Specifying front-end ac supplies for telecommunication systems
To ensure smooth, reliable power delivery, designers must consider additional
safety and performance factors beyond general-purpose supplies
BY ALBERTO de LEON
Todd Products Corp.
Brentwood, NY
Telecommunication systems have traditionally used low-voltage distribution
because it provides two major advantages: meeting the extra-low-voltage
requirements of safety agencies and elegantly integrating with batteries for
backup. With the proliferation of cost-effective high-density dc/dc converters,
the low-voltage distributed-power scheme is expanding into applications for
information technology and telecommunication equipment on customers' premises.
Off-the-shelf considerations
In telecommunication systems, the power conversion from the standard 115- or
230-Vac line to the 48-Vdc bus can be accomplished by several means, from
ferroresonant transformer and rectifier units to commercial switching power
supplies. Ongoing pressure for both low cost and high performance has pointed
the power supply industry toward using standard, off-the-shelf, front-end
topologies. These ac supplies address the same requirements for power factor
correction, protection, and safety common with general-purpose switching power
supplies. However, subtle issues make a telecommunication front-end power
supply different from a regular switching power supply.
The dc/dc converters used to generate voltages required by telecom ICs
generally demand constant power (negative resistance) input. When their input
voltage decreases, the input current must increase. The effects of this
negative resistance can undergo both quasi-static and dynamic analysis.
In a quasi-static analysis, three conditions are intuitively problematic for
general-purpose switching power supplies that feature foldback current
limiting. These conditions occur at start-up, input failure, or output failure.
Consider a single-unit switching power supply feeding a 48-Vdc bus without
batteries attached. At turn-on, the output voltage of the front end increases
as its output capacitor is charged. The load's current demand may exceed the
maximum available current, and the output may latch at an intermediate voltage.
Similarly, a drop-out that exceeds the hold-up time or a momentary short
circuit in the bus can also cause latchup because the voltage will drop below
this intermediate, critical point.
To eliminate these problems and to ensure power integrity, the front end should
feature straight current limiting. That's because foldback, latching, or
“hiccup” current-limiting protection can create a system failure. The diagram
shows the load line comparison for foldback and straight current-limiting
characteristics.
Using dynamic analysis, the designer should consider the typical load changes
presented by dc/dc converters to the front-end power supply. A working system
could have circuit boards removed or added while in operation, or use new
generation processors which draw reduced current while idle and high current
when running. In both cases, the output current must change rapidly to
accommodate the load. A response that's too slow causes an output voltage sag
that may reach values outside the on-board converter's range. The magnitude of
the sag depends on the reactance of the power distribution scheme and the loop
compensation.
Power distribution is almost always accomplished with bus bars for low
impedance. Additionally, decoupling capacitors form a low-Q, second-order,
low-pass filter. In addition to the effects of the buss bars, one should
consider the power supply response. The compensation of the loop must provide
good transient response. The best results are obtained with two-stage topology
having programmed droop current sharing. This simple topology combines
acceptable current balance and optimum transient response.
Agency approvals
The requirements of safety agencies are also a factor when selecting front-end
power supplies for communications systems. CSA, UL, and FCC in North America,
as well as European BABT, FTZ, and EN standards, and AUSTEL for Australia,
should be considered depending on the target and market for the system.
Most safety requirements are covered by UL 1950 and CSA 950, which are written
for information-technology equipment. Every day, the differentiation between
telecom and information technology equipment blurs. When the equipment connects
to the telephone network, other standards such UL 1459 and CSA 225 also apply.
The new effort for a National Standard of Information Technology and
Telecommunication Equipment, known as UL 2950/CSA 950, comprises elements of
the four standards mentioned above.
Only a few differences in construction and performance exist between UL 1950
and UL 1459. For power supply construction, it may be advisable to consider the
flammability rating of internal plastics and printed-circuit boards. For
example, support parts and pc boards must be rated UL 94V-0, and capacitors
must be UL 1414 approved. But the major changes introduced in the UL 2950/CSA
950 standards are specific for creepages and clearances between SELV circuits
and Telephone Network Voltage (TNV) circuits. This is a requirement beyond the
basic isolation described in both UL 1950 and UL 1459. This new standard is
expected to be effective in the year 2000, but older products are likely to be
covered by a grandfather clause until 2005.
In Europe, EN60950 has been supplemented with EN41003, which provides
requirements that protect the user of equipment connected to a
telecommunication network. TNV circuits need only basic isolation from SELV
circuits.
The need for redundancy
Most telecommunication systems require redundancy to increase fault tolerance
and reliability. One way to measure long-term reliability of repairable
redundant systems is the concept of system availability. Power system
availability can be defined either as the percentage of time in which the
system is functional or as the probability that a system is functional measured
after a long time from initial turn on. The mean time between failures (MTBF)
of the power supply is the inverse of the failure rate (lambda), while the
mean time to repair (MTTR) is the inverse of the repair rate. For a single
power supply, the availability equals
mu MTBF
A = ————- = ————–
lambda + mu MTBF + MTTR
When a failure occurs in one of two power supplies in a redundant system, the
system must remain functional. In some applications where the supplies are
hardwired and the system can tolerate scheduled down time, the power supplies
can be straight paralleled. In such cases, a short-circuit failure of an output
rectifier or output capacitor could bring down the 48-Vdc bus. Other failures
would be benign, and the bus would not be disturbed. Such configurations may be
attractive in the distribution of low voltages such as 3 or 5 Vdc, since the
lack of added loses caused by an “ORing” diode (which may add as much as 60%
more heat), could increase the overall reliability of the system.
However, in telecom applications the ORing diode is always present. First,
because in the case of an output rectifier failure, the ORing diode will
prevent any battery connected to the system from discharging. Second, the ORing
diode blocks current from the bus to the output capacitors (a desirable feature
for hot-plug capability). Third, the efficiency of the system is not reduced as
dramatically as before, because comparatively lower currents flow through the
diode.
Failure-tolerant equipment must accept repair without shutting down the system.
The table compares the availability of full-on redundant, maintained,
hot-pluggable power systems and power systems without hot-pluggable capability.
For simplicity, these examples assume a defective front-end power module is
replaced immediately after failing.
When ORing diodes are not used in a hardwired redundant system, the system must
be shut down during power module replacement. From the system's point of view,
this is equivalent to having only a single power supply. For the examples shown
in the table, we assume that the modules are changed only after both units have
failed. This means that the failure rate used in the example is equal to
two-thirds of that of a single module.
The table shows that with a hot-pluggable system, the probability of down time
is dramatically reduced by a factor of more than 66,000, when compared with
hardwired systems (Example 1). Furthermore, if mean replacement time is reduced
by slide-in modules (as is the case with hot-pluggable power supplies) the
probability of down time is reduced by a factor above 1 million, when compared
to the longer replacement time required for hardwired systems (Example 2).
CAPTIONS:
Opening shot:
Careful attention to performance, safety, and reliability ensures smooth
delivery of power from front-end ac supplies in telecommunication systems.
Shown here is the TEL-450 from Todd Products.
Figure:
Straight current limiting should be used in a power supply front end instead of
foldback current limiting. The latter technique can cause the power supply to
latch, increasing the likelihood of a system failure.
Advertisement
