Once upon a time, I had to design a power supply that would operate from as little as 12VDC and as much as 240VAC, all at the same power input. There were no separate inputs for the low voltage DC/high voltage AC, so input voltage went to a bridge rectifier and a filter capacitor. Output voltage was specified as 6VDC at 10 to 100mA. We had a discrete component design already, but there was concern over parts going obsolete. Also, the marketing guys wanted to extend the input operating voltage a bit lower and, while we were at it, take a little of the cost out of the design.
I wanted to be the good soldier here, so I said, “Sure, I can do that.” I figured with the input voltage maximum of 240VAC (340VDC at the output of the bridge rectifier and filter capacitor) we’d want to use a buck switcher. That’s a stepdown ratio of about 1.8 percent. Child’s play. What could go wrong with such a design? Well, as it turns out, plenty. Keep track of your scoring at home to see if you can spot all the traps I laid for myself.
John Betten and Robert Kollman from Texas Instruments published a design idea almost 10 years ago that described a simple buck regulator that could operate from a pretty wide input voltage range. I assumed that if it could operate across the published range, it could probably be tweaked a bit to operate over the range that I needed.
Here’s their circuit:
I was sure I knew at least as much as (and maybe more than) these guys about power supply design, so I’d just make my own hysteretic controller to use in place of the TL5001. That would give me more flexibility over supply voltage limitations of their controller IC.
I used a section of an LM393 comparator, a few Zener diodes, and some discrete passive components. I built up a prototype, and on the bench it worked pretty well. I passed the schematic and BOM to my PC board layout guy. We made PC boards, stuffed them, and bench tested a few. I did a quick visual inspection of the board — all seemed well. Then we went into a large-scale production run, and things started going poorly.
This circuit suffered from several problems based on my oversights. The primary problem is the huge stepdown ratio. At the high-line input voltage, the on-time of the switching FET is so small that turn-on and turn-off time of the FET become significant. At low output current draw, it gets worse. Unit-to-unit variations, ambient temperature, and electrical noise muck things up.
I should have run a simulation and done a sensitivity study (intentionally varying component values to see what would go wrong). I probably should not have assumed I knew more than the guys who had been doing switching regulators for years. I might also have considered a voltage mode PWM controller with more predictable response and more controllable loop parameters. Finally, a two-stage design might have been better in this application.
The icing on the cake though was that we had what we thought were 1.0k-Ohm resistors. They were mislabeled: They were actually 100k. Other than that, what could go wrong?