Florham Park, New Jersey 07932
Published by
Volume 6, Number 8
File Index: Automatic Transfer Switches
Transfer Switching of Critical Computer Loads
Automatic Switch Company
Critical applications using computers and electronic equipment
require the design and construction of elaborate power systems
for continuity of service. These systems are designed to provide the highest quality power, without interruption, to these
sensitive loads.
Redundant power sources and distribution, with downstream
transfer switches are typical of such systems. Such a system is
shown in figure 1. Here the redundant power sources are UPSs.
UTILITY
UPS 1
UPS 2
PDU
failure, or an upstream breaker trip caused by overcurrent, malfunction, fault, or operator error. Regardless of the cause, the
unplanned transfer requires fast voltage sensing and switching.
The resulting outage must be short enough for the sensitive
equipment to remain unaffected.
The single most critical element in the this type of application
is the transfer switch. It must be reliable, and, it must be fast.
A new switching technology, referred to here as the solid state
hybrid transfer switch, was developed specifically for these
critical applications. This new switch maintains the reliability
and cool, efficient, operation of an electro-mechanical switch,
while providing the fast operating speed required in most critical applications.
The Information Technology Industry
Council* recommends the voltage tolerance envelope for computer equipment (see figure 2). Accordingly, computer equipment must operate properly with a complete interruption of
voltage for 20 milliseconds.
PDU
Transfer
Switch
Critical
Load
Figure 1. Redundant UPSs and a transfer switch provide reliable
power to the critical load.
The primary role of the transfer switch is to connect loads to an
acceptable power source. This is accomplished by continually
sensing the availability of both sources and providing instantaneous source selection. Transfer to the alternate source may be
required for one of two reasons:
1. Planned Transfer
System maintenance of the uninterruptible power supply (UPS)
or other upstream equipment. In this instance the transfer is
planned. Front panel controls are typically used to initiate a
transfer between the two live sources.
2. Unplanned Transfer
When the connected source fails, or is interrupted, the transfer
switch must sense these conditions and automatically transfer
the load to select the alternate (acceptable) source. The
unplanned transfer may be the result of a UPS failure, utility
Figure 2. CBEMA Curve Revised February 1996. This curve
indicates that typical information technology equipment (ITE) is
unaffected by voltage interruptions of up to 20 milliseconds, and
voltage sags of 20% for 10 seconds and 30% for 0.5 seconds. See
CBEMA curve application note for more complete information.
Until now, the only transfer switch technology fast enough for
these applications was static switch technology.
This paper describes the performance of three transfer
switching technologies:
a) Electro-mechanical switches
b) Static switches
c) Hybrid solid state switches
A comparison of features and functions of each technology
follows:
Electro-mechanical Switches:
Electro-mechanical transfer switches employ double throw
devices containing time-proven silver alloy contacts supported
by electrically operated and mechanically held operators. The
reliability of this approach is considered to be very high.
The latest transfer controllers are microprocessor type and are
usually designed for an MTBF of at least 300,000 to 500,000
hours. The life of the mechanical switch mechanism is a function of operational cycles rather than time, and is typically
designed for a minimum of 12,000 transfers. In most applications* (two transfers per week) this amounts to 1,000,000 hours
(115 years).
Since the electro-mechanical switch is mechanically locked in
either of two "on" positions (see figure 3), the control power
supply or other control components can malfunction and still
not affect the load. The switch contacts remain closed and continue to supply power from the connected source to the load.
The controls must be functional when power outages occur and
a transfer is required. Monitoring of the control panel indicators, or communications with the controls through the communication interface, can verify that the control is operational, and
further enhance the reliability.
WEIGHT
STATIONARY
CONTACT
TO LOAD
Figure 3. Example of ASCO electrically operated mechanically held
transfer switch mechanism (one source and contact shown). This
switch requires no control signals or electronics during the "idle"
time between transfers in order to stay closed. When the switch is
closed on a source (as shown) the moveable contact and yoke cannot
exert an opening torque on the weight due to the geometry. The controls are only needed to sense and signal the solenoid to transfer
upon an upstream failure.
Volume 6, Number 8
A closed transition transfer switch (CTTS) is usually specified
when planned transfers are frequent and the predominant concern. For planned transfers, this variation of the electromechanical transfer switch operates in a make before break
mode. The controls inhibit transfer until the sources are in synchronism before initiating transfer. This results in zero power
interruption to the load. The response to an outage (unplanned
transfer) is open transition (break before make), however, and
is much slower (typically in the order of 50 to 100 milliseconds
or more).
Static Switches:
Static switches use SCRs (silicon controlled rectifiers) to conduct current from the sources to the load (figure 4). These
devices turn on when gated, and turn off at the first current zero
crossing following the removal of the gate signal. The major
advantage of this type of switching is fast transfer without
source interconnection. The response to an outage is, in the
worst case, less than 8 milliseconds, provided the second
source is available.
The SCRs require the semiconductor junctions to be forward
biased. This results in a voltage drop of approximately 1.5
volts per line (or 225 watts per phase in a 150 amp switch).
This power loss generates considerable heat which must be dissipated by fans. The fans run continuously, and over-temperature sensors are usually required to set an alarm if there is a
cooling malfunction. The switch may not be able to support the
load under an over temperature condition and automatic
bypassing may be required to protect the SCRs.
MOVEABLE
CONTACT
TO SOURCE
Mechanical switches exhibit very low heat dissipation. The
silver alloy contacts in a 150 Amp switch would typically result
in a 75 millivolt drop (approximately 11.25 watts per phase).
Thus the power losses are low and the unit runs cool without
the need for fans.
This SCR technology should be considered when an 8 millisecond outage is the maximum that can be tolerated. It should
be noted that most modern computing equipment can withstand
a 20 millisecond outage as described in the February 1996 revision of the CBEMA curve (see figure 2).
PIN
YOKE
There is an important difference between mechanical switches
and static switches. A static switch control malfunction will
likely result in loss of power to the load (even if the power
sources are acceptable). Since the electro-mechanical transfer
switch is "mechanically held", it will remain connected to the
source, and not affect the load (provided the connected source
is acceptable). In virtually all applications, the time that a
transfer switch is idle (conducting from one source), is much
greater than the time spent transferring from a failed source
(typically by a ratio in excess of 108:1). During this "idle"
time, when all else is in good working order, a transfer switch
failure must not become the cause of a load outage.
Page 2
Detection circuits are employed to prevent transfer if an SCR
shorts (which would result in a source to source connection). If
a shorted SCR on the connected source is detected, the opposite source breaker is shunt tripped (as protection against a
source to source connection), and transfer to the opposite
source is inhibited. Similarly, if circuitry detects open SCRs,
load is transferred to the opposite source to restore power (providing the opposite source is available). Transfer back to the
original source is then inhibited.
Unlike the mechanical switch, the controller must be functioning, and providing continuous gate pulses to the SCRs to keep
them on, or the static switch will, itself, cause the load to lose
power.
Source 1
Source 2
CB1
CB2
CB4
CB5
CB3
Output
Figure 4. Typical one-line diagram of a static switch
Hybrid Solid State Transfer Switches:
This switching technology takes advantage of both static and
mechanical switching elements. It utilizes the best of each technology and virtually eliminates the shortcomings. A standard
electro-mechanical switch provides power to the load through
mechanical contacts. The switch is electrically operated and
mechanically held. The reliability, and cool operation, of
mechanically locked contacts are maintained. SCRs are added
to transfer to the opposite source with fast and precise timing
(see figure 5). The result is a transfer switch that meets the total
outage requirement of the February 1996 CBEMA curve
(Computer Business Equipment Manufacturers Association,
now known as Information Technology Industry Council).
Source 2
Source 1
CN
TS
Planned transfers, with the ASCO hybrid solid state transfer
switch result in less than a 3 millisecond disruption to the load.
An unanticipated loss of the source supplying the load will
result in an unplanned transfer in less than 20 milliseconds
without source to source interconnection (break before make).
The SCR conduction is limited to approximately 15 milliseconds during transfer, and the mechanically locked silver alloy
contacts support the load at all other times. This results in an
average voltage drop in the range of 75 millivolts, and cool
operation with no requirement for fans or over-temperature
alarms.
The unit has two control power supplies, one on the normal
source and one on the alternate source. The control can completely function on either power supply and for 6 seconds after
loss of both power supplies. A third and fourth supply can be
added for additional redundancy. This is not recommended
since the switch, like an electro-mechanical transfer switch,
only relies on the control circuit for transfer. A lost power supply will not result in a lost load (assuming the connected source
is acceptable). Even if both power supplies fail, the switch will
not disconnect the load. Static switches require power supply
redundancy to reduce the likelihood of a control failure resulting in a transfer switch induced outage.
The ASCO hybrid switch operates as follows:
Planned transfers are manually initiated from the front panel.
The controller will check for synchronism (voltage, frequency,
and phase angle), signal the mechanical switch to operate, and
fire the SCRs 3 milliseconds after the mechanical contacts part.
This will immediately extinguish any arc (the sources are in
synchronism which forces zero arc voltage), and connect the
load to the opposite source until the mechanical contacts complete their transfer. The result is a 3 millisecond load disruption
with no current between sources(see figure 6). The SCRs are
then gated off. The extremely short on-time results in almost no
heat burden.
In an unplanned transfer the switch will sense the outage in 3
milliseconds, signal the mechanical transfer switch to operate,
wait until the contacts just disconnect (from the unacceptable
source), wait 9 more milliseconds to insure arc extinction, and
then fire the solid state devices to the opposite source to power
the load. The transfer switch will complete it's mechanical
transfer to the good source, parallel the SCRs, and support the
load until the next transfer. The SCR's are then turned off. The
result is open (break before make) transition, and an outage of
less than 20 milliseconds(see figure 7).
Shorted and open SCR detection and open fuse detection are
incorporated to prevent cross connection. It should be noted
that, even if the SCRs short (or open), the switch can still
respond automatically to an outage and transfer mechanically in
approximately 32 milliseconds.
Figure 5. The Hybrid Solid State Transfer Switch maintains the reliability and cool operation of a mechanical switch. The SCRs are
used for fast transfer to the opposite source. This approach meets the
requirements of the Supply Voltage Tolerance Envelope for
Information Technology Equipment (Feb. 1996).
Volume 6, Number 8
Page 3
Figure 6. The top trace above represents the solenoid current. The
other traces represent the three, phase to neutral, load voltages during
a manually initiated test transfer between two live sources. Note the
maximum load interruption is approximately 3 msec.
Figure 7. The top trace above represents the solenoid current.
The other traces represent the three, phase to neutral, load voltages during an unplanned transfer. The transfer resulted when
the connected source breaker tripped. The total load interruption in this case is less than 20 msec.
A performance and feature summary of these three types of switches is shown below:
Static Switch
(typical)
ElectroMechanical
Switch (CTTS)
Hybrid Solid
State Switch
Ampacity (amps)
30 to 2000
30 to 4000
Up to 150
Line Voltage (VAC)
208-480
208-600
208
Worst case response to a power
outage without source to source
connection (synchronism not required)
8 msec
50-500 msec
20 msec
Load interruption during planned
transfer (no source to source
connection)
<1 msec
0 msec (source
to source
connection)
3 msec
Relative Enclosure Size
Large
Compact
Compact
Overlapping Neutral
No
Yes
Yes
Voltage Drop in 3 phase wye
1.5 volts per phase
75 millivolts
75 millivolts
Power Loss @ 150 amps
700 watts
35 watts
35 watts
Cooling Fans Required
2
None required
None required
Over-temperature alarm
1
None required
None required
SCR Failure Detection
Open and shorted
None required
Open and shorted
Can device automatically
transfer if SCRs fail?
No
N/A
Yes
No
Yes
Yes
Will transfer switch maintain load if
the controller or SCRs malfunction
when connected upstream source
is acceptable
Volume 6, Number 8
Page 4
The selection of transfer switches, for critical applications, is
difficult. The tradeoffs between characteristics such as
mechanically held contacts, high speed operation, power losses, heat, fans, and the reliability under different conditions must
be considered. If the loads can handle interruptions as defined
in the CBEMA curve, both the solid state hybrid switch, and the
static switch, should be given serious consideration. The
ASCO hybrid technology meets the latest standards of performance in speed without sacrificing the cool operation and reliability of mechanical transfer switches.
CBEMA Curve Application Note*
The CBEMA Curve and this Application Note are published by
Environment & Safety Committee #3 (ESC-3) of the
Information Technology Industry Council (ITI), formerly
known as the Computer & Business Equipment Manufacturer's
Association (CBEMA).
2) APPLICABILITY
The CBEMA Curve and this Application Note are applicable to
120V nominal voltages obtained from 120V, 208Y/120V, and
120/240V 60Hz systems. Other nominal voltages and frequencies are not specifically considered and it is the responsibility
of the user to determine the applicability of these documents for
such conditions.
3) DISCUSSION
This section provides a brief description of the individual conditions which are considered in the CBEMA Curve. For all
conditions, the term "nominal voltage" implies an ideal condition of 120V RMS, 60Hz.
Seven types of events are described in this composite curve.
Each event is briefly described in the following sections, with
two similar line voltage sags being described under a single
heading. Two regions outside the shaded portion of the curve
are also noted. All conditions are assumed to be mutually
exclusive at any point in time, and with the exception of steadystate tolerances, are assumed to commence from the nominal
voltage.
3.1) Steady-State Tolerances
The steady-state range describes an RMS voltage which is
either very slowly varying or is constant. The subject range is
+/-10% from the nominal voltage. Any voltages in this range
may be present for an indefinite period, and are a function of
normal loadings and losses in the distribution system.
3.2)
Figure 8. Supply voltage tolerance envelope provided by
Information Technology Industry Council.
This region describes a voltage swell having an RMS amplitude
of up to 120% of the RMS nominal voltage, with a duration of
up to 0.5 seconds. This transient may occur when large loads
are removed from the system or when voltage is supplied from
sources other than the electric utility.
3.3)
1) SCOPE
The CBEMA Curve and this Application Note describe an AC
input voltage boundary which typically can be tolerated (no
interruption in function) by most Information Technology
Equipment (ITE). The CBEMA Curve and this Application
Note comprise a single document and are not to be considered
separately from each other. The CBEMA Curve and this
Application Note are not intended to serve as a design specification for products or AC distribution systems. The CBEMA
Curve and this Application Note describe both steady-state and
transitory conditions
Volume 6, Number 8
Line Voltage Swell
Low-Frequency Decaying Ringwave
This region describes a decaying ringwave transient which typically results from the connection of power-factor-correction
capacitors to an AC distribution system. The frequency of this
transient may range from 200Hz to 5KHz, depending upon the
resonant frequency of the AC distribution system. The magnitude of the transient is expressed as a percentage of the peak
60Hz nominal voltage (not the RMS value). The transient is
assumed to be completely decayed by the end of the half-cycle
in which it occurs. The transient is assumed to occur near the
peak of the nominal voltage waveform The amplitude of the
transient varies from 40% for 200Hz ringwaves to 200% for
5KHz ringwaves, with a linear increase in amplitude with
increasing frequency.
Page 5
3.4) High-Frequency Impulse and Ringwave
3.6) Dropout
This region describes the transients which typically occur as a
result of lightning strikes. Wave shapes applicable to this transient and general test conditions are described in ANSI/IEEE
C62.41-1996. This region of the curve deals with both amplitude and duration (energy), rather than RMS amplitude. The
intent is to provide a 40 Joule minimum transient immunity.
A voltage dropout includes both severe RMS voltage sags and
complete interruptions of the applied voltage, followed by
immediate re-application of the nominal voltage. The interruption may last up to 20 milliseconds. This transient typically
results from the occurrence and subsequent clearing of faults in
the AC distribution system.
3.5) Voltage Sags
3.7) No-Damage Region
Two different RMS voltage sags are described. Generally, these
transients result from application of heavy loads, as well as fault
conditions, at various points in the AC distribution system.
Sags to 80% of nominal (maximum deviation of 20%) are
assumed to have a typical duration of up to 10 seconds, and sags
to 70% of nominal (maximum deviation of 30%) are assumed to
have a duration of up to 0.5 seconds.
Events in this region include sags and dropouts which are more
severe than those specified in the preceding paragraphs, and
continuously applied voltages which are less than the lower
limit of the steady-state tolerance range. The normal functional state of the ITE is not typically expected during these conditions, but no damage to the ITE should result.
3.8) Prohibited Region
This region includes any surge or swell which exceeds the upper
limit of the boundary. If ITE is subjected to such conditions,
damage to the ITE may result.
Volume 6, Number 8
Page 6