycle |
| Battery autonomy at design load | Bridges to the generator | Per spec, often minutes to about 15 min |
Witnessed sign-off and the hold-point register
A hold point is only worth the witness behind it, so the spine of the whole process is the hold-point register: the tracked list of every checkpoint, what proves it, who has to witness it, and whether it is closed. Each entry names the check, the acceptance criterion, the test record that backs it, the date, and the signatures of the parties whose witness the spec requires, commonly the commissioning agent, the owner or their representative, and the manufacturer's field engineer for the startup items.
The register is what enforces the cold-to-hot sequence. Energization is gated on the cold hold points being closed and signed. The functional transfer tests are gated on a successful energization and burn-in. Turnover is gated on every functional and performance hold point being witnessed and the open items resolved. A register that lets a hot test happen with a cold hold point still open has failed at the one job it has.
Track every item to closure with its result, and where something was adjusted and re-tested, record the before and after so the next person can see what changed and that it was proven after the change. A register with items quietly marked complete and never witnessed is worse than no register, because it looks like proof and is not. The signature at the end is only worth what the tracking behind it is worth.
What to document
The commissioning record is what a future operator and the warranty both rely on, so it has to let someone who was not on site reconstruct the test and check the result against the spec. Capture the cold checks, the startup, the burn-in, the battery baseline and discharge, every transfer test, the STS checks, and the hold-point register with its sign-offs, all tied to the specific equipment by serial number and rating.
Record enough that the result is defensible years later: the torque and insulation values with the witness marks, the battery float voltages, ohmic baseline, and connection resistances, the capacity and runtime results with the load and temperature they were taken under, the transfer waveforms or power-quality captures that prove no break, the STS transfer time and threshold settings, the efficiency measurements, and who witnessed each acceptance. If anything was re-torqued, re-set, or re-tested, record the before and after.
| Field to record | Why it matters |
|---|---|
| Equipment IDs and ratings (kVA, kW, V, PF) | Ties the record to the specific UPS, STS, and battery |
| Torque and insulation values, with witness marks | Documents the cold electrical integrity |
| Battery float voltage, ohmic baseline, connection resistance | The reference future capacity tests trend against |
| Capacity and runtime at design load, with conditions | Proves the autonomy that was bought |
| Transfer waveforms or power-quality captures | Evidence the load rode through with no break |
| STS transfer time, thresholds, and retransfer settings | Proves the source swap and the logic behind it |
| Efficiency, double-conversion and eco-mode | The energy performance against the data sheet |
| Hold-point register: criterion, result, witness | The tracked gates and the sign-off |
Common mistakes
- Skipping a hold point to keep the schedule, then energizing or transferring on an unverified path.
- Calling a transfer successful from the front-panel alarm without watching the output waveform against the ITIC curve.
- Accepting the battery on faith because it charged, instead of running the discharge capacity test.
- Running the runtime test at light load and reporting an autonomy the battery cannot make at design load.
- Proving the maintenance bypass interlock for the first time hot, on a live load, instead of cold.
- Transferring an STS into an out-of-sync source because the sync-check was never verified or was set too loose.
- Treating the battery DC terminals as dead because breakers are open, when a charged string holds full voltage regardless.
- Leaving the static or maintenance bypass path unproven, so the escape route is unknown when it is needed.
- Skipping the post-load infrared scan, so a high-resistance joint only announces itself under production load.
- Testing the UPS and STS standalone and never running the integrated systems test that proves the real transfer.
Field checklist
Want this checklist to run itself on every job — with photo proof and a signed record crews can hand the customer? That's FieldOS.
Standards and references
Several bodies govern different parts of UPS and STS commissioning, and naming the right one for the point is what separates a credible record from a guess. For acceptance testing of the installed electrical equipment, ANSI/NETA ATS gives the field test and inspection requirements, and recent editions added coverage specific to UPS and battery energy storage systems. ANSI/NETA ECS covers the commissioning process and recent editions added source-specific commissioning sections for UPS and transfer switches. Use the edition the project specifies.
On the battery side, IEEE recommended practices frame the work by chemistry and topic: IEEE 1184 for selection and sizing of UPS batteries, IEEE 1188 for maintenance, testing, and replacement of valve-regulated lead-acid, the vented lead-acid practice for flooded cells, and IEEE 1106 for vented nickel-cadmium. The IEEE 1679 series covers characterization and evaluation that lithium systems lean on. The NEC, NFPA 70, covers the installation, with the storage-battery and emergency and standby power articles applying by topic, and NFPA 70E covers the arc-flash and shock work practices that apply to the DC battery and the AC gear.
Where the UPS is part of an emergency or legally required standby system, NFPA 110, the standard for emergency and standby power systems, applies to that scope and its testing, so confirm whether the installation falls under it. The Uptime Institute frames the integrated systems test as proof of the design intent for critical facilities. Above all of these sit the manufacturer's instructions and the project specification, which set the actual numbers, and the authority having jurisdiction, which has the final say on what is enforceable. When a standard and the spec disagree, the stricter controlling document wins.
Units and terms
Critical power work runs on a handful of units and acronyms, and reading the wrong one is how a result gets accepted that should not be. A UPS is rated in kVA for apparent power and kW for real power, and the ratio is the power factor; a load that draws more kW than the kVA rating allows at the unit's power factor overloads it even if the kVA looks fine. Autonomy is the runtime in minutes the battery carries the load. Transfer time is in milliseconds and cycles, with one 60 Hz cycle equal to about 16.7 milliseconds.
Output quality shows up as voltage regulation in percent, frequency in Hz, and total harmonic distortion (THD) in percent. Eco-mode is the high-efficiency mode that runs the load on bypass with the inverter on standby. Make-before-break (MBB) describes the maintenance bypass, where the new path is made before the old is broken; break-before-make (BBM) describes the STS source swap, where the old is broken before the new is made. Keep MBB and BBM straight, because they describe two different transfers with two different reasons.
- UPS
- Uninterruptible power supply, which conditions power and carries the load through a source disturbance from stored energy
- STS
- Static transfer switch, a solid-state switch that moves the load between two sources sub-cycle
- kVA / kW
- Apparent power and real power; their ratio is the power factor that the UPS rating depends on
- Autonomy
- Battery runtime in minutes carrying the load, verified at design load
- THD
- Total harmonic distortion, the output waveform's distortion as a percentage
- Eco-mode
- High-efficiency mode running the load on bypass with the inverter on standby
- MBB / BBM
- Make-before-break for the maintenance bypass; break-before-make for the STS source swap
FAQ
What is a commissioning hold point on a UPS?
A commissioning hold point is a witnessed checkpoint where the work stops until a specific verification passes and is signed. On a UPS it gates the sequence: cold checks before energization, energization and burn-in before transfer tests, and every transfer proven before turnover. The register tracks each gate and its sign-off.
How fast does a static transfer switch transfer?
A static transfer switch transfers in roughly 2 to 4 milliseconds, sub-cycle and under a quarter cycle, because it switches on SCRs instead of mechanical contacts. That speed keeps the load up, since server power supplies ride through about 10 to 20 milliseconds per the ITIC curve. The spec and manufacturer set the rated time.
VRLA or lithium-ion for a UPS battery?
VRLA lead-acid costs less up front but lasts shorter, is heat-sensitive, and off-gasses hydrogen. Lithium-ion, usually LFP, lasts 10 to 15 years, takes less space and weight, tolerates higher temperatures, and carries an integral BMS, at higher first cost. Either way the autonomy is only real once a discharge capacity test proves it.
What if a transfer test drops the load?
A dropped load on a transfer test is a failed hold point, not a footnote. Stop, find the cause, often a synchronization fault between inverter and bypass, a control miswire, or an interlock out of sequence, fix it, and re-run the transfer with the output watched against the ITIC curve. Do not energize the room on an unproven transfer.
Why do data centers use double-conversion UPS?
Double-conversion (online) UPS runs the load through the rectifier and inverter continuously, so the IT load never sees the utility and loss of utility is a non-event, since the battery is already on the DC bus. Standby and line-interactive topologies leave a brief transfer break on a real outage, which double-conversion avoids at the cost of efficiency.
How do you test UPS battery runtime?
Charge the battery fully, apply the design load with a load bank, and time the discharge to the end-of-discharge voltage, comparing the result to the design autonomy. Runtime scales nonlinearly with load, so test at design load, not light load. Record the full discharge curve as the baseline future capacity tests trend against.
What is the difference between static bypass and maintenance bypass?
Static bypass is the automatic solid-state path that moves the load to the bypass source sub-cycle when the inverter cannot support it. Maintenance bypass is the manual, interlocked path that isolates the whole UPS for service, operated make-before-break so the load stays up. Commissioning proves both: the automatic transfer and the manual sequence.
Does NETA cover UPS and STS commissioning?
ANSI/NETA ATS gives the field acceptance test and inspection requirements for the electrical equipment, and recent editions added UPS and battery storage coverage. ANSI/NETA ECS covers the commissioning process with source-specific sections for UPS and transfer switches in recent editions. The project specification and manufacturer set the actual acceptance numbers, with the AHJ governing enforceability.
Why is an STS break-before-make and a maintenance bypass make-before-break?
An STS is break-before-make so it never parallels two independent, possibly out-of-phase sources, which would dump cross-current; it relies on synchronized sources to stay smooth to the load. A maintenance bypass is make-before-break so the load path overlaps and the load never loses power while the UPS is isolated. The two transfers solve different problems.
Is a UPS battery safe to work on once the breakers are open?
No. A charged battery string sits at full DC voltage at its terminals regardless of breaker position, and a large string can deliver an arc-flash-grade fault current. Treat the DC terminals as live, use insulated tools and the study's PPE, and follow the manufacturer's procedure. Lead-acid also off-gasses hydrogen, so confirm the room ventilation works.
People also ask
Codes cited in this guide
This guide is written and reviewed against the published standards below. Always confirm the current adopted edition with the authority having jurisdiction.