June 26th, 2009

ronin

So you think you're smart enough to design a switch-mode power supply?


Read this and weep.

I like this one particularly:

The first power supply I ever designed for production was a transistor series regulator. Out of the first 100 fielded, 18 failed within three months. The problem was a path through semiconductor junctions that depended on back-biased junctions for impedance. In an all-parts change, a 10 ohm resistor was added into the collector of a transistor to break up the path. With several thousand units in the field for the ten year life of the product, there was never another failure.

This rather traumatic experience motivated me to become an expert in reliable power supply design. I hope to share much of what I have learned throughout this hypertext. This was one of my first and most important lessons - learned the hard way. Needless to say, looking for low impedance circuit paths through semiconductors is one of the first thing I look for in a design or design review. It has saved myself and many others much grief.

Besides damaging circuit components, inadvertently forward-biased junctions can cause system malfunctions. I received a letter of commendation for solving this problem with a hydrofoil patrol boat. The boat would crash down off the hydrofoils onto the hull anytime the crew transmitted a message or flushed the toilet. The problem with the transmitter was that the high-frequency-ac radiated signal was picked up and rectified by a diode to a dc voltage. This voltage appeared across the gate of an SCR in the hydrofoil control system protection circuit, activating the circuit. The toilet flushing problem was an unsuppressed inductive kick when the pump motor flushing the toilet turned off, thereby activating the control system protections circuit. Desensitizing the control system protection circuit solved both problems. The toilet pump motor was also interesting. An aircraft type toilet, it was suppressed assuming the grounding system used in aircraft, not ships. Adding suppression techniques suitable for ships also solved the problem with the toilet pump motor. The lesson learned? Normally back-biased junctions can be turned on by the environment and you have to protect against the consequences of this happening.


http://www.smpstech.com/junction.htm


See also, tutorial on switch-mode PSes at the same site.