Which UPS?

By Kfir L. Godrich, Dir. of Technology Development, EYP MCF, New York September 1, 2006

It’s no surprise that reliability, energy efficiency and high density are dominating discussions about data centers. High-density server facilities have quickly become a widespread reality. Not only are they adding nearly 1 GW of power demand to the U.S. load, many of the major data centers are using this power in a way that exceeds all market expectations. Past predictions for high-density server racks assumed loads in the neighborhood of 8 kW, but today it is in the magnitude of 500% higher—and in some cases even more, as racks of 40 kW are installed and commissioned.

Why are we so far beyond the forecast for energy needs in data center environments? Because of power-hungry processors and super-charged servers.

Paralleling the increasing demand for data center power is a surging UPS market that’s expected to grow in the next two years at about 10% per year. And the markets for large systems and “green” energy-efficient products are expected to grow even more.

Power densities are skyrocketing, as well, with numbers like we’ve never seen before. Witness the fact that today, there are financial data centers being designed for up to 200 watts per sq. ft., government mega-centers up to 900 watts per sq. ft. and supercomputer centers up to an astounding 1,200 watts per sq. ft.

Hitherto, unheard of power densities bring with them cooling concerns, running the range from the typical data center scheme to multi-story air-handling unit design, to water—and even CO2 cooled racks. In fact, cooling has become such an issue that the industry is only about 50 watts away from the 200-watt air-cooling CPU limitation that will necessitate the transformation of server racks within coolant-enclosed racks.

What is the impact of these trends on UPS design for data centers? Many technologies are available today and need to be carefully considered when designing modern data centers. With an eye toward anticipating and meeting the new market needs, let’s take a look at the advantages of the three major UPS technologies: classic static UPS, offline rotary hybrid and online rotary hybrid.

Classic static UPS

The classic online double-conversion class or solid-state UPS systems have been in use for more than 30 years. Evolution of powerconversion technology has resulted in transistorized design with use of improved insulated gate bipolar transistors (IGBT) with microprocessor-based control and logic circuitry as well as 100% solid-state bypass switches. The transistorized version has a faster response time, higher efficiency and fewer components, and is therefore more reliable than the older silcon-controlled rectifier types.

Additional improvements have been made in inverter technology using the knowledge acquired from vector control techniques in adjustable-speed drive applications, causing the output response to be improved by means of time response and output tracking. Inverter space vector modulation (SVM) is one of the new generations of the pulse-wave modulation (PWM), which some manufacturers find attractive due to the fast-tracking abilities. Certainly, this requires an adequate rectifier/charger performance.

We’ve also seen huge efforts directed into the extension of the range and development of large UPS’ with transformers. Ideally, this class of UPS would include two back-to-back identical power blocks so that the maintenance would be easier and more cost-effective. The transformerless technologies are definitely a step toward that goal.

In mission-critical data centers, where redundancy is a given, the loading of the UPS system is relatively low. Efficiency is the name of the game and the perpetual effort is to push the efficiency curve up and flatten it to achieve lower operational costs at a wider band of load utilization points. It’s an ongoing process. Manufacturers of the big systems are claiming that, starting at 20% to 30% of load, they can achieve a flat efficiency curve in the range of more than 94%. Testing is the key to verifying these claims. From the design point of view, for mission-critical applications, it would be ideal to have UPS systems with high efficiency on low loading. For some, however, this is not desirable.

The battery autonomy considered is typically eight minutes per side of the 2N topology system, with an average of 15 minutes for topologies with less redundancy. High-speed flywheels may be considered on the DC link in order to not discharge batteries when not absolutely necessary, and in that way prolong their lifetime.

The main advantages of the classic static UPS technology are as follows:

• Established technology with a long operating history.

• Static UPS manufacturers have a larger service force and can provide better field support to a large equipment base.

• Static modules exhibit better harmonic isolation between output and input.

• With large chemical batteries supporting the load, there is less concern regarding the ability of multiple diesel engines to start quickly and parallel to the load. In fact, studies have shown that most generator failures are related to starting.

• Battery autonomy. This is one of the strongest reasons why this technology still leads in the mission-critical market, even though some claim it’s only due to psychological comfort.

• Maintenance is relatively simpler. Most components are standard blocks and cards with relatively fast charging times and built-in diagnostics.

Offline rotary hybrid

The second major option in UPS’ is offline rotary hybrid equipment. This UPS class utilizes either an induction coupling for storage and retrieval of kinetic energy, or a more conventional flywheel, sized such that in the event of normal utility failure, sufficient stored energy allows a generator set to start (12-14 seconds). This provides a means to supply continuous power to the critical load and avoids the use of batteries. A subset of these systems includes the ride-through category. This category doesn’t include generator backup. Such units are very suitable for supercomputer applications due to the attractive price and the modus operandi of these facilities.

The offline rotary hybrid UPS consists of a choke (filter), synchronous machine (power-conditioning device), induction coupling (energy-storage device), free-wheel clutch and a diesel engine. The flywheel technologies included in this category might be high-speed and low-speed. Care should be applied when sizing the critical load with these systems, as their critical output might be different from their nominal rating.

The main advantages of the offline rotary hybrid UPS technology are as follows:

• A lower number of components lead to relatively high manufacturer published MTBF numbers for the systems.

• Power factor improvement is built-in, since the three-phase AC machine used is a synchronous machine, acting as both motor and generator depending on the situation (normal or diesel operation), when applied as a no-load device, and acting as a synchronous condenser.

• The rotary system has lower output impedance and a higher fault-clearing capability. The choke design makes it possible to clear output faults up to 14 times the nominal current and up to twice the nominal current upstream while supplying the load under normal conditions.

• The battery-less design is the leading feature that differentiates this category from the classic online solid-state UPS. This is a major advantage in the mind of those believing that batteries equal disaster.

• A large portion of these categories of installation is outdoor in weatherproof enclosures requiring ventilation only.

• The operating efficiency of this category is about 1.5% higher than the static systems.

In our view, this type of system has the best power electronics utilization in the whole UPS market. EYP MCF calculated main power blocks as dense as 3.18 kW/cu. ft. for large systems used in mission-critical data centers, and we see this factor continuing to improve.

Online rotary hybrid

The third major UPS option—online rotary hybrid equipment—uses a combination of solid-state and rotary technology, and therefore, is categorized as a hybrid system. The base unit is the motor-generator (MG), which is always online. When normal power fails, the alternate source of power consists of a set of batteries—with the same autonomy rules as the classic online solid-state systems. High-speed flywheels may be considered, once again, on the DC link-to-limit battery discharge and cycling.

The main advantages of the online rotary hybrid UPS are as follows:

• Hybrid rotary UPS systems are designed with a high overload capability, which allows for large intermittent loading of the system without going to bypass.

• The ability to supply sub-cycle currents of 12 to 17 times full load current to clear a distribution breaker fault.

• Improved I/O isolation. Due to its overload and overcurrent capability and its inertia, the hybrid rotary UPS employs an electro-mechanical switching device for its bypass, which also provides backfeed protection.

• The input current to the UPS is linear with low distortion and is nearly in-phase with the input voltage, resulting in a relatively high input power factor.

• The efficiency of this type of rotary system is about the same as the classic static UPS systems (up to 95%).

• The M/G galvanic I/O isolation is definitely a major advantage.

Topology

Propelled by the growth of power densities, total critical power has also increased in data centers. Utilities are likening the increases in power to what was seen during the telecom boom era, but the scale of power increase is much more than that; there are major data centers that are now exceeding 40 MW of critical power, with changing internal topologies due to limitations or optimization loops. For example, most data centers are moving from the concept of mega-center to “sectionized” facilities in units called “data holes” or “pods.” These spaces are in the range of 10,000-20,000 sq. ft. The pod designs have a good correlation with the critical powering techniques and the UPS systems. With static systems, the limitation of the 5,000-amp circuit breakers is something to think about. To overcome that limitation, offline rotary systems technologies and topologies become relevant. The fact that fewer units are required to fulfill the critical powering function means increased reliability for the whole system.

One interesting topic in the design of high-density facilities might be defined as the “technology vs. topology” debate. What is more important, a highly available—and more expensive—module or a lower-availability module in a more sophisticated topology? Using reliability engineering software tools, we find that the topology argument is stronger than the technology case. That’s a very important conclusion, and a debate that will continue to have a major role in the design of new data centers.

Near future

As long as the utility bill is 30% to 50% of the total operational costs in data centers, we are going to see more and more pressure on efficiency improvement. Total direct-current data centers may play a big role in that picture by changing the UPS internal and external topologies drastically, but this is not going to happen in the very near future. Instead, the near future will see UPS manufacturers focusing on creating larger-power UPS systems for data centers with improved efficiency.

Reliability requirements will still be driving most of the critical systems design with no real sensitivity to other parameters. Green alternative energy technologies are being required more often in the new world energy market, and also should be incorporated as a strategic national target. In the small-to-medium UPS ranges, the technology and the price are there. In order to have these technologies developed for the large systems in data centers, green technologies such as fuel cells, super-capacitors, photovoltaics and wind farms must be promoted and integrated in the UPS and critical systems designs under LEED considerations or the equivalent. This is an important national effort that needs to be made.

As I see it, the future for the UPS market is to provide solutions for high-availability, high-density and sustainable data centers.

‘Cool’ Flywheel Technology

An issue that’s often overlooked in keeping data centers running when power goes out is cooling. But one manufacturer believes it has a solution.

Austin, Texas-based Active Power, a manufacturer of battery-free flywheel UPS’, recently launched a new version of its backup power product, one that generates cooling as a side effect. The product, the CoolAir three-phase flywheel UPS, not only generates power for outage ride-throughs, but also cooling via its TACAS—thermal and compressed air energy storage—technology. According to Active CEO and President Jim Chishem, the idea is not only to bridge power, but cooling as well. “Essentially it’s a battery replacement product, but if a data center’s air conditioning goes down, this provides AC as well.”

The way the system works, according to Chishem, is that stored compressed air is released during an outage, passing through a heated thermal storage unit. That heated energy drives a turbine to produce AC power. During the process a stream of cold, 59°F air is also released, generating supplemental cooling.

Besides offering a novel technology, the company is also offering a novel usage plan. It reached an agreement this summer with Georgia State University to try out the equipment, where the school pays a flat annual fee that allows it to eliminate a large up-front capital cost.

The CoolAir units provide 80 kW of power for up to 15 minutes. Up to eight units can be paralleled. —Jim Crockett