Not all UPS designs are the same. As a result of cost savings in the very competitive UPS market, some designs may not have the highest level of component reliability.

Capacitors are an integral part of the power conversion process and are found in significant numbers in larger rated UPS designs. As such, manufacturers must manage all aspects of their usage and extract the highest levels of reliability from each type and application.

In fact, power capacitors can be even more troublesome than batteries or flywheels if not managed properly. Certainly manufacturers must avoid the type of failure that caused the capacitor failure in the Photo and contain it to prevent migration into other critical components or loss of the UPS output. During the design process, manufacturers must consider many application factors. Many of these factors are not inherent to the capacitor and include worst-case scenarios of ambient conditions and inter-system interactions. All of the stress limits governed by the design will have a conspicuous impact on the actual failure rate and life of the capacitor.

When selecting a UPS topology, you should consider the power capacitor application practice used by the manufacturer. Investigate the common practice to assure maximum capacitor component life. Also verify that after a capacitor failure, the UPS design contains that failure to prevent disturbance of the critical output of the UPS.

Here are some important UPS selection guidelines:

  • Avoid UPS designs with capacitors applied at or near their voltage rating. Although this may reduce space and lower cost, the cost in the long run can be devastating.

  • Demand flexible connectors with properly applied current-limiting impedance instead of bus bars with no impedance.

  • Always require mechanical wire (pressure type) interrupters with sufficient mechanical clearance and freedom of movement for isolation of a failed capacitor and prevention of fire.

  • Verify that the UPS design provides for adequate cooling of the capacitor bank. Remember, capacitors, like all electrical and electronic components, in particular DC capacitors, will decay with heat.

  • Ask the UPS vendor for its capacitor failure history, and always check references.

Also remember that DC capacitors, like batteries, have a limited shelf life. This comes into play with spare capacitors and excess inventory or bargain UPS equipment obtained from facilities never commissioned. This is particularly true of those UPS systems that have been in storage without power applied continuously to the power conversion circuits.

Output Voltage Harmonic Filter Capacitor Application

To better understand the specific UPS designs and how they relate to capacitor life, let's look at the most common large power UPS topologies and compare each for maximum power capacitor life and avoidance of catastrophic failures.

Most, if not all, UPS designs use AC capacitors on their output. This includes UPS systems based on flywheel energy storage or battery storage. AC capacitors form a part of the output voltage harmonic filter in all these UPS topologies. Their purpose is to provide a low-impedance path for higher harmonic voltages created within the inverter as it transforms a DC source of power into usable AC power.

These capacitors tend to be one of the components closest to the output bus and therefore have a profound effect on the load if they fail. Generally, you will find them applied in a line-to-neutral connection for lower voltage stress and lower fault impact. In some designs, however, you may find them in a line-to-line configuration, which means that a typical capacitor failure will create a short circuit across two output lines.

Capacitor designs should always also use internal interrupters to clear faults within the capacitor “roll.” This pressure interrupter is a stretched wire between one of the capacitor can terminals and extended foil of the capacitor. An internal failure of the cap results in the formation of a large amount of gas. The can, however, is designed to expand in response to the gas in the vertical direction; that is, toward the terminals. This expansion will mechanically break the stretched wire.

The prevention of fires is the intent of this pressure interrupter. A design that does not allow enough room for the capacitor expansion may lead to a fire because new capacitor oils (non-PCBs) are not as robust and benign as the original PCB-based oils.

Parallel-for-capacity or parallel-for-redundancy systems

If you use UPS units in either of these two systems, the output capacitors in each of the UPS modules are effectively connected across the UPS system due to the common output bus. A capacitor failure in this application can potentially be more difficult to isolate, since the failure would appear similar to each of the parallel-connected UPS units. What's needed in this scenario is the ability to quickly determine which UPS module has the failed capacitor. Then you must take appropriate action to clear the fault and continue operation, or isolate the entire UPS module. This is called selective tripping. UPS manufacturers use different methodologies for fault determination and selective tripping. Some are more robust than others, effectively decreasing the potential for a system load loss during a capacitor failure.

Although a robust protection and selective tripping system can mitigate the potential for a system crash or load loss, it is better to design an output filter that seldom challenges the protection system and selective tripping function to work.

AC capacitors in double-conversion units

UPS designers use AC capacitors in the input of double-conversion UPS systems (see Fig. 1) to lower harmonic current flow between the UPS and its source, whether it is the utility or a local generator. Double-conversion units typically use six- or 12-pulse rectifiers on their input. These are generally robust power conversion devices that do their job well. However, they are also a well-recognized source of harmonic current flow, with total harmonic distortions (THDs) that approach 30% for the six-pulse and less for the 12-pulse. Operation of the rectifier is the cause of these harmonic currents. Since these currents must flow through the effective series impedance of the source, this creates harmonic voltages that distort the AC input sine wave (see Fig. 2, on page 24).

The degree of input voltage distortion is proportional to the source impedance. Whereas a utility-fed source with a 5% impedance transformer will most certainly be under 10% when connected to a six-pulse rectifier, the same set of harmonics can cause a local generator (with much higher impedances) voltage distortion to be 30% or higher.

Both the harmonic current flow into the controlled rectifier of the double conversion UPS and the effects of a badly distorted sine wave voltage on the output of a local generator can cause significant interface issues between the generator and the UPS. Likewise, these same currents and voltage distortions can cause problems with other loads also connected to the output of the local generator. This is why the majority of UPS customers demand the addition of an input filter to the input of the double conversion UPS.

Although this addition will substantially lower the demand of harmonic current flow from the input power source, it is not without issues. Like the output capacitors, UPS designers use the input filter capacitors in line-to-neutral or in higher stress line-to-line configurations. And again, the system must recognize a capacitor failure and remove it from the circuit managed by the UPS protection system. If not done, the shorted capacitor will effectively remove one or two phase voltages from the input, until a protective device opens.

The input filter capacitor size is also significant because it depends on reducing (supplying) lower order harmonics to the double conversion UPS. The relative size of the capacitor bank means if the double conversion UPS is lightly loaded, the overall power factor looking into the UPS/filter combination will most likely be leading. This means the profile for the double conversion UPS appears to be capacitive to the utility or local generator.

If we examine the interface between the digitally controlled UPS and the local generator, opportunity exists for the two independently controlled devices to begin to circulate reactive currents, due to cyclic phase changes. These currents are generally at higher harmonic levels and therefore add to the voltage distortion on the generator output and double conversion UPS input. This generally exacerbates the inter-system controls, forcing engineering assistance at best and disconnection of the filter at worst.

Still another system design issue is when the UPS/filter is connected to the utility. Here, the filter will “see” harmonic voltages that exist on the grid due to other non-linear loads. The filter components, including the capacitors, may simultaneously have to supply harmonic currents both to the UPS rectifier and to other demands within the utility grid. The filter design must take this into account. If not done, the filter components will be undersized for the “real” application. This will result in establishing a secondary failure rate, one that significantly adds to the normal life cycle rates expected.

A third cause of increased filter/capacitor failures may be unconsidered resonance that can occur between the capacitor bank used in the double conversion UPS filter and other utility impedances during transient conditions. A common occurrence, which has caused problems over time, is the change of utility apparent impedance as power factor capacitor banks automatically switch in and out of the utility grid. Voltage transients caused by the switching power factor capacitors can initiate a ringing that will overtax the UPS input filter capacitor capability and cause filter failures.

The double conversion UPS input rectifier/charger represents a high source of harmonic current distortion for the utility and engine generator. Utilities, though much stiffer than the engine generators, generally call for the input current distortion to be less than 5% THD at full load. This requires adding an input harmonic filter designed for full load. This creates a problem in that, at lighter loads, the benefits of the filter become a serious problem due to their excess capacitance. This can lead to resonance on the utility or prevention of the engine generator from picking up the UPS load. At the same time, the UPS cannot synchronize to the engine generator and is unable to go to bypass.

Not All UPSs Must Have Input Harmonic Filter Capacitors

Though almost all UPS systems use capacitors in their output, one UPS topology demonstrates that reliance on capacitors in the input is not required. Delta Conversion UPS configurations are now gaining wide acceptance as a means to assure maximum compatibility and stability of engine generators without adding large input capacitor-based harmonic filters.

Instead, this design (see Fig. 3, on page 24) applies two bidirectional power converters so that the primary delta converter acts like an active harmonic filter, maintaining less than 5% THD current distortion from no load to full load, without the use of capacitor-based filters. The UPS maintains power factor at an ideal unity PF, never falling leading or causing voltage problems on the generator.

The double conversion UPS consists of a one-direction rectifier/battery charger driving a one-direction inverter that powers the critical load.

The two converters of the Delta Conversion UPS are bidirectional, enabling the UPS to aid in the engine-generator's stability. These two converters are controlled so that, as a load, the delta conversion is a constant unity power factor load of a sine wave form from no load to full load (see Fig. 4, on page 24). And, the Delta Conversion UPS achieves this ideal unity power factor and low input harmonic distortion without the need of a large input capacitor bank.

Henry Lengefeld is senior staff engineer for the Enterprise Systems & Services Group of American Power Conversion. You can reach him at

John Messer is president and CEO of Gordon Associates. You can reach him at

Note: The authors wish to thank Dr. Martin Hudis, senior vice president, technology development at Aerovox for his assistance and technical guidance in preparing this article.

Stress Derating for Capacitors

Stress derating can be extremely importance in achieving a power protection system capable of five or six nines of availability. Reliability of the DC caps is similar to the AC caps in that voltage stress, temperature stress, and current stress are all factors that can lead to higher or lower failure rates, depending on the degree of derating the UPS designer uses.

The film capacitor industry uses a two-parameter Weibull model to predict field reliability under the actual operating conditions. The industry has experimentally verified that actual voltage stress applied to the capacitor, as compared to design limit stress, can be factored at a highly leveraged power of 7 to 9.4. Operating temperature, as compared to the design limit, can be factored at a power of two. Therefore, the amount of stress (both voltage and temperature) derating can have a profound effect on the predicted (and actual) reliability of the overall system.

As a comparative example, UPS Company “X” applies an AC capacitor in a line to neutral configuration resulting in a voltage stress of approximately 41V/micrometer across a 6.8-micrometer-thick film. Based on this stress level and a worse-case operating temperature of 35°C, one can assume a capacitor characteristic life of approximately 52 million hours, based on the predictive expressions. Taking into consideration the total population of capacitors over the life of the product, a UPS designer can further predict this survival probability.

UPS Company “Y” uses a similar AC capacitor construction in a line to neutral configuration resulting in a voltage stress of approximately 54.5V/micrometer across an 8.8-micrometer-thick film. Using the same expressions and operating temperature, one can estimate a characteristic life of approximately 5.01 million hours. Therefore it is quite obvious that the degree of derating is highly leveraged, with just 13.5V/micrometer difference in applications having a 10:1 improvement (or degradation) in overall characteristic life and predicted survival probability.