Examining the 2009 offerings by six of the largest uninterruptible power supply (UPS) manufacturers in the 500kVA and larger capacity range, Mission Critical West, Redondo Beach, Calif., recently unveiled its “Tier IV Critical Facilities Infrastructure” study. Analytical comparisons in the report are presented in the following areas: dynamic performance, reliability/availability, scalability and redundancy issues, green issues, logistical concerns, service and support and price.
Highlights from the 58-page study, which incorporates vendor factory input (including substantial factory tests), end-user interviews, in-house testing data, commissioning data, and product technical specifications, are outlined below:
UPS reliability/availability is the single most important (highest weighted) parameter in any comparative UPS analysis. In all practical cases, users can live with higher utility bills, less space, or slightly dampened transient performance. However, if systems — or their subsystem modules — are prone to failure, the impact can be catastrophic if the cost of downtime is high for that user.
Certain design changes in recent years have had reciprocal impacts on the reliability and efficiency of UPS systems. For example, as some new modular and/or transformer-less designs emerge, care must be taken to ensure equivalent reliability to tested designs using isolation transformation. Similarly, technology changes in rectification will impact efficiency, reliability, and dynamic performance.
Module mean time between failure (MTBF) rating will vary significantly with parts count, operating temp, component design and rating, and other parameters. For instance, The MTBF rating of a given UPS module operating at full load rating at maximum rated temp (40°C), will routinely be four or five times lower than the same module operating at half load in a 25°C ambient. MTBF calculations are normally published at full load rating, which does not apply precisely in Tier II-IV applications at lighter loading but can be used for weighting. MTBF values can easily be misconstrued and should only be used as a relative guide. This not only applies to calculated MTBFs, but also to some field or empirical MTBF data. For example, one vendor’s claim of “over 2 million hours actual” MTBF may be valid over its entire database, perhaps primarily made up of thousands of smaller single-module lightly loaded UPS systems with transformer isolation from service entrance disturbances; however, it has little to do with the expected performance of a fully loaded, multi-module parallel system in a low-impedance/high-power environment. Therefore, it’s important to remember that most vendors apply liberal data exclusion for whatever they deem to constitute infant mortality, commissioning/test stress, design changes, and/or other issues they feel may have caused certain premature failures. The report’s authors attempt to balance all of this in their analysis.
Large module UPS systems can have a series of dynamic performance challenges over the course of their operating lifetime. These depend on input source impedance, output distribution and load specifics, interrupting device design, redundancy, and other criteria, in addition to UPS design issues. A UPS module may see a large step load as other modules or systems drop out. It may be asked to transfer to alternate source or bypass during an unstable utility or generator supply period. It might even see a bolted short circuit directly on its output. UPS systems also can have impacts on diesel electric generation systems as well as contribute to neighbor and in-house distribution harmonic content.
There are a series of frequently cited performance specifications for large UPS systems, some of which are very important (i.e., fault clearing) and others that are relatively unimportant for most or all applications yet stated by vendors to create a perception of importance (i.e., output V THD distortion). Some are important up to a certain specification limit, and then become much less so. Some specifications are unimportant for some applications, yet critical for others. One example is step loading, which may never become critical for 7 x 24 data centers where "N" (as in "N+1") is a high number, yet be critical for some chip fabs or pharmaceutical processes.
UPS efficiency, a function of its “green” advantage (or disadvantage), has become more important in recent years for data center applications. This trend has been driven by LEED and GREEN GRID initiatives, new federal guidelines, and industry standards aimed at reducing energy consumption, load on grids, and greenhouse gases.
It has become an element of pervasive power utilization effectiveness (PUE) and data center infrastructure efficiency (DCiE) metrics. According to the report, there are certainly real savings to be had by making good choices in terms of UPS efficiency so long as reliability and performance are not impacted.
The most efficient large power UPS systems (excluding DVR & similar technologies) are line interactive or delta conversion. There is still some debate as to whether or not these technologies really match double conversion in terms of reliability and performance. Many of these designs also use 5- to 15-sec flywheels in place of batteries, which furthers the debate. In addition, some very large line interactive flywheel systems have unique service issues relating to large bearings and couplings — and usually are very high in initial capital cost. Mid-size systems avoid some of these issues.
Solid-state (static) UPS module efficiency vary with component design and rating, amount of inductance and capacitance, PF of load, metering accuracies, and other parameters. One major item is amount of transformation. While transformation is good for isolation, and overall reliability (in most cases), this system buffering can be bad for efficiency. Newer UPS designs generally use smaller (or fewer) magnetics due primarily to higher speed inverter switching. Therefore, they are theoretically more efficient. However, that design hope does not always work out, as you can see in the actual test results of the report. “Smart” designs also recognize that Tier III and IV UPS systems run normally at 30% to 45% rather than 60% to 90%. As a result, systems are designed to optimize for that.
Products in large UPS power ranges might be equipped with or without optional input filtering, and may use either 6-pulse or 12-pulse rectification, diode bridge, or IGBT rectification, all of which have pluses and minuses on performance and efficiency. With or without special input filtering, all products should not exceed 10% current THD at any load level from 30% to 100%, and would be at or near 5% at full load and 0.85 to 0.95 (or unity) input PF. Actual data center site PF typically approaches unity, and product efficiencies can vary plus or minus 0.2% absolute or more from rated 0.9 PF depending on design. For more information, visit the company’s Web site.