Redundancy with Modular UPS Architecture

Compared to a single UPS, parallel redundant UPS systems can achieve far higher system availability. In this article we analyze the redundancy of three different types of modular UPS systems.

Before deepening the topic of redundancy, it may be useful to refresh our ideas about the purpose of redundancy. Redundancy is built in a system to insure normal operation is one of the redundant elements fails. The purpose of redundancy is to increase the reliability of the system as a whole by duplicating critical elements. In other words, the system becomes fault-tolerant.In modular UPS architecture, this is only true however when important elements and features are implemented. It is important to remember at this point that each single function is a “single point of failure”.

Comparison of 3 different types of architectures:

Standalone N+1
Modular UPS N+1 with “Centralized Parallel Architecture”
Modular UPS N+1 with “De-Centralized Parallel Architecture”
Frame / non critical
Non redundant
Non redundant
Distribution / non critical
Redundant / external
Redundant / integrated
Redundant / integrated
Control HW and SW / critical
Non redundant, single HW in frame
Redundant, included in each module
Rectifier / critical
Redundant, some cases not
Inverter / critical
Static bypass / critical
Non redundant
Battery / critical
Redundant possible
Non redundant
Redundant possible

The purpose of this paper is not to compare Standalone with modular system – this type of comparison has been done already extensively – but to compare the different types of modular systems.

Let’s analyze each of the point listed above.


A frame is not critical.


At some point in time, the inverter output of the modules must be connected to a busbar. In the modular UPS architecture, the common busbar of the modules is in the frame. In case several frames must be paralleled together to reach the required power rating, then a connection point or busbar must be installed externally.

Compared to standalone UPS, modular UPS have the advantage that the number connection to an external busbar is generally less than for a standalone system.

Example: 3 x 200kVA (N+1) standalone UPS in parallel require 3 connection to the external busbar. An advanced modular system built with 50kVA modules need in total 8 + 1 = 9 modules. Nowadays such system can be built of 2 frames in which up to 5 x 50kVA modules can be inserted. In this particular case a module space is left empty for future power upgrading.

When comparing different modular systems, it is important to evaluate carefully the additional cost of connecting several frames in parallel. Some UPS manufacturer propose frames with in-built parallel option, others offer this option on top for a non-negligible amount of money.


In the case of decentralized parallel architecture, each module includes a CPU that is fully autonomous. All CPUs are connected to one another via a communication bus. In case of a CPU failure of a state of the art decentralized parallel architecture, only the faulty module will be safely disconnected to the load, and the system will run normally on the other modules.

Fig.1: Decentralised Parallel Architecture

In case of a centralized parallel architecture, if the central CPU fails, the complete frame will not be able to supply the load from the modules inverters. In this particular case it is important to be assured that the complete load will be switched to bypass. This is not obvious in the case of a CPU failure.

Fig. 2: Centralised Parallel Architecture

From a practical point of view, centralized parallel architecture is – in case of a CPU failure – certainly less reliable than a decentralized parallel architecture. To a certain extend it may even be less reliable that a traditional parallel system of standalone UPS


Rectifier and inverter:

Most of the modular systems have a rectifier and an inverter in each module. These elements are fully redundant in both centralized and decentralized architecture.

In case of a rectifier or inverter failure, it is crucial to isolate the faulty module from the parallel system. Unfortunately, modular UPS systems using centralized parallel architecture do not necessarily includes this vital feature. Since all inverter transistors are connected in parallel, it is likely that a faulty module will take all others with it. Therefore it is very important to check if such isolating feature exists and performs correctly. If not, any reliability analysis based on centralized parallel architecture is wrong.

Static bypass:

A modular UPS is a particularly interesting alternative to a system of parallel standalone UPSs in a N+1 configuration. The redundancy is given by the increment of a single module which is of proportional small power compared to the whole system. For instance a 300kVA N+1 system can be realized with 7 modules of 50kVA, where the 7th module is bringing indeed the system redundancy. Similarly, a 120kVA N+1 system may be build with 7 modules of 20kVA.

Referring back to the table 1 above, the centralized parallel architecture uses a single static bypass per frame whereas each module in a system with de-centralized parallel architecture has its own static bypass. The inconvenient of the centralized architecture comes forth when the system need to switch to bypass, for instance in the simple case of a system overload.

Fig.3: (n+1) parallel redundant UPS with static bypass – electrical and reliability block diagram

If at this moment the centralized static switch has a failure, the load will be dropped since there is no redundancy in the centralized system. In the case of a de-centralized modular architecture, if at this moment a module static switch fails, it will be immediately disconnected from the load and the other modules will take over.

Even in the case of parallelable frames, centralized modular system will still show this inconvenient.

Fig.4: A typical example of a 100kVA redundant system

The picture left is a typical example of a 100kVA redundant system where 2 frames are installed in parallel and where additional modules can be added in case of system extension.

If one of the centralized static bypasses fails, the complete unit is lost. Assuming that the second unit is still functioning, the complete load will be transferred to it. Due to the typical N+ 1 redundancy provided by the system, the remaining unit will be very likely in overload, even on bypass.

Units of modular systems with de-centralized architecture can be also connected in parallel. However, this type of UPS will isolate the faulty module and not a whole unit. There is no risk of overload.

In conclusion, it is important in the design phase of the UPS to evaluate all possible critical events in order to choose the right UPS configuration and UPS system architecture.


When discussing about the reliability of static UPS system, experts generally agree that the batteries are a weak link. In this particular case, this weakness must be taken into account during system planning. The configuration of the overall modular UPS system must be such that the risk of load drop is minimized in the case of a battery failure.

In traditional UPS systems made of standalone UPS in parallel (N+1), the battery are often distributed among the UPS. That is each UPS has its own battery pack. In case of a battery failure, even during a mains outage, the load will be securely fed by the batteries of the N remaining units.

What are the differences among modular UPS systems?  The major difference is given again by the 2 different architectures: Centralized Parallel or De-centralized Parallel.

In the case of a centralized parallel architecture, the batteries must be “in common”. Similarly to the CPU, all modules are connected to the same (common) battery pack. This is also a “single point of failure”. In case of a battery failure, the load is not secured in case of a mains outage. The load will be dropped.

The de-centralized parallel architecture of modular UPS systems allows connecting separate battery pack to each module. In case of a battery failure, only one single module will be unable during an outage of the mains. This feature is particularly important if one assume the weaknesses of the battery as explained above. Only de-centralized parallel architecture can offer full availability.

Figure 5: Parallel UPS System with Dual Battery Strings

Another advantage of de-centralized parallel architecture when looking at separate batteries is the possibility to add or replace modules and batteries without mixing old and new batteries. In common battery configuration a mix a battery will be necessary, unless the complete pack is exchanged.


Figure 5 shows a two-module parallel UPS system. Each UPS module has two strings of batteries connected in parallel. If any one battery cell in such a configuration becomes open-circuit then both UPS modules will continue to operate. As both UPS modules equally share the load the UPS module with only one healthy battery string will discharge its good batteries quicker than the other UPS module, but with appropriately rated batteries the required battery autonomy can be maintained. In addition, in parallel redundant configuration UPS redundancy is not lost.

Related posts:

This entry was posted in ABB UPS, Applications, Efficiency, Technology and tagged , , , , , , , , . Bookmark the permalink.

One Response

Leave a Reply

Your email address will not be published. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>