What N means, and why every other notation depends on it
Start with N, because everything else derives from it. N is not a number of cabinets. N is the capacity required to carry the design load — the smallest set of elements that supports it with nothing spare. Everything after the plus sign is capacity the load does not need.
N therefore only exists once a design load exists. Two 250 kW units carrying 400 kW are not N+1; they are N plus some headroom, and if either stops the survivor is overloaded. The same two units carrying 200 kW are N+1. The hardware never changes; the load does.
So a tender line reading "N+1" with no stated design load has not specified redundancy. It has invited each bidder to choose their own N, and read as nameplate rather than as measured load that choice alone can double the system being priced.
N+1, 2N and 2N+1: what each configuration adds
Three notations cover almost every quotation you will receive. What separates them is whether they duplicate capacity or the whole path.
- N+1 — one spare capacity element
- One capacity element more than the load needs. That element can be a whole UPS on a parallel bus or a power module inside one frame, and the difference matters. Granularity is the commercial argument for hot-swap modular UPS systems: unit-level N+1 buys a redundant machine sized for the whole load, module-level N+1 buys one module. With CumulusPower's 10/20/25 kVA=kW modules and 1 to 10 modules per frame, a 200 kW load takes eight 25 kW modules for N and nine for N+1.
- 2N — two independent systems
- Two complete systems, each able to carry the whole load alone, each with its own input feed, batteries, output and bypass. What you buy is not extra capacity but a second path: a whole side can be de-energised — switchboard, cabling, bypass — while the load stays on the other.
- 2N+1 — two systems and a spare, or two N+1 systems
- The notation that causes the most argument, because it is used for two different designs. It can mean two independent systems plus one additional capacity element, so one side can be out for maintenance while an element fails on the other. It can also mean 2(N+1): two independent systems, each internally N+1 — considerably more equipment. Write out what you mean: "two independent systems, each internally N+1, each able to carry 100% of the design load with one module out of service."
Maintaining without dropping the load is not surviving a fault
Two questions get collapsed into one word. Maintaining equipment without dropping the load, and surviving a fault without dropping it, are different properties — and N+1 delivers the first far more readily than the second.
Maintenance is scheduled: you choose the hour, the load level and the spares on the shelf. The only question is whether the work can be done with the load on inverter — which depends on whether it touches something shared.
A fault is not scheduled, and "a module stops" is the friendly version of it. The unfriendly versions are a short circuit on the common output, a control fault that mis-shares current between paralleled elements, a battery earth fault, and any failure that propagates instead of isolating. Spare capacity answers the first kind only. What answers the rest is a failing element's ability to take itself out of circuit cleanly. In the StratusPower modular UPS each module is a complete UPS in its own right, with three independent power converters, its own static bypass and the hardware needed to safely isolate a fault without impacting the load. That property, not the module count, turns spare capacity into fault tolerance.
What N+1 never buys is work on anything the modules share: the frame's output connection, the manual bypass, the input breaker, the output cabling. Opening any of them still means transferring or dropping the load. Two independent systems buy that freedom — so list the tasks you must perform live and see which configuration covers them.
Where UPS redundancy still leaves a single path
Redundancy is a property of a path, not of a box, running from the incoming supply to the load's own power supply. In most designs called redundant, several links in it are still single — and a system that survives a module failure but not a distribution board failure has been optimised for the wrong event.
- The static bypass and the manual bypass
- A centralised UPS has one static bypass on the common output; distributed-module architectures give each module its own. The maintenance path does not follow. StratusPower's Universal Rack build list specifies DC battery MCB protection one per module, bypass fuses three per module and the output parallel isolator one per module — but the system manual bypass is one per system.
- Downstream distribution
- Most events that reach a load happen below the UPS: one board, one breaker, one cord, one power supply inside a single-corded machine. Two UPS systems feeding one board is 2N upstream and N where it counts. Dual-corded loads on separate boards turn two sources into two real paths; anything single-corded needs a transfer device, which becomes the new common element.
- The transfer switch
- A static transfer switch gives a single-corded load two sources, and is itself a component with failure modes. The ENTS Series answers that internally with up to four independent internal power supplies and hot-swappable power units, and its datasheet cites IEC62310-1/2/3. Read the figures literally: transfer type is Break Before Make, at <5ms (Sync) & ~10ms (Unsync) @ 50Hz and <4ms (Sync) & ~10ms (Unsync) @ 60Hz. The fast figure needs synchronised sources; the load rides through the break either way.
- The battery string
- Module redundancy on the AC side says nothing about the DC side. Where several modules draw on one string, that string is shared — and a lead-acid string is 20-50 series-connected blocks, so one open cell opens all of it, usually only on discharge. The counter-design is a battery per module: StratusPower's system diagram shows a separate battery block per module, and DC battery MCB protection is one per module. Whichever you buy from an industrial battery range, the discharge-test regime matters more than the notation on the drawing.
- Control, communications and load sharing
- Paralleled elements must agree how much current each delivers, and that agreement travels over a communication path. Master-slave control concentrates the decision in one place — a single point of failure in the logic even when every power stage is redundant. Distributed decision-making removes it: in the DARA architecture used by both modular families, modules decide collaboratively over a fully redundant triple-mode parallel bus. Ask what happens when that bus is interrupted.
- Cooling, ambient and the room
- 2N electrical on N cooling is N: redundant equipment in a room with one cooling path or one suppression zone inherits that room's single points. Ambient also changes the rating you get — 0-40°C carries no power derating, and above 1000 m the derating is 1% for each additional 100 m. A second complete system also doubles footprint and heat rejection — which is where StratusPower's 1MW per M² earns its keep.
What module-level redundancy covers, and what it does not
A modular frame delivers real redundancy inside itself. Capacity arrives in modules — 10/20/25 kVA=kW or 50/60 kVA=kW on the CumulusPower modular UPS, 10 / 20 / 25 / 30 or 50 / 62.5 kVA=kW on StratusPower — and a frame accepts from 1 to 10 of the smaller modules, or 1-24 in the largest StratusPower cabinet at 1500 kVA=kW nominal per frame. The redundant element is therefore a fraction of the load, and it is fitted hot.
There is often more margin than the module count suggests: the inverter is rated 124% continuous, and a 750 kVA=kW StratusPower frame is rated 900 kVA=kW continuous overload, so a frame below nameplate can lose a module without the survivors reaching their own limit.
Redundancy also has a running cost, because a redundant system permanently operates below nameplate. Published inverter efficiency is 97,6% for StratusPower and 97,1% for CumulusPower; what you want from a bidder is the figure at day-one load in the offered configuration with the redundant modules online. Maximum Efficiency Management puts into Active-Sleep only those modules no longer required to maintain redundancy — the redundancy rule constrains the saving, not the reverse.
What one frame cannot cover is what the modules share: one input connection, one output connection, one manual bypass, one enclosure, one location. Module redundancy is not system redundancy. To survive a switchboard outage, a feeder fault or a full shutdown you need two frames or two systems — which these platforms support, scaling to 3750 kVA=kW per system for StratusPower and 3600 kVA=kW for CumulusPower's IM50/IM60 frames. A 2N design of modular frames gives both: a second path between frames, and N+1 inside each.
N+1 or 2N? The questions that actually decide it
The notation cannot supply the answer. For datacenter and telecom infrastructure the dual-corded case usually pushes toward two independent systems; for a single-corded industrial process the money is often better spent on module-level redundancy plus a properly specified transfer device. Four questions decide it.
- Which tasks must be done with the load live? Module, fan and capacitor work is covered by module-level N+1. A switchboard, a feeder cable, the manual bypass or the room is not.
- How is the load corded? Dual-corded IT collects the full benefit of 2N; single-corded loads do not, and the specification effort moves to the transfer device.
- Is redundancy a fixed rule or a temporary surplus? A site that quietly grows into its spare module is N again, so fix the rule to final load, not commissioning load.
- How long does the system stay degraded after a failure? N+1 is redundancy only until the failed element is back in service, which depends on spares held on site rather than on the drawing.
How to specify UPS redundancy in a tender so quotes are comparable
Comparable quotes come from a specification that removes the bidder's freedom to choose N. Issue identical load, ambient, runtime and redundancy assumptions to every bidder, and add these lines.
- State the design load and horizon load in kW and define N against the horizon load: "N+1 at 100% of the design load in clause X" leaves nothing to interpretation.
- Write the configuration in words as well as notation. "Two independent systems, each internally N+1" cannot be misread; "2N+1" routinely is.
- List the tasks that must be possible with the load on inverter: power module, fan, capacitor, static bypass, manual bypass, input breaker, output breaker, battery string, control card.
- List the faults to be tolerated with no interruption: one power module, one input feed, one battery string open-circuit, loss of communication between modules or frames, one distribution board.
- Require a single-line diagram naming every element common to both paths. If no such list exists, the design has not been reviewed for redundancy.
- Require the published inverter efficiency, the efficiency at day-one load in the offered configuration, and the derating basis of site ambient and altitude.
- Ask how load sharing behaves on loss of the control or communication bus, and whether control is master-slave or distributed.
- Specify the battery arrangement per module or per system, strings and blocks per string, the discharge-test regime, and the spares holding and replacement time for each hot-swappable element.
