Last month, we discussed low power factor and how resonance can disrupt facility operations. This month, we'll talk about harmonic filter design and various system parameters that affect filter performance.

What do harmonic filters look like? The Photo shows the input harmonic filter of a large UPS. The green arrow highlights one of the three separate inductors, and the yellow arrow highlights the associated capacitors. Why use three separate inductors rather than a single common core? The answer lies in flux density developed in the steel of the core. With a common core (EI lamination), it's difficult to achieve consistent flux density in all three legs. Because inductance in a steel core inductor varies with the permeability of the steel, the rate of change of the current flowing through the inductor, and the resulting flux density, it's easier to achieve and maintain consistent filter performance with three separate components.

What should the current through the filter look like? Hopefully, the frequency of the current will be close to the harmonic that it is intended to address, and the magnitude of current in each leg/phase will be very close. Figure 1 shows harmonic currents flowing through an UPS input harmonic filter: Phase A is 89Arms; Phase B is 86Arms; Phase C is 87Arms; and the filter current frequency is 660Hz (11th harmonic). This filter is working correctly.

Any good power monitor with harmonic analysis capability can perform these measurements. Fourier analysis of the waveform data is also very important. The monitor should be set to record the current flowing through each phase of the filter. If desired, you can use voltage channels to examine the voltage developing across individual inductors and capacitors as well as the voltage across both.

Filter design variations and performance results

What happens if a harmonic filter design uses a very small value of inductance along with a very large value of capacitance (inductors are expensive and bulky, whereas capacitors are smaller and cheaper). The following equation describes the resulting tank circuit:

hr = 1÷2ϖ√(Lcircuit Cfilter)

Potentially, we are back to the same unwanted resonance problems that existed with simple shunt capacitance.

Filter designers usually design for common applications, acknowledging that 5% impedance is fairly common for service entrance electric utility transformers. However, sites with source impedance (Z) much greater than 5% may experience more problems because the higher the source impedance — and the larger the filter capacitance — the worse the problems become. To help address this potential scenario, designers commonly add a blocking inductance. (click here to see Fig. 2) shows the use of a blocking inductance to lower upstream frequency response characteristics.

Figure 3 (click here to see Fig. 3) depicts a UPS application in a facility, and Fig. 4 (click here to see Fig. 4)on page 18 shows a plot of an unintended, momentary resonance that occurs in the facility, despite the best design efforts. The load is a 500kVA UPS that has an input harmonic filter with added blocking inductance to detune the filter against upstream harmonic sources. However, the combination of facility inductance creates a new tank circuit that is excited by the actuations of electric utility PF correction capacitors. The voltage waveform in Fig.4 shows one very severe event recorded at the UPS. The transient voltage reaches nearly 1,400Vpk and lasts milliseconds. This type of event severely strains and damages UPS input components.

Not only is the harmonic filter acting like a sink, but the circuitry also tends to amplify the transient voltages. At the moment the electric utility capacitor bank excites the transient resonant tank, the reactive (harmonic) voltages developed across all of the various inductances equals the reactive (harmonic) voltages developed across the capacitors. Where the voltages developed across each inductive element are individually small, the collective voltage developed across the capacitive elements are concentrated at a single point. As a result, it becomes very large and damaging.

There are several possible solutions to this problem, such as moving the electric utility bank farther away or increasing the ability of the UPS to sustain these events. The easiest solution in this case was to disengage the harmonic filter. There still is a transient voltage, but the magnitude drops from 1,400Vpeak to roughly 700Vpeak.