S 1.70 Central uninterruptible power supply

Initiation responsibility: IT Security Officer, Head of IT, Building Services Manager

Implementation responsibility: Building Services, Administrator

With an uninterruptible power supply (UPS) it is possible to survive brief power failures or maintain the supply of power until it is possible to properly shut down the computers connected to it. Using a UPS is particularly useful,

• if large amounts of data are stored temporarily in a computer (in the cache memory of the network server, for example) before being transferred to non-volatile storage,
• if large amounts of data would be lost during a power failure and would then be need to be re-entered after power returns,
• if the stability of the power supply is not adequately guaranteed.

There are three basic classes of UPSs:

• VFD-UPS (Voltage and Frequency Dependent)
  
  With this type of UPS, the consumers connected during normal operations are fed directly from the power supply grid. If the power fails, the UPS starts automatically and supplies power. A VFD-UPS needs up to 10ms to power up (switchover time), which may already be too long for some IT devices. Since a VFD-UPS does not supply power to the consumers connected to it during normal operations, it used to be referred to as an offline UPS.
  
  VFD stands for Voltage and Frequency Dependent, which means that the voltage and frequency of the output of the UPS depend directly on the input voltage and frequency during normal operation. This means that minor disruptions in the power supplied by the grid may propagate directly to the consumers supplied by a VFD-UPS.
• VI-UPS (Voltage Independent)
  
  A VI-UPS compensates for minor variations in the supply voltage (VI stands for Voltage Independent) without having the UPS completely take over supplying power to the connected consumers. The frequency at the output of a VI-UPS depends directly on the frequency at the power grid, just like with a VFD UPS. A short interruption in the power supply can also occur when using a VI-UPS when it switches over to the battery mode.
• VFI-UPS
  
  On a VFI-UPS (Voltage and Frequency Independent), there is usually no direct dependency between the input and output of the UPS. All electrical energy at the input is rectified and fed into the intermediate circuit. The intermediate circuit then keeps the batteries fully charged and supplies the inverter with power.
  
  The inverter then generates the AC power needed to supply the connected consumers.
  
  The VFI-UPS is therefore always connected between the power grid and the consumers. This means there is no switchover time, which is why only the VFI-UPS is truly uninterruptible. Since all power supplied goes through this type of UPS at all times, it was previously referred to as an online UPS.

When these three types of UPS are compared, there is no question that the VFI-UPS has the best output response and should be preferred, at least for the supply of sensitive IT systems. When additional quality features not addressed in this safeguard are taken into account, a UPS classified according to DIN IEC 62040-3 pursuant to VFI-SS-111 represents the best choice for supplying IT with power.

In contrast to a common assumption of many people, a UPS does not provide overvoltage protection in the truest sense, regardless of its design. A UPS is able to prevent high voltages from reaching the connected consumers in the context of its normal operation. However, a UPS is no help at all to protect against the overvoltages that technical overvoltage protection equipment is designed to protect against. Quite the contrary, a UPS must be protected against overvoltage using suitable safeguards just like all other electrical equipment (see S 1.25 Overvoltage protection).

Two aspects are important when dimensioning a UPS: the backup time and the output power rating.

When specifying the backup time, it is necessary to take into account the purpose of the UPS, the type of IT to be supplied with power, and any other power backup measures available.

If the IT supplied with power by the UPS is able to restart and resume operation without any problems after the power suddenly switches off and then switches back on again, it is sufficient for the UPS to be designed to handle brief power failures. Since most power failures only last a few minutes until power is returned, a backup time of 10 to 15 minutes is adequate.

However, if the IT needs to be shut down properly, such a short backup time will not suffice. In this case, it is best to wait a while after the power failure has occurred before shutting down the systems and not shut them down immediately. A waiting time of about 10 minutes is appropriate. The time required to shut down varies greatly between systems and must be determined for each of the connected IT systems. A rule of thumb for the backup time for such cases is as follows:

Backup time = Waiting time plus twice the shutdown time

The backup time values typically range from 30 to 60 minutes. Doubling the shutdown time creates an additional safety buffer.

In special cases (e.g. telecommunication systems), the required backup time could be several hours. The backup time must be checked every time after changing or adding devices to be supplied by a UPS to make sure the current backup time is sufficient.

Changes to the required backup time can be compensated relatively easily by expanding the battery capacity. This is not the case, however, with the output power rating. The maximum output power is determined by the electronic components built into the rectifiers and inverters. Increasing the output power rating simply by adding extra equipment is usually impossible or only possible with comprehensive alterations. When specifying the output power rating, an adequate power reserve should be planned.

The most sensitive component of a UPS is the battery. The battery can only provide maximum output and reach its maximum service life when it is installed in an environment maintained at the optimal temperature (usually around 20°C) specified by the manufacturer. The output and service life decrease by about 50% for every 10 K below the target temperature. This clearly shows that the battery, which operates best when cool, should never be installed in a room together with heat-generating power electronics, especially in the case of large UPS systems. In order to ensure that the UPS will provide power for the backup time specified, the actual backup time should be determined once per year. Some UPS systems have built-in test mechanisms for this purpose. If this is not the case, the backup time can be determined by a load test.

Since the UPS is the last bastion protecting the IT hardware from power failures, ensuring its availability is particularly important. The UPS therefore has the same protection requirements as the IT it supplies power to. If the IT systems supplied by the UPS are designed redundantly, the UPS system also must be designed redundantly. See S 1.52 Redundancy, modularity, and scalability in the technical infrastructure for more information on this topic.

Furthermore, special attention should be paid to protecting a UPS from fire, water, and access by unauthorised persons. Proper protection against fire makes it almost essential to place mutually redundant UPS units in separate fire zones. This is the only way to prevent the other units from failing during a fire shortly after one of the units catches fire.

As with all other electrical devices, it is also necessary with UPS systems to make sure they are operated within the temperature ranges specified by the manufacturer. This must be taken into account when determining the required cooling capacity.

In order to maintain the protective effect of a UPS, maintenance must be performed regularly. The maintenance schedule specified by the manufacturer of the UPS must be followed for this reason.

Review questions:

• Has it been ensured that the battery is maintained within the required temperature range?
• Is the maintenance schedule of the UPS followed?
• Is the actual capacity of the battery, and therefore the backup time of the UPS, tested regularly?
• Is the UPS backup time checked again after every change to its consumers?