Many years ago, I provided field engineering for a medical ultrasound company in the mid-Atlantic area. One day, I received a call to go to New York City to repair a system. I wondered why the company would pay me to go to an area that had several qualified field engineers. I'm always up for a good challenge, so I headed to NYC.
When I entered the room, I was immediately pounced on by the doctor and the technician. I asked the doctor to have a cup of coffee, calm down, and explain the situation. The doctor said he loved the machine, but this problem was unacceptable, since there was no way to predict a failure. It may work correctly for several hours, or it may fail every few minutes. The doctor was ready to throw the machine out the window and did not want the local engineers back in his office. There was no way to predict the on time.
The symptoms described didn't seem to be related to time, time of day, or other environmental influence. When it shut down, patients with full bladders would have to be rescheduled, and basically everyone associated with the machine was inconvenienced. Other well-qualified engineers were never there when the problem occurred and had monitored the machine for hours at a time. They had promised results, replaced components, and declared victory, but they were unable to solve this intermittent problem.
I started the machine to wait for the failure. Sure enough, after 20 minutes, the machine shut off. This was a large machine in a wooden cabinet that had an access door in the rear almost large enough to get inside. So I started looking for some obvious reason for the failure. I thought that, since the entire machine appeared to be off, the problem must be related to the main power supply, which fed all areas and subsystems of the machine. I was on my knees halfway inside the cabinet when I stuck my hand in a wiring jumble to get a closer view of the main PCB. Next, I accidentally touched the 1,500V CRT supply on the PCB, which was still on. I instinctively jerked my hand back and raised my head right into the corner of a large, sharp heat sink. At that exact instant, the machine turned on. The doctor screamed, "That's it! Whatever you did solved the problem!" As I was rubbing my head and trying to contain the small amount of blood, I explained that I may have found the area of the problem and not the specific cause. So I shut off the machine and restarted it.
After the machine was operational for 10 minutes, sure enough, it shut off again. But before I could get a meter or scope to start looking for clues, it turned back on. This time, to test the area of interest, I used a small hammer to tap the heat sink, instead of my head, and the machine turned off. The heat sink contained only a single TO-3 device that was the pass element for the main power supply. If the main supply was off, everything shut down. I decided to remove this heat sink and get a closer look. There were only three wires to the heat sink, and they all seemed to be soldered properly to a TO-3 socket. There was nothing obviously wrong. I powered the machine again and waited. About 30 minutes later, I saw an arc on the TO-3 socket, and the machine died. Closer inspection of the socket in the area of the spark indicated a high resistance connection between the socket and the transistor pin that was heating, and when it heated enough, it would expand and break contact with the pin. I forced the connection with pliers, and the machine came back on. Since I didn't have a new socket with me and time was of the essence, I soldered both of the pins directly to the socket to stop this problem.
It took me two hours to diagnose and repair the problem to the satisfaction of the doctor. That evening, he bought me a fine dinner, including an expensive bottle of wine. I feel sorry for the other engineers who babysat this machine for hours at a time to get a clue where the problem was. I guess they weren't using their heads to solve the problem.
Michael E Tompkins is vice president of technology development with Hanger.
Tell us your experience in solving a knotty engineering problem. Send stories to Lauren Muskett for Sherlock Ohms.