BROADBAND CHOICES

The common thread through all deployment options is the increase in the base power per living unit passed. This increase does not always follow conventional POTS planning methodologies that are wrapped around standard network traffic modules. Power demands imposed by digital subscriber line (DSL) fluctuate only slightly as customer traffic changes (Figure 1). The increase in base power per living unit is thus driving up the base load demand from power plants, which in turn increases heat dissipation demands in outdoor enclosures or other facilities. Both of these problems, power and heat transfer, must be effectively managed to achieve reliable results.

To avoid early obsolescence associated with rapid technology advancement, broadband infrastructure strategies should be progressive enough to last 15 to 20 years. With some foresight and planning, network power architectures and remote terminal enclosures can remain adaptable and technologically stable as access technologies evolve.

For DSL, the power required is nearly constant, regardless of whether modems are communicating or sitting idle. This can represent from 2 to 6 W per line depending on the specific type of DSL technology being deployed. Along with this realization comes another side effect: Unlike a traditional POTS current loop where significant power is lost to the copper plant, a DSL line does not require a signaling current loop. The line card dissipates nearly 100% of its power demand. The facility that houses the DSL line card must be capable of dissipating the resultant heat without inducing elevated temperatures in the equipment.

For example, a 192-line subscriber line interface card (SLIC) terminal site upgraded with a 96-line asymmetrical DSL access multiplexer (DSLAM) node and driving a 6 to 8 kilofoot loop will draw slightly less than one watt while off-hook. Most of this loss is associated with the voltage drop across the tip and ring, power dissipated in the twisted pair and phone. Thus, if all 192 lines are simultaneously off-hook, the circuits will draw up to 192 W from its power source to drive the network. It is very unlikely that every line will be off-hook at once, but there are also other power demands to consider, such as ringing.

Commonly, network power engineers plan their SLIC power plant capacities assuming each of the 96 lines will be active 600 to 900 seconds per hour, or 17% to 25% of the time. The power plant is sized for a 48 W SLIC load plus some additional overhead power, 50 to 75 W, for ringing, pair gain equipment, housekeeping electronics and battery recharge capacity. Thus, a 150 W power plant and a terminal enclosure that can safely dissipate 100 W is suitable.

ADSL modems will require 5 to 7 W per line continuously. Assuming that half of the POTS lines will be coupled to ADSL modems, the SLIC/DSLAM site will require an additional constant power plant capacity of 576 W. This cabinet will have to house the power plant and will have to dissipate 576 W or 1728 BTU.

The most straightforward and common architecture for upgrading exiting networks with DSL is placing DSL equipment in an existing central office or controlled environment vault (CEV) (Figure 1, upgrade A). The CO's DSL modems are powered by the resident -48 V plant. User modems are powered by the terminal equipment power system. Reliable CO power uses large battery plants and generator systems. As long as the AC power system feeding the customer equipment is functioning, the network operates normally.

If the customer's DSL connection must survive the loss of AC power, the DSL modems can be optionally powered via one or more powering loops originating from the CO. In some cases, the DSL signal can be superimposed on the powering loop. The loop power voltage is -138 V and can be produced by a -48 V bus-fed DC-DC converter driving the loop via a 100 W power limiter.

A remote DSLAM also can be co-located with an existing copper- or fiber-fed DLC remote terminal or cross-connect hub (Figure 1, upgrades B and D). The simplest approach for powering the DSL equipment is to tap into the remote terminal's -48 VDC power plant via an automatic disconnect if rectifier capacity is available. In the event of utility failure, the automatic disconnect will drop the DSL load after a predetermined time, commonly 15 to 30 minutes. Alternatively, if DC capacity is not available, AC capacity may be available from the remote terminal's AC load center, saving costs associated with an additional service drop and service entrance equipment.

If the remote terminal cannot provide power to the DSLAM cabinet, some options are available. One is a direct AC feed from the AC utility without AC backup. This is the least expensive to implement if the digital services being offered are not required to be available during a utility outage and the DSLAM equipment can operate directly from AC utility voltages.

In another case, standby or uninterruptible power supply (UPS) operation during an AC outage can be implemented using a conventional rectifier/battery plant. If the DSLAM's services are not lifeline, a creative low-cost alternative to short-term backup is co-locating a small AC UPS. A UPS used to power cable TV systems delivers 90 VAC at 1000 W for up to 2 hours. It operates off-line with its loads operating from the utility via a transient isolation transformer. Upon loss of the utility, an inverter operating from batteries picks up the load via an auxiliary transformer winding. The transformer ensures that the DSLAM is protected from hostile utility voltage transients due to lightning strikes or other events.

A third option is to power the cabinet remotely with a -138 V source located in a nearby CEV or remote power node. Depending upon DSLAM deployment method, this may not be practical. The approach calls for several DSLAM locations to be remotely powered from a common power node. A DSLAM can serve over 200 living units, and in suburban applications the design area radius can easily exceed 6 kilofeet.

In this case, finding a practical common location to power multiple DSLAMs is made more difficult by large powering voltage drops over the necessarily long loops. Only when the nodes are small, or backup power is required for a fraction of the node's total load, is this option viable.

Remote digital loop carrier host digital terminals (HDTs) can be upgraded or installed to provide a number of broadband DSL services (Figure 1, upgrade C and Figure 2, approach A). The corresponding optical network units (ONUs) are equipped with DSL modem line cards as services are sold. The key side effects of this deployment are the increased demand for power by both the HDT and ONUs along with the associated increase in heat dissipation. As DSL service is deployed, both factors have to be monitored so the entire network does not develop power-related troubles.

Commonly, the ONUs are mechanically designed to handle a range of service upgrades. As new services are added, the additional thermal load at the HDT can be problematic. The HDT typically must supply the additional power demanded by both the ONUs and broadband optical cards. As a result, an HDT's power, battery plant capacity and heat dissipation demands can increase substantially. Often, this cannot be accommodated without replacing the HDT or adding more HDTs.

A lower-cost alternative to HDT replacement is a co-located power node directly adjacent to the HDT cabinet. The node would prrovide all of the -48 VDC and -130 VDC power demanded by the HDT and the ONUs by housing both the power supplies and batteries. This reduces the HDT thermal load from the rectifiers and converters as well as the physical rack space they occupy.

This strategy creates a permanent power infrastructure for new fiber in the loop/FTTC plants using remote HDT cabinets. The power node becomes as much of the plant infrastructure as the fiber, twistedpair and associated cross-connects.

Access equipment deployed in high-density urban areas may not require the HDT to be located in a remote cabinet, but rather in the CO or existing CEV (Figure 2, approach B). This eliminates concerns about HDT real estate, power and heat dissipation demands. The only remaining issue is powering the associated ONUs. The simplest approach is locating a -130 VDC DC-DC converter power plant and cross-connect somewhere in the CO or nearby CEV. Typically the power plant operates from an existing -48 VDC bus. Thus a powering network or infrastructure is created directly from the CO, logistically similar to traditional POTS services.

As powering loop lengths exceed 6000 feet from the CO or nearest available CEV, remote -130 VDC power systems may be necessary. These systems can support from 50 to 200 ONUs depending on ONU power demand and placement density. They are available in many configurations and have -48 VDC power plants, batteries, DC-DC converters and power limiters. The rectifier plant is sized to provide the base network load plus battery recharge capacity. Depending upon the nature of the deployed access equipment, the battery plant will be sized for either the entire base load over an eight-hour period or a reduced base load if the access equipment can shed broadband functions after an AC power failure.

LOCAL POWERING VS. NETWORK POWERING A family tree of access powering architectures (Figure 3) branches to two principal powering paradigms, local powering or powering network. Normally, local powering has been preferable for powering remote terminal or hub equipment. Maintenance of hubs and subsequent power plants across a serving district is manageable.

However, extending fiber deeper into the network increases the number of terminal locations or ONUs dramatically. The number of power plants becomes unmanageable, and the consolidation of power plant equipment becomes necessary. Consolidation leads to the second paradigm, in which a centralized power plant drives a host of access devices via a copper-based powering network.

AC utility service usually is assumed to be 120/240 V or 208/480 V provided by electric utilities. However, both the communications companies and the power utilities serving their networks are aggressively investigating several alternatives.

The actual utility service entrance apparatus (Figure 4) used for the application may be encumbered with a host of local code authority and utility requirements, making it more difficult to select the appropriate equipment configuration. It may be more cost-effective to purchase a pre-approved integrated power pedestal.

A key reliability concern in outdoor applications is protection from AC line disturbances caused by either utility operations or lightning. These disturbances can be significantly more severe with outdoor equipment because the typically low-transient impedance direct utility grid taps have compared with that of equipment within a building. An AC distribution system's capacity for transient suppression considerably dampens this phenomenon. The suppression method should be capable of repetitively withstanding a 200 kA surge while allowing a transient voltage to let through less than that specified for the access equipment or power node electronics (typically between 330 and 440 V). The method should also be able to communicate to the network that it has failed and needs to be replaced.

Network planners must consider three primary alternatives when powering access equipment directly from the AC utility (see Figure 3). Where services are not related to POTS, it is tempting to follow the least expensive non-standby approach of powering the equipment directly from the AC line. A 200-line DSL hub requiring 1000 W of power can avoid up to $2000 of power plant capital equipment cost and another $2000 to $3000 in battery replacement cost over 10 years. Unfortunately, network reliability strangled by power outages is a telephony carrier's worst nightmare-such outages cause 85% of customer-perceived cable TV network failures.

Non-standby powering is not viable when reliability is a top customer concern, unless certain mitigating factors exist. For one, an AC utility grid must service the network with a 99.99% yearly availability. Also, if customers' principal use of the network is passive-not for pay video or subscription information services-an occasional perceivable interruption may not damage service value.

From another perspective, great care and expense is poured into qualifying and deploying advanced access equipment that by itself is expected to provide thousands of hours of trouble-free service. For a modest percentage of the overall deployment cost, typically less than 2%, a solid powering infrastructure can reduce power-related customer trouble reports by as much as 400% over non-standby direct AC powering.

In the event that DSL access equipment is powered non-standby, alternatives should be incorporated. This can be done inexpensively when measures are taken early in the planning stages. One measure would be to specify that the access equipment's internal power supply accepts an alternative backup supply voltage input. This input may be configured to remain passive as long as AC line voltage is present, and in the event that the AC line is lost, power will be drawn from this input. Another alternative is to leave some room in the equipment enclosure-or adjacent to it-for a small battery plant with two hours of energy reserve. Additionally, equipment rack space and a possibly a portion of the enclosure's thermal budget should be reserved for powering equipment. Also, equip the enclosure's service entrance with a generator receptacle and manual transfer switch.

LOCAL POWERING, STANDBY The rectifier alternative for local powering (see Figure 3) can be based on the conventional performance and implementation requirements associated with traditional DLC remote terminal applications. Power supply redundancy may be justified only when telephony services are also being provided.

Once the AC line has returned, basing the rectifier plant size on the desired battery plant recovery time can save costs in applications where the access equipment derives only its backup power from the DC bus. Otherwise the rectifier plant must be sized to cover both the average equipment and battery recovery power demands. In this case, the rectifier will impose a 10% to 15% energy efficiency penalty on the system. This additional energy will be wasted as heat, and if the rectifier is co-located within the access equipment enclosure it must be able to accommodate this additional thermal load.

In situations where the access equipment can accept a 60 or 90 VAC input for backup power, the AC UPS option is preferable. In this case, the UPS must be able to support the maximum or peak power demand. Unlike the rectifier approach, where the battery plant covers the transient peak power demands, the UPS must serve all possible load conditions. Again, the UPS will incur an energy efficiency penalty, about 10% to 18%.

An AC UPS can provide two unique strategic benefits that result in better system reliability and use of capital. Many AC UPS products use a single line-frequency transformer concept that offers a mean time between failure of more than 300,000 hours without redundancy. The transformer has a considerable amount of series inductance and parallel capacitance between the AC line and load connections. The combination provides over 60 dBV of line transient isolation from the load. For example, a metallic 10,000 VAC line transient will produce only a 10 V transient across the access equipment power supply's terminals. This greatly improves system robustness against adverse line conditions and correspondingly reduces engineering performance requirements and subsequent failures of AC service surge protection units and access equipment power supplies.

For DSLAMs equipped with either the DC plant or AC UPS, the battery plant should be sized to provide at least one hour of standby energy. If extended outage coverage is planned via placement of a portable generator at the site, the battery plant should include an additional hour of energy reserve to allow maintenance crews time to get the generator to the site. Narrowband applications providing POTS service will require at least eight hours of standby energy. The rectifier or UPS must be specified with these battery maintenance features, such as temperature compensation and alarms for high-voltage shutdown, discharge and low-voltage disconnect.

Valve-regulated gel or absorbed glass mat batteries are best suited for outdoor applications. Physically, the battery plant can be equipped with optional battery heating pads in colder climates to ensure that the required battery maintains capacity as temperatures drop. Heating pads are required where POTS services are involved. It is also advisable to leave at least one inch of space around each battery to maintain good outside airflow around the batteries during the warmer months.

NETWORK POWERING Network powering is accomplished over a network of twisted-pair wires connected to a power supply with an output voltage that cannot exceed -140 VDC during normal operation. Each conductor pair can supply up to 100 W measured at the power system's terminals (Figure 5). The size and number of powering pairs must be selected to ensure that under peak load conditions the terminal voltage at the network load does not fall below 60% of the supply voltage and that the loop's wattage limit is not being exceeded.

To meet these two conditions, additional or larger pairs may be required for long loops, and additional 100 W outputs are required for larger loads. Be aware that the 100 W limit includes the wattage lost in the resistance of the wire pairs.

In cases where the powering loops length exceed 12,000 feet, it may be it may be necessary to use two outputs of opposite polarity with respect to ground, or 280 VDC between terminals. The network load is then connected across the most positive and most negative wire pair of a two pair combination. This reduces the loop current, voltage drop and loop power loss.

Network power systems typically are based on a rectifier/battery plant very similar to those found in many other telecom applications. The battery plant is usually designed for -48 VDC operation. This voltage is transformed or converted to something slightly less than -140 VDC by a DC-DC converter. The converter accepts the 42 to 60 V battery voltage and delivers an output that remains stable under all normal load conditions. This output stability is very important to the integrity of the powering network's performance, specifically with regard to the loop lengths and the corresponding minimum end of line voltage. Network planners will design powering loops based on staying above this voltage-if the supply's voltage drops, the ONUs become unstable.

There have been a number of attempts to use a -130 V battery system directly without conversion, avoiding the cost of DC-DC converters altogether. However, for an ONU load, the widely varying battery terminal voltage reduces the possible loop powering length by 40% compared with a constant output voltage. Therefore this can only be considered when the cost associated with 40% of the installed powering cable is less than the cost of DC-DC converters. Power limiters will still be required and must also be adapted to wide input voltage range. Typically a constant voltage is assumed, and a limiter simply monitors the loop current. However, when connected directly to a 130 V battery bus, the input voltage will vary from 105 to 140 VDC. The limiter must then respond to a continuously changing current trip to maintain a 100 W maximum. This will be critical when considering that the ONU power supply represents a constant power load.

Network power is often one of the last considerations while network business models are being assembled but one of the first hurdles when actual deployment begins. This phenomenon is at least partly based in the reality that business planners do not readily understand the technical and financial effects involved with powering fiber-based networks. Correspondingly, engineers and construction managers do not fully appreciate their effect on an aggressive business plan when specifying equipment based on outdated techniques or industry standards.