Static Transfer Switches come in 2 classes:
1. Large STSs - Stand up static transfer switches (also referred to as STS) come in NEMA 1 enclosure for larger applications ranging between 200A and 4000A at 208-240V (rated at 22KA) or 380-480V (rated at 65KA) at 50Hz or 60Hz. Higher capacity fault ratings are also available if specified. These STSs always have a built-in bypassing system made up of 6 plug-in or draw-out circuit breakers to enable routine or emergency maintenance of the switching devices & electronics. These systems are used at most all super critical applications worldwide where continuity of power is mandatory to keep the critical loads on at all times. Data centers, air traffic control sites, telecommunications centers, and process control (petrochemical & power plant applications) make up the majority of the use. They are also used on board critical Navy vessels to maintain power to radar and weapons systems. In the US, they are normally always listed to UL1008 Standards for Automatic Static Transfer Switches. Their physical size ranges from 2' - 10' wide by 3' - 5' deep depending on the rating of 200-4000A. The larger ones (800-4000A) are generally installed in electrical switchgear rooms while the 200-600A versions are installed in computer rooms near the loads they serve (combined with distribution panels). The smaller 200-800A versions can also be a part of a dual PDU/STS system if switching is to be done at 208V. If switching is preferred to be done at 480V ahead of a PDU transformer then you have to be careful that if the UPSs will be out of phase by more than 20-30 degrees then you need to get a type of STS that has built-in out-of-phase transfer capabilities so the total inrush into the downstream PDU transformer for whatever out-of-phase angle (from 0-180 degrees) never exceeds the instantaneous trip rating of the STS CBs or the upstream CBs feeding the STS. Example - if you have a 400A STS it will have 400A CBs with an instantaneous trip of 10X = 4000A. The maximum inrush drawn during transfer at any phase angle cannot exceed this or the feeder breakers will trip and drop the load. If a major UPS fails and it is out of phase, then the combinations of multiple inrushes when multiple 400A STSs transfer can also trip the UPS main output and bypass CBs so some system level computations are required to know if your design has covered this eventuality. Remember that these issues are not a matter of “if” but are a matter of “when”. If a failure can occur then it will, someday, so your design should be able to handle it.
Because of the critical nature of these transfer switches these automatic static transfer switches come with a comprehensive list of status and alarms to make sure any kind of a problem is detected in advance to head off a real load failure.
2. Point of Use STSs - These are point of use rackmount static transfer switches (RMSTS) that are in every respect the same as the above stand up STS units but shrunk in size to fit the rack-mounting dimensions for point of use applications. They have two major advantages over large fixed STSs in that they are mobile, and they are closest to the load. There are a lot of devices that can fail between a large STS and the load such as distribution panels, distribution breakers, feeders or operators causing an operational error. None of these can be prevented by a larger STS since the problem is after the STS (much like if a problem is after a UPS - the UPS can't do anything about it). These small rack level devices can also be moved around from one rack to rack with ease. No special conduit runs are required to be relocated as they use whatever redundant sources of power feeders you would normally bring to all racks (except for the super critical dual cord loads that would require a 3rd feed - please refer to the news release on this subject). Therefore, no advance planning is required to know where the single corded loads will be. Wherever they may be, you would install one rack-mount static transfer switch. If the plans change that can be handled too with ease.
Moreover, many users do not like using the larger switches because if one fails then the impact is felt by many racks all at once whereas the use of a rack-mount static transfer switch limits any loss to only one rack. It is always less disruptive to lose 30 Amperes worth of loads in one rack than 600 Amperes worth in 20-30 racks all at once. To enable use of rack-mounted static transfer switches, they have to have a very resilient design with built-in redundancies not offered in the larger STS systems. A properly designed rack-mount static transfer switch must have a reliability at least 20X that of larger systems due to the quantity of systems used.
These products also always come with a built-in bypass and unlike the relay type rack-mounted transfer switches have extensive monitoring and user communications options. Unlike the large STSs above, in the case of rack-mounted static transfer switches, the bypass isolation is always on an outer enclosure so as to enable the complete withdrawal of the electronics section that includes the logic and the SCRs. A nice thing to have which is physically not practical with large switches or everyone would offer it.
Rack-mounted static transfer switches provide protection to single and triple corded loads as well as dual corded super critical loads referenced above. They offer the most flexibility among all designs as well as the highest level of protection by their point of use location. This assures that potential failure of all electrical devices upstream of the load are accounted for.
General Overview on Static Transfer Switches (STS)
With advent of too many single large UPS failures at large data centers, static transfer switches, also known as STSs or solid state transfer switches came into the picture because the earlier mechanical transfer switches referenced above were not fast enough to maintain the load between transfers. The “2N” facility configuration was born in the 90s to achieve total redundancy of the UPSs with STSs transferring in between. Out went the single large UPSs or dual large UPSs with tie bus capabilities and in came two or more completely isolated UPS systems with STSs to jump from one to another in case one failed. The new static transfer switches enabled properties not possible before such as a 1/2 cycle transfer time and transferring between sources irrespective of phase angle difference which made them ideal for use on switching between UPS systems. To do this they use SCRs (Silicon Controlled Rectifiers) which are very robust high reliability switching devices ideally suited for this application. SCRs are great devices to use as switches because they have a very low voltage drop (1-2 volts) while conducting and have an excellent overload and fault capability.
Majority of data centers have now been converted to the 2N design with multiple automatic static transfer switches transferring the critical loads between the two or more UPS systems. A typical data center could consist of 2 independent 2000KVA UPSs each feeding a critical bus gear consisting of 10 x 400A circuit breakers that feed 10 x 400A STSs so the STSs are fed from both sides and they can transfer instantly if one side fails. The failure does not necessarily mean a UPS failure. Statistics have shown that in fact 75% of UPS power failures are caused by human error during maintenance or operations activities. Maintenance access is another important consideration in data centers. You have to be able to offload any of your UPSs so you can maintain them. Going to bypass does not do the job because the UPS bypassing capability itself has to also be tested. Also the UPS Output Switchgear is a source of data center failures and up until the availability of STSs they were inaccessible. You could do all the maintenance on the UPS but you could not touch the critical load bus fed from it. Using STSs enabled the data center operators to transfer all the loads away from one output gear to another so they could do bus maintenance and fix loose connections and such before they led to a massive failure. So STSs are also a major maintenance tool in a data center. Aside from the critical output bus the major feeders also require remedial maintenance if a breaker fails or a feeder cable fails. Once again the STS can be used to reroute the power away from the failed segment so it can be fixed without affecting the load.
Use of Redundant Servers to Do Away with the Rack-Mount Static Transfer Switch
Some users try to eliminate the need for STSs or RMSTSs and instead use two servers with one as back up to the other. This option is risky when you consider that if the power supply in one of the two servers happens to be bad and power failure occurs at the other back up server then the complete functionality of that server and its back up are both lost. For this reason most mission critical sites rather consider using two servers but protect both against power failures by powering the racks both are in from a RMSTS. This way if a power supply in one servers has failed at least the other server has redundancy of source of power and a glitch won't take it down.
Also consider that the RMSTS is a very efficient device. Its power consumption with only a 1V drop across it is very low and it takes a very small amount of space and is considerably cheaper than the cost of computing equipment going into racks that can cost from $80k to $100k in single phase computer equipment per rack. Doubling up on servers, setting aside the immense hardware costs and space (one switch can feed an entire rack so to avoid it you would need another rack full of servers) also doubles your electrical load as well as your heat load. All things considered this option is a lot more expensive than one might think. The comparison is a rack full of added hardware, a rack full of added electrical power, and a rack full of added heat vs. avoiding the cost of a single RMSTS. The estimated excess cost is over 20:1 (1 being the cost of the switch) if space and operational costs are also considered. If on-line availability is the top consideration, then to accomplish the highest level of resiliency use the above two racks with the redundant servers but power each rack through a RMSTS thus making both UPSs available to both racks. That would yield the highest reliability because consider that while the RMSTS can protect you against power issues the servers themselves fail too. To cover against such failures then you have to provide redundant servers. Many global corporations with financial transactions, disruption of which can cost millions per occurrence, do just this. Indeed they also use the RMSTSs on the dual cord loads and power both cords through one of 2 switches at each rack. If UPS A fails the switches transfer to B and if B fails they transfer to A. If the A side power supplies of some of these servers are burned and the B side UPS fails there is a major failure. The 2 switches solve this problem because they allow the A side UPS to feed the B side of the servers so there is no issue. If you want to ratchet this one level higher in reliability, you can also survive if both UPSs fail but you'll need a 3rd feeder to the rack to do this. That feeder would be connected to a bus that is fed from either the utility or the generators. Generators would come on when either UPS or the utility fail. Now imagine your rack has two power strips, A and B, feeding your dual cords. A is fed from UPSA and B is fed from our RMSTS which in turn is fed from UPSB and the bus referenced above. Let us say UPSA fails. No problem as power strip B is still fine but in the mean time generators come on and are feeding the second source of the RMSTS. Let us now assume a UPSB output distribution switchboard feeder CB overloads and trips. At this point your whole rack would be gone, but the RMSTS transfers to the back up bus fed from generators and an outage is avoided.
2 Pole versus 1 Pole Switches
Always use a 2 pole rack-mounted switch (relay or static) requiring both legs of the source to switch, be it hot or neutral. The transfer operation must electrically appear as if you are standing in front of two outlets and unplugging your load from one outlet and re-plugging it back into another completely independent one in 1/2 cycle. Everything is switched including the neutral lead. Single pole transfer switches require that you connect the neutrals of the sources together and make that your load neutral. This is allowed only when used between two small UPSs within the same enclosure so that the neutrals of the UPSs are tied together and grounded together on one point within the same enclosure. This is not allowed on independently derived sources such as when large UPSs in different enclosures are involved as it would create multiple grounds in the system, because the returning load neutral current would split at the switch and go back to two grounded transformers thus forming a double ground system which is against code. With two grounds you create a ground path in the facility earth for your load current to flow. Unfortunately, many sites have installed single pole transfer switches already without watching for this safety hazard issue. If inspection authorities discover this, they would tag the installation to be corrected before further use. This also creates a single point failure to the switch in that a line to neutral fault, if it disturbs the neutral voltage, will be seen on both sources of the switch since both neutrals are connected and will disable the switch from transfer (both sources will appear unacceptable at the same time). Moreover, this type of a fault can also fail both UPSs at the same time.