Large UPS Technology report: Balancing green and high reliability

By Dennis P. DeCoster Mission Critical West Inc., Redondo Beach, Calif. December 16, 2009

Twenty five years ago, the large UPS market—500 kVA and above—was served by two distinctly different technologies: rotary and static. At that time, a rotary UPS for data center applications was a rather kluge-like combination of motors and generators, while a static UPS was a step-wave inverter-based technology (Figure 1). The slow switching statics of the day had many reliability and application issues, while the rotaries were very inefficient but, in many cases, more reliable.

I had the opportunity to help introduce to the Western U.S. market a new technology, which I coined “hybrid rotary” because it embodied the best of both static and rotary approaches. In its day, that product actually was the most reliable as well as the most efficient UPS available for data center service, and it was very successful despite a high price tag. Decisions were easier if the money was there.

Flash forward to late 2009. Today, there is no single UPS product that can claim both titles. There are many new technologies wrought with both promise and confusing complications. Even as UPS technologies have evolved over the past two decades, so too has Mission Critical West . Beginning as a UPS sales and applications engineering firm with both static and rotary UPS principals, we moved into turnkey design-build installations of UPS and critical facilities infrastructure, then into reliability/availability assessments and large UPS evaluations for end-user clientele.

In the process, we have witnessed hundreds of factory and site performance tests of large UPS, and amassed a substantial database of end-user and third-party service firm inputs on actual field reliability and repair issues. We decided in mid-2009 to summarize the salient pieces of this work on latest generation large UPS systems in “The 2009/2010 Large UPS Evaluation Report,” a comprehensive study giving end users an unbiased, user-oriented and comprehensive analysis of popular large UPS systems available in the U.S. This article is based on the report.


Lower utility usage and reduced carbon footprint are desirable for all aspects of facilities operations; the critical power environment is no exception. Data centers in particular have come under fire as a major power user and a target for power reductions in the national interest. The basis for this is the rapid growth of power consumption for data centers, now projected to approach 100 billion kWh within the next two years. In reaction to pressure from the U.S. Environmental Protection Agency/EnergyStar , Dept. of Energy, and other federal and state agencies, most UPS manufacturers have released, or will release, green products, or perhaps re-marketed existing products with a new, eco-friendly face. The goal in all cases is to tout higher efficiency and lower carbon footprint.

But is efficiency all that is important? Most data center managers would agree certainly not. For decades, high UPS availability (and therefore UPS reliability) was paramount. Realistically, it still is. There will be little sympathy for the manager who saved $100,000 in utility savings while being named the responsible party for the load loss that cost the firm $1 million in revenue and lost accounts. And making a UPS system with proven reliability highly efficient by chopping out components that gave it that reliability is risky. So the challenge is to balance utility cost reduction with required high critical power availability for typical large UPS applications. In “The 2009/2010 Large UPS Evaluation Report,” we did that in our evaluation of 10 popular large UPS and continuous power supply (CPS) products available in the U.S. in electrical capacities of 500 kVA and up. The discoveries we made were, in some cases, surprising.


To keep things simple for the required length of this article, all UPS/CPS systems for critical load service are grouped into two basic categories. First is the double conversion type, where all of the power is converted from ac to dc, and back to ac. The second type includes line interactive (which includes most rotaries) and delta conversion topologies, both of which shunt power around the rectifier/inverter to save on losses.

The many new developments within these technologies are explored in “The 2009/2010 Large UPS Evaluation Report”; only some of the general differences and a few key new developments are addressed in this article. Although a variety of voltages are used in the United States, such as 400, 480, 600, 4,160 V, and others, this article primarily addresses 400 to 480 V systems. All of the configuration and transformation possibilities exceed the space of this article.

Double conversion technologies still dominate the high end of the market and historically have been regarded by some as the most reliable of UPS topologies, while also seen as the least efficient (Figure 2). Our evaluations show that this is, in fact, true for certain products, but not true for some newly developed products, despite claims to the contrary. Line interactive products, on the other hand, have suffered under a sense that they have lower reliability/availability than double conversion, but are more efficient. We have found both of these assumptions to be flawed in actual field deployments, depending upon the product in question.


There have been several notable developments in both UPS rectification and inversion (Table 1). For years, large double conversion UPS systems used either phase-controlled, silicon-controlled rectifiers (SCRs), 6- or 12-pulse, or uncontrolled diode bridges with separate battery charging. Insulated gate bipolar transistor (IGBT) rectifiers were used at lower power levels but not at high power levels (hundreds of kW) due to device constraints (1,200 V IGBTs are typically required at 480 Vac) and, to a lesser extent, transient susceptibility from proximity to very high power service entrances.

Today, there are several large UPS products available with IGBT front ends that purport to have solved these issues. If considering double conversion topology, you now have a choice of rectification: 6-pulse SCR, 12-pulse SCR, diode bridge with dc chopper, or IGBT. The standard for high reliability large UPS has been the 12-pulse SCR. Although 1% to 2% less efficient than 6-pulse, it offers lower input THD and lower required capacitance, which help to minimize leading power factor, which is potentially responsible for standby generator failures under light load conditions. It also offers faster response for transient load fluctuations, easing the hits on the UPS battery.

Diode bridge rectifiers are simple, exceptionally reliable, and flood the inverter with all the power it needs for transient events. But they require separate devices and circuits for battery charging, which are less reliable. Efficiency is good, but not best. IGBT rectifiers offer excellent active control of reflected harmonics and excess capacitance issues, while giving excellent transient (inverter) load performance due to their much faster switching speeds. But IGBTs are multiplexed transistor arrays, and therefore more susceptible to high energy transients than their hockey puck SCR or diode equivalents.

Additionally, conventional IGBT rectifiers demonstrate better than average efficiency at high loadings, but this tails off sharply at light loads. One newer IGBT rectifier technology features a three-level approach—rather than conventional two-level—enabling lower rated IGBTs and better controls. This carrier-stored trench-gate bipolar transistor chip technology (available to UPS builders from a variety of rectifier manufacturers) also provides reduced switching and conduction losses compared to conventional IGBTs. The result of using these “trench” IGBTs is a dramatic improvement in low load efficiency, albeit at higher parts count and perhaps cost. Our evaluation showed this efficiency improvement to be exceptional.

Inverters have changed somewhat less radically perhaps, but there have certainly been positive changes. Most manufacturers have adopted some form of digital signal processing in their latest product offerings. This, along with the evolving maturation of pulse-width modulation IGBT inversion in its various configurations, has produced excellent reliability with improved efficiency virtually across the board. Switching speeds and modulation technologies have been optimized in late generation products to produce an excellent combination of load handling and efficiency across a wide range of loads.


Traditionally, output transformation has been used on all double conversion UPS. However, several UPS manufacturers are now offering (or planning to offer) large UPS models that can operate without output transformation. This obviously saves space and weight, lowers manufacturing cost, and would improve system efficiency.

The tradeoff is in isolation at several levels. Most significantly perhaps is the buffering effect between inverter and dynamic loads, and between parallel modules themselves in very large systems. This puts tremendous pressure on designers of transformerless systems to perfect inverter speeds and controls such that no bumps occur that can take a module offline during dynamic events (load steps, outages, faults, etc.) under all conditions. This task is rather daunting, and in our evaluations, some have had issues getting this right on the first pass.

One surprise we found is that a transformerless design does not always translate to lowest UPS efficiency in class. In side-to-side manufacturer test results, one transformed design actually had lower overall losses than a transformerless design in target partial load ranges. There is no question that a transformerless approach is higher efficiency than a transformed approach on the same basic module configuration, but all components of the UPS must be optimized to render the highest efficiency in class.


Line interactive and delta conversion UPS systems divert the bulk of load power around energy-robbing rectifiers and inverters through magnetics designed to electrically slow power flow enough to afford voltage regulation if needed (Figure 3). Generally, these systems work quite well despite absence of steady-state frequency regulation found in double conversion systems. In generator-backed applications in the United States, there is very little need for steady-state frequency regulation due to typical grid stability and switch mode power supply loads.

With few exceptions, line interactive UPS are more efficient than double conversion UPS by several points at full load. However, depending on the type of energy storage, some line interactive efficiencies can tank to as little as 80% or less at partial loads (25% to 35%). Once relegated to niche sales opportunities, line interactive technologies today are quite definitely mainstream with more than $1 billon in operation. This is underscored by a reference list, which—despite not nearly as large as that of double conversion users—is substantial with a hundred or more Fortune 1,000 firms and government entities.


Offline UPS systems dominate the very small, sub-2 kVA UPS market, but are virtually never considered for large, high criticality service. This is due in part to the absence of steady-state power conditioning or voltage regulation, and to the relative under-sizing of inverters that are not continuous-duty rated. Logically, shunting virtually all power around conversions devices (inverters, rectifiers) will lead to extremely high efficiencies, and they do—around 99% for those brave enough to consider the approach.

But many manufacturers are now considering or actually offering offline alternatives relabeled “green” or “eco” or “energy saver,” as additional operating modes in their double conversion UPS. The benefit of this development is that the systems will use the same continuous duty inverter they use for normal operation, which is a reliability plus. The regulation issue remains, but for some less critical applications in very expensive utility regions, we see this as a potential plus.


Rotary UPS systems, including CPS systems, have evolved over the years to mean primarily flywheel UPS systems today. Many flywheels by themselves are actually somewhat inefficient relative to battery alternatives, due to frictional losses, support system losses (vacuum pump), rotor windage losses, and/or bearings not present in battery systems.

Once charged, battery systems draw little current and generate no heat to speak of. But the flywheel UPS system may be quite efficient, particularly at high loads. This is because virtually all manufacturer-integrated flywheel UPS/CPS systems employ line interactive technology. So, overall system efficiency becomes a composite of both flywheel losses (high or low, depending on manufacturer), and ac-ac power conversion losses.

The most efficient flywheels are those using pure magnetic (air) bearings. However, many high speed flywheel products are deployed in double conversion UPS products, resulting in relatively low ac-ac efficiencies (91% to 94%, similar to double conversion battery UPS) and often somewhat lower reliability at increased cost. However, in selected space-critical applications, the footprint advantage for flywheels used in their sweet spot can counterbalance these issues.


As already noted, we see a tremendous variation in UPS efficiencies as technologies and loads vary. Figures 4 and 5 illustrate how wide these variances can be in some cases. The impact of low efficiency can be dramatic. Assume one particular system has, perhaps, a 6% advantage over another at a given loading (which is quite possible), and that UPS was air conditioned. You could then reasonably assume an 8% swing when cost of air conditioning is considered. At average U.S. utility rates, this amounts to more than $300,000 in extra utility costs (and more than 1,500 tons of CO2 equivalent) for every 5-year operating period, for every 1,000 kW of load carried. This is without rebates.


Statistically, an N+1 parallel system is most reliable/available when N is a smaller number. Therefore, multi-MW critical power designers favor lower numbers of large UPS modules over higher numbers of smaller ones, if all other criteria are equal. This has moved 750/825 kVA, 1,000/1,100 kVA, and even 1.3/1.6 MVA modules into high use in mega-campuses. But higher levels of redundancy—and therefore availability—in N+1, or System + System (where each system is N+1) UPS configurations, necessarily means lower normal mode load levels, and potentially lower efficiencies. Loadings of 30% to 45% are typical in redundant System + System configurations.

UPS paralleling today in most systems is by way of a main system control cabinet with master static bypass, module-level static bypasses and controls, and generator paralleling with electromechanical switching in some rotary systems. Until recently, in this country discrete module level bypass and paralleling was thought to be less reliable due to the number of switching events. However, we have not found that to be the case. Rather, the incidents of parallel failure across the board are considerably lower than found a decade ago, regardless of technology.


Designing UPS at the medium voltage level (i.e., 4,160 V) has advantages in some applications. At very large power levels, cost of distribution is lower, systems in many incidences can be or are deployed in outside locations (good and bad), and efficiencies can be improved since one level of transformation can be eliminated as well as UPS air conditioning (product-dependent). A case might even be made for reduced arc flash hazard using latest generation interrupting devices.

On the minus side, very few UPS products are made for direct medium-voltage use, and many of those that are have significant downsides such as multi-ton flywheels with extreme Mean Time to Repair issues, offline inverters, very low partial load efficiencies, or other issues. Still, it is worth doing a cost-benefit analysis in less critical multi-MW commercial or industrial applications for medium voltage.


Utility costs have risen quite sharply in some U.S. regions recently. While the average U.S. utility cost is now $0.10/kWh, in some areas it could easily be as high as $0.15/kWh or more. Not withstanding discussions of green grids and infrastructure overhauls, utilities have realized it is often more cost-effective to offer rebates for high efficiency customer renovations than to build new infrastructure. These rebates vary from region to region depending primarily on capacity and reserve constraints, but one thing is crystal clear: Look into and secure rebates before—rather than after—any critical power equipment purchase.

Table 1: Large UPS rectifier types

Rectification Type Reliability Input THD Capacitance Issues Efficiency (Partial Load) Relative Cost
This table compares reliability, harmonics, capacitance, efficiency, and relative cost of five different large UPS rectifier types.
6-pulse SCR Mid Highest High Mid Low
12-pulse SCR Mid Mid Mid Low Mid
Diode bridge Highest Mid Mid Mid Low
IGBT Mid Very low None Low Mid
CSTBT Low Very low None Highest High


Author Information
DeCoster is executive principal at Mission Critical West Inc., a California-based consulting service. With 25 years of experience in the UPS industry, he has consulted for many Fortune 500 firms, and has authored articles on UPS and dc storage technologies in Consulting-Specifying Engineer and Pure Power , and papers for IEEE-IAS, the Power Quality Conference, and battery conferences. He is the principal author of “The 2009/2010 Large UPS Evaluation Report,” an annual evaluation and comparison of major UPS manufacturers and products.