The convergence of information technology (IT) equipment and telecommunications networks has brought about the need to power equipment with a variety of input requirements. Designers of today's telecommunications facilities are faced with the challenge of configuring a power system that supports these requirements as reliably and economically as possible.
Traditional telecommunications equipment generally requires -48VDC input power. Such power systems consist of multiple parallel-redundant rectifiers that convert AC power to -48VDC power, charge lead-acid storage batteries, and supply power to critical-load equipment. Converters or inverters are used to derive other required voltages from the -48VDC power plant. Long battery support times are necessary to support the equipment during AC power or rectifier failures. Sometimes plant personnel use engine-generator systems to supplement the AC power during sustained power failures. Battery support times range from a minimum of 1 hr to more than 24 hr, with typical ones lasting between 3 hr and 8 hr. Fig. 1 shows a typical telecommunications power system using rectifiers and -48VDC battery systems to support the critical load equipment.
Traditional IT equipment requires AC input power that generally matches the available AC power source configurations — typically 120V, 208V, or 240V single-phase AC in 60-Hz countries and 220V to 240V single-phase AC in 50-Hz countries. These power configurations use AC UPS systems with battery systems sized to provide necessary time for an orderly shutdown or to start standby engine-generators. Virtually all critical IT facilities include permanently sited engine-generator systems (and associated automatic transfer switches) to protect against sustained AC power failures. Fig. 2, on page 23, shows a typical IT power configuration using AC UPS systems to support critical loads.
Both traditional telecommunications and IT facilities generally include other critical support equipment, such as lighting and air conditioning. This equipment generally tolerates brief interruptions in power without adverse consequences on the operation of the telecommunications or IT equipment.
For telecommunications systems without permanently sited engine generators, it is common to power support equipment with inverters connected to the -48VDC power system. IT systems run such equipment with AC power. An IT system without permanently sited engine generators provides orderly shutdown time, while one with engine generators provides extended run time.
With the convergence of telecommunications and IT equipment, it is likely that critical electronic loads in any one facility will require both -48VDC power and one or more of the AC power voltages. Often a site's system cannot operate unless both the DC- and AC-powered loads are operational. This codependence demands equally high reliability and availability for the DC and AC power systems.
Why have 8 hr of battery backup for the DC-powered loads and only 15 min for the AC loads? The answer lies in the historical biases of AC and DC electrical power system designers. DC power engineers believe batteries are more reliable, but AC power engineers believe power conversion equipment, engine-generators, and AC power are more dependable. Understanding each perspective requires an analysis of both systems' application.
The AC/DC Debate
DC power systems use a multitude of parallel rectifiers to derive -48VDC from AC power. For example, a 4000A rectifier system is often built using 21 or more 200A rectifier modules. The typical mean time between failure (MTBF) of these rectifiers is between 100,000 and 200,000 hr. With so many rectifiers used in parallel, the observed failure rate of the power conversion equipment is substantially increased.
From the AC power engineer's perspective, the most reliable system would use N+1 parallel redundant rectifiers, with N as small as possible, to reduce parts and the probability of a rectifier failure. In other words, the more components, the higher probability of any one component failing (Fig. 3).
DC power systems use -48VDC battery plants with long-duration backup times (i.e., low rate of discharge). They use only 24 cells in series and apply them at a relatively low discharge rate. Furthermore, the end-of-discharge and float voltages are carefully controlled.
On the other hand, AC UPS systems typically use 120 to 240 cells in series and apply them at very high discharge rates (i.e., 10- to 20-min backup times), with deep discharge voltages down to 1.65V per cell or less. With these striking application differences, it's no wonder that DC power engineers believe batteries are reliable, while AC power engineers believe they are the most unreliable component of the AC UPS system.
DC systems don't always include permanently sited engine-generator systems. If they do, they often lack regular maintenance and testing. In addition, a serial dependency with reduced observed reliability exists because both the rectifier and engine-generator must be operational.
High-availability AC power systems routinely include large, permanently sited, redundant engine-generators that receive careful maintenance and regular testing to assure availability when needed. The AC load equipment can use the output of the engine generator even when the AC power conversion equipment is in bypass.
Fig. 4, below, and Table 1, on page 24, show first-order reliability calculations comparing the relative calculated unavailabilities of DC and AC powering options. The unavailability of an AC UPS system (with 10-min battery backup) is 3.5×10-10 . For a DC power system (with 8-hr battery backup), the unavailability is approximately the same at 9×10-10 .
The unavailability of the DC power system is dependent on the size of the battery plant. For example, if you use a 3-hr battery, the unavailability of the DC power system increases to 3.45×10-7 . The AC UPS power system is relatively unaffected by battery size. Increasing the battery backup time to 3 hr yields the same unavailability as the 10-min battery. These first-order calculations don't include real-world issues such as failures in the power distribution system, lack of maintenance, and human error. They do, however, point out the relative dependencies on the various subcomponents.
One of the arguments often posed in favor of DC over AC power is that it is more straightforward with fewer power conversions. The source of AC-powered electronic equipment is the public utility. The UPS converts AC power to DC power to charge the batteries and power the inverter. The inverter converts DC power back to AC power. This AC power supplies the load equipment, which in turn converts the power back to DC.
In DC power systems, the rectifiers convert the AC power to DC and directly supply the DC-powered load equipment. But an often-ignored fact is the load equipment uses power converters to convert the -48VDC power to other voltages required by the electronics. The expected efficiency advantage of a DC powering approach is often not realized.
The AC-to-DC efficiency of the multiple, smaller rectifiers that typify the DC system averages 89%, while the AC-to-AC efficiency of large UPS systems generally ranges from 92% to 94%. Since the efficiencies of the different power supplies are approximately the same, the overall efficiencies of the two systems produce nearly identical measurements (Fig. 5).
In the case of DC power systems, the mitigating reliability factor is the battery is directly connected to the load bus. For AC UPS systems, the mitigating factor is the ability to provide an alternate source of power with the UPS bypass circuit. Despite the philosophical differences in powering perspectives, both approaches are feasible. And when implemented properly, both demonstrate high reliability.
Input Power Requirements
When you select a power system configuration, the input power requirement of the loads is the most important consideration. Other factors include size, installation costs, operating efficiency, and maintenance costs. Most IT equipment requires AC input power, just as most telecom equipment requires -48VDC input power. The mix of the two generally dictates the configuration. For example, a central office that uses -48VDC power for most of its critical equipment will lean toward a conventional DC approach. Data centers with equipment that mainly uses AC power will follow a conventional AC UPS approach.
If the manufacturer offers a choice of input power, the percentage mix becomes a variable, depending on the particular preferences. For typical Internet hosting sites and server-based telecommunications sites, the mix of AC-powered loads (in kW) ranges from 85% to 95%, and DC-powered loads are 5% to 15%. Typical site sizes tend to be in the 10,000-to-100,000-sq-foot range, with power densities of the load equipment measuring 35W to 80W per square foot. With these requirements, the critical power system sizes tend to fall between 350 kW and 8MW. At -48VDC, these power levels range from 7000A to an astounding 160,000A. At 480VAC 3-phase, these same power levels fall within a more manageable 420A to 9600A. Controlling voltage drop in -48VDC power systems at higher power levels becomes extremely difficult and expensive. Clearly, larger power systems favor AC power distribution systems or smaller DC power systems located close to the load equipment.
Distributed DC Power Configuration
Typical IT rooms use packaged, commercially available power distribution units (PDUs) to distribute AC power to the load equipment. The PDU performs the conditioning, distribution, and monitoring of the load equipment power. The 480VAC output of large UPS systems is distributed to a number of PDUs located throughout the data center. A PDU typically contains an isolation transformer, which provides voltage stepdown to 208/120VAC, common-mode noise isolation, local voltage adjustment, and ground referencing. It also contains output distribution panelboards with output circuit breakers, cables, and receptacles to match load equipment requirements. PDUs should be located close to the load equipment to minimize the output distribution circuit length and voltage drop.
For -48VDC power systems, a similar PDU approach is possible. Point-of-use N+1 redundant rectifier systems (without batteries and with integral output power distribution) are located close to the load equipment. The rectifiers are powered from the same AC UPS systems as the other associated electronic load equipment. The footprint of these batteryless rectifier systems (DC PDUs) is approximately the same as the secondary DC distribution bays normally required with conventional centralized DC power systems.
To mitigate potential failures and allow maintenance within the AC or DC systems without load shutdown, dual UPS systems with redundant power paths, dual rectifier systems, and dual-input load equipment are used (Fig. 6). This high-availability approach is known as a “hybrid distributed redundant power system.” On the surface, this sounds expensive, but reduced installation costs and a smaller footprint make it cost-effective.
Telecom and IT equipment that requires AC and -48VDC power provides new challenges for system designers. Conventional configurations may not be the most appropriate choice for powering this type of equipment, particularly for Internet applications with high-load power densities, codependency of the AC- and DC-powered load equipment, and a mixture of loads that require AC input power.
Alternative, cost-effective configurations can create systems that have the necessary reliability, availability, and maintainability to meet today's nonstop processing requirements.
Gruzs, T.M., and James Hall, “AC, DC or Hybrid Power Solutions for Today's Telecommunications Facilities,” INTELEC® 2000 Conference, Phoenix, AZ, Sept. 10-14, 2000.
“Powering the Internet, Datacom Equipment in Telecom Facilities: The Need for a DC Powering Option,” Technical Subgroup on Telecommunications Energy Systems of the Power Electronics Society of the IEEE, INTELEC Web site: http://www.pels.org/Comm/Telecom/Telecom.html.
Cost Comparison — A Case Study
A large Internet hosting company was building a number of sites nationwide. Each site was approximately 10,000 sq ft with a planned load density of up to 75W per square foot. The exact mix of AC- and DC-powered load equipment was unknown since it was dependent on the tenant's selected equipment. However, the design criteria allowed DC-powered equipment to comprise up to 10% of the total load. To assure maximum availability, the design required redundant, dual-distribution power systems to support load equipment having dual-input power connections.
The selected AC UPS configuration was distributed redundant, 750kVA systems (two independent systems with three 375kVA modules, each for N+1 redundancy) with 15 min of battery backup at full load. Permanently sited, redundant (N+1) standby engine generators were included. The ultimate maximum total -48VDC power load to be accommodated was 1600A. The initial design required the ability to support half the ultimate load. All of the loads were expected to be dual-input DC- or AC-powered equipment. The company considered several different DC power-system configurations.
Two designs were selected for engineering and cost analysis. One configuration used centralized 1600A DC power plants. The alternative configuration used distributed batteryless DC rectifier systems powered from the AC UPS systems. These rectifier systems were deployed in capacities of 400A each. The rectifiers and output distribution were combined and located in the same footprint as the centralized system's secondary distribution bays. The dual redundant AC UPS systems and redundant standby generator systems provided outage protection, thus eliminating the 1-hr DC system battery requirement. The AC UPS size was the same for either alternative because both the AC- and DC-powered load equipment required accommodation. Furthermore, the incremental increase in UPS capacity to accommodate the additional 10% DC load did not cause the UPS module size to change.
The hybrid distributed DC power system configuration met the high-availability requirements of the Internet hosting company by using the dual redundant AC UPS with redundant power distribution paths. A high degree of fault tolerance was obtained without any single failure points in the AC or DC system. Any component within the AC or DC system could fail or be maintained without disrupting the critical load equipment operation.
A cost comparison of the two designs is shown in Table 2. Significant savings come with the distributed batteryless DC rectifier approach, mainly due to the lower engineering costs, reduced installation materials, and lower overall installation costs. Other benefits include reduced footprint, ease of expansion, and greater flexibility in accommodating varying mixes of AC- and DC-powered load equipment.