Yogi Berra, the New York Yankees’ legendary ­catcher, has been credited with many unusual phrases. Perhaps the most well-known came during the 1973 pennant race: “It ain’t over till it’s over,” which applies to the continuing saga of the tripping AFCI breaker I described in my October column (see “Follow the Current”).

The tripping breaker was initially attributed to a faulty outdoor sensor spotlight. That conclusion proved incorrect a week later when the breaker tripped after the light had been replaced. The discredited observation of a flickering light bulb in a ceiling fan fixture also turned out to not be a coincidence.

More investigating is needed

When the breaker tripped again, it was time to determine if there was a fault in the wiring or in a device plugged into that circuit. Anything easily unplugged was unplugged. The majority of the previous knob-and-tube wiring had been replaced, with accessible junction boxes in the attic used for every three or four runs to outlet boxes. Each of these was opened and examined for loose connections or any damage indicative of arcing. With none found, the cables in the attic and basement were examined, also with no problems.

The electrical panel had an adjacent breaker on the same side, which was a 20A AFCI. The wires to each breaker were swapped to see if the problem followed the breaker or the circuit. After a week, the problem reappeared when the newly connected breaker tripped. While it could have been a systematic problem with that model of AFCI breaker, it was more likely a problem in the circuit.

The occasional trips were only noticed in the evening or at night. The only other equipment on that circuit that was temporal were the ceiling fan with the flickering light and the overhead kitchen light, both of which were turned on in the evening. When it occurred while the kitchen light was off, the ceiling fan with three LED light bulbs became the prime suspect, especially after two bulbs began flickering. Checking the wiring to the fan and its switch found nothing once again. It was time to bring out the PQ monitor.

PQ monitor to the rescue

A simple test fixture was made to test the candelabra-shaped 12W LED bulbs with E12 base, recording the current and voltage waveforms when the bulb was flickering. Graph A above shows a highly distorted current, which also affected the voltage. Graph B is from a nonflickering bulb and is typical of findings for single-phase loads with a rectifier, capacitor and switching regulator circuit. 

The base of the defective bulb where the electronic components were housed was taken apart. The regulator integrated circuit marking was visible, which made it possible to look up the data sheet. The data sheet included a design circuit that almost matched what was traced out on the printed circuit board. But several highly recommended components were missing, most likely to save on costs.

The current waveform of the defective bulb has a signature not unlike an arcing waveform. The RMS current was twice the level of a good bulb, but still under a quarter of an amp, which one wouldn’t think was enough to trip the breaker. 

However, something was amiss, as the actual current being drawn was high enough to significantly distort the voltage waveform. Spectrum analysis of the distortion showed many interharmonic signal components, not just harmonic components as with the good bulb. The current waveform of the good bulb has little, if any, effect on the voltage.

This points out another lesson and a possible Yogi-ism, if he ever dabbled in the PQ realm. Just because your instrument shows it, doesn’t mean it’s really like that. The problem is likely a limitation attributed to Nyquist’s Theorem. The instrument had a sampling rate of 15 kHz, which allows signals up to 7.5 kHz to be accurately measured. Signal components with frequencies above that are either filtered out by the instrument or “folded back,” where the higher-frequency components show up as lower-frequency components. 

Without a high-speed sampling instrument, the exact frequency spectrum and amplitudes can’t be determined. However, the random pattern of distortion on the voltage waveform is not unlike arcing, so a logical conclusion is that this voltage and current waveforms could cause an AFCI breaker to trip. This was reinforced when the LED bulbs were replaced with CFLs, and the problem has not occurred for a month since (and no flickering).

To the reader who asked if the smoking gun was found, it seems more likely now that it was the defective LED bulbs and not the motion-sensor light that caused the trips. This reminds us not to dismiss any clues until the investigation is complete, which this one wasn’t until the true spectrum of the waveforms was analyzed.

Richard P. Bingham