Co in the balance

Computers are finding a new home in COs, and the resulting convergence is affecting the environment's powering needs he convergence of computing and telephony brings carriers new business opportunities, but it also gives central office planners and equipment manufacturers new architectural hurdles to overcome. This convergence has blurred the distinct differences and boundaries between voice and data networks.

Data networking equipment is moving into the CO to gain necessary large-scale, closely coupled access to the telephone network. Computing and data networking equipment, previously powered from AC mains, must now be powered from DC central office power.

Power systems engineers who were experts in AC power characteristics, standards, and regulations and requirements now must learn the same about power in the CO. Equipment designers must familiarize themselves with new requirements, standards, practices and implementation options to successfully power equipment in the CO.

Protection, distribution, redundancy

Battery plants are a good place to start. CO power runs on a DC voltage supplied by a battery plant consisting of one or more rectifiers, a bank of batteries, and battery charge control and protection equipment (Figure 1).

Rectifiers convert power from an AC mains supply to a regulated DC voltage, and the output is used to power CO equipment during normal operation. Rectifiers also maintain the batteries at full charge-a condition known as floating the batteries. When the AC mains power is interrupted, the batteries power the equipment. Upon restoration, the rectifiers supply power for the CO equipment and recharge the batteries.

The first concern with a battery plant is protecting the load equipment and power distribution from the batteries. The batteries harness so much energy that a short circuit is always dangerous. The battery plant's output is always protected with some form of current-limiting device-typically a fuse rather than a circuit breaker.

Protecting batteries from the load is another concern. The lead acid batteries used in COs can be damaged by being excessively discharged, but using a low-voltage disconnect prevents this. This device disconnects the batteries from the power distribution network when the battery voltage becomes dangerously low.

Regarding distribution, the -48 V return of the battery plant is passed through a shunt and then connected to the single point ground for the CO. This is the only point in the CO power distribution grid that the -48 V or -48 V return are connected to earth ground.

The -48 V return from the single point ground and the -48 V from the battery plant are then connected to the main (or primary) power distribution frame (PDF). Within this PDF, the main power bus from the battery plant is broken up into several distribution branches, each with its own fuse or circuit breaker.

Each of these distribution buses from the main PDF is routed to next-level PDFs. Each of these second-level PDFs is typically associated with a given set or row of equipment (Figure 2). Power cables are run from a secondary PDF down the row to each piece of equipment being powered.

Each of the branch circuits from a secondary PDF is limited by fuse or circuit breaker to a maximum of 60 A. As a result, any one piece of equipment can draw no more than about 2400 W from a single branch circuit (60 A at the minimum voltage of -40 VDC).

With typical DC-DC converter efficiencies of 80% to 85%, the load circuitry can draw no more than about 1900 to 2000 W. If the load equipment needs more power, special cabling must be run back to the next-highest PDF level. This can be too expensive and complex for many users.

Redundancy is another element in CO power. There have been two DC power buses, often known as the A and B buses, preventing faults or maintenance on one bus from affecting the other, and maintaining DC power availability.

How the redundant buses are created depends on the type of CO. Local exchange offices typically use a single battery plant, with output routed to two main PDFs, each with a main input fuse. Each main PDF distributes power to a set of secondary PDFs, indicating "four wire" distribution with separate -48 V returns and no connection to earth within the distribution network (Figure 3).

This arrangement of one battery plant with redundant distribution protects against many, but not all, faults. In long-distance networks, with revenue depending on toll collection, offices have more fault tolerance. Typically, there are two independent battery plants, each with its own power distribution network. Thus, even a fault in one of the battery plants will not bring down the CO equipment.

Power entry and distribution within equipment bays also has a "four wire" nature (Figure 4). Each major piece of equipment in a bay has its own set of fuses, which often serve as power disconnects.

How far to carry the redundant buses within power equipment depends on the situation. One approach is to carry them all the way through the back plane to the individual circuit cards. The other approach is to combine the redundant buses as soon as they enter the equipment (Figure 5). Both approaches depend on the strictness of fault-tolerance requirements.

The first approach, combining the buses on each circuit card, requires each card to have two diodes and two fuses. Also, each module needs to provide suppression that can absorb the worst-case transient on the CO distribution bus, a requirement that adds expense and complexity.

The second approach, combining the buses at the system level, simplifies the circuit boards but requires two DC input modules, one for each of the power distribution buses. These modules have to be designed to carry the current required by a fully loaded system. However, they typically contain all of the circuitry and components needed to support redundant operation with hot swap as well as transient suppression and EMI filtering.

One advantage to this approach is that it makes possible power system designs in which input modules can be interchanged with AC/DC front-end power supplies. Configuring such a system to operate from either CO or AC mains power in such a system is simply a matter of installing the correct type of input module or supply.

The trade-off? A DC distribution bus system fault will bring down the entire system. This is an unlikely event. Except for systems needing the highest possible fault tolerance, blending the CO distribution buses at the equipment input is the most common approach.

The mysteries of -48 V A CO's DC voltage is often said to be "-48 VDC." This refers to two strings of 24 2-VDC battery cells in series. However, the actual voltage of a fully charged cell at room temperature is more typically 2.17 to 2.19 VDC (-52 to -52.5 VDC for a 24-cell string).

Smaller systems such as wireless base stations away from a CO use sealed, or valve-regulated, lead acid batteries. The float voltage of an individual cell in one of these batteries is typically 2.25 to 2.27 VDC (54 to 54.5 VDC in a 24-cell string).

Traditional POTS is actually an electric power utility as well as a communications service. The power for each telephone is delivered from the battery plant over the same wires that carry the voice signals.

If the telephone power were configured as +48 V, the potential difference between the positive conductor and the earth attracts negatively charged ions. These ions migrate out of the earth and accumulate on the positive conductor and connections, corroding them and causing major reliability problems.

At -48 V, the most positive conductor has no difference of potential with the earth. There is no electric field to cause the migration of corrosive ions to the -48 V return conductors.

There is, however, a potential difference between the -48 V conductor and the earth. In this situation, the electric field from the earth to the -48 V conductor draws copper ions from the wire and connections-an "anti-corrosion." This will eventually cause wire degradation, but the problem is much more manageable than accumulated corrosion.

As described earlier, typical operating voltage ranges from -52 to -52.5 VDC. However, when operating on batteries, the voltage may go as low as -40.5 VDC at the equipment connection before the batteries are disconnected.

Allowance must be made for voltage drops through the local frame's fuse panel and distribution. It is not unusual for the voltage at the back plane to be as low as -38 VDC at the end of a battery discharge.

From time to time, The bus voltage may also be significantly higher than the nominal. This happens during equalization, in which battery voltage is raised to as much as -58 VDC to force the cells of the battery string to equalize their charge.

This condition may persist for minutes or hours, so the equipment must be able operate indefinitely. ETSI standards specify that equipment must operate without damage at input voltages up to -60 VDC.

For -60 V systems, the limits per ETSI standards for normal operating voltages are from -50 to -72 VDC. Abnormal voltages are defined as being the range of 0 to -50 VDC and from -72 to -75 VDC. The system does not have to meet specifications when the input voltage is abnormal, but it must not be damaged.

Faults, transients and outages Consider the entire CO power distribution system-from the batteries to the final circuit card. A short circuit anywhere in the distribution network, usually caused by servicing accidents, will draw a fault current limited only by the resistance and inductance of the distribution network. This current may be a few amperes or thousands of amperes.

Voltages downstream from the fault-from the short circuit to the end of that distribution branch-will be greatly reduced for as long as the short circuit lasts. Also, the large fault currents will reduce the voltages in the distribution network between the point of fault to the batteries.

Depending on the short's location and severity, faults in one of the redundant buses may even cause an outage on the non-faulted bus. When the fault clears, the voltage will be restored on the non-faulted bus.

The short circuit will last until either the cause of the fault is removed or an upstream fuse or circuit breaker clears. When the fuse or circuit breaker interrupts the large fault currents, the inductance of the power distribution network will ring with the various deliberate and parasitic capacitors of the distribution bus. This will create a voltage spike of varying amplitude, ring frequency and duration.

Standards are not clear on what is required for immunity to transients. Bellcore's EMI specification,

GR-1089, does not address transient requirements. Bellcore standard TR-TSY-001003 requires that converters be able to withstand 2500 Vpk ring waves. With the lack of an acknowledged standard, this one is often arbitrarily chosen.

Requirements for transient immunity should be set at the system, rather than at the circuit module level.

Before the short circuit fault can be cleared, some portion of the power distribution network will experience an outage. The magnitude and duration of the outage are somewhat unpredictable.

If uninterrupted operation is required, each piece of equipment must have its own local "hold-up" energy for up to 20 milliseconds. The most stringent availability criteria require that this hold-up time be met even when the bus voltage is at or near its minimum value. This can lead to large values of hold-up capacitance.

Even if the end equipment were completely isolated from bus outages, some local, internal hold-up capacitance is usually required. Consider a system that uses a distributed power architecture within the sub-rack (or card cage). A short circuit on the power input of one module will cause the voltage within the equipment to collapse until the fuse on that circuit card clears. By using the smallest, fastest fuse possible, this time can usually be kept to less than one millisecond.

Obvious causes of CO power loss are AC mains outages that outlast the capacity of the batteries or fuel supply of any local generators. In the typical CO, batteries are sized to support operation for eight hours after AC power is lost. If the CO has only battery backup, the voltage will decrease slowly over that time from the nominal float voltage to the point where the low-voltage disconnect is opened.

At that time, the bus voltage is instantaneously removed without warning. Power converters operating from CO power must be able to operate for extended periods with input voltages as low as 38 VDC.

If the CO uses a generator to provide backup during extended outages, then power may be lost without warning at any time when the generator runs out of fuel.

Design issues Today, networking and computing equipment must be able to work either on a customer's premises or in a CO. These environments trigger certain requirements.

Just before shipment from a manufacturer, equipment needs to be easily and quickly configured for operation from either AC or DC power. How this is done depends on the architecture of the power system.

With a central power system, both AC/DC power supply and DC-DC converters, hot-swappable and in parallel, must be designed. For a 1+1 configuration, the DC-DC converters can operate from the A and B buses to provide the input power redundancy necessary for the CO. This requires two active power converters to be designed-one for AC input and another for DC input. Designing two power supplies for one system is generally not desirable, especially if the production volume of one version will be low.

With a distributed power system, a front-end power supply is used to create the 48 VDC from mains power for AC input applications. For DC input operation, passive DC input modules are used. This is a much less expensive solution than an active converter. When examining the total cost of ownership for systems that must operate from either CO or mains AC power, the low cost of such a DC input module often sways the system toward using distributed power.

If the system uses distributed power with passive input modules, it must use a positive ground system within the chassis. Fuses must be in the ungrounded leg of the circuit.

If a central power system architecture is used, there is no choice in transient suppression strategy. The converter, be it AC-DC or DC-DC, must absorb the entire transient. Any part of the transient passed to the output is also passed directly to the load circuits-generally unacceptable.

In a distributed power system, the transient, in principle, can be absorbed over several stages: the input power to internal bus interface (either an AC input front-end power supply, an active DC-DC converter or a passive input module), the input power filtering on each circuit board and the on-board converters.

In practice, if an active converter is used between the CO power bus and the internal system bus, the converter is designed to absorb the entire transient.

When a passive DC input module is used, the first attempt is to budget the transient across all three power stages between the input bus and load circuits. In practice, the input module generally absorbs the entire transient.

The next consideration is the on-board filter, usually small-to-conserve board space. Typically, they provide little in the way of transient blocking, meaning the input module must block the entire transient. This often requires multiple stages of filtering and transient absorption.

Further, substantial hold-up capacitance may be required to ensure continuous operation. One way of doing this is to place all of the capacitance to hold up a given circuit module on that circuit board.

The other way is to split the hold-up capacitance between the circuit modules and the equipment level power input module. This is particularly attractive if the equipment is to be powered by both AC and DC inputs.

If a system is powered by an AC input power supply, that power supply will have the hold-up capacitance needed for the entire system (excluding the hold-up capacitance needed on each card against faults within the equipment). For DC systems, the capacitance required to hold up against input outages is placed in the DC input module.

Bus voltage and system cooling When the system is operating on battery power, the bus voltage may be as low as -38 VDC. And because the battery discharge takes place over hours, the equipment must be designed to operate indefinitely at these low input voltages.

With loads that are essentially constant power, current in the distribution buses and converter inputs is higher than normal, and system losses are higher than normal. While the extra losses at low line are not very problematic, they are worsened by the widespread use of DC fans.

If the fans that cool the system operate directly from the distribution bus, they will be running slower and providing less cooling at low bus voltages-just when the system needs more cooling. This problem can be addressed in three different ways: * First, constant-speed fans can provide adequate cooling at low bus voltage, though this can be noisy.

* Second, fans with noise-minimizing speed control circuits can be used. The penalty is the addition of control circuitry and a DC-DC converter to provide the adjustable voltage to the fans.

* Third, a slightly more popular solution: A fixed output voltage converter can maintain the fans at constant speed regardless of bus voltage. However, an extra converter is needed, and this solution also is slightly noisy.

Designing CO power supply frameworks in an era of industry convergence is a demanding and precise job. There are important decisions to make at every level, for every process, including distribution, redundancy, fault tolerance, outage restoration, bus voltage and even system cooling. Becoming familiar with the options presented by these issues is a significant step in learning how to build the CO of the future.

Bob White is the Engineering Manager for Artesyn Technologies' High Power Standard Product DC-DC Converters Group in Broomfield, Colo.