The Singing Telephone Switch

I worked for a company that designed and manufactured electronic telephone switches for landline phones. Some years ago our technical support department was contacted by a customer who he was getting complaints from his customers about a singing or whistling sound on the telephone line. The company made a number of attempts to locate the source of the noise but couldn't pin it down. They asked for help from the design department, so I was assigned to the task.

We set up a conference call with the customer and I started asking a few questions. The first thing I asked was how many customers were experiencing the problem. I didn't expect the answer I got -- all 3,000 of them! Not only that, but the local tech said you don't need to be on the phone to hear it -- just walk into the equipment room and it is audible. This was obviously not a problem that could be solved over the phone, so we set up a field visit.

When I got to the site, I couldn't believe my ears. As soon as I walked in the door I heard this piercing whistle. I hooked an oscilloscope between the frame ground and the 48V DC power bus and observed a 3Khz sine wave. It appeared that the entire system, which consisted of 12 racks of equipment, had turned itself into an audio oscillator.

After a few minutes of talking to the local tech, I ascertained that the switch had been going through an expansion in the number of lines. It was during this time that the problem started to occur. It was intermittent at first but eventually became permanent. This gave me a clue as to where to start looking, but I was baffled as to what the cause might be.

I started removing line cards that had not been assigned to customers, and sure enough, eventually I got the oscillation to stop. I could start it again by reinserting the cards. The curious thing was that it didn't matter which cards I removed. It appeared that there was a critical number of line cards that would cause the system to go unstable. Then another piece of the puzzle fell into place: the original line cards in the switch were electromechanical with a relay-driven line interface.

The new ones were the more recent electronic line interface cards. I tried getting below the critical number by removing electromechanical cards. The oscillation continued. So somehow the electronic line interface cards were causing the instability. But they had been in production several years, and we had never experienced this before.

Still baffled, I kept trying combinations by pulling and inserting cards to try and identify a pattern. Maybe it was line circuits that served a particular area or were assigned to the same cable. Nothing came up until the tech decided to power down one of the line frames. There are two power battery feeds to the line frames, both -48V DC.

One battery supplies the DC/DC converters to generate the 5V for the electronics; the other feeds the line circuits out to the customers. He pulled the fuse to the line circuits first and the oscillation immediately stopped. Another clue! But it got even better than that. There were two large 15A cartridge fuses for the line feed. That's because the line feed is routed through an LC filter at the top of the frame to supply all of the line circuits in the frame.