Engineers (hopefully) learn a lesson about how an unspecified, but critical parameter can eventually jump up and bite you.
By Larry Coyle, Contributing Writer
I was acting as a consultant designing electronic products for a company who manufactured home entertainment gear. On one occasion, I was asked to look at a problem that had cropped up in one of their old product lines. The item in question was a fairly conventional audio amplifier based on an old design. They had been selling these for years, and parts procurement and manufacturing had been farmed out to a Pacific Rim outfit.
It seemed that after several years into production, customers began returning the amps, complaining of “noise.” I got my hands on one of the troubled units, set myself up in a small test lab and hooked up a couple of speakers. I turned on the power switch and ran some music through the amplifier. I’m not a “golden ears” guy, but it sounded OK to me. I turned the music off and ran the loudness control up and down. Same with the tone controls. Still nothing happened that would lead anybody to return the unit.
I reported this back to the technical manager of the department and was told: “Leave the power switch on, don’t feed in any signal and wait.” I went back to my setup and followed these instructions. After several minutes of silence — a pistol shot rang out! At least, that’s what it sounded like. I could only imagine the effect on customers who tried to lull themselves to sleep with some soothing music, but had gone to bed without turning the power switch off.
And so I started in to work. Several hours with an oscilloscope and ear plugs led to the unexpected conclusion that the problem seemed to originate in the output stage.
The answer turned out to be surprisingly simple. The original designers, in an attempt to add some marketing pizzazz, included an “auto power down” feature. A threshold comparator monitored the audio level present in the amplifier, and after several minutes of no signal, released a good old electromechanical relay which turned off the power to the system’s output stages. The output stage was a push-pull AB circuit driving the speakers directly. Separate plus-and-minus supply rails powered the two output transistors. The aforementioned relay, a double-pole model, disconnected the two supply rails when de-energized.
All well and good, as long as the two poles of the relay released at nearly the same time. But, apparently, the offshore manufacturer had switched relay vendors without notifying anyone. The newer relays had a lot more “slop,” and there was a large difference in the release time of the two poles. Accordingly, there could be several tens of milliseconds when one side of the output stage was powered on and the other was not. This unbalance momentarily put the full supply rail across the speaker with the resulting gunshot effect.
While the answer was simple, a good solution was not. The best we could do for the irate customers was to squeeze some large capacitors onto the supply rails on all the returned units. This had the effect of reducing the gunshot noise to a dull thud, which upper management — in an attempt to avoid any refunds — deemed acceptable.
My consulting gig ended shortly after this debacle, so I never knew how it all worked out. I believe they just dropped this model from their product line. I hope the engineering staff absorbed the lesson on how an unspecified but critical parameter can eventually jump up and bite you. I know I did.
Contributing Writer Larry Coyle holds an MSEE degree and is now retired from an active career which began and ended as a freelance consultant in the field of consumer electronics. In between, he earned his living designing electronics for scientific instrumentation and space applications.
Tweet This
29 
