Bad crimp, bad news
Recently, I attended a presentation on connector reliability and what happens when connectors are incorrectly used. J. Duncan Glover, PhD, gave the presentation to the Boston Chapter of the IEEE Reliability Society. Given my recent blogs on how contact resistance can cause heating, and on dishwasher fires, this presentation came at just the right time.
Glover, a consulting engineer specializing in electrical failures, cited a case he investigated. Some heating/cooling units manufactured by Goodman Co. of Houston, Tex., would occasionally catch fire, something that started in December 2007. The units were typically installed in motels and apartment complexes.
After several fires had occurred, Goodman engineers believed that they were caused by faulty power cords. The heating/cooling units had been in production for several years with no fires prior to Goodman switching to power cords made by a company called Tower in 2007. A recall, started August 27, 2008, resulted in the Goodman company replacing the faulty cords –- with other faulty cords manufactured by Tower. But many owners did know about the recall, and the defective power cords remained in service.
Fires appeared to start at the unit's control board, but the boards were too badly burned for Glover to tell where the fire actually started. Was the fire caused by the board, or by the power cord? Figure 1 shows a photo of a burned control board, manufactured by Everex, shown during Glover's presentation.
The clues came from inspecting heating/cooling units that had not started fires. All of the fires started in the winter, when the heating/cooling units were drawing the most current. Glover discovered that many of the units had damaged power-cord connectors and wires at the point where they were inserted into tabs on the control board.
These units had some damage, but still functioned. Figure 2 shows one such partial damage: The insulation on the power cord's black wire is pulled back from overheating. On other units, the same black wires were severed, but no fire occurred.
The power cords were rated for 20A. Engineers at Everex tested the cords made by Tower with 40A and found failures. The flag connector, made by AMP (now TE Connectivity), heated and failed. Tests on power cords from the previous supplier turned up no failures. The investigations went on.
The Exerex control board was UL listed and the AMP connector for the black wire in question was properly specified. Still, Tower's lawyers accused Everex of faulty boards, but documentation showed that the boards did in fact meet UL standard 873 for construction and spacing. Although the spacing between the faulty wire and other wires was just 2mm, it met UL 873 because the adjacent wires didn't carry voltage of opposite polarity to that on the failed black wire.
The investigation moved to the connector, where Glover and others found that improper crimping was the root cause of the problem.
Figure 3 shows a cross section of two crimps. The connector on the right was properly crimped. The wires are packed closely together in the connector. The left cross section shows an improper crimp. Spacing between the wires resulted in higher contact resistance, which caused heating, which caused even higher resistance, which caused thermal runaway, and eventually, fire.
The flag connector's specifications for crimping clearly specify the maximum height of the resulting crimp. The cross section on the left in Figure 3 exceeds that height. Something was wrong in the crimping at the time of assembly in China. Not enough force was applied to the connectors during power-cord manufacturing because Tower's factory wasn't using the specified AMP crimping machine - they were using a knockoff.
Tower changed to AMP crimping machines and the crimps then met specifications. Problem solved. Goodman replaced Tower as its power-cord supplier in 2008.