Report details engineering's role in GM Cobalt switch defect

GM received a report on Thursday critical of its handling of ignition-switch problems in Chevy Cobalt and related vehicles. With a slight nudge, the ignition switch could move from the run to accessory position, inhibiting power steering and braking and disabling airbag deployment. Thus far, GM has linked 13 deaths and 54 crashes to rolling stalls caused by the defective switches.

There is plenty of blame to go around, according to the report, prepared by Anton R. Valukas of Jenner & Block. The New York Times offers an overview of the failure of GM's lawyers. “The automaker’s legal department took actions that obscured the deadly flaw, both inside and outside the company,” the Times writes. And a reading of the original report (a redacted version has been posted on the NHTSA site) paints a fascinating picture of the role of engineering, and in particular the role of GM engineer Raymond DeGiorgio, in the design and production of the switch.

The story began in 1997 when GM engineers designed what they called a discrete logic ignition switch (DLIS), which communicates switch position (such as crank, run, accessory, and off) to a body control module. The switch was supposed to be more reliable and less costly than existing switches, and it was to serve as a common switch that could be used across vehicle platforms. GM selected Eaton Corp., subsequently acquired by Delphi in 2001, to make the switch.

The original target specification for the tactile characteristics of the switch called for a torque of 20 N-cm to turn the switch from run to accessory.

DeGiorgio, then the design release engineer for the switch, in March 2001 finalized the switch specification, despite Eaton representatives having previously said the switch was not meeting specified torque values. At this point DeGiorgio specified a tolerance of ±5 N-cm on the original 20-N-cm specification yet ultimately accepted a switch that fell below the 15-N-cm acceptable lower limit.

And in February 2002, a Delphi product engineer signaled that the accessory detent torque was as low as 9.5 N-cm. The Delphi engineer said a fix could cause cracking of rotors and premature detent wear-out, and it could compromise electrical functions. DeGiorgio responded that if increasing the detent torque “will destroy this switch” then do nothing. “Under no circumstances do we want to compromise the electrical performance of this switch nor PPAP [production part approval process] status.” He referred to the device as “the switch from hell.”

“Ultimately,” says the report, “with the knowledge and approval of DeGiorgio, the ignition switch went into production despite having a torque value well below the requirements set forth in the specification.” The report notes that the fact that an out-of-spec switch had been used was not known to engineers investigating problems with the switch for over a decade.

Problems with the switch soon became apparent when it was implemented in the Saturn Ion, but no-crank/no-start problems overshadowed intermittent stalling. As an interesting aside, the report says the no-crank problems were related to cold temperatures and were hard to diagnose at the comparatively warm Spring Hills test facility. Once the problem was identified with the help of a refrigerated trailer, the switch was modified and received a new part number.

The report says the no-crank/no-start problem was an embarrassment for DeGiorgio, and he and other engineers failed to focus on the rolling-stall problem, which had different causes that weren't resolved by the redesign that addressed the no-start problem.

In 2004 GM began the Cobalt vehicle production program, using the switch designed for the Ion. The report says GM would have produced 80 to 90 captured-test-fleet (CTF) vehicles test-driven by GM employees. Although GM personnel interviewed by the report's authors recalled no reports of CTF problems, in fact program quality manager Joseph Taylor personally experienced a moving stall due to the ignition switch being inadvertently turned off. Taylor and others did not consider rolling stalls to constitute a safety issue but rather a convenience issue. The problem was also experienced by at least one automotive journalist reviewing the new car.

In addition, in November 2004 engineers in GM's High Performance Vehicle Operations group reported to DeGiorgio that the Cobalt SS had experienced stalls on the test track. An engineer in the group forwarded a complaint to DeGiorgio asking if the switch was meeting spec and what the options were—questions that DeGiorgio did not directly respond to. Draft responses suggested that there would be no easy fix for ensuring adequate key-cylinder retention forces during high-g maneuvers.

Earlier in the year, in May, and not related to the Cobalt, GM had presented a demonstration to NHTSA officials that was intended to show that rolling stalls and attendant loss of driver-assist capabilities did not constitute a safety issue. Notes from that period show that NHTSA indicated that manufacturers would not be immunized from conducting a safety recall to address rolling stall problems.

In March 2005 engineers presented possible solutions to the Cobalt rolling stall problem, including changing the key head from a slot design to a hole design—minimizing the torque other items on a key ring could apply to the switch via the key. Another option was to change the position of the switch on the steering column. No action was taken. The report cites “…a troubling disavowal of responsibility made possible by a proliferation of committees.” (Ultimately the hole design was adopted as an interim solution.)

The report states that in late 2005 (months after bad press and customer complaints), DeGiorgio discussed with Delphi putting a stronger spring and plunger into the switch. Delphi produced a new version of the switch with a new PCB design (which addressed the still present though reduced no-crank/no-start problem) and a new detent spring plunger. DeGiorgio approved the new design in April 2006, and it went into production sometime after June 26, 2006.

Inexplicably, the redesigned switch retained the part number of the old switch. The report notes that had others within GM been aware of the redesign of the switch, they could have more easily tied safety problems to the old switches already in service. The change became widely known only when a plaintiff's expert X-rayed switches from early and late model years and determined they were different.

Meanwhile in 2006, litigation related to fatalities had begun. In 2007 investigations by a Wisconsin state trooper and Indiana University researchers showed that movement of a switch from run to accessory would disable the airbag deployment—”the correct answer that GM engineers did not accept until January 2014,” the report says.

“As a whole,” the report states, “from beginning to end, the story of the Cobalt is one of numerous failures leading to tragic results for many.”