SCSI-2 socket bevel http://people.csail.mit.edu/jaffer/CNS/HWV

1997 Outbreak of Virus Infecting .050 100-pin D-connectors


An article in Electronics Australia entitled "A Hardware Virus?" poses the question:
I am only glad the virus was contained and did not spread to the rest of the world! Can you imagine if this sort of thing happened in a big computer assembly plant?
Such an outbreak did occur at ColorAge, a small manufacturer of print servers for color-laser-copiers, circa 1997.

For interfacing cables between our boards and the color-laser-copiers we used 100-pin models (AMP 749111-8/787169-9) in the same family of connectors as SCSI-2 uses. Because the balanced (2-wire) signaling the copiers use is somewhat redundant, the virus caused system function to degrade in multiple ways.

Ailments would be seen to accumulate as a board was tested with more copiers: unstable edges of light colored regions in printed pictures; color channels would become posterized; loss of horizontal or vertical sync; and serial communication between the copier and our board would become erratic or fail entirely.

To determine whether the problem was limited to a recent production problem versus component aging affecting all boards, we scoured our plant for old boards; which then also caught the malady. Our sharp-eyed technician finally noticed the connector damage and resolved the mystery.

When a male connector with a bent pin was mated to the female, it would, if pressed hard enough, create its own channel through the female's plastic block. That bent pin would either contact a neighboring socket contact or be open circuit.

That distorted plastic could then guide a straight pin to the wrong position, bending it in the process. Straightening the bent pins weakened them and the deformed plastic could not be repaired; as with animal epidemics, the infected connectors had to be destroyed and replaced.

Why would someone push hard enough to deform the connector?

100 pins have a lot of friction. Moreover, additional force is required the first time a female connector is mated. The backshell comes with jackscrews which are easily capable of driving several pins through the plastic.

Is it exaggeration to call this a virus?

Because the spontaneous origination of a virus is a rare event to observe, the particulars of this case deserve more scrutiny.

How did the pin get bent in the first place?

The 100-pin connector is over 6.cm long. If it is not pulled straight apart, the pins at one end will still be in contact with the female. If the cable is then tugged, it will bend those pins at the end outward.

The female sockets have a bevel down into their centers. So this first bend of the pins will probably be corrected by the bevel. But the bent pins will have more friction in their sockets (than unbent pins), making that end tend to stick at separation, bending the pins further. Eventually, the pins bend beyond the bevel and the virus is born.

This syndrome could be seen as a failure of the connector design to scale to spans over a couple centimeters. With their shorter lever arms, shorter connectors will experience less torque, and will be less likely to bend pins when uncoupling. Also, with the smaller number of pins, the insertion force is less; making the resistance to mating easier to discern when a pin is bent.

Copyright © 2004 Aubrey Jaffer

I am a guest and not a member of the MIT Computer Science and Artificial Intelligence Laboratory.  My actions and comments do not reflect in any way on MIT.
Computer Natural Science
agj @ alum.mit.edu
Go Figure!